1: \documentclass[11pt,leqno]{article}

2: \usepackage{amsthm,amsfonts,amssymb,epsfig,graphics,amsmath,

3: eufrak,oldgerm}

4: \usepackage[latin1]{inputenc}\relax

5: \usepackage[active]{srcltx}

6: 

7: \newcommand{\BB}{{\mathbb B}}

8: \newcommand{\CC}{{\mathbb C}}

9: \newcommand{\DD}{{\mathbb D}}

10: \newcommand{\EE }{{\mathbb E}}

11: \newcommand{\FF}{{\mathbb F}}

12: \newcommand{\GG}{{\mathbb G}}

13: \newcommand{\HH}{{\mathbb H}}

14: \newcommand{\II}{{\mathbb I}}

15: \newcommand{\JJ}{{\mathbb J}}

16: \newcommand{\KK}{{\mathbb K}}

17: \newcommand{\LL}{{\mathbb L}}

18: \newcommand{\MM}{{\mathbb M}}

19: \newcommand{\NN}{{\mathbb N}}

20: \newcommand{\PP}{{\mathbb P}}

21: \newcommand{\QQ}{{\mathbb Q}}

22: \newcommand{\RR}{{\mathbb R}}

23: \newcommand{\SSS}{{\mathbb S}}

24: \newcommand{\TT}{{\mathbb T}}

25: \newcommand{\XX}{{\mathbb X}}

26: \newcommand{\ZZ}{{\mathbb Z}}

27: 

28: 

29: \newcommand{\DDelta}{\text{\underline{$ \Delta $}}}

30: 

31: \def\a{{\rm\bf a}}

32: \def\cc{{\rm\bf c}}

33: \def\f{{\rm\bf f}}

34: \def\g{{\rm\bf g}}

35: \def\h{{\rm\bf h}}

36: \def\p{{\rm\bf p}}

37: \def\r{{\rm\bf r}}

38: \def\u{{\rm\bf u}}

39: \def\v{{\rm\bf v}}

40: \def\w{{\rm\bf w}}

41: 

42: \def\uu{\text {\rm \bf \c{u}}}

43: \def\pp{\text {\rm \bf \c{p}}}

44: \def\pb{\text {\rm \d{p}}}

45: 

46: \def\cA{\mathcal{A}}

47: \def\cB{\mathcal{B}}

48: \def\cC{\mathcal{C}}

49: \def\cD{\mathcal{D}}

50: \def\cE{\mathcal{E}}

51: \def\cF{\mathcal{F}}

52: \def\cG{\mathcal{G}}

53: \def\cH{\mathcal{H}}

54: \def\cI{\mathcal{I}}

55: \def\cJ{\mathcal{J}}

56: \def\cK{\mathcal{K}}

57: \def\cL{\mathcal{L}}

58: \def\cM{\mathcal{M}}

59: \def\cN{\mathcal{N}}

60: \def\cO{\mathcal{O}}

61: \def\cP{\mathcal{P}}

62: \def\cQ{\mathcal{Q}}

63: \def\cR{\mathcal{R}}

64: \def\cS{\mathcal{S}}

65: \def\cT{\mathcal{T}}

66: \def\cU{\mathcal{U}}

67: \def\cV{\mathcal{V}}

68: \def\cW{\mathcal{W}}

69: \def\cX{\mathcal{X}}

70: \def\cY{\mathcal{Y}}

71: \def\cZ{\mathcal{Z}}

72: 

73: 

74: \def\mA{\mathfrak{A}}

75: \def\mB{\mathfrak{B}}

76: \def\mC{\mathfrak{C}}

77: \def\mD{\mathfrak{D}}

78: \def\mE{\mathfrak{E}}

79: \def\mF{\mathfrak{F}}

80: \def\mG{\mathfrak{G}}

81: \def\mH{\mathfrak{H}}

82: \def\mI{\mathfrak{I}}

83: \def\mJ{\mathfrak{J}}

84: \def\mK{\mathfrak{K}}

85: \def\mL{\mathfrak{L}}

86: \def\mM{\mathfrak{M}}

87: \def\mN{\mathfrak{N}}

88: \def\mO{\mathfrak{O}}

89: \def\mP{\mathfrak{P}}

90: \def\mQ{\mathfrak{Q}}

91: \def\mR{\mathfrak{R}}

92: \def\mS{\mathfrak{S}}

93: \def\mT{\mathfrak{T}}

94: \def\mU{\mathfrak{U}}

95: \def\mV{\mathfrak{V}}

96: \def\mW{\mathfrak{W}}

97: \def\mX{\mathfrak{X}}

98: \def\mY{\mathfrak{Y}}

99: \def\mZ{\mathfrak{Z}}

100: 

101: 

102: \def\ma{\mathfrak{a}}

103: \def\mb{\mathfrak{b}}

104: \def\mc{\mathfrak{c}}

105: \def\md{\mathfrak{d}}

106: \def\me{\mathfrak{e}}

107: \def\mf{\mathfrak{f}}

108: \def\mg{\mathfrak{g}}

109: \def\mh{\mathfrak{h}}

110: \def\mi{\mathfrak{i}}

111: \def\mj{\mathfrak{j}}

112: \def\mk{\mathfrak{k}}

113: \def\ml{\mathfrak{l}}

114: \def\mm{\mathfrak{m}}

115: \def\mn{\mathfrak{n}}

116: \def\mo{\mathfrak{o}}

117: \def\mp{\mathfrak{p}}

118: \def\mq{\mathfrak{q}}

119: \def\mr{\mathfrak{r}}

120: \def\ms{\mathfrak{s}}

121: \def\mt{\mathfrak{t}}

122: \def\mu{\mathfrak{u}}

123: \def\mv{\mathfrak{v}}

124: \def\mw{\mathfrak{w}}

125: \def\mx{\mathfrak{x}}

126: \def\my{\mathfrak{y}}

127: \def\mz{\mathfrak{z}}

128: 

129: \def\Eq#1#2{\mathop{\sim}\limits_{#1 \, \rightarrow \, #2}}

130: \def\Tend#1#2{\mathop{\longrightarrow}\limits_{#1 \, \longrightarrow \, #2}}

131: \def\eps {\varepsilon}

132: \def\epsi {\epsilon}

133: \def\part{\partial}

134: \def\un{\underline}

135: \def\u{{\bf u}}

136: \def\v{{\bf v}}

137: \def\w{{\bf w}}

138: \def\p{{\bf p}}

139: \def\r{{\bf r}}

140: \def\s{{\bf s}}

141: \def\X{{\bf X}}

142: 

143: 

144: \def\eu{\text{\rm \small u}}

145: \def\ew{\text{\rm \small w}}

146: 

147: \def\Dive{\md \mi \mv}

148: \def\Grade{\mg \mr \ma \md}

149: 

150: \def\iu{\textsc{\large u}}

151: \def\iv{\textsc{\large v}}

152: \def\iw{\textsc{\large w}}

153: 

154: \renewcommand{\contentsname}{Contents}

155: 

156: \newcommand\id{\mathrm{Id} }

157: \newcommand\Div{\mathrm{div }}

158: \newcommand\Grad{\mathrm{grad }}

159: 

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

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

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

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

164: \newtheorem{defi}{Definition} [section]

165: \newtheorem{ass}{Assumption} [section]

166: \numberwithin{equation}{section}

167: 

168: \newenvironment{preuve}{\noindent {\sl{Preuve.}}}{

169: \begin{flushright} $\Box$ \end{flushright} }

170: 

171: %\usepackage{showkeys} 

172: 

173: 

174: \begin{document} 

175: 

176: \centerline{\LARGE \bf Cascade of phases in turbulent flows}

177: 

178: \bigskip

179: 

180: \centerline{\sc Christophe CHEVERRY

181: \footnote{\footnotesize IRMAR, Universit\'e de Rennes I, Campus de 

182: Beaulieu, 35042 Rennes cedex, France, \text{ }\text{ }\text{ }\text{ }

183: christophe.cheverry@univ-rennes1.fr.}}

184: 

185: 

186: \bigskip

187: \bigskip

188: %----------------- 

189: %------------------ 

190: 

191:  

192: \noindent{\bf \small Abstract.} {\small This article is 

193: devoted to incompressible Euler equations (or to 

194: Navier-Stokes equations in the vanishing viscosity limit). 

195: It describes the propagation of {\it quasi-singularities}. 

196: The underlying phenomena are consistent with the notion 

197: of a {\it cascade of energy}.}

198: 

199: \bigskip

200: 

201: \noindent{\bf \small R\'esum\'e.} {\small Cet article 

202: \'etudie les \'equations d'Euler incompressible (ou de

203: Navier-Stokes en pr\'esence de viscosit\'e \'evanescente).

204: On y d\'ecrit la propagation de {\it quasi-singularit\'es}. 

205: Les ph\'enom\`enes sous-jacents confirment l'id\'ee selon 

206: laquelle il se produit une {\it cascade d'energie}.}

207: 

208: \bigskip

209: 

210: \setcounter{section}{0}

211: 

212: \section{Introduction.} $ \, $

213: 

214: \vskip -5mm

215: 

216: Consider incompressible fluid equations

217: 

218: \medskip

219: 

220: \noindent{$ (\cE) \qquad \part_t \u + (\u \cdot \nabla ) \, \u + 

221: \nabla \p = 0 \, , \qquad \Div \ \u = 0 \, , \qquad (t,x) \in 

222: [0,T] \times \RR^d \, , $}

223: 

224: \medskip

225: 

226: \noindent{where $ \u = {}^t ( \u^1 , \cdots , \u^d) $ is the fluid 

227: velocity  and $ \p $ is the pressure. The structure of {\it weak} 

228: solutions of $ (\cE) $ in $ d - $space dimensions with $ d \geq 2 $ 

229: is a problem of wide current interest \cite{Chem}-\cite{Li}. 

230: The questions are how to describe the phenomena with adequate 

231: models and how to visualize the results in spite of their

232: complexity. We will achieve a small step in these two 

233: directions. 

234: 

235: \medskip

236: 

237: According to the physical intuition, the appearance of 

238: singularities is linked with the {\it increase of the vorticity}.

239: Along this line, we have to mark the contributions \cite{BKM} 

240: and \cite{CF}. Interesting objects are solutions which do

241: not blow up in finite time but whose associated vorticities

242: increase arbitrarily fast. These are {\it quasi-singularities}.

243: Their study is of practical importance.

244: 

245: \medskip

246: 

247: Typical examples of quasi-singularities are oscillations.

248: This is a well-known fact going back to \cite{C}-\cite{MPP}.

249: The works \cite{C} and \cite{MPP} rely on phenomenological 

250: considerations and engineering experiments. Further 

251: developments are related to homogenization \cite{E}-\cite{E2}, 

252: compensated compactness \cite{D}-\cite{Ge} and non linear 

253: geometric optics \cite{Che}-\cite{CGM}-\cite{CGM1}. 

254: 

255: \smallskip

256: 

257: $\, $

258: 

259: \break

260: 

261: $ \, $

262: 

263: In \cite{D}, DiPerna and Majda show the persistence of 

264: oscillations in three dimensional Euler equations. To 

265: this end, they select a parameter $ \eps \in \, ]0,1] $ 

266: and look at

267: \begin{equation} \label{oscidipmaj}

268: \ \u^\eps_s (t,x) \, := \, {}^t \bigl( \g ( x_2 , \eps^{-1} 

269: \, x_2) , 0 , \h \bigl( x_1 - \g ( x_2 , \eps^{-1} \, x_2) 

270: \, t , x_2, \eps^{-1} \, x_2 \bigr) \bigr) \quad

271: \end{equation}

272: where $ \g (x_2,\theta) $ and $ \h (x_1,x_2,\theta) $ 

273: are smooth bounded functions with period $ 1 $ in 

274: $ \theta $. They remark that the functions $ \u^\eps_s $ 

275: are exact smooth solutions of $ (\cE) $ and they let $ \eps $

276: goes to zero. Yet, this construction is of a very special 

277: form. First, it comes from shear layers (steady 2-D solutions) as

278: 

279: \medskip

280: 

281: $ \tilde \u^\eps_s (t,x) \, = \, \tilde \u^\eps_s (0,x) \, 

282: = \, {}^t \bigl( \g ( x_2,\eps^{-1} \, x_2) , 0 \bigr) \, 

283: \in \, \RR^2 \, . $

284: 

285: \medskip

286: 

287: \noindent{Secondly, it involves a phase $ \varphi_0 (t,x) 

288: \equiv x_2 $ which does not depend on $ \eps $.} Of course,

289: this is a common fact \cite{CG}-\cite{G}-\cite{G2}-\cite{Se}

290: when dealing with large amplitude high frequency waves.

291: Nevertheless, this is far from giving a complete idea of what 

292: can happen.

293: 

294: \medskip

295: 

296: Our aim in this paper is to develop a theory which allows to 

297: remove the two restrictions mentioned above. Fix $ \flat = 

298: (l,N) \in \NN^2 $ where the integers $ l $ and $ N $ are such 

299: that $ 0 < l < N $. Introduce the {\it geometrical} phase

300: 

301: \medskip

302: 

303: $ \varphi^\eps_g (t,x) \, := \, \varphi_0 (t,x) \, + \, \sum_{k=1}^{

304: l-1} \, \eps^{\frac{k}{l}} \ \varphi_k (t,x) \, . $

305: 

306: \medskip

307: 

308: \noindent{In the section 2, we state the Theorem \ref{appBKW} which

309: provides with {\it approximate} solutions $ \u^\eps_\flat $ defined 

310: on the interval $ [0,T]  $ with $ T > 0 $ and having the form}

311: \begin{equation} \label{formegenerale}

312: \begin{array} {ll}

313: \u^\eps_\flat (t,x) \! \! \! & = \, {}^t ( \u^{\eps 1}_\flat, 

314: \cdots , \u^{\eps d}_\flat)(t,x) \\

315: \ & = \, \u_0 (t,x) \,  + \, \sum_{k=1}^N \, \eps^{

316: \frac{k}{l}} \ U_k \bigl( t,x, \eps^{-1} \, \varphi^\eps_g (t,x) 

317: \bigr) \qquad \qquad

318: \end{array}

319: \end{equation}

320: where the smooth profiles

321: 

322: \medskip

323: 

324: $ U_k (t,x,\theta) \, = \, {}^t (U^1_k, \cdots, U^d_k) 

325: (t,x,\theta) \in \RR^d \, , \qquad 1 \leq k \leq N \, , $

326: 

327: \medskip

328: 

329: \noindent{are periodic functions of $ \theta \in \RR /

330: \ZZ $.} We assume that 

331: 

332: \medskip

333: 

334: $ \exists \, (t,x,\theta) \in [0,T] \times \RR^d \times 

335: \TT \, ; \qquad \part_\theta U_1 (t,x,\theta) \not = 0 $.

336: 

337: \medskip

338: 

339: \noindent{We say that the family $ \{ \u^\eps_\flat \}_\eps $ 

340: is a {\it weak}, a {\it strong} or a {\it turbulent} 

341: oscillation according as we have respectively $ l=1 $, 

342: $ l = 2 $ or $ l \geq 3 $.}

343: 

344: \medskip

345: 

346: The order of magnitude of the energy of the oscillations 

347: is $ \eps^{\frac{1}{l}} $. Compute the vorticities associated with the 

348: functions $ \u^\eps_\flat $. These are the skew-symmetric 

349: matrices $ \Omega^\eps_\flat = (\Omega^{\eps i}_{\flat j} 

350: )_{1 \leq i,j \leq d} $ where

351: $$ \quad \begin{array} {l}

352: \Omega^{\eps i}_{\flat j} (t,x) \, := \, (\part_j 

353: \u^{\eps i}_\flat - \part_i \u^{\eps j}_\flat)(t,x) \\

354: \ = \, \sum_{k=1}^N \, \eps^{\frac{k}{l}-1} \ ( \part_j 

355: \varphi^\eps_g \ \part_\theta U^i_k - \part_i \varphi^\eps_g \ 

356: \part_\theta U^j_k) \bigl( t,x, \eps^{-1} \, \varphi^\eps_g (t,x) 

357: \bigr) \\

358: \quad \ + \, ( \part_j \u^i_0  - \part_i \u^j_0)(t,x) \, + \, 

359: \sum_{k=1}^N \, \eps^{\frac{k}{l}} \ ( \part_j U^i_k  - 

360: \part_i U^j_k ) \bigl( t,x, \eps^{-1} \, \varphi^\eps_g 

361: (t,x) \bigr) \, .  

362: \end{array} $$

363: The principal term in $ \Omega^\eps_\flat $ is of size 

364: $ \eps^{\frac{1}{l}-1}$. When $ l \geq 2 $, no uniform 

365: estimates are available on the family $ \{ \Omega^\eps_\flat 

366: \}_{\eps \in \, ]0,1] } $. In particular, if $ d = 3 $, there 

367: is no uniform control on the enstrophy

368: 

369: \medskip

370: 

371: $ \int_0^T \int_{\RR^3} \ \vert \omega^\eps_\flat (t,x) \vert^2 \ \, 

372: dt \, dx \, , \qquad \omega^\eps_\flat (t,x) \, := \, (\nabla \wedge 

373: \u^\eps_\flat)(t,x) \, \equiv \, \Omega^\eps_\flat (t,x) \, . $ 

374: 

375: \medskip

376: 

377: \noindent{We see here that strong and turbulent oscillations 

378: are examples of quasi-singularities.} Observe that the 

379: expansion (\ref{formegenerale}) involves a more complicated

380: structure than in (\ref{oscidipmaj}) though the corresponding

381: regime is less singular.

382: 

383: \medskip

384: 

385: The BKW analysis reveals that the phase shift $ \varphi_1 $

386: and the terms $ \varphi_k $ with $ 2 \leq k \leq l-1 $ play

387: different parts. The r\^ole of $ \varphi_1 $ is partly examined 

388: in the articles \cite{Che} and \cite{CGM} which deal with the 

389: case $ l = 2 $. When $ l \geq 3 $, the phenomenon to emphasize 

390: is the creation of the $ \varphi_k $ with $ 2 \leq k \leq l-1 $.

391: Indeed, suppose that

392: 

393: \medskip

394: 

395: $ \varphi_2 (0,\cdot) \equiv \, \cdots \, \equiv \varphi_{l-1} 

396: (0,\cdot) \equiv 0 \, , \qquad l \geq 3 \, . $

397: 

398: \medskip

399: 

400: \noindent{Then, generically, we find}

401: 

402: \medskip

403: 

404: $ \exists \, t \in \, ]0,T] \, ; \qquad \varphi_2 (t,\cdot) 

405: \not \equiv 0 \, , \qquad \cdots \, , \qquad \varphi_{l-1} 

406: (t,\cdot) \not \equiv 0 \, . $

407: 

408: \medskip

409: 

410: \noindent{Now starting with {\it large} amplitude waves (this 

411: corresponds to the limit case $ l = + \infty $) that is}

412: 

413: \medskip

414: 

415: $ \u^\eps_\infty (0,x) \, = \, \sum_{k=0}^\infty \, 

416: \eps^k \ U_k \bigl( 0,x, \eps^{-1} \, \varphi_0 (0,x) 

417: \bigr) \, , \qquad  \part_\theta U_0 \not \equiv 0 \, , $

418: 

419: \medskip

420: 

421: \noindent{the description of $ \u^\eps_\infty (t,\cdot) $ on the 

422: interval $ [0,T] $ with $ T > 0 $ needs the introduction of an {\it

423: infinite cascade} of phases $ \varphi_k $.} The scenario is 

424: the following. Oscillations of the velocity develop spontaneously 

425: in all the intermediate frequencies $ \eps^{\frac{k}{l}-1} $ 

426: and in all the directions $ \nabla \varphi_k (t,x) $. This 

427: expresses {\it turbulent} features in the flow.

428: 

429: \medskip

430: 

431: The family $ \{ \u^\eps_\flat \}_{\eps \in \, ]0,1]} $

432: is $ \eps - $stratified \cite{G2} with respect to the phase

433: $ \varphi^\eps_g $ with in general $ \varphi^\eps_g \not 

434: \equiv \varphi_0 $. The presence in  $ \varphi^\eps_g $ of the

435: non trivial functions $ \varphi_k $ is necessary and sufficient 

436: to encompass the {\it geometrical} features of the propagation. 

437: It has various consequences which are detailed in the section 

438: 3. It brings informations about microstructures, compensated 

439: compactness and non linear geometric optics. It also confirms 

440: observations made in the statistical approach of turbulences 

441: \cite{FMRT}-\cite{L}.

442: 

443: \medskip

444: 

445: The chapter 4 is devoted to the demonstration of Theorem

446: \ref{appBKW}. Because of {\it closure problems}, the use of 

447: the geometrical phase $ \varphi^\eps_g $ does not suffice to 

448: perform the BKW analysis. Among other things,  {\it adjusting 

449: phases} $ \varphi_k $ with $ l \leq k \leq N $ must be 

450: incorporated in order to put the system of formal equations 

451: in a triangular form.

452: 

453: \medskip

454: 

455: The expressions $ \u^\eps_\flat $ are not exact solutions of

456: Euler equations, yielding small error terms $ \f^\eps_\flat $ 

457: as source terms. The matter is to know if there exists exact 

458: solutions which coincide with $ \u^\eps_\flat (0,\cdot) $ at 

459: time $ t = 0 $, which are defined on $ [0,T] $ with $ T > 0 $, 

460: and which are close to the approximate divergence free solutions 

461: $ \u^\eps_\flat $. This is a problem of {\it stability}.

462: 

463: \medskip

464: 

465: \noindent{The construction of exact solutions requires a good 

466: understanding of the different mechanisms of amplifications 

467: which occur.} In the subsection 5.1, we make a distinction 

468: between {\it  obvious} and {\it hidden} instabilities.

469: 

470: \medskip

471: 

472: \noindent{The obvious instabilities can be detected by 

473: looking at the BKW analysis presented before.} They

474: imply the non linear instability of Euler equations

475: (Proposition \ref{euinstap}). They need to be absorbed 

476: a dependent change of variables which induces a defect 

477: of hyperbolicity. The hidden instabilities can be revealed 

478: by soliciting this lack of hyperbolicity. They require to 

479: be controled the addition of dissipation terms.

480: 

481: \medskip

482: 

483: \noindent{In the subsection 5.2, we look at incompressible fluids 

484: with anisotropic viscosity.} This is the framework of \cite{CDGG}

485: though we adopt a different point of view. We consider strong 

486: oscillations. We show (Theorem \ref{ciprin}) that {\it exact}

487: solutions corresponding to $ \u^\eps_{(2,N)} $ exist on some 

488: interval $ [0,T] $ with $ T > 0 $ independent on $ \eps \in 

489: \, ]0,1] $. 

490: 

491: \bigskip

492: \smallskip

493: 

494:  

495: {\small \parskip=-3pt

496: \tableofcontents

497: }

498: 

499: 

500: \break

501: 

502: \section{Euler equations in the variables $ (t,x) $.} $\,$

503: 

504: \vskip -5mm

505: 

506: The description of incompressible flows in turbulent regime 

507: is a delicate question. No systematic analysis is yet available. 

508: However, special appro- ximate solutions with rapidly varying 

509: structure in space and time can be exhibited. Their construction

510: is summarized in this chapter 2.

511: 

512: \subsection{Notations.}

513: 

514: \smallskip

515:   

516: \noindent{$ \bullet $ {\em Variables.}} Let $ T \in \RR^+_* $. The time 

517: variable is $ t \in [0,T] $. Let $ d \in \NN \setminus \{ 0 ,

518: 1 \} $. The space variables are $ (x, \theta) \in \RR^d \times

519: \TT $ where $ \TT := \RR / \ZZ $. Mark the ball

520: 

521: \medskip

522: 

523: $ B(0,R] \, := \, \bigl \{ \, x \in \RR^d \, ; \ \vert x \vert^2

524: := \sum_{i=1}^d \, x_i^{\, 2} \leq R \, \bigr \} \, , \qquad R \in 

525: \RR^+ \, . $ 

526: 

527: \medskip

528: 

529: \noindent{The state variables are the velocity field $ u = {}^t 

530: (u^1,\cdots,u^d) \in \RR^d $ and the pressure $ p \in \RR $. 

531: Given $ (u , \tilde u ) \in (\RR^d)^2 $, define

532: 

533: \medskip

534: 

535: $ u \cdot \tilde u := \sum_{i=1}^d \, u^i \, \tilde u^i \, , 

536: \qquad \vert u \vert^2 := u \cdot u \, , \qquad u \otimes \tilde 

537: u := (u^j \, \tilde u^i )_{1 \leq i,j \leq d} \, . $

538: 

539: \medskip

540: 

541: \noindent{The symbol $ S^d_+ $ is for the set of positive definite 

542: quadratic form on $ \RR^d $.} An element $ \mq \in S^d_+ $ can be 

543: represented by some $ d \times d $ matrix $ ( \mq_{ij} )_{1 \leq 

544: i,j \leq d} $.

545: 

546: \bigskip

547: 

548: \noindent{$ \bullet $ {\em Functional spaces.}} Distinguish the 

549: expressions $ \u (t,x) $ which do not depend on the variable 

550: $ \theta $ from the expressions $ u(t,x,\theta) $ which depend on 

551: $ \theta $. The boldfaced type $ \u $ is used in the first

552: case whereas the letter $ u $ is employed in the second 

553: situation. 

554: 

555: \medskip

556: 

557: \noindent{Note $ C^\infty_b ( [0,T] \times \RR^d ) $ the space 

558: of functions in $ [0,T] \times \RR^d $ with bounded continuous 

559: derivatives of any order.} Let $ m \in \NN $. The Sobolev space 

560: $ H^m $ is the set of functions 

561: 

562: \medskip

563: 

564: $ u(x,\theta) \, = \, \sum_{k \in \ZZ} \, \u_k (x) \ e^{

565: i \, k \, \theta} $ 

566: 

567: \medskip

568: 

569: \noindent{such that}

570: 

571: \medskip

572: 

573: $ \parallel u \parallel_{H^m}^2 \, := \, \sum_{k \in \ZZ} \, (1 +

574: \vert k \vert^2)^m \ \int_{\RR^d} \, (1 + \vert \xi \vert^2)^m \

575: \vert \hat \u_k (\xi) \vert^2 \ d \xi \, < \, \infty $

576: 

577: \medskip

578: 

579: \noindent{where} 

580: 

581: \medskip

582: 

583: $ \cF ( \u )(\xi) \, = \, \hat \u (\xi) \, := \, ( 2 \, \pi)^{- 

584: \frac{d}{2}} \ \int_{\RR^d} \, e^{- i \, x \cdot \xi} \ \u (x) \ 

585: dx \, , \qquad \xi \in \RR^d \, . $

586: 

587: \medskip

588: 

589: \noindent{With these conventions, the condition $ \u \in H^m $ 

590: means simply that}

591: 

592: \medskip

593: 

594: $ \parallel \u \parallel_{H^m}^2 \, := \, \int_{\RR^d} \, (1 + 

595: \vert \xi \vert^2)^m \ \vert \hat \u (\xi) \vert^2 \ d \xi \, 

596: < \, \infty \, . $

597: 

598: \medskip

599: 

600: \noindent{Define}

601: 

602: \medskip

603: 

604: $ H^m_T \, := \, \bigl \{ \, u \ ; \ \part^j_t u \in L^2 ( [0,T];

605: H^{m-j}) \, , \ \forall \, j \in \{0, \cdots,m \} \, \bigr \} \, , $

606: 

607: \medskip

608: 

609: $ \cW^m_T \, := \, \bigl \{ \, u \ ; \ u \in C^j ( [0,T]; H^{m-j}) \, , 

610: \ \forall \, j \in \{0, \cdots,m \} \, \bigr \} \, , $

611: 

612: \medskip

613: 

614: \noindent{with the corresponding norms}

615: 

616: \medskip

617: 

618: $ \parallel u \parallel_{H^m_T}^2 \, := \, \sum_{j=0}^m \, \int_0^T \,

619: \parallel \part^j_t u (t,\cdot) \parallel_{H^m}^2 \ dt \, , $

620: 

621: \medskip

622: 

623: $ \parallel u \parallel_{\cW^m_T} \, := \, \sup_{t \in [0,T]} \ \ 

624: \sum_{j=0}^m \, \parallel \part^j_t u (t,\cdot) \parallel_{H^m} \, . $

625: 

626: \medskip

627: 

628: \noindent{Consider also }

629: $$ \ \left. \begin{array} {lll}

630: H^m_\infty \, := \, \bigcap_{T \in \RR^+} \, H^m_T \, , & \quad 

631: H^\infty_T \, := \, \bigcap_{m \in \NN} \, H^m_T \, , & \quad 

632: H^\infty_\infty \, := \, \bigcap_{T \in \RR^+} \, H^\infty_T \, , \\ 

633: \cW^m_\infty \, := \, \bigcap_{T \in \RR^+} \, \cW^m_T \, , & \quad  

634: \cW^\infty_T \, := \, \bigcap_{m \in \NN} \, \cW^m_T \, , & \quad 

635: \cW^\infty_\infty \, := \, \bigcap_{T \in \RR^+} \, \cW^\infty_T \, . 

636: \end{array} \right. $$

637: When $ m = 0 $, replace $ H^0 $ with $ L^2 $. Any function $ u 

638: \in L^2 $ can be decomposed according to

639: 

640: \medskip

641: 

642: $ u (t,x,\theta) \, = \, \langle u \rangle (t,x) + u^* (t,x,\theta) \, 

643: = \, \bar u (t,x) + u^* (t,x,\theta) $

644: 

645: \medskip

646: 

647: \noindent{where}

648: 

649: \medskip

650: 

651: $ \langle u \rangle (t,x) \, \equiv \, \bar u (t,x) \, := \, \int_\TT \, 

652: u(t,x,\theta) \ d \theta \, . $ 

653: 

654: \medskip

655: 

656: \noindent{Let $ \Gamma $ be the symbol of any of the spaces $  H^m $, 

657: $ H^m_T $, $ \cW^m_T $, $ \cdots $ defined before.} In order to

658: specify the functions with mean value zero, introduce

659: 

660: \medskip

661: 

662: $ \Gamma^* \, := \, \lbrace \, u \in \Gamma \, ; \ \bar u \equiv 0 \, 

663: \rbrace \, . $

664: 

665: \medskip

666: 

667: \noindent{Mark also}

668: 

669: \medskip

670: 

671: $ \text{supp}_x \, u^* \, := \, \text{closure of} \ \bigl \{ \, x 

672: \in \RR^d \, ; \ \parallel u^* (x, \cdot) \parallel_{L^2(\TT)} 

673: \not = \, 0 \, \bigr \} \, . $

674: 

675: \bigskip

676: 

677: \noindent{$ \bullet $ {\em Differential operators.}} Note

678: 

679: \medskip

680: 

681: $ \part_t \equiv \part_0 := \part / \part \, t \, , \qquad \ \

682: \part_\theta \equiv \part_{d+1} := \part / \part \, \theta \, , $

683: 

684: \smallskip

685: 

686: $ \part_j := \part / \part \, x_j \, , \qquad \qquad \ j 

687: \in \{1 , \cdots , d \} \, , $

688: 

689: \smallskip

690: 

691: $ \nabla := (\part_1, \cdots ,\part_d) \, , \qquad \Delta := 

692: \Delta_x + \part^2_\theta = \part^{\, 2}_1 + \cdots + \part^{\, 

693: 2}_{d} + \part^2_\theta \, . $

694: 

695: \medskip

696: 

697: \noindent{Let $ u \in \cW^\infty_T $.} Define

698: 

699: \medskip

700: 

701: $ u \cdot \nabla \, := \, u^1 \ \part_1 + \cdots + u^d \ \part_d \, , $

702: 

703: \smallskip

704: 

705: $ \Div \ u \, := \, \part_1 u^1 + \cdots + \part_d u^d \, , $

706: 

707: \smallskip

708: 

709: $ \Div \ (u \otimes \tilde u) \, := \, \sum_{j=1}^d \, {}^t \bigl( 

710: \part_j ( u^j \, \tilde u^1) \, , \cdots , \part_j ( u^j \, \tilde 

711: u^d) \bigr) \in \RR^d \, . $

712: 

713: \medskip

714: 

715: \noindent{Employ the bracket $ < \cdot , \cdot >_H $ for the scalar 

716: product in the Hilbert space $ H $. Note $ \cL (E;F) $ the space of 

717: linear continuous applications $ T : E \longrightarrow F $ where $ E $ 

718: and $ F $ are Banach spaces. The symbol $ \cL (E) $ is simply for 

719: $\cL (E;E) $. Introduce the commutator

720: 

721: \medskip

722: 

723: $ [A;B] \, := \, A \circ B - B \circ A \, , \qquad (A,B) \in \cL(E)^2 \, . $

724: 

725: \medskip

726: 

727: \noindent{Let $ r \in \ZZ $.} The operator $ T $ is in $ \mL^r $ if

728: 

729: \medskip

730: 

731: $ \parallel T \parallel_{\cL (H^{m+r}_T ; H^m_T)} \, < \, \infty \, , 

732: \qquad \forall \, m \in \NN \, . $

733: 

734: \medskip

735: 

736: \noindent{Let $ \eps_0 > 0 $.} The family of operators $ \{ T^\eps 

737: \}_\eps \in \cL (H^\infty_T)^{]0,\eps_0]} $ is in $ \mU \mL^r $ if 

738: 

739: \medskip

740: 

741: $ \sup_{\eps \in \, ]0,\eps_0]} \ \ \parallel T^\eps \parallel_{

742: \cL (H^{m+r}_T ; H^m_T)} \, < \, \infty \, , \qquad \forall \, m 

743: \in \NN \, . $

744: 

745: \medskip

746: 

747: \noindent{Consider a family $ \{ f^\eps \}_\eps \in (\cW^\infty_T)^{

748: ]0,\eps_0]} $. We say that $ \{ f^\eps \}_\eps = \bigcirc (\eps^r ) $ 

749: if

750: 

751: \medskip

752: 

753: $ \sup_{\eps \in \, ]0,\eps_0]} \ \ \eps^{-r} \ \parallel f^\eps 

754: \parallel_{\cW^m_T} \, < \, \infty \, , \qquad \forall \, m \in 

755: \NN \, . $

756: 

757: \medskip

758: 

759: \noindent{Given a family $ \{ \f^\eps \}_\eps \in (\cW^\infty_T)^{

760: ]0,\eps_0]} $, we say that $ \{ \f^\eps \}_\eps = \bigcirc (\eps^r ) $ 

761: if

762: 

763: \medskip

764: 

765: $ \sup_{\eps \in \, ]0,\eps_0]} \ \ \eps^{-r+m} \ \parallel \f^\eps 

766: \parallel_{\cW^m_T} \, < \, \infty \, , \qquad \forall \, m \in 

767: \NN \, . $

768: 

769: \medskip

770: 

771: \noindent{Observe that the two preceding definitions have very 

772: different significations according as we use the letter $ f $ 

773: or the boldfaced type $ \f $.} In particular, the second inequalities

774: correspond to $ \eps - \, $stratified estimates. The families 

775: $ \{ f^\eps \}_\eps $ or $ \{ \f^\eps \}_\eps $ are $ \bigcirc 

776: (\eps^\infty ) $ if they are $ \bigcirc (\eps^r ) $ for all 

777: $ r \in \RR $.

778: 

779: 

780: \subsection{Divergence free approximate solutions in $ (t,x) $.} 

781: 

782: \noindent{$ \bullet $ {\bf A first result.}} Select smooth functions

783: 

784: \medskip

785: 

786: $ \u_{00} \in H^\infty \, , \qquad \varphi_{00} \in C^1 (\RR^d ) \, ,

787: \qquad \nabla \varphi_{00} \in C^\infty_b (\RR^d ) \, . $

788: 

789: \medskip

790: 

791: \noindent{Suppose that} 

792: 

793: \medskip

794: 

795: $ \exists \, c > 0 \, ; \qquad \vert \nabla \varphi_{00} (x) 

796: \vert \, \geq \, 2 \ c \, , \qquad \forall \, x \in \RR^d \, . $

797: 

798: \medskip

799: 

800: \noindent{For $ T > 0 $ small enough, the equation $ (\cE) $

801: associated with}

802: 

803: \medskip

804: 

805: $ \u_0 (0,x) = \u_{00} (x) \, , \qquad \forall \, x \in \RR^d $}

806: 

807: \medskip

808: 

809: \noindent{has a smooth solution $ \u_0(t,x) \in \cW^\infty_T $.}

810: Solve the eiconal equation

811: 

812: \medskip

813: 

814: \noindent{$ (ei) \qquad \part_t \varphi_0 + (\u_0 \cdot \nabla) \, 

815: \varphi_0 = 0 \, , \qquad (t,x) \in [0,T] \times \RR^d $}

816: 

817: \medskip

818: 

819: \noindent{with the initial data}

820: 

821: \medskip

822: 

823: $ \varphi_0 (0,x) = \varphi_{00} (x) \, , \qquad \forall \, x \in 

824: \RR^d \, . $}

825: 

826: \medskip

827: 

828: \noindent{If necessary, restrict the time $ T $ in order to have}

829: \begin{equation} \label{nonstaou} 

830: \vert \nabla \varphi_0 (t,x) \vert \, \geq \, c \, , \qquad \forall \, 

831: (t,x) \in [0,T] \times \RR^d \, . \qquad \qquad \qquad

832: \end{equation}

833: Call $ \Pi_0 (t,x) $ the orthogonal projector from $ \RR^d $ 

834: onto the hyperplane

835: 

836: \medskip

837: 

838: $ \nabla \varphi_0 (t,x)^\perp \, := \, \bigl \lbrace \, u 

839: \in \RR^d \, ; \ u \cdot \nabla \varphi_0 (t,x) = 0 \, \bigr 

840: \rbrace \, .$}

841: 

842: \smallskip  

843: 

844: \begin{theo} \label{appBKW} Select any $ \flat = (l,N) \in \NN^2_* $ 

845: such that $ 0 < l \, (3 + \frac{d}{2}) \ll N $. Consider the following 

846: initial data

847: 

848: \medskip

849: 

850: $ U_{k0}^* (x,\theta) = \Pi_0 (0,x) \, U_{k0}^* (x,\theta) \in 

851: H^\infty \, , \qquad  1 \leq k \leq N \, , $

852: 

853: \smallskip

854: 

855: $ \bar U_{k0} (x) \in H^\infty \, , \qquad 1 \leq k \leq 

856: N \, , $

857: 

858: \smallskip

859: 

860: $ \varphi_{k0} (x) \in H^\infty \, , \qquad 1 \leq k \leq l-1 \, . $

861: 

862: \medskip

863: 

864: \noindent{First, there are finite sequences $ \{ U_k \}_{1 \leq k \leq N} $ 

865: and $ \{ P_k \}_{1 \leq k \leq N} $ with}

866: 

867: \medskip

868: 

869: $ U_k (t,x,\theta) \in \cW^\infty_T \, , \qquad P_k (t,x,\theta) \in 

870: \cW^\infty_T \, , \qquad 1 \leq k \leq N \, , $

871: 

872: \medskip

873: 

874: \noindent{and a finite sequence $ \{ \varphi_k \}_{1 \leq k \leq l-1} $ 

875: with}

876: 

877: \medskip

878: 

879: $ \varphi_k (t,x) \in \cW^\infty_T \, , \qquad 1 \leq k \leq l-1 \, , $

880: 

881: \medskip

882: 

883: \noindent{which are such that}

884: 

885: \medskip

886: 

887: $ \Pi_0 (0,x) \, U_k^* (0,x,\theta) =  \Pi_0 (0,x) \, U_{k0}^*

888: (x,\theta) \, , \qquad 1 \leq k \leq N \, , $

889: 

890: \smallskip

891: 

892: $ \bar U_k (0,x) = \bar U_{k0} (x) \, , \qquad \! 1 \leq k \leq 

893: N \, , $

894: 

895: \smallskip

896: 

897: $ \varphi_k (0,x) = \varphi_{k0} (x) \, , \qquad \! 1 \leq k 

898: \leq l-1 \, . $ 

899: 

900: \medskip

901: 

902: \noindent{Secondly, there is $ \eps_0 \in \, ]0,1 ] $ and  

903: correctors}

904: 

905: \medskip

906: 

907: $ \cc \u^\eps_\flat (t,x) \in \cW^\infty_T \, , \qquad \cc \p^\eps_\flat 

908: (t,x) \in \cW^\infty_T \, , \qquad \eps \in \,  ]0,\eps_0] \, , $ 

909: 

910: \medskip

911: 

912: \noindent{which give rise to families satisfying}

913: 

914: \medskip

915: 

916: $ \{ \cc \u^\eps_\flat \}_\eps \, = \, \bigcirc(\eps^{\frac{N}{l}-2}) \, , 

917: \qquad \{ \cc \p^\eps_\flat \}_\eps \, = \, \bigcirc(\eps^{\frac{N}{l}}) 

918: \, . $ 

919: 

920: \medskip

921: 

922: \noindent{Then, all these expressions are adjusted so that the functions 

923: $ \u^\eps_\flat $ and $ \p^\eps_\flat $ defined according to}

924: \vskip -4mm

925: \begin{equation} \label{BKWdel}

926: \, \left. \begin{array}{l}

927: \u^\eps_\flat (t,x) := \u_0 (t,x) + \sum_{k=1}^N \, 

928: \eps^{\frac{k}{l}} \ U_k \bigl( t,x, \eps^{-1} \, \varphi^\eps_g 

929: (t,x) \bigr) + \cc \u^\eps_\flat (t,x) \ \, \\

930: \p^\eps_\flat (t,x) := \p_0 (t,x) + \sum_{k=1}^N \, 

931: \eps^{\frac{k}{l}} \ P_k \bigl( t,x, \eps^{-1} \, \varphi^\eps_g 

932: (t,x) \bigr) + \cc \p^\eps_\flat (t,x) \ \ \,

933: \end{array} \right.

934: \end{equation}

935: where $ \varphi^\eps_g (t,x) $ is the geometrical phase 

936: \vskip -4mm  

937: \begin{equation} \label{phasegeo} 

938: \varphi^\eps_g (t,x) \, := \, \varphi_0 (t,x) \, + \, \hbox{$ \sum_{k=1}^{l-1}$}

939: \ \eps^{\frac{k}{l}} \ \varphi_k (t,x) \qquad \qquad \qquad \ \

940: \end{equation}

941: are approximate solutions of $ (\cE) $ on the interval $ [0,T] $.

942: More precisely 

943: 

944: \medskip

945: 

946: $ \part_t \u^\eps_\flat + ( \u^\eps_\flat \cdot \nabla) \u^\eps_\flat 

947: + \nabla \p^\eps_\flat = \f^\eps_\flat \, , \qquad \Div \ \u^\eps_\flat 

948: = 0 \, , \qquad \f^\eps_\flat = \bigcirc(\eps^{\frac{N}{l}-3-\frac{d}

949: {2}}) \, . $

950: 

951: \end{theo}

952: 

953: \smallskip

954: 

955: \noindent{$ \bullet $ {\bf Some comments.}} 

956: 

957: \medskip

958: 

959: \noindent{{\it Remark 2.2.1:}} In what follows, we suppose that 

960: $ U^*_1 $ is non trivial.} In other words, we start with some 

961: initial data satisfying

962: \begin{equation} \label{nonnontri}

963: \ \exists \, (x,\theta) \in \RR^d \times \TT \, ; \qquad U^*_1

964: (0,x,\theta) = U^*_{10} (x,\theta) \not = 0 \, . \qquad \qquad 

965: \qquad \quad

966: \end{equation}

967: 

968: \vskip -9mm

969: 

970: \hfill $ \triangle $

971: 

972: \bigskip

973: 

974: \noindent{{\it Remark 2.2.2:}} Fix any $ l \in \NN_* $. The Borel's

975: summation process allows to take $ N = + \infty $ in the Theorem

976: \ref{appBKW}. It yields BKW solutions $ (\u^\eps_\flat , \p^\eps_\flat) $ 

977: which solve $ (\cE) $ with infinite accuracy

978: 

979: \medskip

980: 

981: $ \part_t \u^\eps_\flat + ( \u^\eps_\flat \cdot \nabla) \u^\eps_\flat + \nabla 

982: \p^\eps_\flat \, = \, \bigcirc(\eps^\infty) \, , \qquad \Div \ \u^\eps_\flat 

983: \, = \, 0 \, . $ \hfill $ \triangle $

984: 

985: \bigskip

986: 

987: \noindent{{\it Remark 2.2.3:}} Suppose that the function $ \u_0 \in 

988: \cW^\infty_\infty $ is a global solution of Euler equations. Suppose 

989: also that the phase $ \varphi_0 \in \cW^\infty_\infty $ is subjected 

990: to (\ref{nonstaou}) on the strip $ [0,\infty [ \times \RR^d $ and that 

991: it is a global solution of the eiconal equation $ (ei) $. Then 

992: the Theorem \ref{appBKW} can be applied with any $ T \in \RR^+_* $. It 

993: means that no blow up occurs at the level of the equations yielding the 

994: profiles $ U_k $, $ P_k $ and the phases $ \varphi_k $. Yet, non linear

995: effects are present. \hfill $ \triangle $

996: 

997: \bigskip

998: 

999: \noindent{{\it Remark 2.2.4:}} The characteristic curves

1000: of the field $ \part_t + \u_0 \cdot \nabla_x $ are obtained

1001: by solving the differential equation 

1002: 

1003: \medskip

1004: 

1005: $ \part_t \, \Gamma (t,x) \, = \, \u_0 \bigl( t, \Gamma (t,x)

1006: \bigr) \, , \qquad \Gamma (0,x) = x \, . $

1007: 

1008: \medskip

1009: 

1010: \noindent{Suppose that the oscillations of the profiles 

1011: $ U^*_{k0} $ are concentrated in some domain $ D \subset 

1012: \RR^d $.} In other words 

1013: 

1014: \medskip

1015: 

1016: $ \text{supp}_x \, U_{k0}^* \, \subset \, D \, , \qquad 

1017: \forall \, k \in \{ 1, \cdots , N \} \, . $

1018: 

1019: \medskip

1020: 

1021: \noindent{The BKW analysis reveals that for all $ t \in 

1022: [0,T] $ we have} 

1023: 

1024: \medskip

1025: 

1026: $ \text{supp}_x \, U_k^* (t,\cdot) \, \subset \, \bigl 

1027: \lbrace \, \Gamma (t,x) \, ; \ x \in D \, \bigr \rbrace 

1028: \, , \qquad \forall \, k \in \{ 1, \cdots , N \} \, . $

1029: 

1030: \medskip

1031: 

1032: \noindent{The phenomena under study have a finite speed of 

1033: propagation.} \hfill $ \triangle $

1034: 

1035: \bigskip

1036: 

1037: \noindent{{\it Remark 2.2.5:}} The influence of dissipation 

1038: terms will be taken into account in the subsection 4.1. The 

1039: viscosity we will incorporate is anisotropic. It is small enough

1040: in the direction $ \nabla \varphi^\eps_\flat $ in order to be 

1041: compatible with the propagation of oscillations.  \hfill $ \triangle $ 

1042: 

1043: \subsection{End of the proof of Theorem \ref{appBKW}.} 

1044: 

1045: The Theorem \ref{appBKW} is a consequence of the Proposition 

1046: \ref{appBKWinter} which will be stated and demonstrated in the 

1047: subsection 4.2. Below, we just explain how to deduce the Theorem 

1048: \ref{appBKW} from the Proposition \ref{appBKWinter} applied with 

1049: $ \nu = 0 $. 

1050: 

1051: \medskip

1052: 

1053: \noindent{$ \bullet $ {\bf Dictionary between the profiles}.} Select

1054: arbitrary initial data for

1055: 

1056: \medskip

1057: 

1058: $ \Pi_0 (0,x) \, \tilde U_k^* (0,x,\theta) \in H^\infty \, , \qquad 

1059: \langle \tilde U_k \rangle (0,x) \in H^\infty \, , \qquad 1 \leq k

1060: \leq N \, , $

1061: 

1062: \medskip

1063: 

1064: \noindent{and arbitrary initial data for}

1065: 

1066: \medskip

1067: 

1068: $ \varphi_k (0,x) \in H^\infty \, , \qquad  1 \leq k \leq l-1 \, . $

1069: 

1070: \medskip

1071: 

1072: \noindent{On the contrary, impose}

1073: \begin{equation} \label{inipreli}

1074: \varphi_k (0,\cdot) \equiv 0 \, , \qquad \forall \, k \in \{l,

1075: \cdots, N \} \, . \qquad \qquad \qquad \qquad \qquad

1076: \end{equation}

1077: The Proposition \ref{appBKWinter} provides with finite 

1078: sequences

1079: 

1080: \medskip 

1081: 

1082: $ \{ \tilde U_k \}_{1 \leq k \leq N} \, , \qquad \{ \tilde P_k \}_{

1083: 1 \leq k \leq N} \, , \qquad \{ \varphi_k \}_{1 \leq k \leq N} \, , $

1084: 

1085: \medskip

1086: 

1087: \noindent{and source terms}

1088: 

1089: \medskip

1090: 

1091: $ \tilde f^\eps_\flat (t,x,\theta) \in \cW^\infty_T \, , \qquad 

1092: \tilde g^\eps_\flat (t,x,\theta) \in \cW^\infty_T \, . $

1093: 

1094: \medskip

1095: 

1096: \noindent{such that the associated oscillations}

1097: $$ \left. \begin{array}{l}

1098: \tilde \u^\eps_\flat (t,x) \, := \, \u_0 (t,x) \, + \, 

1099: \sum_{k=1}^{N} \, \eps^{\frac{k}{l}} \ \tilde U_k \bigl( 

1100: t,x, \eps^{-1} \, \varphi^\eps_\flat (t,x) \bigr) \, , \\

1101: \tilde \p^\eps_\flat (t,x) \, := \, \p_0 (t,x) \, + \, 

1102: \sum_{k=1}^{N} \, \eps^{\frac{k}{l}} \ \tilde P_k \bigl(t,

1103: x, \eps^{-1} \, \varphi^\eps_\flat (t,x) \bigr) \, , \qquad 

1104: \qquad \qquad \ \\

1105: \tilde \f^\eps_\flat (t,x) \, := \, \eps^{-1} \ \tilde 

1106: f^\eps_\flat \bigl(t,x, \eps^{-1} \, \varphi^\eps_\flat 

1107: (t,x) \bigr) \, , \\

1108: \tilde \g^\eps_\flat (t,x) \, := \, \eps^{-1} \ \tilde 

1109: g^\eps_\flat \bigl(t,x, \eps^{-1} \, \varphi^\eps_\flat 

1110: (t,x) \bigr) \, ,

1111: \end{array} \right. $$

1112: are subjected to

1113: 

1114: \medskip

1115: 

1116: $ \part_t \tilde \u^\eps_\flat + ( \tilde \u^\eps_\flat \cdot \nabla) 

1117: \tilde \u^\eps_\flat + \nabla \tilde \p^\eps_\flat = \tilde \f^\eps_\flat

1118: = \bigcirc ( \eps^{\frac{N+1}{l}-1}) \, , \quad \ \Div \ \tilde \u^\eps_\flat 

1119: = \tilde \g^\eps_\flat = \bigcirc ( \eps^{\frac{N+1}{l}-1}) \, . $

1120: 

1121: \medskip

1122: 

1123: \noindent{The oscillations $ \tilde \u^\eps_\flat $ and $ \tilde \p^\eps_\flat $

1124: involve the {\it complete} phase $ \varphi^\eps_\flat (t,x) $ which is

1125: the sum of the geometrical phase $ \varphi^\eps_g (t,x) $ plus some 

1126: {\it adjusting} phase $ \eps \, \varphi^\eps_a (t,x) $. More precisely

1127: 

1128: \medskip

1129: 

1130: $ \varphi^\eps_\flat (t,x) := \varphi^\eps_g (t,x) + \eps \ 

1131: \varphi^\eps_a (t,x) \, , \qquad \varphi^\eps_a (t,x) := 

1132: \sum_{k=l}^N \, \eps^{\frac{k}{l}-1} \ \varphi_k (t,x) \, . $

1133: 

1134: \medskip

1135: 

1136: \noindent{The functions $ \tilde \u^\eps_\flat $ and $ \tilde 

1137: \p^\eps_\flat $ can also be written in terms of the phase 

1138: $ \varphi^\eps_g $.} Indeed, there is a unique decomposition

1139: 

1140: \medskip

1141: 

1142: $ \tilde \u^\eps_\flat = \u^\eps_\flat + \r \u^\eps_\flat =

1143: \u^\eps_\flat + \bigcirc (\eps^{\frac{N+1}{l}}) \, , \qquad 

1144: \tilde \p^\eps_\flat = \p^\eps_\flat + \r \p^\eps_\flat =

1145: \p^\eps_\flat + \bigcirc (\eps^{\frac{N+1}{l}}) \, , $

1146: 

1147: \medskip

1148: 

1149: \noindent{involving the representations}

1150: \begin{equation} \label{BKWdelbis}

1151: \ \u^\eps_\flat (t,x) = u^\eps_\flat \bigl( t,x, \eps^{-1} \, 

1152: \varphi^\eps_g (t,x) \bigr) \, , \qquad \p^\eps_\flat (t,x) = 

1153: p^\eps_\flat \bigl( t,x, \eps^{-1} \, \varphi^\eps_g (t,x) 

1154: \bigr) 

1155: \end{equation}

1156: where the profiles $ u^\eps_\flat ( t,x, \theta ) $ and 

1157: $ p^\eps_\flat ( t,x, \theta ) $ have the form

1158: $$ \left. \begin{array}{l}

1159: u^\eps_\flat ( t,x, \theta ) \, = \, \u_0 (t,x) + \sum_{k=1}^N \, 

1160: \eps^{\frac{k}{l}} \ U_k ( t,x, \theta) \, , \qquad \qquad \qquad 

1161: \qquad \qquad \ \\

1162: p^\eps_\flat ( t,x, \theta ) \, = \, \p_0 (t,x) + \sum_{k=1}^N \, 

1163: \eps^{\frac{k}{l}} \ P_k ( t,x, \theta) \, .

1164: \end{array} \right. $$

1165: The transition from $ \tilde \u^\eps_\flat $ to $ \u^\eps_\flat $ 

1166: is achieved through the phase shift $ \varphi^\eps_a $

1167: 

1168: \medskip

1169: 

1170: $ \tilde U_k ( t,x, \eps^{-1} \, \varphi^\eps_\flat ) \, = \, 

1171: \tilde U_k \bigl( t,x, \eps^{-1} \, \varphi^\eps_g + \varphi_l + 

1172: \sum_{k=l+1}^N \, \eps^{\frac{k}{l} - 1} \ \varphi_k \bigr) \, . $

1173: 

1174: \medskip

1175: 

1176: \noindent{Use the Taylor formula in order to absorb the small 

1177: term in the right.} It furnishes the following explicit link 

1178: between the $ (U_k,P_k) $ and the $ (\tilde U_k , \tilde P_k) $

1179: \begin{equation} \label{diction}

1180: \ \left. \begin{array}{l}

1181: U_k \bigl( t,x,\theta - \varphi_l (t,x) \bigr) \, := \, \tilde U_k 

1182: (t,x,\theta) \, + \, \cG^k (\tilde U_1, \cdots, \tilde U_{k-1}) (t,

1183: x,\theta) \, , \ \ \\

1184: P_k \bigl( t,x,\theta - \varphi_l (t,x) \bigr) \, := \, \tilde P_k 

1185: (t,x,\theta) \, + \, \cG^k (\tilde P_1, \cdots, \tilde P_{k-1}) (t,

1186: x,\theta) \, .

1187: \end{array} \right.

1188: \end{equation}

1189: The application $ \cG^k $ can be put in the form

1190: 

1191: \medskip

1192: 

1193: $ \cG^k (\tilde U_1, \cdots, \tilde U_{k-1}) \, := \, \sum_{p=1}^{k-1} \, 

1194: \partial_\theta^p \cG^k_p (\tilde U_1, \cdots, \tilde U_{k-p}) \, , 

1195: \qquad k \in \{ 1, \cdots , N \} \, .  $

1196: 

1197: \medskip

1198: 

1199: \noindent{The terms $ \cG^k_p $ are given by} 

1200: 

1201: \medskip

1202: 

1203: $ \cG^k_p (\tilde U_1, \cdots, \tilde U_{k-p}) \, := \, 

1204: \frac{1}{p \, !} \ \sum_{\alpha \in \cJ^k_p} \ \varphi_{

1205: l+1+\alpha_1} \times \cdots \times \varphi_{l+1+\alpha_p} 

1206: \ \tilde U_{\alpha_{p+1}} \, , $

1207: 

1208: \medskip

1209: 

1210: \noindent{where the sum is taken over the set}

1211: 

1212: \medskip

1213: 

1214: $ \cJ^k_p \, := \, \bigl \lbrace \, \alpha = (\alpha_1, 

1215: \cdots,\alpha_p, \alpha_{p+1}) \in \NN^{p+1} \, ; $

1216: 

1217: \smallskip

1218: 

1219: \hfill $ 0 \leq \alpha_j \leq N-l-1 \, , \qquad \forall \, 

1220: j \in \{1, \cdots, p \} \, , \qquad \qquad \quad \ \ \, $

1221: 

1222: \smallskip

1223: 

1224: \hfill $ 1 \leq \alpha_{p+1} \leq k-p \, , \qquad \alpha_1 

1225: + \cdots + \alpha_p + \alpha_{p+1} = k-p \, \bigr \rbrace 

1226: \, . $

1227: 

1228: \medskip

1229: 

1230: \noindent{The relation (\ref{diction}) and the definition of 

1231: $ \cG^k $ imply that}

1232: 

1233: \medskip

1234: 

1235: $ \bar U_k (t,x) \, = \, \langle \tilde U_k \rangle (t,x) \, , 

1236: \qquad \forall \, k \in \{ 1, \cdots , N \} \, , \qquad 

1237: \forall \, t \in [0,T] \, . $

1238: 

1239: \medskip

1240: 

1241: \noindent{Therefore, prescribing the initial data for the 

1242: $ \bar U_k $ or the $ \langle \tilde U_k \rangle $ amounts 

1243: to the same thing.} The condition (\ref{inipreli}) yields

1244: 

1245: \medskip

1246: 

1247: $ \cG^k_p (\tilde U_1, \cdots, \tilde U_{k-p}) (0,x,\theta) 

1248: = 0 \, , \qquad \forall \, k \in \{ 1, \cdots , N \} \, . $

1249: 

1250: \medskip

1251: 

1252: \noindent{Since $ \varphi_l (0,\cdot) \equiv 0 $, we have}

1253: 

1254: \medskip

1255: 

1256: $ \Pi_0 (0,x) \, U_k^* (0,x,\theta) \, = \, \Pi_0 (0,x) \, 

1257: \tilde U_k^* (0,x,\theta) \, , \qquad \forall \, k \in \{ 1, 

1258: \cdots , N \} \, . $

1259: 

1260: \medskip

1261: 

1262: \noindent{It is clearly equivalent to specify the initial 

1263: data for the $ \Pi_0 \, U_k^* $ or the $ \Pi_0 \, \tilde 

1264: U_k^* $.

1265: 

1266: \bigskip

1267: 

1268: \noindent{$ \bullet $ {\bf The divergence free relation in 

1269: the variables $ (t,x) $.}} Consider the application 

1270: 

1271: \medskip

1272: 

1273: $ \Div \, : \, H^\infty \, \longrightarrow \, \text{\rm Im} \, 

1274: (\Div) \, \subset \, \bigl \lbrace \, \g \in H^\infty \, ; \ \hat \g

1275: (0) = 0 \, \bigr \rbrace \, . $ 

1276: 

1277: \medskip

1278: 

1279: \noindent{We can select some special right inverse.}

1280: 

1281: \begin{lem} \label{invpartbar} There is a linear operator

1282: $ \text{\rm ridiv} \, : \, \text{\rm Im} \, (\Div) \longrightarrow 

1283: H^\infty $ with

1284: \begin{equation} \label{exainv}

1285: \Div \circ \text{\rm ridiv} \ \g \, = \, \g \, , \qquad \forall \, 

1286: \g \in \text{\rm Im} \, (\Div) \, . \qquad \qquad \qquad \quad \ 

1287: \end{equation}

1288: For all $ \iota > 0 $ and for all $ m \in \NN $, there is a 

1289: constant $ C_m^\iota > 0 $ such that

1290: \begin{equation} \label{estiinv}

1291: \parallel \text{\rm ridiv} \ \g \parallel_{H^m} \, \leq \, C_m \ 

1292: \parallel \g \parallel_{H^{m+1+ \frac{d}{2} + \iota}} \, , \qquad 

1293: \forall \, \g \in \text{\rm Im} \, (\Div) \, . \qquad

1294: \end{equation}

1295: 

1296: \end{lem}

1297: 

1298: \smallskip

1299: 

1300: \noindent{\em \underline{Proo}f \underline{o}f \underline{the} 

1301: \underline{Lemma} \underline{\ref{invpartbar}}.} Introduce a 

1302: cut-off function $ \psi \in C^\infty (\RR^d) $ such that

1303: 

1304: \medskip

1305: 

1306: $ \bigl \{ \, \xi \, ; \ \psi (\xi) \not = 0 \, \bigr \} \, 

1307: \subset \, B (0,2] \, , \qquad \bigl \{ \, \xi \, ; \ \psi 

1308: (\xi) = 1 \, \bigr \} \, \supset \, B (0,1] \, . $

1309: 

1310: \medskip

1311: 

1312: \noindent{For $ g \in \text{\rm Im} \, (\Div) $, take the 

1313: explicit formula}

1314: 

1315: \medskip

1316: 

1317: $ \text{ridiv} \, (\g) : = \cF^{-1} \, \bigl( \, \int_0^1 \, 

1318: \nabla_\xi  (\psi \, \hat \g) (r \, \xi) \ dr \, + \, \vert 

1319: \xi \vert^{-2} \ (1-\psi)(\xi) \ \hat \g (\xi) \times \xi \, 

1320: \bigr) \, . $

1321: 

1322: \medskip

1323:  

1324: \noindent{Since $ \hat \g (0) = 0 $, the relation (\ref{exainv}) 

1325: is satisfied.} For $ s > \frac{d}{2} $, the injection $ H^s 

1326: (\RR^d) \hookrightarrow L^\infty (\RR^d) $ is continuous. It

1327: leads to (\ref{estiinv}). \hfill $ \Diamond $

1328: 

1329: \bigskip

1330: 

1331: \noindent{$ \bullet $ {\bf The Leray projector in the variables $ (t,x) $.}} 

1332: Note $ \Pi (\xi) $ the orthogonal projector from $ \RR^d $ onto the plane 

1333: 

1334: \medskip

1335: 

1336: $ \xi^\perp \, := \, \lbrace \, u \in \RR^d \, ; \ u \cdot \xi 

1337: = 0 \, \rbrace \, . $

1338: 

1339: \medskip

1340:  

1341: \noindent{Introduce the closed subspace}

1342: 

1343: \medskip

1344: 

1345: $ \text{F} \, := \,  \bigl \lbrace \, \u \in L^2 \, ; 

1346: \ \Div \, \u = 0 \, \bigr \rbrace \, \subset \, L^2 \, . $

1347: 

1348: \medskip

1349: 

1350: \noindent{Call $ P $ the orthogonal projector from $ L^2 $ 

1351: onto F.} It corresponds to the Fourier multiplier

1352: 

1353: \medskip

1354: 

1355: $ P \, \u \, = \, \Pi (D_x) \, \u \, := \, ( 2 \, \pi)^{- 

1356: \frac{d}{2}} \ \int_{\RR^d} \, e^{i \, x \cdot \xi} \ \Pi 

1357: (\xi) \, \hat \u (\xi) \ d \xi \, . $

1358: 

1359: \medskip

1360: 

1361: \noindent{The application $ P $ is the Leray projector onto the 

1362: space of divergence free vector fields.} It is a self-adjoint 

1363: operator such that

1364: 

1365: \medskip

1366: 

1367: $ \ker \, \Div \, = \, \text{Im} \, P \, , \qquad \text{Im} \, 

1368: \nabla \, = \, \bigl( \ker \, (\Div) \bigr)^\perp \, = \, \ker \,  

1369: P \, . $

1370: 

1371: \medskip

1372: 

1373: \noindent{Consider the Cauchy problem}

1374: 

1375: \medskip

1376: 

1377: $ \partial_t \u + \nabla \p = \f \, , \qquad \Div \, \u = 0 \, , 

1378: \qquad \u (0, \cdot) = \h $

1379: 

1380: \medskip

1381: 

1382: \noindent{with data $ \f \in L^2_T $ and $ \h \in L^2 $.} 

1383: It leads to the equivalent conditions

1384: 

1385: \medskip

1386: 

1387: $ \partial_t \u = P \, \f \, , \qquad \u (0, \cdot) 

1388: = P \, \h \, , \qquad \nabla \p = (\id - P) \, \f \, . $

1389: 

1390: \bigskip

1391: 

1392: \noindent{Now we come back to the proof of Theorem \ref{appBKW}.} 

1393: It remains to absorb the term $ \tilde \g^\eps_\flat \in 

1394: \text{Im} \, (\Div) $. To this end, take $ \iota = \frac{1}

1395: {2 \, l} $. Define $ \u^\eps_\flat $ and $ \p^\eps_\flat $ 

1396: as in (\ref{BKWdel}) with the $ U_k $ and $ P_k $ of 

1397: (\ref{diction}). Introduce 

1398: 

1399: \medskip

1400: 

1401: $ \cc \u^\eps_\flat \, := \, \r \u^\eps_\flat - \text{ridiv} \, 

1402: \tilde \g^\eps_\flat = \bigcirc (\eps^{\frac{N}{l}-2-\frac{d}{2}}) \, , 

1403: \qquad \cc \p^\eps_\flat \, := \, \r \p^\eps_\flat = \bigcirc 

1404: (\eps^{\frac{N+1}{l}}) \, . $

1405: 

1406: \medskip

1407: 

1408: \noindent{After substitution in $ (\cE) $, we lose again a 

1409: power of $ \eps $.} We find

1410: 

1411: \medskip

1412: 

1413: $ \f^\eps_\flat = \tilde \f^\eps_\flat - (\text{ridiv} \, 

1414: \tilde \g^\eps_\flat \cdot \nabla) \, \tilde \u^\eps_\flat 

1415: - (\tilde \u^\eps_\flat \cdot \nabla) \, \text{ridiv} \, 

1416: \tilde \g^\eps_\flat $

1417: 

1418: \smallskip

1419:  

1420: $ \qquad \quad \ - \, \part_t \text{ridiv} \, \tilde \g^\eps_\flat

1421: + (\text{ridiv} \, \tilde \g^\eps_\flat \cdot \nabla) \, 

1422: \text{ridiv} \, \tilde \g^\eps_\flat \, = \, \bigcirc(\eps^{

1423: \frac{N}{l}-3-\frac{d}{2}}) \, . $

1424: 

1425: \medskip

1426: 

1427: \noindent{The Theorem \ref{appBKW} looks like classical statements in one phase 

1428: non linear geometric optics except that the phase $ \varphi^\eps_g $ does depend 

1429: on $ \eps $.} In the next chapter, we examine the part of the $ \varphi_k $ 

1430: which make up $ \varphi^\eps_g $ and $ \varphi^\eps_a $.

1431: 

1432: 

1433: \section{The cascade of phases.} {\it Turbulence} and {\it intermittency} are topics

1434: which represent extremely different points of view. Two approaches compete:

1435: 

1436: \medskip

1437: 

1438: \noindent{a) }The deterministic approach which study the time evolution

1439: of flows arising in fluid mechanics \cite{B}-\cite{C}-\cite{D}-\cite{E}-\cite{MPP}.

1440: 

1441: \medskip

1442: 

1443: \noindent{b) }The statistical approach in which the velocity of the

1444: fluid is a random variable \cite{FMRT}-\cite{L}.

1445: 

1446: \medskip

1447: 

1448: \noindent{The Theorem \ref{appBKW} is mainly connected with a)}. It brings

1449: various informations related to the propagation of quasi-singularities. These

1450: aspects are detailed at first. Then we briefly explain b) and we draw 

1451: (in the setting of the Theorem \ref{appBKW}) a phenomenological 

1452: comparison between a) and b).

1453: 

1454: \smallskip

1455: 

1456: \subsection{Microstructures.}  The result \ref{appBKW} is concerned

1457: with the convection of microstructures. It is linked with the multiple 

1458: scale approach of \cite{MPP} and \cite{C}. In \cite{MPP} the authors 

1459: look for BKW solutions $  \u^\eps_a $ in the form

1460: 

1461: \medskip

1462: 

1463: $ \u^\eps_a (t,x) \, = \, \u_0 (t,x) \, + \, U_0^* \bigl( t,x, 

1464: \eps^{-1} \, t , \eps^{-1} \, \vec \varphi_0 (t,x) \bigr) \, + \, 

1465: \bigcirc (\eps) \, . $ 

1466: 

1467: \medskip

1468: 

1469: \noindent{In the more recent paper \cite{C}, the selected expansion is}

1470: 

1471: \medskip

1472: 

1473: $ \mu^\eps_a (t,x) \, = \, \u_0 (t,x) \, + \, \eps^{\frac{1}{3}} \ 

1474: U_1 \bigl( t,x, \eps^{-\frac{2}{3}} \, t, \eps^{-1} \, \vec \varphi_0 

1475: (t,x) \bigr) \, + \, \bigcirc (\eps^{\frac{2}{3}}) \, . $ 

1476: 

1477: \medskip

1478: 

1479: \noindent{Both articles  \cite{C} and \cite{MPP} use homogenization 

1480: techniques.} They perform computations involving expressions as 

1481: $ \u^\eps_a $ or $ \mu^\eps_a $. Simplifications (supported by 

1482: engineering experiments) are made in order to get effective 

1483: equations for the evolution of $ (\u_0 , U_0^*) $ or $ (\u_0 ,

1484: U_1 ) $. 

1485: 

1486: \smallskip

1487: 

1488: \noindent{Consider the simple case of one phase expansions (that 

1489: is when $ \vec \varphi_0 \equiv \varphi_0 $ is a scalar valued 

1490: function).} Reasons why a complete mathematical analysis based on 

1491: $ \u^\eps_a $ or $ \mu^\eps_a $ is not available can be drawn from 

1492: the Theorem \ref{appBKW}. For instance, look at $ \mu^\eps_a $.

1493: When $ l = 3 $, the oscillation $ \mu^\eps_a $ involves the same 

1494: scales as $ \u^\eps_{(3,N)} $ since

1495: 

1496: \medskip

1497: 

1498: $ \eps^{-1} \, \varphi^\eps_g (t,x) \, = \, \eps^{-1} \ \varphi_0 

1499: (t,x) + \eps^{- \frac{2}{3}} \ \varphi_1 (t,x) + \eps^{- \frac{1}

1500: {3}} \  \varphi_2 (t,x) \, . $

1501: 

1502: \medskip

1503: 

1504: \noindent{Now the analogy stops here since in general $ \varphi_1 

1505: (t,x) \not \equiv t $ and $ \varphi_2 (t,x) \not \equiv 0 $.} These 

1506: are geometrical obstructions which prevent to describe the propagation 

1507: by way of $ \mu^\eps_a $. The asymptotic expansion $ \mu^\eps_a $ 

1508: is not suitable.

1509: 

1510: \medskip

1511: 

1512: \noindent{Analogous arguments concerning $ \u^\eps_a $ will be 

1513: presented in the paragraph 3.5.}

1514: 

1515: \smallskip

1516: 

1517: \subsection{The geometrical phase.} Let us examine 

1518: more carefully how the expression $ \varphi^\eps_g $ 

1519: is built. Because of the condition (\ref{nonstaou}), 

1520: for $ \eps $ small enough, it is still not stationary

1521: \begin{equation} \label{nondegeat}

1522: \ \exists \, \eps_0 > 0 \, ; \qquad \nabla \varphi^\eps_g 

1523: (t,x) \not = 0 \, , \qquad \forall \, (\eps,t,x) \in \, 

1524: ]0,\eps_0] \times [0,T] \times \RR^d \, . \ \

1525: \end{equation}

1526: In fact, the function $ \varphi^\eps_g $ comes from the 

1527: approximate eiconal equation

1528: 

1529: \medskip

1530: 

1531: $ \partial_t \varphi^\eps_g + ( \bar u^\eps_\flat \cdot 

1532: \nabla) \varphi^\eps_g = \bigcirc ( \eps ) $

1533: 

1534: \medskip

1535: 

1536: \noindent{which is equivalent to}

1537: 

1538: \medskip

1539: 

1540: $ \partial_t \varphi_k + \u_0 \cdot \nabla \varphi_k + \sum_{j=0}^{k-1}

1541: \, \bar U_{k-j} \cdot \nabla \varphi_j = 0 \, , \qquad \forall \, k \in \{ 1,

1542: \cdots, l-1 \} \, .  $

1543: 

1544: \medskip

1545: 

1546: \noindent{The family $ \{ \u^\eps_\flat (t,x) \}_{\eps \in \, ]0,1]} $

1547: has an $ \eps - \, $stratified regularity \cite{G2} with respect to the

1548: phase $ \varphi^\eps_g $.} This is a {\it geometrical} information.

1549: 

1550: \smallskip

1551: 

1552: \subsection{Closure problems.} We have explained why appealing 

1553: only to $ \varphi_0 $ is not sufficient. It turns out that BKW 

1554: computations relying only on the geometrical phase $ \varphi^\eps_g $ 

1555: come also to nothing. This is a subtle aspect when proving the 

1556: Theorem \ref{appBKW}. We lay now stress on it.

1557: 

1558: \medskip

1559: 

1560: \noindent{For all $ N \in \NN_* $, the application $ \cG $ 

1561: defined below is one to one}

1562: $$ \ \left. \begin{array}{rcl}

1563: \cG \ \, : \, (\cW^\infty_T)^N & \longrightarrow & (\cW^\infty_T)^N \\

1564: \\

1565: \left(\begin{array}{c}

1566: \tilde U_1 \\

1567: \tilde U_2 \\

1568: \vdots \\

1569: \tilde U_N

1570: \end{array} \right)(t,x,\theta) & \longmapsto & 

1571: \left(\begin{array}{c}

1572: \tilde U_1 \\

1573: \tilde U_2 + \cG^1 (\tilde U_1) \\

1574: \vdots \\

1575: \tilde U_N + \cG^N (\tilde U_1, \cdots, \tilde U_{N-1}) 

1576: \end{array} \right)(t,x,\theta + \varphi_l (t,x) ) \, .

1577: \end{array} \right. $$

1578: Once the $ U_j $ or the $ \tilde U_j $ are known, it is entirely 

1579: equivalent to use $ \u^\eps_\flat $ or $ \tilde \u^\eps_\flat $. 

1580: Before the $ U_j $ or the $ \tilde U_j $ have been identified, in 

1581: particular when performing the BKW calculus, it is deeply different to 

1582: employ $ \u^\eps_\flat $ or $ \tilde \u^\eps_\flat $. Indeed, there is 

1583: a unique choice of the $ \varphi_k $ with $ l \leq k \leq N $, which 

1584: imposes a {\it specific hierarchy} between the profiles $ \tilde 

1585: U_k $, which makes possible the {\it triangulation} of the equations 

1586: obtained by the formal computations.

1587: 

1588: \medskip

1589: 

1590: \noindent{In the subsection 2.3, we will perform the BKW analysis

1591: with the profiles $ \tilde U_k $.} It yields a sequence of equations

1592: \begin{equation} \label{cloeetilde}

1593: \tilde X^k ( \tilde U_1, \cdots, \tilde U_{k+l} ) \, = \, 0 \, , 

1594: \qquad 1 \leq k \leq N \, . \qquad \qquad \quad \

1595: \end{equation}

1596: As usual in non linear geometric optics, this can be rewritten 

1597: in order to find a sequence of well-posed equations

1598: \begin{equation} \label{cloeq}

1599: \dot X^k ( \dot U_k ) \, = \, \cF ( \dot U_1, \cdots , \dot 

1600: U_{k-1} ) \, , \qquad 1 \leq k \leq N \, , \qquad \qquad

1601: \end{equation}

1602: where the $ \dot U_k $ are made of pieces of the $ \tilde

1603: U_j $. Of course, the equation (\ref{cloeq}) can be interpreted 

1604: in terms of the $ \tilde U_j $ and then in terms of the $ U_j $. 

1605: In this second step, it requires to implement the phase 

1606: shift $ \varphi_l $ and the transformations $ \cG^j_p $ 

1607: with $ 1 \leq j \leq k-1 $ and $ 1 \leq p \leq j $. Now, 

1608: the BKW analysis reveals that $ \varphi_l $ or the various 

1609: coefficients $ \varphi_i $ which appear in the definition 

1610: of such $ \cG^j_p $ do not depend only on $ (\dot U_1, 

1611: \cdots, \dot U_k) $ but also on some $ \dot U_i $ with 

1612: $ i > k $. The resulting system is therefore underdetermined.

1613: Computations involving the functions $ U_j $ lead to a 

1614: sequence of equations which are not closed.

1615: 

1616: \medskip

1617: 

1618: \noindent{The insertion of the phases $ \varphi_k $ with $ 1 

1619: \leq k \leq N $ is an elegant way to introduce $ \cG $.} The 

1620: change of variables $ \cG $, though it is a function of $ (U_1, 

1621: \cdots, U_N) $, is needed to progress. It allows to get round 

1622: {\it closure problems}. 

1623: 

1624: 

1625: \subsection{Compensated compactness.} Dissipation terms can be 

1626: incorporated in the discussion. In the variables $ (t,x) $, the 

1627: addition of some viscosity $ \kappa $ is compatible with the 

1628: propagation of oscillations if for instance $ \kappa = \nu \, 

1629: \eps^2 $. There are approximate solutions $ (\u^\eps_\flat ,

1630: \p^\eps_\flat ) $ of the Navier-Stokes equations. They satisfy

1631: (\ref{BKWdel}) and 

1632: \medskip

1633: 

1634: $ \part_t \u^\eps_\flat + ( \u^\eps_\flat \cdot \nabla) \u^\eps_\flat 

1635: + \nabla \p^\eps_\flat = \nu \ \eps^2 \ \Delta_x \u^\eps_\flat + 

1636: \f^\eps_\flat \, , \qquad \Div \ \u^\eps_\flat = 0 \, , $

1637: 

1638: \medskip

1639: 

1640: \noindent{with $ \f^\eps_\flat = \bigcirc(\eps^\infty) $.}

1641: When $ \nu > 0 $, Leray's theorem provides with global weak 

1642: solutions $ (\u^\eps,\p^\eps) (t,x) $ of the following 

1643: Cauchy problem}

1644: $$ \left \{ \begin{array} {l}

1645: \part_t \u^\eps + ( \u^\eps \cdot \nabla) \u^\eps + \nabla 

1646: \p^\eps = \nu \ \eps^2 \ \Delta_x \u^\eps \, , \qquad \Div \ 

1647: \u^\eps = 0 \, , \qquad \qquad \\

1648: \u^\eps (0,\cdot) \equiv \u^\eps_\flat (0,\cdot) \, . 

1649: \end{array} \right. $$

1650: Suppose now that $ \u_0 \equiv 0 $. Then, we have 

1651: also the uniform controls

1652: \begin{equation} \label{compcomp}

1653: \, \left. \begin{array}{l}

1654: \sup \ \bigl \lbrace \, \parallel \eps^{-\frac{1}{l}} \,

1655: \u^\eps \parallel_{L^2_T} \, ; \ \eps \in \, ]0,1] \, 

1656: \bigr \rbrace \, \leq \, C \, < \, \infty \, , \\

1657: \sup \ \bigl \lbrace \, \nu \ \eps^2 \ \int_0^T \, 

1658: \parallel \eps^{-\frac{1}{l}} \, \u^\eps (t,\cdot) 

1659: \parallel^2_{H^1 (\RR^d) } \ dt \,  ; \ \eps \in \, 

1660: ]0,1] \, \bigr \rbrace \, \leq \, C \, < \, \infty \, . \

1661: \end{array} \right.

1662: \end{equation}

1663: Arguments issued from the theory of compensated compactness 

1664: \cite{Ge} can be employed to study the sequence $ \{ \eps^{

1665: -\frac{1}{l}} \, \u^\eps \}_\eps $. In the spirit of \cite{D} 

1666: or \cite{E}, we can try to exploit the informations contained 

1667: in (\ref{compcomp}) and the equation on $ \u^\eps $ in order 

1668: to describe the asymptotic behaviour when $ \eps $ goes to 

1669: zero of the functions $ \eps^{- \frac{1}{l}} \, \u^\eps $. 

1670: However this approach seems to be not applicable here.

1671: 

1672: \medskip

1673: 

1674: \noindent{Indeed, {\it obvious} instabilities occur.} 

1675: The related mechanisms, which induce the non linear instability 

1676: of Euler equations, are detailed in the paragraph 5.1. Below, we 

1677: just give an intuitive idea of what can happen. Use the representation 

1678: $ \tilde \u^\eps_\flat $ involving the phase $ \varphi^\eps_\flat $. 

1679: The determination of the intermediate term $ \varphi_l $ requires to 

1680: identify $ \langle \tilde U_l \rangle $ and $ \tilde U^*_{l-1} $. This 

1681: is a consequence of the equations (\ref{phik}) and (\ref{moyj+1}). 

1682: 

1683: \medskip

1684: 

1685: \noindent{In view of the formula (\ref{diction}), when $ \varphi_l $ 

1686: is modified by an amount of $ \delta \varphi_l $, the quantity 

1687: $ U_1 (t,x,\theta) $ undergoes a perturbation of the same order 

1688: $ \delta \varphi_l $.} When dealing with quasi-singularities, 

1689: some quantities with $ \eps $ in factor (like $ \langle \tilde 

1690: U_l \rangle $) or with $ \eps^{1 - \frac{1}{l}} $ in factor 

1691: (like $ \tilde U^*_{l-1} $) can control informations of size 

1692: $ \eps^{\frac{1}{l}} $. This fact is expressed by the 

1693: following rules of transformation

1694: \begin{equation} \label{rulesoftr}

1695: \left. \begin{array}{lcl}

1696: \langle \tilde U_l \rangle \ / \ \langle \tilde U_l \rangle \, 

1697: + \, \delta \langle \tilde U_l \rangle \quad & 

1698: \Longrightarrow & \quad \u^\eps_\flat \ / \ \u^\eps_\flat + 

1699: \bigcirc ( \eps^{\frac{1}{l}} ) \ \delta \langle \tilde U_l 

1700: \rangle \, , \\

1701: \tilde U^*_{l-1} \ / \ \tilde U^*_{l-1} \, + \, \delta 

1702: \tilde U^*_{l-1} \quad & \Longrightarrow & \quad \u^\eps_\flat \ 

1703: / \ \u^\eps_\flat + \bigcirc ( \eps^{\frac{1}{l}} ) \ \delta 

1704: \tilde U^*_l \, . \

1705: \end{array} \right. 

1706: \end{equation}

1707: Now reverse the preceding reasoning. To describe features in the

1708: principal oscillating term $ \eps^{\frac{1}{l}} \ U^*_1 

1709: \bigl(t,x,\eps^{-1} \, \varphi^\eps_g (t,x) \bigr) $, we 

1710: must identify $ \varphi_l $ which means to obtain $ \langle 

1711: \tilde U_l \rangle $ and $ \tilde U^*_{l-1} $. In other words, 

1712: we need to know quantities which have respectively $ \eps $ and 

1713: $ \eps^{1- \frac{1}{l}} $ in factor. When $ l \geq 2 $ such 

1714: informations are clearly not reachable by rough controls as 

1715: (\ref{compcomp}). 

1716: 

1717: \medskip

1718: 

1719: \noindent{This discussion indicates that the study of turbulent 

1720: regimes requires to combine at least geometrical aspects, multiphase 

1721: analysis and high order expansions.} The tools of non linear geometric 

1722: optics seem to be appropriate. Some attempts in this direction have 

1723: already been made.

1724: 

1725: \smallskip

1726: 

1727: \subsection {Non linear geometric optics.} We make in this paragraph 3.5

1728: several comments about non linear geometric optics. They concern both old

1729: \cite{G}-\cite{G2}-\cite{Se} and recent \cite{Che}-\cite{CGM}-\cite{CGM1} 

1730: results which all are devoted to one phase expansions of the type

1731: \begin{equation} \label{ogmonop}

1732: \u^\eps_\natural (t,x) \, := \, \u_0 (t,x) \,  + \, \hbox{$ \sum_{k=1}^\infty $}

1733: \, \eps^{\frac{k}{l}} \ U_k \bigl( t,x, \eps^{-1} \, \varphi_0 (t,x) \bigr) 

1734: \, . \quad \

1735: \end{equation}

1736: When $ l=1 $, one is faced with {\it weakly non linear geometric optics}. 

1737: The asymptotic behavior and the stability of $ \u^\eps_\natural $ are

1738: well understood. In fact a complete theory has been achieved in the

1739: general framework of multidimensional systems of conservation laws 

1740: (see \cite{G}-\cite{G2} and the related references). Because of the

1741: formation of shocks, the life span of exact solutions close to 

1742: $ \u^\eps_\natural $ does not go beyond $ T \simeq 1 $. 

1743: 

1744: \medskip

1745: 

1746: \noindent{When $ l=2 $, expressions as $ \u^\eps_\natural $ are called 

1747: {\it strong} oscillations.} The construction of such BKW solutions can 

1748: be undertaken only if the system of conservation laws has a special 

1749: structure. {\it Transparency} conditions are needed to progress. They

1750: can be deduced from the presence of a linearly degenerate field

1751: \cite{CGM}. In the hyperbolic situation the family $ \{ \u^\eps_\natural 

1752: \}_{\eps \in \, ]0,1]} $ is unstable \cite{CGM} on the interval $ [0,T] $.

1753: It becomes stable on condition that a small viscosity is incorporated

1754: \cite{Che}. Applications can be given to describe large-scale 

1755: motions in the atmosphere \cite{Che}.

1756: 

1757: \medskip

1758: 

1759: \noindent{Compressible Euler equations are the prototype of a non 

1760: linear hyperbolic system having a linearly degenerate field.} After

1761: a finite time, singularities appear. These correspond to the generation

1762: of shocks by compression \cite{Si}. The situation is different in the

1763: incompressible setting. There is no genuine shock and the production

1764: of singularities poses a much more subtle problem \cite{BKM}-\cite{CF}

1765: which up to now remains basically open.

1766: 

1767: \medskip

1768: 

1769: \noindent{Incompressible fluid equations lie at an extreme end in the 

1770: sense that they are the most {\it degenerate} (or the most {\it linear})

1771: equations which have just been mentioned. Following the approach of 

1772: \cite{JMR1} related to {\it transparency}, repeating the reasoning 

1773: which goes from \cite{G}-\cite{G2} to \cite{Che}-\cite{CGM}, one 

1774: expects to go further than $ l = 2 $ when dealing with $ (\cE) $. 

1775: Now, this is precisely what says the Theorem \ref{appBKW} since 

1776: it allows to reach any $ l \in \NN_* \, !$

1777: 

1778: \medskip

1779: 

1780: \noindent{To tackle the limit case $ l= \infty $, one is tempted to look

1781: at asymptotic expansions of the form

1782: \begin{equation} \label{ogmonopl}

1783: \ \u^\eps_\infty (t,x) \, := \, \hbox{$ \sum_{k=0}^\infty $} \, \eps^k \ 

1784: U_k \bigl( t,x, \eps^{-1} \, \varphi_0 (t,x) \bigr) \, , \qquad

1785: \partial_\theta U_0^* \not \equiv 0 \, . \qquad \quad  

1786: \end{equation}

1787: The oscillations contained in $ \u^\eps_\infty $ have a {\it large}

1788: amplitude. Modulation equations for $ U_0 $ are proposed in \cite{Se}. 

1789: However these transport equations are not hyperbolic so that they are 

1790: ill posed (in the sense of Hadamard) with respect to the initial value 

1791: problem. It confirms that a BKW construction based on (\ref{ogmonopl}) 

1792: is not relevant\footnote{The singularities are carried here by the velocity 

1793: field. The discussion is very different when the oscillations are 

1794: polarized on the entropy \cite{CGM1}.}. 

1795: 

1796: \medskip

1797: 

1798: \noindent{The contribution \cite{Se} does not explain why the expansion

1799: (\ref{ogmonopl}) is not the good one.} We come back below to this point.

1800: At first sight the Theorem \ref{appBKW} does not include large 

1801: amplitude waves since $ \u^\eps_\flat - \u_0 = \bigcirc ( \eps^{\frac

1802: {1}{l}} ) \ll \bigcirc(1) $. A change of variables leads to recant 

1803: this impression. Suppose that $ \u_0 \equiv 0 $ and $ \part_\theta 

1804: U_1^* \not \equiv 0 $. Then define

1805:  

1806: \medskip

1807: 

1808: $ \dot \u^\eps_\flat (t,x) \, := \, \eps^{-\frac{1}{l}} \ \u^\eps_\flat 

1809: (\eps^{-\frac{1}{l}} \, t,x) \, , \qquad \dot \p^\eps_\flat (t,x) \, :=

1810: \, \eps^{-\frac{2}{l}} \ \p^\eps_\flat (\eps^{-\frac{1}{l}} \, t,x) \, . $

1811: 

1812: \medskip

1813: 

1814: \noindent{Observe that the structure of $ \dot \u^\eps_\flat $ and 

1815: $ \dot \p^\eps_\flat $ is very different from the one in (\ref{ogmonopl}) 

1816: since we have

1817: $$ \quad \left. \begin{array}{l}

1818: \dot \u^\eps_\flat (t,x) = \sum_{k=1}^\infty \, \eps^{ 

1819: \frac{k-1}{l}} \ U_k \bigl( \eps^{-\frac{1}{l}} \, t,x, \eps^{-1} \, 

1820: \varphi^\eps_g (\eps^{-\frac{1}{l}} \, t,x) \bigr) + 

1821: \eps^{-\frac{1}{l}} \ \cc \u^\eps_\flat (\eps^{-\frac{1}{l}} \, t,x) 

1822:  \, , \\

1823: \dot \p^\eps_\flat (t,x) = \sum_{k=1}^\infty \, \eps^{

1824: \frac{k-2}{l}} \ P_k \bigl( \eps^{-\frac{1}{l}} \, t,x, \eps^{-1} \, 

1825: \varphi^\eps_g (\eps^{-\frac{1}{l}} \, t,x) \bigr) + 

1826: \eps^{-\frac{2}{l}} \ \cc \p^\eps_\flat (\eps^{-\frac{1}{l}} \, t,x) \, .

1827: \end{array} \right. $$

1828: The functions  $ \dot \u^\eps_\flat $ and $ \dot \p^\eps_\flat $

1829: satisfy

1830: 

1831: \medskip

1832: 

1833: $ \part_t \dot \u^\eps_\flat + ( \dot \u^\eps_\flat \cdot \nabla) 

1834: \dot \u^\eps_\flat + \nabla \dot \p^\eps_\flat = \dot \f^\eps_\flat \, , 

1835: \quad \ \Div \, \dot \u^\eps_\flat = 0 \, , \quad \ \dot \f^\eps_\flat 

1836: (t,x) = \eps^{-\frac{2}{l}} \ \f^\eps_\flat (\eps^{-\frac{1}{l}} \, t,x) 

1837: \, . $

1838: 

1839: \medskip

1840: 

1841: \noindent{The functions $ \dot \u^\eps_\flat $ are oscillations 

1842: of the order $ 1 $.} They are approximate solutions of $ (\cE) $ 

1843: on the {\it small} interval $ [0, \eps^{\frac{1}{l}} \, T] $.

1844: Indeed, for all $ m \in \NN $, the family $ \{ \dot \f^\eps_\flat 

1845: \}_\eps $ is subjected to the uniform majoration

1846: 

1847: \medskip

1848: 

1849: $ \sup_{\eps \in \, ]0,\eps_0]} \quad \eps^{- \frac{N}{l} +

1850: \frac{2}{l} + 3 + m} \ \parallel \dot \f^\eps_\flat \parallel_{

1851: \cW^m_{\! \text{\tiny $ \eps^{(1/l)} \, T $}} } \ < \, \infty \, . \ $

1852: 

1853: \medskip

1854: 

1855: \noindent{If moreover $ N = + \infty $ and} 

1856: \begin{equation} \label{speini}

1857: \left. \begin{array}{l}

1858: \varphi_1 (0,\cdot) \, \equiv \, \cdots  \, \equiv \, \varphi_{l-1}

1859: (0,\cdot) \, \equiv \, 0 \, , \\

1860: U_{k+1} (0,\cdot) \, \equiv \, 0 \, , \qquad \forall \, k \in \NN 

1861: \setminus ( l \, \NN) \, , \qquad \qquad \qquad \qquad \qquad \

1862: \end{array} \right.

1863: \end{equation}

1864: the trace $ \dot \u^\eps_\flat (0,\cdot) $ has the form

1865: 

1866: \medskip

1867: 

1868: $ \dot \u^\eps_\flat (0,x) \, = \ \sum_{k=0}^\infty \, \eps^k \  

1869: U_{1 + l \, k} \bigl(0,x, \eps^{-1} \, \varphi_0 (0,x) \bigr) \, , 

1870: \qquad \part_\theta U_1^* \, \not \equiv \, 0 \, .$

1871: 

1872: \medskip

1873: 

1874: \noindent{At the time $ t=0$, we recover (\ref{ogmonopl}). Now 

1875: the construction underlying the Theorem \ref{appBKW} reveals that 

1876: in general}

1877: \begin{equation} \label{phasenonn}

1878: \varphi_k(t,\cdot) \, \not \equiv \, 0 \, , \qquad \forall \, t \in \, ]0,T] \, , \qquad 

1879: \forall \, k \in \{ 2, \cdots, l-1 \} \, . \quad 

1880: \end{equation}

1881: The functions $ \varphi_j $ with $ j \in \{2,\cdots, l-1 \} $ are 

1882: not present when $ t = 0 $. But the description of $ \dot \u^\eps_\flat 

1883: (t,\cdot) $ on the interval $ [0, \eps^{1 - \frac{k}{l}} \, T] $ with $ 

1884: k \in \{2,\cdots, l-1 \} $ requires the introduction of the {\it phase 

1885: shifts} $ \varphi_j $ for $ j \in \{2,\cdots,k\} $. More generally, the 

1886: description of $ \dot \u^\eps_\flat (t,\cdot) $ on the whole interval 

1887: $ [0, T] $ needs the introduction of an {\it infinite cascade} of phases 

1888: $ \{ \varphi_j \}_{j \in \NN_*} $.

1889: 

1890: \medskip

1891: 

1892: \noindent{Such a phenomenon does not occur when constructing large 

1893: amplitude oscillations for systems of conservation laws in one space 

1894: dimension \cite{CG}-\cite{E2}.} It is specific to the multidimensional 

1895: framework. It explains why the classical approach of \cite{Se} fails.

1896: 

1897: \medskip

1898: 

1899: \noindent{It seems that the creation of the $ \varphi_j $ is due to 

1900: mechanisms which have not already been studied.} It is not linked with 

1901: {\it resonances}. It is related neither to {\it dispersive} nor to 

1902: {\it diffractive} effects.

1903: 

1904: \bigskip

1905: \vskip -1mm

1906: 

1907: \noindent{{\it Remark 3.5.1 (about $ \varphi_1 $):}} The term $ \varphi_1 $ 

1908: does not appear if $ \varphi_1 (0,\cdot) \equiv 0 $ and $ \bar U_1

1909: (0,\cdot) \equiv 0 $. When these two conditions are not verified, the 

1910: phase shift $ \varphi_1 $ can be absorbed by the technical trick 

1911: exposed in \cite{CGM}. Just replace $ \u_0 (0,\cdot) $ by $ \u_0 

1912: (0,\cdot) + \delta \, \bar U_1 (0,\cdot) $. Perform the BKW calculus 

1913: with a fixed $ \delta > 0 $. Then choose $ \delta = \eps $.  \hfill 

1914: $ \triangle $

1915: 

1916: \bigskip

1917: 

1918: \noindent{{\it Remark 3.5.2 (about $ \varphi_2 $):}} In general, we

1919: have $ \varphi_2 \not \equiv 0 $ even if

1920: 

1921: \medskip

1922: 

1923: $ \varphi_1 (0,\cdot) \equiv \varphi_2 (0,\cdot) \equiv 0 \, , \qquad 

1924: \bar U_1 (0,\cdot) \equiv \bar U_2 (0,\cdot) \equiv 0 \, . $

1925: 

1926: \medskip

1927: 

1928: \noindent{Indeed the time evolution of $ \bar U_2 $ is governed 

1929: by (\ref{moy2}).} It involves the source term $ \Div \, \langle 

1930: U^*_1 \otimes U^*_1 \rangle $ which is able to awake $ \bar U_2 $. This 

1931: influence can then be transmitted to $ \varphi_2 $ through the 

1932: transport equation

1933: \begin{equation} \label{trans2}

1934: \part_t \varphi_2 + (\u_0 \cdot \nabla) \varphi_2 + (\bar U_1 

1935: \cdot \nabla) \varphi_1 + (\bar U_2 \cdot \nabla) \varphi_0

1936: \, = \, 0 \, . \qquad \qquad

1937: \end{equation}

1938: Likewise, the other terms $ \varphi_3 $, $\cdots $, 

1939: $ \varphi_{l-1} $ are in general non trivial even if 

1940: 

1941: \medskip

1942: 

1943: $ \varphi_1 (0,\cdot) \equiv \cdots \equiv \varphi_{l-1} (0,\cdot) 

1944: \equiv 0 \, , \qquad \bar U_1 (0,\cdot) \equiv \cdots \equiv 

1945: \bar U_{l-1} (0,\cdot) \equiv 0 \, . $

1946: 

1947: \medskip

1948: 

1949: \noindent{There is no more trick which allows to get rid of 

1950: $  \varphi_2 $, $ \cdots$, $ \varphi_{l-1} $.} \hfill $ \triangle $

1951: 

1952: \bigskip

1953: 

1954: \noindent{{\it Remark 3.5.3 (why turbulent flows ?):}} The introduction 

1955: of the phase shifts $  \varphi_k $ with $ 2 \leq k \leq l-1 $ cannot 

1956: be avoided. Therefore the difficulties that we deal with appear from 

1957: $ l = 3 $. When $ l \geq 3 $, the characteristic rate $ e $ of eddy  

1958: dissipation is bigger than one \cite{C}. This is the reason why such 

1959: situations are refered to {\it turbulent regimes}. \hfill $ \triangle $

1960: 

1961: \bigskip

1962: 

1963: \noindent{{\it Remark 3.5.4 (about shear layers):}} We have said in 

1964: the introduction that the expression $ \u^\eps_s $ given by formula

1965: (\ref{oscidipmaj}) is of a very special form. Let us explain why.

1966: Change the variable $ t $ into $ \eps^{\frac{1}{l}} \, t $ 

1967: and $ \u^\eps_s $ into $ \dot \u^\eps_s := \eps^{\frac{1}{l}} \, 

1968: \u^\eps_s $. The main phase $ \varphi_0 (t,x) \equiv x_2 $ remains 

1969: the same. Now we are faced with

1970: 

1971: \medskip

1972: 

1973: $ \dot \u^\eps_s (t,x) \, := \, {}^t \bigl( \eps^{\frac{1}{l}} \, 

1974: \g ( x_2,\eps^{-1} \, x_2) , 0 , \eps^{\frac{1}{l}} \, \h \bigl( 

1975: x_1 - \eps^{\frac{1}{l}} \, \g (x_2, \eps^{-1} \, x_2) \, t , x_2, 

1976: \eps^{-1} \, x_2 \bigr) \bigr) \, . $

1977: 

1978: \medskip

1979: 

1980: \noindent{It is still a solution of Euler equations.} Now it falls

1981: in the framework of the Theorem \ref{appBKW}. The constraints 

1982: on $ \bar U_2 = {}^t (\bar U_2^1, \bar U_2^2, \bar U_2^3 ) $ 

1983: reduce to

1984: 

1985: \medskip

1986: 

1987: $ \bar U_2^1 \, \equiv \, \bar U_2^2 \, \equiv \, 0 \, , \qquad 

1988: \part_t \bar U_2^3 \, + \, \langle \g \, \part_1 \h \rangle \,

1989: = \, 0 \, . $

1990: 

1991: \medskip

1992: 

1993: \noindent{The contribution $ \bar U_2 $ is non trivial but it is

1994: polarized so that $ \bar U_2 \cdot \nabla \varphi_0 \equiv 0 $.}

1995: Therefore it does not produce the phase shift $ \varphi_2 $. The

1996: same phenomenon occurs concerning $ \varphi_3 $, $ \cdots , $ 

1997: $ \varphi_{l-1} $. These terms are not present. It turns out that 

1998: the expansion $ \u^\eps_s $ involves only the phase $ \varphi_0 

1999: (t,x) \equiv x_2 $. \hfill $ \triangle $

2000: 

2001: \bigskip

2002: 

2003: \noindent{The choice for the amplitude of the oscillations is very 

2004: important.} It is strongly related to the scale of time $ T $ under

2005: consideration. The idea is to increase the time of propagation $ T $

2006: to reach the regime where non linear effects appear. Starting

2007: with some large amplitude high frequency waves 

2008: 

2009: \medskip

2010: 

2011: $  \u^\eps_\infty (0,x) \, = \, U_0 \bigl( 0,x, \eps^{-1} \, \varphi_0 

2012: (0,x) \bigr) \, + \, \bigcirc (\eps) \, , \qquad \partial_\theta U_0^* 

2013: (0,\cdot) \not \equiv 0 \, , $

2014: 

2015: \medskip

2016: 

2017: \noindent{the preceding discussion can be summarized by the following 

2018: diagram:}

2019: 

2020: \bigskip

2021: 

2022: $$ \left. \begin{array}{ccccccc}

2023: T \simeq 1 &

2024: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2025: & \ & 

2026: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2027: & \ & 

2028: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!- \\

2029: 

2030: \ & \left. \begin{array}{c}

2031: \text{\scriptsize infinite cascade} \\

2032: \text{\scriptsize of phases} \\

2033: \text{\scriptsize  $\varphi_0 - (\varphi_1) - \cdots $}

2034: \end{array} \right. & \left. \begin{array}{c}

2035: \mid \\

2036: \mid \\ 

2037: \mid

2038: \end{array} \right. &  \left. \begin{array}{c}

2039:  \\

2040: \text{\scriptsize turbulent} \\ 

2041: \text{\scriptsize flows}

2042: \end{array} \right.

2043: 

2044: & \left. \begin{array}{c}

2045: \mid \\

2046: \mid \\ 

2047: \mid

2048: \end{array} \right.

2049:  & \left. \begin{array}{c}

2050: \\

2051: \text{\footnotesize incompressible} \\

2052: \text{\footnotesize fluid equations}

2053: \end{array} \right.

2054:   \\

2055: 

2056: T \simeq \eps^{\frac{1}{3}} \ &

2057:  -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2058: & \ &

2059: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2060: & \ & 

2061: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!- \\

2062: 

2063: \ & \text{\scriptsize  $\varphi_0 - (\varphi_1) - \varphi_2 $} & \left. \begin{array}{c}

2064: \mid \\

2065: \mid

2066: \end{array} \right. & \left. \begin{array}{c}

2067: \text{\footnotesize turbulent} \\

2068: \text{\footnotesize flows}

2069: \end{array} \right. & \left. \begin{array}{c}

2070: \mid \\

2071: \mid

2072: \end{array} \right. & \left. \begin{array}{c}

2073: \text{\footnotesize incompressible} \\

2074: \text{\footnotesize fluid equations}

2075: \end{array} \right. \\

2076: 

2077: T \simeq \eps^{\frac{1}{2}} \ & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2078: & \ &

2079: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2080: & \ & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!- \\

2081: 

2082: &\text{\footnotesize $ \varphi_0 - (\varphi_1) $} & \left. \begin{array}{c}

2083: \mid \\

2084: \mid \\ 

2085: \mid

2086: \end{array} \right. & \left. \begin{array}{c}

2087: \text{\small strong}\\

2088: \text{\small oscillations} \\

2089: \cite{Che}-\cite{CGM}

2090: \end{array} \right. & \left. \begin{array}{c}

2091: \mid \\

2092: \mid \\ 

2093: \mid

2094: \end{array} \right. & \left. \begin{array}{c}

2095: \text{\small systems of conservation}\\

2096: \text{\small laws with a linearly} \\

2097: \text{\small degenerate field}

2098: \end{array} \right.

2099:  \\

2100: T \simeq \eps & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2101: & \ &

2102: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2103: & \ & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!- \\

2104: 

2105:  & \varphi_0 & \left. \begin{array}{c}

2106: \mid \\

2107: \mid \\ 

2108: \mid \\

2109: \mid \\

2110: \mid

2111: \end{array} \right. &  \left. \begin{array}{c}

2112: \text{weakly} \\

2113: \text{non linear} \\

2114: \text{geometric}\\

2115: \text{optics} \\ 

2116:  \cite{G}-\cite{G2}

2117: \end{array} \right.

2118:  & \left. \begin{array}{c}

2119: \mid \\

2120: \mid \\ 

2121: \mid \\

2122: \mid \\

2123: \mid

2124: \end{array} \right. &  \left. \begin{array}{c}

2125: \text{\large systems} \\

2126: \text{\large of} \\

2127: \text{\large conservation} \\ 

2128: \text{\large laws}

2129: \end{array} \right. \\

2130: 

2131: T=0 & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2132: & \ &

2133: -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-

2134: & \ & -\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!- \\

2135: 

2136: & \text{\large \bf phases} & \mid & \text{\large \bf regimes} & \mid & \text{\large \bf equations} \\

2137: 

2138: \end{array} \right. \quad $$

2139: 

2140: \bigskip

2141: 

2142: \noindent{This picture allows to understand the position of the

2143: actual paper in comparison with previous results.}

2144: 

2145: \smallskip

2146: 

2147: \subsection{The statistical approach.} It deals mainly with

2148: {\it quantitative} informations obtained at the level of expressions, 

2149: say $ \mu (x) $, which in general do not depend on the time $ t $. 

2150: The introduction of $ \mu $ can be achieved by looking at {\it 

2151: stationary statistical} solutions \cite{FMRT} of the Navier-Stokes  

2152: equations that is

2153: 

2154: \medskip

2155: 

2156: $ \mu(x) \, \equiv \, \lim_{\, T \, \longrightarrow \, \infty} \quad

2157: \frac{1}{T} \ \int_0^T \, \u(t,x) \ dt $

2158: 

2159: \medskip

2160: 

2161: \noindent{or in conjunction with the {\it ensemble average operator} 

2162: (\cite{L}-V-6) marked by the brackets $ < \cdot > $.} We will follow 

2163: this second option. The description below is extracted from the book 

2164: of M. Lesieur \cite{L} (chapters V and VI). We work with $ d = 3 $. 

2165: Interesting quantities are the mean kinetic energy

2166: 

2167: \medskip

2168: 

2169: $ \frac{1}{2} \ < \mu(x)^2 > \ \sim \, \int_{\RR^3} \, \vert \mu(x) \vert^2 \ dx \, , $

2170: 

2171: \medskip

2172: 

2173: \noindent{the enstrophy (that is the space integral of the square norm of the vorticity)}

2174: 

2175: \medskip

2176: 

2177: $ \frac{1}{2} \ < \omega (x)^2 > \ \sim \, \int_{\RR^3} \, \vert \omega (x) 

2178: \vert^2 \ dx \, , \qquad \omega (x) := \nabla \wedge \mu (x) $

2179: 

2180: \medskip

2181: 

2182: \noindent{and the rate of dissipation $ e \, \sim \, \kappa \ < 

2183: \omega (x)^2 > $.} In the setting of {\it isotropic} turbulence, 

2184: these quantities can be expressed in terms of a scalar function 

2185: $ k \longmapsto E(k) $. The real number $ E(k) $ represents the 

2186: density of kinetic energy at wave number $ k $ (or the kinetic 

2187: energy in Fourier space integrated on a sphere of radius $ k $). 

2188: The relations are the following

2189: 

2190: \medskip

2191: 

2192: \noindent{\cite{L}}-V-10-4$ \qquad \ \frac{1}{2} \ < \mu(x)^2 > \ = \, \int_0^{+ \infty} \, 

2193: E(k) \ dk \, . $ 

2194: 

2195: \medskip

2196: 

2197: \noindent{\cite{L}}-V-10-15$ \qquad \! \frac{1}{2} \ < \omega(x)^2 > \ = \, \int_0^{+ \infty} \, 

2198: k^2 \ E(k) \ dk \, . $ 

2199: 

2200: \medskip

2201: 

2202: \noindent{\cite{L}}-VI-3-15$ \qquad e \, = \,  2 \ \kappa \ \int_0^{+ \infty} \, k^2 \  

2203: E(k) \ dk \, . $ 

2204: 

2205: \medskip

2206: 

2207: \noindent{Kolmogorov's theory assumes that}

2208: 

2209: \medskip

2210: 

2211: \noindent{\cite{L}}-VI-4-1 $ \qquad \exists \ c > 0 \, ; \qquad E(k) \,= \, c \ e^{2/3} \ 

2212: k^{-5/3} \, , \qquad \forall \, k \in [k_i,k_d] \, . $

2213: 

2214: \medskip

2215: 

2216: \noindent{This law is valid up to the frequency $ k_d $ with}

2217: 

2218: \medskip

2219: 

2220: \noindent{\cite{L}}-VI-4-2 $ \qquad \, k_d \ \sim  \ ( \, e \, / \, \kappa^3 \, )^{1/4} \, . $

2221: 

2222: \medskip

2223: 

2224: \noindent{The small quantity $ \eps := k_d^{-1} $ is the Kolmogorov 

2225: dissipative scale.} The relations \cite{L}}-VI-3-15 and \cite{L}}-VI-4-2 

2226: imply that the rate of injection of kinetic energy $ e $ is linked to 

2227: the number $ l $ according to $ e \sim \eps^{-1+\frac{3}{l}} $. We 

2228: recover here that $ e \sim 1 $ when $ l =3 $ (see \cite{C}).

2229: 

2230: \medskip

2231: 

2232: \noindent{A starting point for the conventional theory of turbulence is

2233: the notion that, on average, kinetic energy is transfered from low wave

2234: numbers modes to high wave numbers modes. A recent paper \cite{FMRT} put 

2235: forward the following idea: in the spectral region below that of injection 

2236: of energy, an inverse (from high to low modes) transfer of energy takes 

2237: place. At any rate, it is a central question to determine how the kinetic

2238: energy is distributed.

2239: 

2240: \subsection{Phenomenological comparison.} The statistical 

2241: approach is concerned with the spectral properties of solutions.

2242: Below, we draw a parallel with the propagation of quasi-singularities

2243: as it is described in the Theorem \ref{appBKW}. 

2244: 

2245: \medskip

2246: 

2247: \noindent{Let us examine how the square $ \cF(\u^\eps_\flat) (t,\xi)^2 $ 

2248: of the Fourier transform of $ \u^\eps_\flat(t,x) $ is distributed.} To 

2249: this end, consider the application

2250: \vskip -5mm

2251: $$ \left. \begin{array} {rcl}

2252: \tilde E(t,\cdot) \, : \, \RR^+ & \longrightarrow & \RR^+ \\

2253: k \ & \longmapsto & \tilde E(t,k) \, := \, \int_{\{ \xi 

2254: \in \RR^d \, ; \, \vert \xi \vert = k \}} \ \vert \cF 

2255: (\u^\eps_\flat) (t,\xi) \vert^2 \ \, d \sigma (\xi) \, . \quad

2256: \end{array} \right. \ $$

2257: \vskip -1mm

2258: \noindent{The initial data $ \u^\eps_\flat (0,\cdot) $ has a {\it spectral gap}.} 

2259: In another words, the graph of the function $ k \longmapsto \tilde

2260: E(0,k) $ appears concentrated around the two characteristic wave 

2261: numbers $ k \simeq 1 $ and $ k \simeq\eps^{-1} = k_d $.} In view of 

2262: (\ref{phasenonn}), this situation does not persist. At the time 

2263: $ t = \eps^{\frac{1}{l}} $, the concentration is around $ l $ 

2264: characteristic wave numbers which are intermediate between the 

2265: two preceding ones. This corresponds to a {\it discrete} cascade 

2266: of energy.

2267: 

2268: \medskip

2269: 

2270: \noindent{Suppose now (\ref{speini}) and consider $ \dot \u^\eps_\flat $.} 

2271: The life span of $ \dot \u^\eps_\flat (t,\cdot) $ is $ \eps^{\frac{1}{l}} 

2272: \, T $.} There are various manners to get a family $ \{\dot \u^\eps_\flat(t,\cdot)

2273: \}_{\eps \in \, ]0,1] } $ which is defined on some interval $ [0,\tilde T] $ with 

2274: $ \tilde T > 0 $ independent on $ \eps $.} In particular, we can  

2275: 

2276: \medskip

2277: 

2278: \noindent{a) Select any $ \tilde T > 0 $ when $ T = + \infty $.} However 

2279: nothing guarantees that the functions $ \dot \u^\eps_\flat $ are still 

2280: approximate solutions on the interval $ [0,\tilde T] $. Indeed, since   

2281: $ t $ is replaced by $ \eps^{- (1/l)} \, t $, the size of the error 

2282: terms $ \dot \f^\eps_\flat $ depends on the increase of $ \f^\eps_\flat $  

2283: with respect to $ t $. At this level, we are faced with secular growth 

2284: problems \cite{La}.

2285: 

2286: \medskip

2287: 

2288: \noindent{b) Use a convergence process\footnote{When performing

2289: the formal analysis, arbitrary values can be given to the 

2290: parameters $ \eps \in \, ]0,1] $ and $ l \in \NN_* $. For instance 

2291: $ \eps $ can be fixed whereas $ l $ goes to $ \infty $. Or 

2292: $ l = - (\ln \eps) / (\ln 2) $ so that $  \eps^{\frac{1}{l}} \, 

2293: T = \frac{1}{2} \, T > 0 $.} which needs the introduction of 

2294: an {\it infinite} cascade of phase shifts. The intuition\footnote{

2295: Even at a formal level, difficulties occur in order to justify 

2296: the different convergences. Rigorous results in this direction

2297: seem to be a difficult task.} is that the graph of $ \tilde E $ 

2298: becomes continuous (no more gap). This corresponds to the 

2299: impression of an {\it infinite} cascade of energy. This 

2300: remark is consistent with engineering experiments and 

2301: the observations reported in the statistical approach.

2302: 

2303: \bigskip

2304: 

2305: The turbulent phenomena which we study are very complex in their

2306: realization. When $ t > 0 $, the description of $ \dot \u^\eps_\flat 

2307: (t,\cdot) $ involves an infinite set of phases so that computations 

2308: and representations are hard to implement. It gives the impression 

2309: of a chaos. Nevertheless, our analysis reveals that these phenomena 

2310: contain no mystery in their generation. On the contrary quantitative

2311: and qualitative features can be predicted in the framework of non 

2312: linear geometric optics. 

2313: 

2314: \smallskip

2315: 

2316: $\, $

2317: 

2318: \section{Euler equations in the variables $ (t,x,\theta) $.} 

2319: 

2320: As explained in the previous chapter, the demonstration of

2321: the Theorem \ref{appBKW} is achieved with the representation

2322: \begin{equation} \label{BKWdelbis}

2323: \ \tilde \u^\eps_\flat (t,x) = \tilde u^\eps_\flat \bigl( 

2324: t,x, \eps^{-1} \, \varphi^\eps_\flat (t,x) \bigr) \, , \quad 

2325: \ \tilde \p^\eps_\flat (t,x) = \tilde p^\eps_\flat \bigl( 

2326: t,x, \eps^{-1} \, \varphi^\eps_\flat (t,x) \bigr) \, . \ \

2327: \end{equation}

2328: Recall that the complete phase $ \varphi^\eps_\flat (t,x) $ is 

2329: \begin{equation} \label{phasecomplete}

2330: \varphi^\eps_\flat (t,x) \, = \, \varphi^\eps_g (t,x) + \eps \ 

2331: \varphi^\eps_a (t,x) \, = \, \varphi_0 (t,x) + \hbox{$ \sum_{

2332: k=1}^N $} \ \eps^{\frac{k}{l}} \ \varphi_k (t,x) \

2333: \end{equation}

2334: and that the profiles $ \tilde u^\eps_\flat ( t,x, \theta ) $ 

2335: and $ \tilde p^\eps_\flat ( t,x, \theta ) $ have the form

2336: \begin{equation} \label{feclate}

2337: \left. \begin{array}{l}

2338: \tilde u^\eps_\flat ( t,x, \theta ) \, = \, \u_0 (t,x) + 

2339: \sum_{k=1}^N \, \eps^{\frac{k}{l}} \ \tilde U_k ( t,x, 

2340: \theta) \, , \qquad \qquad \qquad \qquad \, \\

2341: \tilde p^\eps_\flat ( t,x, \theta ) \, = \, \p_0 (t,x) + 

2342: \sum_{k=1}^N \, \eps^{\frac{k}{l}} \ \tilde P_k ( t,x, 

2343: \theta) \, . 

2344: \end{array} \right. 

2345: \end{equation}

2346: 

2347: \subsection{Preliminaries.}

2348: 

2349: $ \bullet $ {\bf Anisotropic viscosity.} Mark the 

2350: abbreviated notations

2351: 

2352: \medskip

2353: 

2354: $ X^\eps_\flat (t,x) := \nabla \varphi^\eps_\flat (t,x) = 

2355: \sum_{k=0}^N \, \eps^{\frac{k}{l}} \ X_k (t,x) \, , \qquad 

2356: X_k (t,x) := \nabla \varphi_k (t,x) \, , $

2357: 

2358: \medskip

2359: 

2360: $ \mX^\eps_{\flat 1} (t,x) := \vert X^\eps_\flat (t,x) 

2361: \vert^{-1} \ X^\eps_\flat (t,x) \, . $

2362: 

2363: \medskip

2364: 

2365: \noindent{Complete the unit vector $ \mX^\eps_{\flat 1} (t,x) $  

2366: into some orthonormal basis of $ \RR^d $} 

2367: 

2368: \medskip

2369: 

2370: $ \mX^\eps_{\flat i} (t,x) \cdot \mX^\eps_{\flat j} (t,x) =

2371: \delta_{ij} \, , \qquad \forall \, (i,j) \in \{1, \cdots , 

2372: d \}^2 \, , $

2373: 

2374: \medskip

2375: 

2376: \noindent{so that all the vector fields $ \mX^\eps_{\flat i} $

2377: are smooth functions on $ [0,T] \times \RR^d $.} The 

2378: corresponding differential operators are denoted

2379: 

2380: \medskip

2381: 

2382: $ \mX^\eps_{\flat i} (\part ) := \mX^\eps_{\flat i} (t,x) 

2383: \cdot \nabla \, , \qquad i \in \{1, \cdots , d \} \, . $

2384: 

2385: \medskip

2386: 

2387: \noindent{Their adjoints are} 

2388: 

2389: \medskip

2390: 

2391: $ \mX^\eps_{\flat i} (\part )^* := \mX^\eps_{\flat i} (t,x) 

2392: \cdot \nabla + \Div \, ( \mX^\eps_{\flat i} ) (t,x) \, , 

2393: \qquad i \in \{1, \cdots , d \} \, . $

2394: 

2395: \medskip

2396: 

2397: \noindent{Select $ \mq \in C^\infty_b ( [0,T] \times \RR^d ;

2398: S^d_+ ) $ be such that}

2399: 

2400: \medskip

2401: 

2402: $ \exists \, c > 0 \, ; \qquad \mq (t,x) \geq c \, , \qquad 

2403: \forall \, (t,x) \in [0,T] \times \RR^d \, . $

2404: 

2405: \medskip

2406: 

2407: \noindent{Let $ (m,n ) \in \NN^2 $. Consider the elliptic 

2408: operator $ E^{\eps m}_{\flat n} (\part) $ defined according

2409: to}

2410: 

2411: \medskip

2412: 

2413: $ E^{\eps m}_{\flat n} (\part) \, := \,  \bigl( \,

2414: \eps^{\frac{m}{l}} \, \mX^\eps_{\flat 1} (\part )^* \, ,

2415: \, \eps^{\frac{n}{l}} \, \mX^\eps_{\flat 2} (\part )^* \, ,

2416: \, \cdots \, , \, \eps^{\frac{n}{l}} \, \mX^\eps_{\flat d} 

2417: (\part )^* \, \bigr) $

2418: $$ \qquad \qquad \qquad \qquad 

2419: \left( \begin{array} {cccc}

2420: \mq_{11} (t,x) & \mq_{12} (t,x) &\cdots & \mq_{1d} (t,x) \\

2421: \mq_{21} (t,x) & \mq_{22} (t,x) & \cdots & \mq_{2d} (t,x) \\

2422: \vdots & \vdots & \ & \vdots \\

2423: \mq_{d1} (t,x) & \mq_{d2} (t,x) & \cdots & \mq_{dd} (t,x) 

2424: \end{array} \right)

2425: \left( 

2426: \begin{array} {c}

2427: \eps^{\frac{m}{l}} \, \mX^\eps_{\flat 1} (\part ) \\

2428: \eps^{\frac{n}{l}} \, \mX^\eps_{\flat 2} (\part ) \\

2429: \vdots \\

2430: \eps^{\frac{n}{l}} \, \mX^\eps_{\flat d} (\part )

2431: \end{array} \right) . $$

2432: The introduction of the operator $ E^{\eps m}_{\flat n} (\part) $

2433: in the right of $ (\cE) $ is compatible with the propagation

2434: of oscillations only if $ m \geq l $ and $ n \geq 0 $. We

2435: retain the limit case $ l=m $ and $ n = 0 $. The other 

2436: situations are easier to deal with, at least when performing 

2437: formal computations.

2438: 

2439: \bigskip

2440: 

2441: \noindent{$ \bullet $ {\bf Interpretation in $ (t,x,\theta) $.}} 

2442: To deal with the variables $ (t,x,\theta) $, define

2443: 

2444: \medskip

2445: 

2446: $ \md_{j,\eps} \, := \, \eps \ \part_j \, + \, \part_j 

2447: \varphi^\eps_\flat \times \part_\theta \, , \qquad j \in 

2448: \{0, \cdots, d \} \, , $

2449: 

2450: \medskip

2451: 

2452: $ \md_\eps \, := \, ( \md_{1,\eps} , \cdots , \md_{d,\eps} ) 

2453: \, , $

2454: 

2455: \medskip

2456: 

2457: $ \Grade^\eps_\flat \, := \, {}^t ( \md_{1,\eps} , \cdots ,

2458: \md_{d,\eps} ) \, = \, \eps \ \nabla + X^\eps_\flat \times

2459: \partial_\theta \, , $

2460: 

2461: \medskip

2462: 

2463: $ \Dive^\eps_\flat \, := \, ( \Grade^\eps_\flat)^\star \,

2464: = \, \eps \ \Div + X^\eps_\flat \cdot \partial_\theta \, . $

2465: 

2466: \medskip

2467: 

2468: \noindent{The derivatives $ \mX^\eps_{\flat \dag} $ become}

2469: 

2470: \medskip

2471: 

2472: $ \eps \ \mX^\eps_{\flat 1} ( \md_\eps ) \, := \, \eps \

2473: \mX^\eps_{\flat 1} (\part ) \, + \, \vert X^\eps_\flat 

2474: (t,x) \vert \times \part_\theta \, , $

2475: 

2476: \medskip

2477: 

2478: $ \eps \ \mX^\eps_{\flat j} ( \md_\eps ) \, := \, \eps \

2479: \mX^\eps_{\flat j} (\part ) \, , \qquad \forall \, j \in 

2480: \{2, \cdots, d \} \, . $

2481: 

2482: \medskip

2483: 

2484: \noindent{The action of $ E^{\eps l}_{\flat 0} (\part) $ 

2485: expressed in the variables $ (t,x,\theta) $ gives rise to

2486: some negative differential operator of the order two, noted

2487: $ E^{\eps l}_{\flat 0} ( \md_\eps ) $.} The coefficients

2488: of the derivatives in $ E^{\eps l}_{\flat 0} ( \md_\eps ) $

2489: are of size one, except in front of $ \mX^\eps_{\flat 1} 

2490: (\part ) $. To avoid technicalities and to simplify the

2491: notations, we substitute the Laplacian $ \nu \, \Delta $

2492: for $ E^{\eps l}_{\flat 0} ( \md_\eps ) $. 

2493: 

2494: \medskip

2495: 

2496: \noindent{When $ \nu = 0 $, we recover Euler equations.} 

2497: When $ \nu > 0 $, the action $ \nu \, \Delta $ can be 

2498: viewed as the `trace' in $ (t,x,\theta) $ of the anisotropic 

2499: viscosity $ E^{\eps l}_{\flat 0} (\part) $. Now, consider 

2500: the Cauchy problem

2501: \begin{equation} \label{cauchypr}

2502: \ \left \lbrace \begin{array} {ll}

2503: \md_{0,\eps} \, \tilde u^\eps_\flat + ( \tilde u^\eps_\flat

2504: \cdot \Grade^\eps_\flat) \, \tilde u^\eps_\flat \! \! \! & 

2505: + \, \Grade^\eps_\flat \, \tilde p^\eps_\flat \\

2506: \ & = \, \nu \ \eps \ \Delta \, \tilde u^\eps_\flat + \tilde 

2507: f^\eps_\flat \, , \qquad \Dive^\eps_\flat \, \tilde 

2508: u^\eps_\flat = \tilde g^\eps_\flat \, , \qquad \\ 

2509: \tilde u^\eps_\flat (0,\cdot) \, = \, \tilde h^\eps_\flat 

2510: (\cdot) \, , & \

2511: \end{array} \right. 

2512: \end{equation}

2513: with given data 

2514: 

2515: \medskip

2516: 

2517: $ \tilde f^\eps_\flat \in \cW^\infty_T \, , \qquad \tilde 

2518: g^\eps_\flat \in \cW^\infty_T \, , \qquad \tilde h^\eps_\flat 

2519: \in H^\infty \, . $ 

2520: 

2521: \medskip

2522: 

2523: \noindent{Suppose that $ \nu = 0 $ and select some smooth 

2524: solution $ ( \tilde u^\eps_\flat , \tilde p^\eps_\flat ) $ 

2525: of (\ref{cauchypr}). The expressions $ \tilde \u^\eps_\flat $ 

2526: and $ \tilde \p^\eps_\flat $ given by the formula (\ref{BKWdelbis}) 

2527: are subjected to

2528: \begin{equation} \label{cauchyprbf}

2529: \left \lbrace \begin{array} {l}

2530: \part_t \tilde \u^\eps_\flat + ( \tilde \u^\eps_\flat \cdot \nabla) 

2531: \tilde \u^\eps_\flat + \nabla \tilde \p^\eps_\flat \, = \, \tilde 

2532: \f^\eps_\flat \, , \qquad \Div \, \tilde \u^\eps_\flat = \tilde 

2533: \g^\eps_\flat \, , \qquad \ \, \\

2534: \tilde \u^\eps_\flat (0,\cdot) \, = \, \tilde \h^\eps_\flat (\cdot) \, , 

2535: \end{array} \right. \qquad 

2536: \end{equation}

2537: where the functions $ \tilde \f^\eps_\flat (t,x) $, $ \tilde 

2538: \g^\eps_\flat (t,x) $ and $ \tilde \h^\eps_\flat (t,x) $ are 

2539: obtained by replacing the variable $ \theta $ by $ \varphi^\eps_\flat 

2540: (t,x) $ in the expressions $ \eps^{-1} \ \tilde f^\eps_\flat 

2541: ( t,x, \theta ) $, $ \eps^{-1} \ \tilde g^\eps_\flat ( t,x, 

2542: \theta ) $ and $ \tilde h^\eps_\flat (t,x, \theta) $. In other 

2543: words, any solution of (\ref{cauchypr}) with $ \nu = 0 $ yields 

2544: a solution of (\ref{cauchyprbf}). From now on, we proceed directly 

2545: with the relaxed system (\ref{cauchypr}).

2546: 

2547: \smallskip

2548: 

2549: \subsection{The BKW analysis.}

2550: 

2551: Select a smooth solution $ \u_0(t,x) \in \cW^\infty_T $ of

2552: 

2553: \medskip

2554: 

2555: $ \part_t \u_0 + (\u_0 \cdot \nabla ) \, \u_0 + \nabla \p_0 \,

2556: = \, \nu \ \Delta_x \u_0 \, , \qquad \Div \ \u_0 = 0 \, . $

2557: 

2558: \medskip

2559: 

2560: \noindent{Choose a phase $ \varphi_0 (t,x) \in C^1 ([0,T] 

2561: \times \RR^d) $ with $ \nabla \varphi_0 (t,x) \in C^\infty_b 

2562: ([0,T] \times \RR^d) $. Suppose moreover that it satisfies 

2563: the eiconal equation $ (ei) $ and the condition (\ref{nonstaou}).} 

2564: The main step in the construction of approximate solutions is the 

2565: following intermediate result.

2566: 

2567: \smallskip

2568: 

2569: \begin{prop} \label{appBKWinter} Select any $ \flat = (l,N) \in \NN^2 $ 

2570: such that $ 0 < l < N $. Consider the following initial data

2571: 

2572: \medskip

2573: 

2574: $ \tilde U_{k0}^* (x,\theta) = \Pi_0 (0,x) \, \tilde U_{k0}^* 

2575: (x,\theta) \in H^\infty \, , \qquad  1 \leq k \leq N \, , $

2576: 

2577: \smallskip

2578: 

2579: $ \langle \tilde U_{k0} \rangle (x) \in H^\infty \, , 

2580: \qquad 1 \leq k \leq N \, , $

2581: 

2582: \smallskip

2583: 

2584: $ \varphi_{k0} (x) \in H^\infty \, , \qquad  \quad \!

2585: 1 \leq k \leq N \, . $

2586: 

2587: \medskip

2588: 

2589: \noindent{There are finite sequences $ \{ \tilde U_k \}_{1 \leq 

2590: k \leq N} $ and $ \{ \tilde P_k \}_{1 \leq k \leq N} $ with}

2591: 

2592: \medskip

2593: 

2594: $ \tilde U_k (t,x,\theta) \in \cW^\infty_T \, , \qquad \tilde 

2595: P_k (t,x,\theta) \in \cW^\infty_T \, , \qquad 1 \leq k \leq N \, , $

2596: 

2597: \medskip

2598: 

2599: \noindent{and a finite sequence $ \{ \varphi_k \}_{1 \leq k \leq N} $ 

2600: with}

2601: 

2602: \medskip

2603: 

2604: $ \varphi_k (t,x) \in \cW^\infty_T \, , \qquad 1 \leq k \leq N \, , $

2605: 

2606: \medskip

2607: 

2608: \noindent{which are such that}

2609: 

2610: \medskip

2611: 

2612: $ \Pi_0 (0,x) \, \tilde U_k^* (0,x,\theta) =  \Pi_0 (0,x) \, 

2613: \tilde U_{k0}^* (x,\theta) \, , \qquad 1 \leq k \leq N \, , $

2614: 

2615: \smallskip

2616: 

2617: $ \langle \tilde U_k \rangle (0,x) = \langle \tilde U_{k0} 

2618: \rangle (x) \, , \qquad \! 1 \leq k \leq N \, , $

2619: 

2620: \smallskip

2621: 

2622: $ \varphi_k (0,x) = \varphi_{k0} (x) \, , \qquad \quad

2623: \ 1 \leq k \leq N \, . $ 

2624: 

2625: \medskip

2626: 

2627: \noindent{Define $ \varphi^\eps_\flat $ as in (\ref{phasecomplete}).}

2628: All the preceding expressions are adjusted so that the functions 

2629: $ \tilde u^\eps_\flat $ and $ \tilde p^\eps_\flat $ associated with 

2630: the expansions in (\ref{feclate}) are approximate solutions 

2631: on the interval $ [0,T] $. More precisely, they satisfy  

2632: (\ref{cauchypr}) with 

2633: \begin{equation} \label{conclufi}

2634: \tilde h^\eps_\flat (x, \theta) \, = \, \u_0 (0,x) + \hbox{$ \sum_{

2635: k=1}^N $} \, \eps^{\frac{k}{l}} \ \tilde U_{k0} (x,\theta) \qquad 

2636: \qquad \qquad

2637: \end{equation}

2638: and we have

2639: \begin{equation} \label{conclufiaussi}

2640: \{ \tilde f^\eps_\flat \}_\eps = \bigcirc( \eps^{\frac{N+1}{l}}) 

2641: \, , \qquad \{ \tilde g^\eps_\flat \}_\eps = \bigcirc( \eps^{

2642: \frac{N+1}{l}}) \, . \qquad \qquad \qquad \quad

2643: \end{equation}

2644: 

2645: \end{prop}

2646: 

2647: \medskip

2648: 

2649: \noindent{$ \bullet $ {\bf Proof of the Proposition \ref{appBKWinter}.}}

2650: For convenience, we will drop in this paragraph the tilde '$\, \tilde \ \, $' 

2651: on the profiles $ u^\eps_\flat $, $ p^\eps_\flat $, $ U_k $ and $ P_k $.

2652: This modification concerns only this demonstration. We hope that it 

2653: will not induce confusions: we still work here with the complete phase 

2654: $ \varphi^\eps_\flat $.

2655:  

2656: \medskip

2657: 

2658: \noindent{Because of (\ref{nondegeat}) we can define the application 

2659: $ \Pi^\eps_\flat (t,x) $ which is the orthogonal projector on the 

2660: hyperplane $ \nabla \varphi^\eps_\flat (t,x)^\perp \subset \RR^d $.} 

2661: We adopt the convention

2662: 

2663: \medskip

2664: 

2665: $ \Pi^\eps_\flat (t,x) = \sum_{k=0}^\infty \, \eps^{\frac{k}{l}} \ 

2666: \Pi_k (t,x) \, , \qquad \Pi_k \in \cW^\infty_T \, , \qquad \eps 

2667: \in \, ]0, \eps_0] \, . $ 

2668: 

2669: \medskip

2670: 

2671: \noindent{The access to $ \Pi_k $ needs only the knowledge of the $ X_j $ 

2672: for $ j \leq k $.} Introduce

2673: 

2674: \medskip

2675: 

2676: $ v^\eps_\flat := X^\eps_\flat \cdot u^\eps_\flat = \sum_{k=0}^\infty \,

2677: \eps^{\frac{k}{l}} \ V_k \, , \qquad \! V_k = X_k \cdot \u_0 + 

2678: \sum_{j=0}^{k-1} \, X_j \cdot U_{k-j} \, , $

2679: 

2680: \medskip

2681: 

2682: $ w^\eps_\flat := \Pi^\eps_\flat \, u^\eps_\flat = \sum_{k=0}^\infty \, 

2683: \eps^{\frac{k}{l}} \ W_k \, , \qquad W_k = \Pi_k \, \u_0 + \sum_{j=0}^{k-1} 

2684: \, \Pi_j \, U_{k-j} \, . $

2685: 

2686: \medskip

2687: 

2688: \noindent{By construction}

2689: 

2690: \medskip

2691: 

2692: $ u^\eps_\flat \, = \, v^\eps_\flat \ \vert X^\eps_\flat \vert^{-2} 

2693: \ X^\eps_\flat + w^\eps_\flat \, , \qquad U_k \, = \, V_k \ \vert X_0 

2694: \vert^{-2} \  X_0 \, + W_k + h $

2695: 

2696: \medskip

2697: 

2698: \noindent{where $ h $ depends only on the $ X_j $ for $ j \leq k $ 

2699: and on the $ U_j $ for $ j \leq k-1 $.} 

2700: 

2701: \medskip

2702: 

2703: \noindent{The conditions prescribed in the Proposition \ref{appBKWinter} 

2704: on the initial data $ U_{k0}^* $ allow to fix the functions $ \nabla 

2705: \varphi_0(0,x) \cdot U_k^* (0,x,\theta) $ as we want.} Since

2706: 

2707: \medskip

2708: 

2709: $ V_k^* \, = \, \nabla \varphi_0 \cdot U_k^* \, + \, \sum_{j=1}^{k-1} \, 

2710: X_j \cdot U_{k-j}^* \, , $ 

2711: 

2712: \medskip

2713: 

2714: \noindent{the same is true (by induction) for the components $ V_k^* 

2715: (0,x,\theta) $.} To begin with, we impose the polarization conditions

2716: \begin{equation} \label{polafirst}

2717: P_k^* \, \equiv \, V_k^* \, \equiv \, 0 \, , \qquad \forall \, k \in 

2718: \{ 1,\cdots,l \} \qquad \qquad \qquad \qquad \

2719: \end{equation}

2720: and we adjust {\it a priori} the geometrical phase $ \varphi^\eps_g $  

2721: so that

2722: \begin{equation} \label{shiftfirst}

2723: \part_t \varphi_k + \bar V_k = 0 \, , \qquad \forall \, k \in \{ 1,

2724: \cdots, l-1 \} \qquad \qquad \qquad \qquad \quad

2725: \end{equation}

2726: which implies that

2727: 

2728: \medskip

2729: 

2730: $ \part_t \varphi^\eps_g + (\bar u^\eps_\flat \cdot \nabla)

2731: \varphi^\eps_g \, = \, \sum_{k=l}^\infty \, \eps^{\frac{k}{l}} 

2732: \ \bar V_k \, = \, \bigcirc (\eps) \, . $

2733: 

2734: \medskip

2735: 

2736: \noindent{It amounts to the same thing to look at the equations in

2737: (\ref{cauchypr}) or at the following singular system (we drop here 

2738: the indices $ \eps $ and $ \flat $ at the level of $ u^\eps_\flat $, 

2739: $ v^\eps_\flat $, $ w^\eps_\flat $, $ p^\eps_\flat $ , $ \Pi^\eps_\flat $

2740: and $ \varphi^\eps_\flat $)}

2741: \begin{equation} \label{singu}

2742: \left \{ \begin{array}{ll}

2743: \part_t u + ( u \cdot \nabla ) \, u + \nabla p & \! \! \! + \, 

2744: \eps^{-1} \ (\part_t \varphi + v) \ \partial_\theta u \quad \\

2745: & \! \! \! + \, \eps^{-1} \ \part_\theta p \ \nabla \varphi \, 

2746: = \, \nu \ \Delta \, u \, , \\

2747: \Div \ u + \eps^{-1} \ \part_\theta v = 0 \, . & \

2748: \end{array} \right. \qquad \qquad \

2749: \end{equation}

2750: The functions $ v $ is subjected to

2751: \begin{equation} \label{v}

2752: \left. \begin{array}{rl}

2753: \part_t v \! \! \! \! & + \, (u \cdot \nabla) \, v + X \cdot \nabla p 

2754: + \eps^{-1} \ (\part_t \varphi + v) \ \partial_\theta v  \\

2755: \ & + \, \eps^{-1} \ \part_\theta p \ \parallel X \parallel^2 

2756: - \bigl( \part_t X + (u \cdot \nabla) \, X \bigr) \cdot u  = 

2757: \nu \  X \cdot \Delta \, u \, .

2758: \end{array} \right.

2759: \end{equation}

2760: The functions $ w $ satisfies

2761: \begin{equation} \label{w}

2762: \left. \begin{array}{rl}

2763: \part_t w \! \! \! \! & + \, (u \cdot \nabla) \, w + \Pi \, \nabla 

2764: p + \eps^{-1} \ (\part_t \varphi + v) \ \partial_\theta w \qquad 

2765: \qquad \quad \ \ \ \\

2766: \ & - \, \bigl( \part_t \Pi + (u \cdot \nabla) \, \Pi \bigr) \, u 

2767: = \nu \ \Pi \, \Delta \, u \, .

2768: \end{array} \right.

2769: \end{equation}

2770: Substitute the expressions $ u^\eps_\flat $ and $ p^\eps_\flat $ 

2771: given by (\ref{feclate}) into (\ref{singu}). Then arrange the terms 

2772: according to the different powers of $ \eps $ which are in factor. 

2773: The contributions coming from the orders $ \eps^{\frac{1}{l}-1} $, 

2774: $ \cdots $, $ \eps^{-\frac{1}{l}} $ and $ \eps^0 $ are eliminated 

2775: through (\ref{polafirst}), (\ref{shiftfirst}) and the constraints 

2776: imposed on $ (\u_0,\p_0) $. 

2777: 

2778: \medskip

2779: 

2780: \noindent{Now, look at the terms in front of $\eps^{\frac{j}{l}} $ 

2781: with $ j \in \NN_* $.} It remains

2782: \begin{equation} \label{eqgj}

2783: \left \{ \begin{array}{l}

2784: \part_t U_j + \sum_{k=0}^j \, ( U_k \cdot \nabla) \, U_{j-k} + 

2785: \nabla P_j + \sum_{k=0}^j \, \part_\theta P_{l+k} \ \nabla 

2786: \varphi_{j-k} \\

2787: \qquad \, + \, \sum_{k=0}^{j-1} \, (\part_t \varphi_{l+k} + V_{l+k}) \ 

2788: \part_\theta U_{j-k} \, = \, \nu \ \Delta \, U_j \, , \\

2789: \Div \ U_j + \part_\theta V_{j+l} = 0 \, .

2790: \end{array} \right. \quad

2791: \end{equation}

2792: Proceed in a similar manner with (\ref{v}). Just arrange the terms

2793: which come from $ X \cdot \Delta \, u $ and which do not involve 

2794: $ U_j $ in a source term $ \cH^j_V $

2795: \begin{equation} \label{eqgjeg}

2796: \left. \begin{array}{rl}

2797: \, \part_t V_j \! \! \! \! & + \, \sum_{k=0}^j \, (U_k \cdot \nabla) \, 

2798: V_{j-k} + \sum_{k=0}^{j-1} \, (\part_t \varphi_{l+k} + V_{l+k}) \

2799: \part_\theta V_{j-k} \\

2800: \ & + \, \sum_{k=0}^j \, X_k \cdot \nabla P_{j-k} \, - \, \sum_{k=0}^j \, 

2801: \part_t X_k \cdot U_{j-k} \\

2802: \ & - \, \sum_{k=0}^j \, \bigl( \sum_{l=0}^k \, (U_{k-l} \cdot \nabla) \, 

2803: X_l \bigr) \cdot U_{j-k} \\

2804: \ & + \, \sum_{k=1}^{j-1} \, \bigl( \sum_{l=0}^k \, X_l \cdot X_{k-l} 

2805: \bigr) \ \part_\theta P_{j+l-k} + \vert X_0 \vert^2 \

2806: \part_\theta P_{j+l} \\

2807: \ & = \, \nu \ X_0 \cdot \Delta \, U_j + \cH^j_V (t,x,\theta,U_1,X_1, 

2808: \cdots, U_{j-1},X_{j-1}, X_j) \, . \ \

2809: \end{array} \right.

2810: \end{equation}

2811: The same operation with (\ref{w}) yields 

2812: \begin{equation} \label{eqgjel}

2813: \, \left. \begin{array}{rl}

2814: \part_t W_j \! \! \! \! & + \, \sum_{k=0}^j \, (U_k \cdot \nabla) \, 

2815: W_{j-k} \, + \, \sum_{k=0}^{j-1} \, (\part_t \varphi_{l+k} + V_{l+k}) \ 

2816: \part_\theta W_{j-k} \quad \\

2817: \ & + \, \sum_{k=0}^j \, \Pi_k \, \nabla P_{j-k} \, - \, \sum_{k=0}^j \, 

2818: \part_t \Pi_k \, U_{j-k} \\

2819: \ & - \, \sum_{k=0}^j \, \bigl( \sum_{l=0}^k \, (U_{k-l} \cdot \nabla)

2820: \, \Pi_l \bigr) \,U_{j-k} \\ 

2821: \ & = \, \nu \ \Pi_0 \ \Delta \, U_j + \cH^j_W (t,x,\theta,U_1,X_1, 

2822: \cdots, U_{j-1},X_{j-1}, X_j) \, .

2823: \end{array} \right.

2824: \end{equation}

2825: Then extract the mean value of (\ref{eqgj}) 

2826: \begin{equation} \label{moyj}

2827: \ \left \{ \begin{array}{l}

2828: \part_t \bar U_j + (\u_0 \cdot \nabla) \, \bar U_j + (\bar U_j \cdot 

2829: \nabla) \, \u_0 + \nabla \bar P_j \qquad \qquad \qquad \qquad \qquad \\

2830: \qquad \, + \, \sum_{k=1}^{j-1} \, \langle ( U_k \cdot \nabla) \, 

2831: U_{j-k} \rangle \\

2832: \qquad \, + \sum_{k=1}^{j-1} \, \langle V^*_{l+k} \ \part_\theta 

2833: U_{j-k} \rangle \, = \, \nu \ \Delta_x \, \bar U_j \, , \\

2834: \Div \ \bar U_j = 0 \, .

2835: \end{array} \right.

2836: \end{equation}

2837: Observe also that

2838: \begin{equation} \label{divj}

2839: V^*_{j+l} \, = \, - \, \Div \ \partial_\theta^{-1} U^*_j \, , \qquad 

2840: \forall \, j \in \NN_* \, . \qquad \qquad \qquad \qquad \qquad

2841: \end{equation}

2842: Consider the inductive reasoning based on

2843: 

2844: \medskip

2845: 

2846: \noindent{\bf Hypothesis} $ (H_j) \, $:

2847: 

2848: \medskip

2849: 

2850: {\em

2851: \noindent{i) The expressions $ U_1 $, $ \cdots $, $ U_j $ and  $ P_1 $, 

2852: $ \cdots $, $ P_j $ are known.}

2853: 

2854: \medskip

2855: 

2856: \noindent{ii) The phases $ \varphi_1 $, $\cdots $, $ \varphi_j $ are 

2857: identified. The same is true for the vectors $ X_1 $, $\cdots $, $ X_j $ 

2858: and the projectors $ \Pi_1 $, $\cdots $, $ \Pi_j $. Moreover, the 

2859: following relations are satisfied}

2860: \begin{equation} \label{phik}

2861: \part_t \varphi_{j+k} + \bar V_{j+k} = 0 \, , \qquad \forall \, k \in 

2862: \{ 1, \cdots, l-1 \} \, . 

2863: \qquad \end{equation}

2864: iii) The correctors $ V^*_{j+1} $, $ \cdots $, $ V^*_{j+l} $ and 

2865: $ P^*_{j+1} $, $ \cdots $, $ P^*_{j+l} $ are identified and

2866: \begin{equation} \label{vk+n}

2867: V^*_{j+k} = \, - \, \Div \ \part_\theta^{-1} U^*_{j+k-l} \, , \qquad 

2868: \forall \, k \in \{1, \cdots, l \}  \, . \qquad \quad

2869: \end{equation}

2870: }

2871: 

2872: \noindent{- \underline{Verification} \underline{of} $ (H_1) $}. The mean

2873: value $ \bar U_1 $ is obtained by solving 

2874: 

2875: \medskip

2876: 

2877: $ \part_t \bar U_1 + (\u_0 \cdot \nabla) \, \bar U_1 + (\bar U_1 \cdot 

2878: \nabla) \, \u_0 + \nabla \bar P_1 = \nu \ \Delta_x \, \bar U_1 \, , 

2879: \qquad \Div \ \bar U_1 = 0 \, . $

2880: 

2881: \medskip

2882: 

2883: \noindent{Using (\ref{shiftfirst}), it allows to determine $ \varphi_1 $.}

2884: Now look at the oscillating part of (\ref{eqgjel}) with the indice $ j = 1 $.

2885: The constraint on $ W_1^* \equiv U_1^* $ writes

2886: 

2887: \medskip

2888: 

2889: $ \part_t W_1^* + ( \u_0 \cdot \nabla) \, W_1^* + ( \part_t \varphi_l 

2890: + \bar V_l) \ \part_\theta W_1^* = M \, W_1^* + \nu \ \Pi_0 \ \Delta

2891: \, W^*_1 $

2892: 

2893: \medskip

2894: 

2895: \noindent{where $ M $ is the linear application}

2896: 

2897: \medskip

2898: 

2899: $ M \, U \, := \, (\part_t \Pi_0) \, U + \bigl( (\u_0 \cdot \nabla) 

2900: \Pi_0 \bigr) \, U - \Pi_0 \, (U \cdot \nabla ) \u_0 \, . $

2901: 

2902: \medskip

2903: 

2904: \noindent{We impose (\ref{phik}) for $ j=1 $.} In view of 

2905: (\ref{shiftfirst}), it reduces to

2906: 

2907: \medskip

2908: 

2909: $ \part_t \varphi_l + \bar V_l = 0 \, . $

2910: 

2911: \medskip

2912: 

2913: \noindent{The link between $ W_1^* $ and $ \bar V_l $ is 

2914: removed.} It remains the linear equation

2915: 

2916: \medskip

2917: 

2918: $ \part_t W_1^* + ( \u_0 \cdot \nabla) \, W_1^* = M \, W_1^* + \nu \ 

2919: \Pi_0 \ \Delta \, W^*_1 \, . $

2920: 

2921: \medskip

2922: 

2923: \noindent{When $ \nu = 0 $, the profile $ W^*_1 $ is obtained 

2924: by integrating along the characteristic curves $ \Gamma (\cdot,x) $.

2925: This justifies the remark 2.2.4 for $ k = 1 $. Observe also that the 

2926: polarization condition $ W^*_1 = \Pi_0 \, W^*_1 $ is conserved since 

2927: the equation given for $ W_1^* $ is equivalent to}

2928: \begin{equation} \label{eqW} 

2929: \ \left \{ \begin{array}{l}

2930: \Pi_0 \, \bigl \lbrack \part_t W_1^* + ( \u_0 \cdot 

2931: \nabla) \, W_1^* + (W_1^* \cdot \nabla) \u_0  \bigr \rbrack \, 

2932: = \, \nu \ \Pi_0 \ \Delta \, W^*_1 \, , \qquad  \quad \\

2933: W_1^* \, = \, \Pi_0 \, W^*_1 \, .   

2934: \end{array} \right. 

2935: \end{equation}

2936: Introduce the linear form 

2937: 

2938: \medskip

2939: 

2940: $ \ell \, U \, := \ \vert X_0 \vert^{-2} \ \bigl \lbrack \, 

2941: \part_t X_0 \cdot U + \bigl( (\u_0 \cdot \nabla) X_0 \bigr) 

2942: \cdot U - X_0 \cdot \bigl( (U \cdot \nabla ) \u_0 \bigr) \, 

2943: \bigr \rbrack \, . $

2944: 

2945: \medskip

2946: 

2947: \medskip

2948: 

2949: \noindent{The constraint $ V_1^* \equiv 0 $ is equivalent to}

2950: 

2951: \medskip

2952: 

2953: $ P^*_{l+1} \, = \, \ell \ \part_\theta^{-1} W_1^* \, + \, 

2954: \nu \ \vert X_0 \vert^{-2} \ X_0 \cdot \part_\theta^{-1} \, 

2955: \Delta \, W^*_1 \, , $ 

2956: 

2957: \medskip

2958: 

2959: \noindent{We have also}

2960: 

2961: \medskip

2962: 

2963: $ V^*_{l+1} = \, - \, \Div \ \part_\theta^{-1} W_1^* \, . $

2964: 

2965: \medskip

2966: 

2967: \noindent{At this stage, we know who is $ U_1 \equiv \bar U_1 

2968: + W_1^* $ and $ P_1 \equiv 0 $.} Moreover, we have the relations 

2969: (\ref{phik}) and (\ref{vk+n}). Thus, the hypothesis $ (H_1) $ 

2970: is verified.

2971: 

2972: \bigskip

2973: 

2974: \noindent{- \underline{The} \underline{induction}.} Suppose 

2975: that the conditions given in $ (H_j )$ are satisfied. The 

2976: question is to obtain $ (H_{j+1} )$. Consider first (\ref{moyj}) 

2977: with the indice $ j+1 $. The relation (\ref{vk+n}) induces 

2978: simplifications. It remains

2979: \begin{equation} \label{moyj+1}

2980: \left \{ \begin{array}{l}

2981: \part_t \bar U_{j+1} + (\u_0 \cdot \nabla) \, \bar U_{j+1} 

2982: + (\bar U_{j+1} \cdot \nabla) \, \u_0 \qquad \qquad \\

2983: \qquad + \, \sum_{k=1}^{j} \, ( \bar U_k \cdot \nabla) \, 

2984: \bar U_{j+1-k} + \, \sum_{k=1}^j \, \Div \ \langle 

2985: U^*_k \otimes U^*_{j+1-k} \rangle \\

2986: \qquad + \, \nabla \bar P_{j+1} \, = \, \nu \ \Delta_x 

2987: \, \bar U_{j+1} \, , \qquad \Div \ \bar U_{j+1} = 0 \, .

2988: \end{array} \right.

2989: \end{equation}

2990: This system gives access to $ \bar U_{j+1} $ and $ \bar P_{j+1} $. 

2991: For $ j=1 $, it yields

2992: \begin{equation} \label{moy2}

2993: \ \left \{ \begin{array}{l}

2994: \part_t \bar U_2 + (\u_0 \cdot \nabla) \, \bar U_2 + (\bar U_2 \cdot 

2995: \nabla) \, \u_0 + \nabla \bar P_2 \qquad \qquad \qquad \quad \ \ \, \\

2996: \quad \ + \,( \bar U_1 \cdot \nabla) \, \bar U_1 + \Div \ \langle 

2997: U^*_1 \otimes U^*_1 \rangle = \, \nu \ \Delta \bar U_2 \, , \quad

2998: \ \Div \ \bar U_2 = 0 \, .

2999: \end{array} \right.

3000: \end{equation}

3001: Because of (\ref{nonnontri}), the source term $ \langle U^*_1 \otimes

3002: U^*_1 \rangle $ is sure to be non trivial. We recover here that in

3003: general $ \bar U_2 \not \equiv 0 $ even if $ \bar U_1 (0,\cdot) 

3004: \equiv \bar U_2  (0,\cdot) \equiv 0 $. The term $ \bar U_2 $ 

3005: excites $ \varphi_2 $ through (\ref{trans2}). Generically, we 

3006: have $ \varphi_2 \not \equiv 0 $ even if

3007: 

3008: \medskip

3009: 

3010: $ \bar U_1 (0,\cdot) \, \equiv \, \bar U_2  (0,\cdot) \, \equiv \, 0 \, , 

3011: \qquad \varphi_1 (0,\cdot) \, \equiv \, \varphi_2  (0,\cdot) \, \equiv 

3012: \, 0 \, . $

3013: 

3014: \medskip

3015: 

3016: \noindent{Observe however that exceptions can happen (see the 

3017: remark 3.5.4).} The information (\ref{phik}) for $ k=1 $ means 

3018: that

3019: 

3020: \medskip

3021: 

3022: $ \part_t \varphi_{j+1} + (\u_0 \cdot \nabla) \, \varphi_{j+1} 

3023: + X_0 \cdot \bar U_{j+1} + \sum_{l=1}^j \, X_l \cdot \bar U_{j

3024: +1-l} \, = \, 0 \, . $

3025: 

3026: \medskip

3027: 

3028: \noindent{Deduce $ \varphi_{j+1} $ from this equation, and 

3029: therefore $ X_{j+1} $ and $ \Pi_{j+1} $.} Complete with the 

3030: triangulation condition

3031: \begin{equation} \label{phibi}

3032: \part_t \varphi_{j+l} + \bar V_{j+l} = 0 \, . \qquad \qquad \qquad

3033: \qquad \qquad \qquad \qquad \qquad \quad \quad

3034: \end{equation}

3035: Then extract the oscillating part of (\ref{eqgjel}) written 

3036: with $ j+1 $. Use $ (H_j) $ and (\ref{phibi}) in order to 

3037: simplify the resulting equation. It yields

3038: \begin{equation} \label{oscij+1}

3039: \ \part_t W_{j+1}^* + ( \u_0 \cdot \nabla) \, W_{j+1}^* = M \, 

3040: W_{j+1}^* + \nu \ \Pi_0 \ \Delta \, W^*_{j+1} + f \qquad \qquad \

3041: \end{equation}

3042: where $ f $ is known. We get $ W^*_{j+1} $ by solving (\ref{oscij+1}).

3043: Therefore we have $ U^*_{j+1} $ and we can deduce $ V^*_{j+l+1} = \, 

3044: - \, \Div \ \part_\theta^{-1} U^*_{j+1} $.

3045: 

3046: \medskip

3047: 

3048: \noindent{Now look at the constraint (\ref{eqgjeg}) for the 

3049: indice $ j+1 $.} Extract the oscillating part. It allows to 

3050: recover $ P^*_{j+l+1} $. Thus we have $ (H_{j+1}) $. 

3051: 

3052: \medskip

3053: 

3054: \noindent{Apply the induction up to $ j = N -l $.} It yields 

3055: $ U_1 $, $ \cdots $, $ U_N $. Construct oscillations 

3056: $ \tilde u^\eps_\flat $ and $ \tilde p^\eps_\flat $ by way of 

3057: (\ref{feclate}). It furnishes source terms $ \tilde f^\eps_\flat $ 

3058: and $ \tilde g^\eps_\flat $ through (\ref{cauchypr}). By 

3059: construction, we recover (\ref{conclufiaussi}). \hfill $ \Box $

3060: 

3061: \subsection{Divergence free approximate solutions in $ (t,x, \theta) $.} 

3062: 

3063: In this subsection 4.3, we impose on $ \varphi_0 $ a constraint 

3064: which is more restrictive than (\ref{nonstaou}). We suppose that 

3065: we can find a direction $ \zeta \in \RR^d \setminus \{ 0 \} $ such

3066: that

3067: \begin{equation} \label{nonstaoubis} 

3068: \exists \, c > 0 \, ; \qquad \nabla \varphi_0 (t,x) \cdot \zeta \, 

3069: \geq \, c \, , \qquad \forall \, (t,x) \in [0,T] 

3070: \times \RR^d \, . \quad \

3071: \end{equation}

3072: 

3073: \begin{prop} \label{complement} The assumptions are as in

3074: the Proposition \ref{appBKWinter}. The profiles $  \tilde 

3075: U_k $, $  \tilde P_k $, and the phases $  \varphi_k $ are

3076: defined in the same way. Then, there are correctors

3077: 

3078: \medskip

3079: 

3080: $ \tilde {c u}^\eps_\flat (t,x,\theta) \in \cW^\infty_T \, , 

3081: \qquad \{ \tilde{c u}^\eps_\flat \}_\eps \, = \, \bigcirc(

3082: \eps^{\frac{N}{l}}) $ 

3083: 

3084: \medskip

3085: 

3086: \noindent{such that the functions $ \tilde u^\eps_\flat $ 

3087: and $ \tilde p^\eps_\flat $ defined according to} 

3088: $$ \, \left. \begin{array}{l}

3089: \tilde u^\eps_\flat (t,x) := \u_0 (t,x) + \sum_{k=1}^N \, 

3090: \eps^{\frac{k}{l}} \ \tilde U_k \bigl( t,x, \eps^{-1} \, 

3091: \varphi^\eps_g (t,x) \bigr) + \tilde {c u}^\eps_\flat (t,x)  

3092: \qquad \ \ \\

3093: \tilde p^\eps_\flat (t,x) := \p_0 (t,x) + \sum_{k=1}^N \, 

3094: \eps^{\frac{k}{l}} \ \tilde P_k \bigl( t,x, \eps^{-1} \, 

3095: \varphi^\eps_g (t,x) \bigr) 

3096: \end{array} \right. $$

3097: satisfy the Cauchy problem

3098: \begin{equation} \label{eclacpa}

3099: \ \left \lbrace \begin{array} {l}

3100: \md_{0,\eps} \, \tilde u^\eps_\flat + ( \tilde u^\eps_\flat

3101: \cdot \Grade^\eps_\flat) \, \tilde u^\eps_\flat + 

3102: \Grade^\eps_\flat \, \tilde p^\eps_\flat \\

3103: \qquad \qquad \qquad \qquad \qquad \! = \, \nu \ \eps \ 

3104: \Delta \, \tilde u^\eps_\flat + \tilde f^\eps_\flat \, , 

3105: \qquad \Dive^\eps_\flat \, \tilde u^\eps_\flat = 0 \qquad \\ 

3106: \tilde u^\eps_\flat (0,x,\theta) \, = \,  \u_0 (0,x) + 

3107: \sum_{k=1}^N \, \eps^{\frac{k}{l}} \ \tilde U_{k0} (x,\theta)

3108: \end{array} \right.

3109: \end{equation}

3110: and we still have $ \{ \tilde f^\eps_\flat \}_\eps \, = \, 

3111: \bigcirc(\eps^{\frac{N+1}{l}}) $.

3112: 

3113: \end{prop}

3114: 

3115: \noindent{We need some material before proving the Proposition

3116: \ref{complement}.}

3117: 

3118: \medskip

3119: 

3120: \noindent{$ \bullet $ {\bf The divergence free relation in 

3121: the variables $ (t,x,\theta) $.}} We can select some special 

3122: right inverse of the application $ \Dive^\eps_\flat \, : \, 

3123: H^{\infty *}_T \longrightarrow H^{\infty *}_T $.} 

3124: 

3125: \begin{lem} \label{invpart*} There is a linear operator

3126: $ \mr \mi \Dive^\eps_\flat \, : \, \text{\rm Im} \, ( 

3127: \Dive^\eps_\flat) \longrightarrow H^{\infty *}_T $ with

3128: \begin{equation} \label{exainv'}

3129: \Dive^\eps_\flat \circ \mr \mi \Dive^\eps_\flat \ g \, = \, 

3130: g \, , \qquad \forall \, g \in \text{\rm Im} \, (\Dive^\eps_\flat) 

3131: \, . \qquad \qquad \qquad \qquad 

3132: \end{equation}

3133: For all $ m \in \NN $, there is a constant $ C_m > 0 $ such that

3134: \begin{equation} \label{estiinv'}

3135: \parallel \mr \mi \Dive^\eps_\flat \ g \parallel_{H^m} \, \leq \, 

3136: C_m \ \parallel g \parallel_{H^{m+1+ \frac{d}{2}}} \, , 

3137: \qquad \forall \, g \in \text{\rm Im} \, (\Dive^\eps_\flat) \, . 

3138: \quad \ 

3139: \end{equation}

3140: 

3141: \end{lem}

3142: 

3143: \medskip

3144: 

3145: \noindent{\em \underline{Proo}f \underline{o}f \underline{the} 

3146: \underline{Lemma} 4\underline{.1}.} Let $ n \in \NN_* $. Note

3147: 

3148: \medskip

3149: 

3150: $ t_j := j \, T / n \, , \qquad x_j = k /n \, , \qquad 1 \leq 

3151: j \leq n-1 \, , \qquad k \in \ZZ^d \, . $

3152: 

3153: \medskip

3154: 

3155: \noindent{Consider a related partition of unity}

3156: 

3157: \medskip

3158: 

3159: $ \chi_{(j,k)} \in C^\infty ([0,T] \times \RR^d) \, , \qquad 

3160: \qquad (j,k) \in \{ 1, \cdots ,n-1 \} \times \ZZ^d \, , $

3161: 

3162: \smallskip

3163: 

3164: $ \sum_{j=1}^{n-1} \, \sum_{k \in \ZZ^d} \, \chi_{(j,k)} (t,x) \, 

3165: = \, 1 \, , \qquad \forall \, (t,x) \in [0,T] \times \RR^d \, , $

3166: 

3167: \smallskip

3168: 

3169: $ \bigl \{ \, (t,x) \, ; \ \chi_{(j,k)}(t,x) \not = 0 \, \bigr \} \, 

3170: \subset \, [t_j - \frac{2}{n} , t_j + \frac{2}{n} ] \times B (x_j,

3171: \frac{2}{n} ] \, , $

3172: 

3173: \smallskip

3174: 

3175: $ \bigl \{ \, (t,x) \, ; \ \chi_{(j,k)}(t,x) = 1 \, \bigr \} \, 

3176: \supset \, [t_j - \frac{1}{n} , t_j + \frac{1}{n} ] \times B 

3177: (x_j,\frac{1}{n} ] \, . $

3178: 

3179: \medskip

3180: 

3181: \noindent{By hypothesis, there is a function $ v \in H^{\infty *}_T $ 

3182: such that $ g = \Dive^\eps_\flat \, v $.} Introduce

3183: 

3184: \medskip

3185: 

3186: $ v_{(j,k)} \, := \, \chi_{(j,k)} \ v \in H^{\infty *}_T \, , 

3187: \qquad g_{(j,k)} := \Dive^\eps_\flat \, v_{(j,k)} \, . $

3188: 

3189: \medskip

3190: 

3191: \noindent{It suffices to exhibit $ \mr \mi \Dive^\eps_\flat \, 

3192: g_{(j,k)} $ and to show (\ref{estiinv'}) with a constant $ C_m $ 

3193: which is uniform in $ {(j,k)} $.} The problem of finding $ \mr 

3194: \mi \Dive^\eps_\flat \, g_{(j,k)} $ can be reduced to a model 

3195: situation. This can be achieved by using a change of variables 

3196: in $ (t,x) $, based on (\ref{nonstaoubis}). From now on, the 

3197: time $ t $ is viewed as a parameter, the space variable is 

3198: $ x = (x_1,\hat x) \in \RR \times \RR^{d-1} $, and we work with 

3199: 

3200: \medskip

3201: 

3202: $ g = g^* = \Dive^\eps_\flat \, v = ( \eps \, \part_1 + \part_\theta) 

3203: v_1 + \part_2 v_2 + \cdots + \part_d v_d \, , $

3204: 

3205: \medskip

3206: 

3207: $ \bigl \{ \, x \, ; \ g(x,\theta) \not = 0 \, \bigr \} \, \subset \,

3208: \bigl \{ \, x \, ; \ v(x,\theta) \not = 0 \, \bigr \} \, \subset \,

3209: B(0,\frac{1}{2}] \, . $

3210: 

3211: \medskip

3212: 

3213: \noindent{Let $ \psi \in C^\infty (\RR^{d-1};\RR^+) $ be such that

3214: $ \int_{\RR^{d-1}} \, \psi(\hat x) \ d \hat x = 1 $ and

3215: 

3216: \medskip

3217: 

3218: $ \bigl \{ \, \hat x \, ; \ \psi (\hat x) \not = 0 \, \bigr \} \, 

3219: \subset \, B (0,1] \, , \qquad \bigl \{ \, \hat x \, ; \ \psi 

3220: (\hat x) = 1 \, \bigr \} \, \supset \, B (0,\frac{1}{2}] \, . $

3221: 

3222: \medskip

3223: 

3224: \noindent{Decompose $ g $ according to}

3225: 

3226: $ g \, = \, (g - \breve g) \ \psi + \breve g \ \psi \, , \qquad 

3227: \breve g (x) := \int_{\RR^{d-1}} \, g(x_1,\hat x) \ d \hat x \, 

3228: = \, ( \eps \, \part_1 + \part_\theta) \breve v_1 \, . $

3229: 

3230: \medskip

3231: 

3232: \noindent{Seek a special solution $ u $ having the form}

3233: 

3234: \medskip

3235: 

3236: $ u \, = \, \mr \mi \Dive^\eps_\flat \, g \, = \, {}^t \bigl( \,

3237: a \, , \, \text{\rm ridiv} \, [ (g - \breve g) \ \psi ] \, \bigr) \, , 

3238: \qquad a \in H^{\infty *}_T $

3239: 

3240: \medskip

3241: 

3242: \noindent{where 'ridiv' is the operator of Lemma \ref{invpartbar}

3243: applied in the dimension $ d-1 $. It remains to control the

3244: scalar function $ a $ which satisfies the constraint}

3245: 

3246: \medskip

3247: 

3248: $ \eps \ \part_1 a + \part_\theta a \, = \, h \, := \, \breve g \ 

3249: \psi \, = \, (\eps \ \part_1 + \part_\theta) (\breve v_1 \, 

3250: \psi ) \, . $

3251: 

3252: \medskip

3253: 

3254: \noindent{Take the explicit solution}

3255: 

3256: \medskip

3257: 

3258: $ a (x_1, \hat x, \theta) \, = \, \int_{- \infty}^\theta \, h

3259: \bigl( x_1 + \eps \, (s-\theta) , \hat x , s \bigr) \ ds $

3260: 

3261: \smallskip

3262: 

3263: $ \qquad \qquad \ \ = \, \eps^{-1} \ \int_{- \infty}^0 \, 

3264: h \bigl(x_1 + r , \hat x , \theta + \eps^{-1} \, r \bigr) 

3265: \ dr \, . $

3266: 

3267: \medskip

3268: 

3269: \noindent{By construction}

3270: 

3271: \medskip

3272: 

3273: $ a (x,\theta + 1) = a (x,\theta) \, , \qquad \int_\TT \, a 

3274: (x,\theta) \ d \theta \, = \, 0 \, , \qquad \forall \, (x, 

3275: \theta) \in \RR^d \times \TT \, . $

3276: 

3277: \medskip

3278: 

3279: \noindent{For $ \vert x_1 \vert + \vert \hat x \vert \geq 2 $, 

3280: we find}

3281: 

3282: \medskip

3283: 

3284: $ a (x,\theta) \, = \, \int_{- \infty}^\theta \, \frac

3285: {d}{ds} \, \bigl[ (\breve v_1 \, \psi) \bigl( x_1 + 

3286: \eps \, (s-\theta) , \hat x , s \bigr) \bigr \rbrack \ 

3287: ds \, = \, (\breve v_1 \, \psi ) (x,\theta) \, = \, 0 \, . $

3288: 

3289: \medskip

3290: 

3291: \noindent{It implies that}

3292: 

3293: \medskip

3294: 

3295: $ \bigl \{ \, (x,\theta) \, ; \ a (x,\theta) \not = 0 \, 

3296: \bigr \} \, \subset \, B(0;2] \, . $

3297: 

3298: \medskip

3299: 

3300: \noindent{Note $ \mh := \part^{-1}_\theta h \in H^{\infty *} $.} 

3301: Obviously

3302: 

3303: \medskip

3304: 

3305: $ \parallel \mh \parallel_{H^m} \, \leq \, C_m \ \parallel h 

3306: \parallel_{H^m} \, , \qquad \forall \, m \in \NN \, , $

3307: 

3308: \medskip

3309: 

3310: $ \bigl \{ \, (x,\theta) \, ; \ \mh (x,\theta) \not = 0 \, 

3311: \bigr \} \, \subset \, B(0;1] \, , $

3312: 

3313: \medskip

3314: 

3315: \noindent{and we have the identity}

3316: 

3317: \medskip

3318: 

3319: $ a (x_1,\hat x,\theta) \, = \, \mh (x_1,\hat x,\theta) \, 

3320: - \, \int_{ - x_1 - 1}^{- x_1 + 1} \ \part_1 \mh \bigl(

3321: x_1 + r , \hat x , \theta + \eps^{-1} \, r \bigr) \ dr \, . $

3322: 

3323: \medskip

3324: 

3325: \noindent{The term on the right is supported in $ B(0,2] $.}

3326: Use Fubini and Cauchy-Schwarz inequality to control the 

3327: integration of $ \part_1 \mh $. It yields (\ref{estiinv'}).} 

3328: \hfill $ \Diamond $

3329: 

3330: \bigskip

3331: 

3332: \noindent{$ \bullet $ {\bf The Leray projector interpreted in 

3333: the variables $ (t,x,\theta) $.}} Introduce the closed subspace

3334: 

3335: \medskip

3336: 

3337: $ \text{F}^\eps_\flat \, := \,  \bigl \lbrace \, u^* \in L^{2*}_T \, ; 

3338: \ \Dive^\eps_\flat \, u^* = 0 \, \bigr \rbrace \, \subset \, 

3339: L^{2*}_T \, . $

3340: 

3341: \medskip

3342: 

3343: \noindent{Note $ \mP^\eps_\flat $ the orthogonal projector from 

3344: $ L^{2*}_T $ onto F$ {}^\eps_\flat $.} This is a self-adjoint 

3345: operator such that

3346: 

3347: \medskip

3348: 

3349: $ \ker \, \Dive^\eps_\flat = \text{Im} \, \mP^\eps_\flat \, , \qquad 

3350: \text{Im} \, \Grade^\eps_\flat = ( \ker \, \Dive^\eps_\flat)^\perp 

3351: = \ker \,  \mP^\eps_\flat \, . $

3352: 

3353: \medskip

3354: 

3355: \noindent{Expand the function $ u^* \in L^{2*}_T $ in Fourier series 

3356: and decompose the action of $ \mP^\eps_\flat $ in view of the Fourier

3357: modes}

3358: 

3359: \medskip

3360: 

3361: $ u^*(t,x,\theta) = \sum_{k \in \ZZ_*} \, \u_k (t,x) \ e^{i \, 

3362: k \, \theta} \, , \qquad \mP^\eps_\flat \, u^* = \sum_{k \in \ZZ_*} \, 

3363: \mP^\eps_{\flat k} \, \u_k (t,x) \ e^{i \, k \, \theta}\, . $ 

3364: 

3365: \medskip

3366: 

3367: \noindent{Simple computations indicate that}

3368: 

3369: \medskip

3370: 

3371: $ \mP^\eps_{\flat k} \,\u_k \, := \, e^{- \, i \, \eps^{-1} \, k \, 

3372: \varphi^\eps_\flat} \ \Pi (D_x) \ \bigl( e^{i \, \eps^{-1} \, k \, 

3373: \varphi^\eps_\flat} \, \u_k \bigr) \, . $

3374: 

3375: \medskip

3376: 

3377: \noindent{The following result explains why the projector 

3378: $ \mP^\eps_\flat $ is replaced by $ \Pi_0 $ when performing 

3379: the BKW calculus.}

3380: 

3381: \begin{lem} \label{projeler} $ \, $

3382: 

3383: \smallskip

3384: 

3385: \noindent{i) The family $ \{ \mP^\eps_\flat \}_\eps $ is in 

3386: $ \mU \mL^0 $. We have $ [ \part_\theta ; \mP^\eps_\flat ] 

3387: = 0 $ and}

3388: 

3389: \medskip

3390: 

3391: $ [ \md_{j,\eps} ; \mP^\eps_\flat ] = 0 \, , \qquad \forall \, 

3392: j \in \{ 0 , \cdots , d \} \, . $

3393: 

3394: \medskip

3395: 

3396: \noindent{ii) The projector $ \Pi^\eps_\flat (t,x) $ is an 

3397: approximation of $ \mP^\eps_\flat $ in the sense that}

3398: 

3399: \medskip

3400: 

3401: $ \bigl \{ \, \mP^\eps_\flat - \Pi^\eps_\flat \, \bigr \}_\eps 

3402: \, \in \, \eps \ \mU \mL^{2+\frac{d}{2}} \, , \qquad \bigl \{ \, 

3403: \mP^\eps_\flat \ (\id - \Pi^\eps_\flat) \, \bigr \}_\eps \, \in \, 

3404: \eps \ \mU \mL^1 \, . $

3405: 

3406: \end{lem}

3407: 

3408: 

3409: \noindent{\em \underline{Proo}f \underline{o}f \underline{the} 

3410: \underline{Lemma} 4.\underline{2}.} Since $ \mP^\eps_\flat $ 

3411: is a projector, we are sure that

3412: 

3413: \medskip

3414: 

3415: $ \parallel \mP^\eps_\flat \, u \parallel_{L^2_T} \ \leq \ 

3416: \parallel u \parallel_{L^2_T} \, , \qquad \forall \, (\eps,u) 

3417: \in \, ] 0, \eps_0 ] \times L^2_T \, . $

3418: 

3419: \medskip

3420: 

3421: \noindent{It shows that $ \{ \mP^\eps_\flat \}_\eps \in \mU \mL^0 $.}

3422: Compute

3423: 

3424: \medskip

3425: 

3426: $ [ \md_{j,\eps} ; \mP^\eps_\flat ] \, u^* (t,x,\theta) \, 

3427: = \, \sum_{k \in \ZZ_*} \, [ \, \eps \ \part_j + i \ k \ 

3428: \part_j \varphi^\eps_\flat \, ; \, \mP^\eps_{\flat k} \, ] \, 

3429: \u_k (t,x) \ e^{i \, k \, \theta} \, . $ 

3430: 

3431: \medskip

3432: 

3433: \noindent{Observe that}

3434: 

3435: \medskip

3436: 

3437: $ ( \eps \ \part_j + i \ k \ \part_j \varphi^\eps_\flat ) \

3438: \mP^\eps_{\flat k} \, \u_k \, = \, e^{- \, i \, \eps^{-1} \, 

3439: k \, \varphi^\eps_\flat} \ \Pi (D_x) \ \eps \, \part_j \ 

3440: \bigl( e^{i \, \eps^{-1} \, k \, \varphi^\eps_\flat} \ 

3441: \u_k ) $

3442: 

3443: \smallskip

3444: 

3445: \hfill $ = \, \mP^\eps_{\flat k} \ ( \eps \ \part_j + 

3446: i \ k \ \part_j \varphi^\eps_\flat ) \, \u_k \, . \qquad 

3447: \qquad \qquad \qquad \, $

3448: 

3449: \medskip

3450: 

3451: \noindent{All these informations give access to the first 

3452: assertion i).} Now consider ii). The asymptotic expansion

3453: formula for pseudodifferential operators say that for 

3454: all $ \u_k $ in $ C^\infty_0 ( \RR^d_T ) $ we have

3455: $$ \quad \forall \, (t,x) \in \RR^d_T \, , \quad \ 

3456: \lim_{\eps \, \longrightarrow \, 0} \ \ \bigl \{ \,

3457: ( \mP^\eps_{\flat k} \, \u_k) (t,x) \, - \, \Pi 

3458: \bigl( \nabla \varphi^\eps_\flat (t,x) \bigr) \, 

3459: \u_k (t,x) \, \bigr \} \, = \, 0 \, . $$

3460: Since $ \Pi^\eps_\flat = \Pi ( \nabla \varphi^\eps_\flat ) $,

3461: it indicates that $ \mP^\eps_\flat $ is close to 

3462: $ \Pi^\eps_\flat $. We have to make this information 

3463: more precise. To this end, proceed to the decomposition

3464: 

3465: \medskip

3466: 

3467: $ u^* \, = \, v^* + \eps \ \nabla p^* + \part_\theta p^* 

3468: \times X^\eps_\flat \, , \qquad v^* = \mP^\eps_\flat \, 

3469: u^* \, . $

3470: 

3471: \medskip

3472: 

3473: \noindent{We seek a solution $ ( v^*, p^*) $ of these

3474: constraints such that}

3475: 

3476: \medskip

3477: 

3478: $ v^* \, = \, \Pi^\eps_\flat \, u^* + \eps \ \tilde v^* \, , \qquad 

3479: p^* \, = \, \parallel X^\eps_\flat \parallel^{-2} \ X^\eps_\flat \cdot

3480: \part_\theta^{-1} u^* + \eps \ \tilde p^* \, . $

3481: 

3482: \medskip

3483: 

3484: \noindent{After substitution, we find the relation}

3485: 

3486: \medskip

3487: 

3488: $ - \, \nabla \bigl( \parallel X^\eps_\flat \parallel^{-2} \

3489: X^\eps_\flat \cdot \part_\theta^{-1} u^* \bigr) \, = \, 

3490: \tilde v^* + \eps \ \nabla \tilde p^* + \part_\theta \tilde 

3491: p^* \times X^\eps_\flat $

3492: 

3493: \medskip

3494: 

3495: \noindent{which must be completed by the condition}

3496: 

3497: \medskip

3498: 

3499: $ - \, \Div \, ( \Pi^\eps_\flat \, u^* ) \, = \, \eps \ 

3500: \Div \, \tilde v^* + X^\eps_\flat \cdot \part_\theta 

3501: \tilde v^* \, . $

3502: 

3503: \medskip

3504: 

3505: \noindent{It follows that}

3506: 

3507: \medskip

3508: 

3509: $ \tilde v^* \, = \, - \, \mP^\eps_\flat \, \bigl \lbrack

3510: \nabla ( \parallel X^\eps_\flat \parallel^{-2} \ X^\eps_\flat 

3511: \cdot \part_\theta^{-1} u^* ) \bigr \rbrack + ( \mP^\eps_\flat 

3512: - \id ) \ \mr \mi \Dive^\eps_\flat \ \bigl( \Div \, ( \Pi^\eps_\flat 

3513: \, u^* ) \bigr) \, . $ 

3514: 

3515: \medskip

3516: 

3517: \noindent{In view of this relation, the point ii) becomes clear.

3518: \hfill $ \Diamond $

3519: 

3520: \medskip

3521: 

3522: \noindent{Consider the Cauchy problem}

3523: 

3524: \medskip

3525: 

3526: $ \md_{0,\eps} u^* + \eps^{-1} \ \Grade^\eps_\flat \, p^* \, 

3527: = \, f^* \, , \qquad \Dive^\eps_\flat \, u^* = 0 \, , \qquad 

3528: u^* (0,\cdot) = h^* (\cdot) $

3529: 

3530: \medskip

3531: 

3532: \noindent{with data $ f^* \in L^{2*}_T $ and $ h^* \in L^{2*} $.} 

3533: Compose on the left with $ \mP^\eps_\flat $. It yields 

3534: 

3535: \medskip

3536: 

3537: $ \md_{0,\eps} u^* = \mP^\eps_\flat \, f^* + [\md_{0,\eps} ; 

3538: \mP^\eps_\flat ] \, u^* \, , \qquad u^* (0,\cdot) = \mP^\eps_\flat 

3539: \, h^* (\cdot) \, . $

3540: 

3541: \medskip

3542: 

3543: \noindent{The Cauchy problem can be solved in two steps.} 

3544: First extract $ u^* $ from the above equation. Then recover 

3545: $ p^* $ from the remaining relations.

3546: 

3547: \bigskip

3548: 

3549: \noindent{$ \bullet $ {\bf Proof of the Proposition 

3550: \ref{complement}.}} It remains to absorb the term $ \tilde 

3551: g^\eps_\flat $. Use the decomposition

3552: 

3553: \medskip

3554: 

3555: $ g^\eps_\flat = \langle \tilde g^\eps_\flat \rangle + 

3556: \tilde g^{\eps *}_\flat \, , \qquad \langle \tilde g^\eps_\flat 

3557: \rangle \in \text{Im} \, (\Div) \, , \qquad \tilde g^{\eps *}_\flat 

3558: \in \text{Im} \, (\Dive^\eps_\flat) \, . $

3559: 

3560: \medskip

3561: 

3562: \noindent{It suffices to choose}

3563: 

3564: \medskip

3565: 

3566: $ c \u^\eps_\flat \, := \, - \, \text{ridiv} \, \langle \tilde 

3567: g^\eps_\flat \rangle - \mr \mi \Dive^\eps_\flat \, \tilde 

3568: g^{\eps*}_\flat = \bigcirc (\eps^{\frac{N+1}{l}}) \, . $

3569: 

3570: \section{Stability of strong oscillations} $ \, $

3571: 

3572: \vskip -5mm

3573: 

3574: \noindent{The case of turbulent regimes ($ l \geq 3 $) 

3575: will not be undertaken here.} From now on, fix $ l = 2 $ 

3576: and $ N \gg ( 6 + d ) $.} Consider the Cauchy problem

3577: \begin{equation} \label{CPdep}

3578: \left \{ \begin{array}{l}

3579: \part_t \u^\eps + ( \u^\eps \cdot \nabla ) \, \u^\eps + 

3580: \nabla \p^\eps \, = \, \nu \ E^{\eps l}_{\flat 0} ( \part )

3581: \, \u^\eps \, , \qquad \Div \ \u^\eps \, = \, 0 \, ,

3582: \quad \\

3583: \u^\eps (0,x) \, = \, \u^\eps_\flat (0,x) \, . 

3584: \end{array} \right.

3585: \end{equation}

3586: Let $ T_\eps $ be the upper bound of the $ T \geq 0 $ such 

3587: that (\ref{CPdep}) has a solution $ \u^\eps \in \cW^0_T $. 

3588: Classical results \cite{Chem} for fluid equations imply that 

3589: $ T_\eps > 0 $. Our aim in this chapter 5 is to investigate 

3590: the singular limit `$ \eps $ {\it goes to zero}'. Such an 

3591: analysis must at least contain the two following parts.

3592: 

3593: \medskip

3594: 

3595: \noindent{a) An {\it existence} result for a time $ T_0 $ which is 

3596: independent on the small parameter $ \eps \in \, ]0,\eps_0] $.} 

3597: It is required that

3598: 

3599: \medskip

3600: 

3601: $ \inf \ \{ \, T_\eps \, ; \ \eps \in \, ]0,1] \, \} \, \geq \, 

3602: T_0 \, > \, 0 \, . $

3603: 

3604: \medskip

3605: 

3606: \noindent{When $ \nu > 0 $, or when $ \nu = 0 $ and $ d = 2 $,

3607: we know \cite{Chem}-\cite{Li} that $ T_\eps = + \infty $ so 

3608: that $ T_0 = + \infty $.} When $ \nu = 0 $ and $ d \geq 3 $, 

3609: nothing guarantees that $ T_0 > 0 $. To our knowledge, this

3610: is an open question.

3611: 

3612: \medskip

3613: 

3614: \noindent{b) A {\it convergence} result.} The exact solution 

3615: $ \u^\eps $ is not sure to remain close on the whole interval

3616: $ [0,T_0] $ to the approximate solution $ \u^\eps_\flat $ 

3617: given by the Theorem \ref{appBKW}. Proving estimates on 

3618: $ \u^\eps - \u^\eps_\flat $ is a delicate matter.

3619: 

3620: \subsection{Various types of instabilities}

3621: 

3622: $ \bullet $ {\bf Obvious instabilities.} The obvious instabilities

3623: are the mechanisms of amplifications which can be detected by 

3624: looking directly at the formal expansions $ \u^\eps_\flat $. 

3625: They imply the non linear instability of Euler equations. 

3626: Indeed, fix any $ T > 0 $, any $ \u_0 \in \cW^\infty_T 

3627: (\RR^d) $ which is solution of $ (\cE) $, and any 

3628: $ \delta > 0 $.} Work on the balls

3629: 

3630: \medskip

3631: 

3632: $ B_0 ( \u_0 ; \delta ] \, := \, \bigl \{ \, \u \in L^2 \, ;

3633: \ \parallel \u (\cdot) - \u_0 (0,\cdot) \parallel_{L^2 (\RR^d)} 

3634: \, \leq \, \delta \, \bigr \} \, . $ 

3635: 

3636: \medskip

3637: 

3638: $ B_T ( \u_0 ; \delta ] := \, \bigl \{ \, \u \in L^2_T \, ;

3639: \ \parallel \u - \u_0 \parallel_{L^2 ([0,T] \times \RR^d)} \, 

3640: \leq \, \delta \, \bigr \} \, . $ 

3641: 

3642: \begin{prop} \label{euinstap} For all constant $ C > 0 $, there 

3643: are small data 

3644: 

3645: \medskip

3646: 

3647: $ (\h,\tilde \h) \in \bigl( \, B_0 ( \u_0 ; \delta ] 

3648: \cap H^\infty \, \bigr)^2 \, , \qquad (\f, \tilde \f) 

3649: \in  \bigl( \, B_T ( \u_0 ; \delta ] \cap \cW^\infty_T 

3650: \, \bigr)^2 $ 

3651: 

3652: \medskip

3653: 

3654: \noindent{so that the Cauchy problems}

3655: 

3656: \medskip

3657: 

3658: $ \part_t \u + ( \u \cdot \nabla ) \, \u + \nabla \p \, =

3659: \, \f \, , \qquad \Div \ \u \, = \, 0 \, , \qquad \u (0 , 

3660: \cdot) = \h(\cdot) \, , $

3661: 

3662: \smallskip

3663: 

3664: $ \part_t \tilde \u + ( \tilde \u \cdot \nabla ) \, \tilde \u 

3665: + \nabla \tilde \p \, = \, \tilde \f \, , \qquad \Div \ \tilde 

3666: \u \, = \, 0 \, , \qquad \tilde \u (0 , \cdot) = \tilde \h (\cdot) \, , $

3667: 

3668: \medskip

3669: 

3670: \noindent{have solutions $ (\u , \tilde \u) \in B_T (\u_0;\delta]^2 $ 

3671: and there is $ t \in \, ]0,T] $ such that}

3672: \begin{equation} \label{contrsta}

3673: \begin{array}{rl}

3674: \ \parallel (\u - \tilde \u)(t, \cdot) \parallel_{L^2 (\RR^d)} \ 

3675: \geq \, C & \! \! \! \bigl( \, \parallel \h - \tilde \h \parallel_{L^2 

3676: (\RR^d)} \\

3677: \ & + \, \hbox{$\int_0^t$} \, \parallel (\f - \tilde \f) (s,\cdot) 

3678: \parallel_{L^2 (\RR^d)} \ ds \, \bigr) \, . 

3679: \end{array} 

3680: \end{equation}

3681: 

3682: \end{prop}

3683: 

3684: \medskip

3685: 

3686: \noindent{Inequalities as (\ref{contrsta}) are well-known.} In 

3687: general \cite{CGM}-\cite{FSV}-\cite{Gr}, the demonstration is

3688: achieved in two steps. First detect equilibria where instability

3689: arises in the discrete spectrum. Then establish that linearized

3690: instability implies non linear instability. The procedure we adopt 

3691: below is different.} We just look at approximate solutions like  

3692: $ \u^\eps_\flat $. It follows a more simple proof of (\ref{contrsta}).

3693: 

3694: \medskip

3695: 

3696: \noindent{\em \underline{Proo}f \underline{o}f \underline{the} 

3697: \underline{Pro}p\underline{osition} \underline{\ref{euinstap}}.}  

3698: Take $ l=2 $ and $ N \geq (8+d)  $. Consider two deals of initial data

3699: 

3700: \medskip

3701: 

3702: $ \tilde U^1_k (0,x,\theta) \, , \qquad \varphi^1_k (0,x) \, , 

3703: \qquad 1 \leq k \leq N \, , $

3704: 

3705: \medskip

3706: 

3707: $ \tilde U^2_k (0,x,\theta) \, , \qquad \varphi^2_k (0,x) \, , 

3708: \qquad 1 \leq k \leq N \, . $

3709: 

3710: \medskip

3711: 

3712: \noindent{Fix these expressions in the following way}

3713: 

3714: \medskip

3715: 

3716: $ \tilde U^1_1 (0,\cdot) \equiv \tilde U^2_1 (0,\cdot) \, , 

3717: \quad \ \varphi^1_1 (0,\cdot) \equiv \varphi^2_1 (0,\cdot)

3718: \equiv 0 \, , \quad \ \varphi^1_2 (0,\cdot) \equiv \varphi^2_2 

3719: (0,\cdot) \equiv 0 \, . $

3720: 

3721: \medskip

3722: 

3723: \noindent{It implies that}

3724: 

3725: \medskip

3726: 

3727: $ \tilde U^1_1 (t,\cdot) \equiv \tilde U^2_1 (t,\cdot) \, , 

3728: \qquad \varphi^1_1 (t,\cdot) \equiv \varphi^2_1 (t,\cdot) 

3729: \equiv 0 \, , \qquad \forall \, t \in [0,T] \, . $

3730: 

3731: \medskip

3732: 

3733: \noindent{Adjust $ \tilde U^1_2 (0,\cdot) $ and $ \tilde 

3734: U^2_2 (0,\cdot) $ so that}

3735: 

3736: \medskip

3737: 

3738: $ \part_t \, ( \varphi^1_2 - \varphi^2_2 ) (0,\cdot) \, = \, - \,

3739: \nabla \varphi_0 \cdot \langle \tilde U^1_2 - \tilde U^2_2 \rangle

3740: (0,\cdot) \, \not \equiv \, 0 \, . $

3741: 

3742: \medskip

3743: 

3744: \noindent{Therefore, we are sure to find some $ t > 0 $ 

3745: such that $ ( \varphi^1_2 - \varphi^2_2 ) (t,\cdot) 

3746: \not \equiv 0 $.} It follows that

3747: \begin{equation} \label{difama} 

3748: \begin{array} {ll}

3749: U^1_1 (t,x,\theta) \! \! \! & = \, \tilde U^1_1 \bigl(

3750: t, x, \theta + \varphi^1_2( t, x) \bigr) \\

3751: \ & \not \equiv \, U^2_1 (t, x,\theta) \, = \, \tilde 

3752: U^1_1 \bigl(t, x,\theta + \varphi^2_2 (t,x) \bigr) \, . 

3753: \qquad \qquad \

3754: \end{array}

3755: \end{equation}

3756: Note $ \u^{\eps 1}_\flat $ and $ \u^{\eps 2}_\flat $ 

3757: the approximate solutions built with the profiles 

3758: $ \{ U^1_k \}_k $ and $\{ U^2_k \}_k $. The associated 

3759: error terms are $ \f^{\eps 1}_\flat $ and $ \f^{\eps 

3760: 2}_\flat $.

3761: 

3762: \medskip

3763: 

3764: \noindent{Proceed by contradiction.} Suppose that the 

3765: Proposition \ref{euinstap} is wrong. Then, there is 

3766: $ C > 0 $ and $ \eps_1 \in \, ]0,\eps_0 ] $ such 

3767: that for all $ \eps \in \, ] 0 ,\eps_1 ] $, we have

3768: $$ \ \begin{array}{rl}

3769: \parallel (\u^{\eps 1}_\flat - \u^{\eps 2}_\flat)(t, \cdot) 

3770: \parallel_{L^2 (\RR^d)} \ \leq \, C & \! \! \! \bigl( \, \parallel 

3771: (\u^{\eps 1}_\flat - \u^{\eps 2}_\flat)(0, \cdot) \parallel_{

3772: L^2 (\RR^d)} \\

3773: \ & + \, \hbox{$ \int_0^t $} \, \parallel (\f^{\eps 1}_\flat - 

3774: \f^{\eps 2}_\flat) (s,\cdot) \parallel_{L^2 (\RR^d)} \ d s \, 

3775: \bigr) \, .  

3776: \end{array} $$

3777: Divide this inequality by $ \sqrt \eps $. By construction, 

3778: we have

3779: 

3780: \medskip

3781: 

3782: $ \eps^{- \frac{1}{2}} \ \parallel (\u^{\eps 1}_\flat -

3783: \u^{\eps 2}_\flat)(0, \cdot) \parallel_{L^2 (\RR^d)} \, = \, 

3784: \bigcirc ( \sqrt \eps ) \, , $

3785: 

3786: \medskip

3787: 

3788: $ \eps^{- \frac{1}{2}} \ \parallel (\f^{\eps 1}_\flat -  

3789: \f^{\eps 2}_\flat)(s, \cdot) \parallel_{L^2 (\RR^d)} \, 

3790: = \, \bigcirc ( \sqrt \eps ) \, , \qquad \forall \, s

3791: \in [0,t] \, . $

3792: 

3793: \medskip

3794: 

3795: $ \eps^{- \frac{1}{2}} \ \parallel (\u^{\eps 1}_\flat -  

3796: \u^{\eps 2}_\flat)(t, \cdot) \parallel_{L^2 (\RR^d)} $

3797: 

3798: \smallskip

3799: 

3800: \hfill $ = \, \parallel (U^1_1 - U^2_1) (t,\cdot, \eps^{-1} \, 

3801: \varphi^\eps_g (t, \cdot) \parallel_{L^2 (\RR^d)} \, + \, 

3802: \bigcirc ( \sqrt \eps ) \, . \qquad $

3803: 

3804: \medskip

3805: 

3806: \noindent{It follows that}

3807: $$ \ \lim_{ \eps \, \longrightarrow \, 0 } \ \  \eps^{- 

3808: \frac{1}{2}} \ \parallel (\u^{\eps 1}_\flat - \u^{\eps 

3809: 2}_\flat)(t, \cdot) \parallel_{L^2 (\RR^d)} \, = \, 

3810: \parallel (U^1_1 - U^2_1)(t, \cdot) \parallel_{L^2 

3811: (\RR^d \times \TT)} \, = \, 0 $$

3812: which is inconsistent with (\ref{difama}). \hfill $ \Box $

3813: 

3814: \bigskip

3815: 

3816: \noindent{{\it Remark 5.1.1:}} In the demonstration presented 

3817: above, the amplification is due to $ \varphi_2 $ which is the 

3818: principal term in the adjusting phase. The presence of $ \varphi_2 $

3819: becomes efficient in comparison with the other effects when 

3820: 

3821: \medskip

3822: 

3823: $ \vert \, \tilde U^1_1 \bigl(t,x,\theta + \varphi^1_2 (t,x) \bigr) -  

3824: \tilde U^2_1 \bigl(t,x,\theta + \varphi^2_2 (t,x) \bigr) \, \vert \, 

3825: \sim \, c \ t \, \gg \, \sqrt \eps \, . $

3826: 

3827: \medskip

3828: 

3829: \noindent{This requires to wait a lapse of time bigger than 

3830: $ \sqrt \eps $.} This delay can be reduced by adapting the 

3831: above procedure to the cases $ l > 2 $. \hfill $ \triangle $

3832: 

3833: \medskip

3834: 

3835: \noindent{Obvious instabilities have an important consequence.}

3836: To describe the related amplifications, it is necessary to introduce 

3837: new quantities which correspond to the phase shifts. In other words, 

3838: the only way to get $ L^2 \, - $estimates is to {\it blow up} the 

3839: state variables. This principle is detailed in \cite{Che} in the 

3840: case of compressible Euler equations.

3841: 

3842: \bigskip

3843: 

3844: \noindent{$ \bullet $ {\bf Hidden instabilities.}} Hidden 

3845: instabilities are the amplifications which are not detected by 

3846: the monophase description of the section 4. On the other hand, 

3847: they can be revealed by a multiphase analysis. Introduce a second 

3848: phase $ \psi_0 (t,x) \in \cW^\infty_T $ such that

3849: 

3850: \medskip

3851: 

3852: $ \part_t \psi_0 + (\u_0 \cdot \nabla) \, \psi_0 = 0 \, , 

3853: \qquad \nabla \psi_0 \wedge \nabla \varphi_0 \not \equiv 0 $

3854: 

3855: \medskip

3856: 

3857: \noindent{and disturb the Cauchy data of (\ref{CPdep}) 

3858: according to}

3859: 

3860: \medskip

3861: 

3862: $ \u^\eps (0,x) \, = \, \u^\eps_\flat (0,x) \, + \, \eps^{\frac

3863: {M}{l}} \ U \bigl(x, \eps^{-1} \ \psi_0 (0,x) \bigr) \, , \qquad 

3864: M \gg N \, . $

3865: 

3866: \medskip

3867: 

3868: \noindent{The small oscillations contained in the perturbation 

3869: of size $ \eps^{\frac{M}{l}} $ are not always kept under control. 

3870: They interact with $ \u^\eps_\flat $ and with themselves. They can 

3871: be organized in such a way to affect the leading oscillation $ \u^\eps_\flat $. 

3872: Concretely (see \cite{CGM}), we can adjust $ U $ and $ \psi_0 $ so that

3873: there is a constant $ C > 0 $ and times $ t_\eps \in \, 

3874: ]0,T_\eps[ $ going to zero with $ \eps $ such that

3875: 

3876: \medskip

3877: 

3878: $ \parallel (\u^\eps -   \u^\eps_\flat) (t,\cdot) \parallel_{

3879: L^2 (\RR^d)} \, \geq \, C \ \eps^{\frac{1}{2}} \, , \qquad 

3880: \forall \, \eps \in \, ]0,\eps_0] \, . $

3881: 

3882: \medskip

3883: 

3884: \noindent{The power $ \eps^{\frac{M}{l}} $ at the time $ t=0 $ 

3885: is turned into $ \eps^{\frac{1}{2}} $ at the time $ t=t_\eps $.}

3886: Such amplifications occur whatever the selection of $ l \geq 2 $.

3887: They imply minorations like (\ref{contrsta}). However, the underlying 

3888: mechanisms are distinct from the preceding ones. They are implemented 

3889: by oscillations which are transversal to $ \varphi_0 $ and whose 

3890: wavelengths are $ \bigcirc(\eps ) $. They are cancelled by the 

3891: addition of the anisotropic viscosity $ \nu \ E^{\eps l}_{\flat 0} $.

3892: 

3893: 

3894: \subsection{Exact solutions}

3895: 

3896: $ \bullet $ {\bf Statement of the result.} The first information 

3897: brought by the BKW construction is that mean values $ \bar U_k $ and 

3898: oscillations $ U^*_k $ of the profiles $ U_k $ do not play the same

3899: part. This fact is well illustrated by the rules of transformation 

3900: (\ref{rulesoftr}). It means that we have to distinguish these quantities 

3901: if we want to go further in the analysis. This can be done by involving 

3902: the variables $ (t,x,\theta) $ that is by working at the level of 

3903: (\ref{eclacpa}). To deal with $ (u^\eps , p^\eps) (t,x,\theta) $ 

3904: instead of $ (\u^\eps , \p^\eps) (t,x) $ is usual in non linear 

3905: geometric optics \cite{Sc}. It allows to mark the terms apt to 

3906: induce instabilities.

3907: 

3908: \medskip

3909: 

3910: \noindent{Select some approximate solution $ (u^\eps_{(2,N)} , 

3911: p^\eps_{(2,N)} ) $ with source term $ f^\eps_{(2,N)} $ given by 

3912: the Proposition \ref{complement} and look at

3913: \begin{equation} \label{eclacpasui}

3914: \ \left \lbrace \begin{array} {l}

3915: \md_{0,\eps} \, u^\eps + ( u^\eps \cdot \Grade^\eps_\flat) \, 

3916: u^\eps + \Grade^\eps_\flat \, p^\eps \, = \, \nu \ \eps \ \Delta \, 

3917: u^\eps \, , \qquad \Dive^\eps_\flat \, u^\eps = 0 \, , \\

3918: u^\eps (0,x,\theta) \, = \, u^\eps_{(2,N)} (0,x,\theta) \, .

3919: \end{array} \right.

3920: \end{equation} 

3921: 

3922: 

3923: \begin{theo} \label{ciprin} Fix any integer $ N > d + 8 $. There 

3924: is $ \eps_N \in \, ]0,1] $ and $ \nu_N > 0 $ such that for all 

3925: $ \eps \in \, ]0,\eps_N] $ and for all $ \nu > \nu_N $ the Cauchy 

3926: problem (\ref{eclacpasui}) has a unique solution $ (u^\eps , p^\eps) $ 

3927: defined on the strip $ [0,T] \times \RR^d \times \TT $. Moreover

3928: 

3929: \medskip

3930: 

3931: $ \bigl \lbrace u^\eps -  u^\eps_{(2,N)}  \bigr \rbrace_\eps \, 

3932: = \, \bigcirc (\eps^{\frac{N}{2}-d-4}) \, . $

3933: 

3934: \end{theo}

3935: 

3936: \smallskip

3937: 

3938: \noindent{\em \underline{Proo}f \underline{o}f \underline{the} 

3939: \underline{Theorem} \underline{\ref{ciprin}}.} The system

3940: (\ref{eclacpasui}) amounts to the same thing as

3941: \begin{equation} \label{singudepmoyos}

3942: \left \{ \begin{array}{l}

3943: \part_t \bar u^\eps + (\bar u^\eps \cdot \nabla) \bar u^\eps 

3944: + \Div \, \langle u^{\eps*} \otimes u^{\eps*} \rangle + 

3945: \nabla \bar p^\eps = \nu \ \Delta_x \, 

3946: \bar u^\eps \, , \\ 

3947: \part_{0,\eps} u^{\eps*} + (\bar u^\eps \cdot \Grade^\eps_\flat) \, 

3948: u^{\eps*} + \eps \ (u^{\eps*}  \cdot \nabla ) \, 

3949: \bar u^\eps \\

3950: \qquad \quad \, + \, \left \lbrack ( u^{\eps *} \cdot 

3951: \Grade^\eps_\flat) \, u^{\eps*} \right \rbrack^* + 

3952: \Grade^\eps_\flat \, p^{\eps *} = \nu \ \eps \ 

3953: \Delta \, u^{\eps *} \, , \qquad \ \\

3954: \Div \, \bar u^\eps = \Dive^\eps_\flat \, u^{\eps *} = 0 \, . 

3955: \end{array} \right.

3956: \end{equation}

3957: The equation (\ref{singudepmoyos}) is also equivalent to solve

3958: the Cauchy problem

3959: \begin{equation} \label{singudepmoyosmod}

3960: \left \{ \begin{array}{l}

3961:  P \, \part_t \bar u^\eps + P \, \left \lbrack (\bar u^\eps \cdot \nabla)

3962: \bar u^\eps \right \rbrack + P \, \left \lbrack \Div \, \langle 

3963: u^{\eps*} \otimes u^{\eps*} \rangle \right \rbrack = 

3964: \nu \ \Delta_x \, \bar u^\eps \, , \qquad \\ 

3965: \mP^\eps_\flat \, \part_{0,\eps} u^{\eps*} + \mP^\eps_\flat \, \left \lbrack 

3966: (\bar u^\eps \cdot \Grade^\eps_\flat) \, u^{\eps*} \right 

3967: \rbrack + \eps \ \mP^\eps_\flat \, \left \lbrack (u^{\eps*} 

3968: \cdot \nabla ) \, \bar u^\eps \right \rbrack \\

3969: \qquad \quad \, + \, \mP^\eps_\flat \, \left \lbrack 

3970: ( u^{\eps *} \cdot \Grade^\eps_\flat) \, u^{\eps*} 

3971: \right \rbrack^* = \nu \ \eps \ \mP^\eps_\flat \ 

3972: \Delta \, u^{\eps *} \, , 

3973: \end{array} \right.

3974: \end{equation}

3975: associated with the compatible initial data

3976: 

3977: \medskip

3978: 

3979: $ \bar u^\eps (0, \cdot) = P \, \bar u^\eps_\flat (0, \cdot) \, ,

3980: \qquad u^{\eps*} (0, \cdot) = \mP^\eps_\flat \, u^{\eps*}_\flat 

3981: (0, \cdot) \, . $

3982: 

3983: \medskip

3984: 

3985: \noindent{$ \bullet $ {\bf Blow up.}} Introduce the new unknown 

3986: 

3987: \medskip

3988: 

3989: $ d^\eps \, = \, {}^t ( \bar d^\eps , d^{\eps*}) \, = \, {}^t 

3990: ( \, P \, \bar d^\eps \, , \, \mP^\eps_\flat \, d^{\eps*} \, ) $

3991: 

3992: \smallskip

3993: 

3994: $ \quad \ := \, 

3995: \eps^{- \iota} \ \bigl( \, \eps^{-\frac{1}{l}} \ ( \bar u^\eps 

3996: - \bar u^\eps_\flat) \, , \, ( u^{\eps *} - u^{\eps *}_\flat )

3997: \bigr) \, , \qquad \flat = (2,N) \, . $ 

3998: 

3999: \medskip

4000: 

4001: \noindent{This transformation agrees with (\ref{rulesoftr}).} 

4002: The weight $ \eps^{-\frac{1}{l}} $ in front of $ ( \bar u^\eps 

4003: - \bar u^\eps_\flat)  $ induces a shift on the indice $ l $. 

4004: Functions $ \bar U_l $ and $ U^*_{l-1} $ play now the same 

4005: part related to the  amplifications. To write the equation 

4006: on $ d^\eps $ in an abbreviated form, we need notations. 

4007: Quasilinear terms 

4008: $$ \ \left. \begin{array}{l}

4009: \cL_{11}^\eps \, \bar d \, := \, P \, \bigl \lbrack \, 

4010: (\bar u^\eps_\flat \cdot \nabla) \bar d \, \bigr \rbrack \, , 

4011: \qquad \qquad \quad \ \\

4012: \cL_{12}^\eps \, d^* \, := \, P \, \bigl \lbrack \, \Div 

4013: \, \langle \eps^{- \frac{1}{2}} \ u^{\eps *}_\flat \otimes 

4014: d^* \, + \, d^* \otimes \eps^{- \frac{1}{2}} \ u^{\eps 

4015: *}_\flat \rangle \, \bigr \rbrack \, , \\

4016: \cL_{21}^\eps \, \bar d \, := \, \eps^{\frac{1}{2}} \ 

4017: \mP^\eps_\flat \ \bigl \lbrack \, ( u^{\eps *}_\flat 

4018: \cdot \nabla) \bar d \, \bigr \rbrack \, , \\

4019: \cL_{22}^\eps \, d^* \, := \, \mP^\eps_\flat \, \left 

4020: \lbrack \, (\bar u^\eps_\flat \cdot \nabla ) d^* \, 

4021: \right \rbrack \, + \, \eps^{-1} \ \mP^\eps_\flat \, \left 

4022: \lbrack \, ( u^{\eps *}_\flat \cdot \Grade^\eps_\flat) \, 

4023: d^* \, \right \rbrack^* \qquad \qquad \qquad \\

4024: \qquad \qquad \ + \, \eps^{-1} \ \mP^\eps_\flat \, 

4025: \left \lbrack \, (\partial_t \varphi^\eps_\flat + \bar 

4026: u^\eps_\flat \cdot \nabla \varphi^\eps_\flat ) \ 

4027: \partial_\theta d^* \, \right \rbrack \, . 

4028: \end{array} \right. $$

4029: Semilinear terms

4030: $$  \ \left. \begin{array}{l}

4031: A_{11}^\eps \, \bar d \, := \, P \, \bigl \lbrack \,

4032: ( \bar d \cdot \nabla) \bar u^\eps_\flat \, \bigr 

4033: \rbrack \, , \qquad \\

4034: A_{21}^\eps \, \bar d \, := \, \mP^\eps_\flat \, 

4035: \bigl \lbrack \, ( \bar d \cdot \Grade^\eps_\flat) \, 

4036: ( \eps^{- \frac{1}{2}} \ u^{\eps *}_\flat ) \, \bigr 

4037: \rbrack \, , \\

4038: A_{22}^\eps \, d^* \, := \, \mP^\eps_\flat \, \left 

4039: \lbrack \, ( d^* \cdot \nabla) \, \bar u^\eps_\flat \, 

4040: \right \rbrack \, + \, \eps^{-1} \ \mP^\eps_\flat \, 

4041: \left \lbrack \, ( d^* \cdot \Grade^\eps_\flat) \, 

4042: u^{\eps*}_\flat \, \right \rbrack^* \, . \qquad 

4043: \qquad \quad

4044: \end{array} \right. $$

4045: Small quadratic terms 

4046: $$  \ \left. \begin{array}{l}

4047: Q_1^\eps \, := \, \eps^{\frac{3}{2}} \ P \, \bigl 

4048: \lbrack \, \Div \ ( \bar d \otimes \bar d ) \bigr 

4049: \rbrack \, + \, \eps^{\frac{1}{2}} \ P \, \bigl 

4050: \lbrack \, \Div \ \langle d^* \otimes d^* \rangle \, 

4051: \bigr \rbrack \, , \\

4052: Q_2^\eps \, := \, \eps^{\frac{1}{2}} \ \mP^\eps_\flat \, 

4053: \left \lbrack \, (\bar d \cdot \Grade^\eps_\flat) \, d^* \,

4054: \right \rbrack \, + \, \eps^{\frac{3}{2}} \ \mP^\eps_\flat \, 

4055: \left \lbrack \, (d^* \cdot \nabla ) \, \bar d \, 

4056: \right \rbrack \qquad \qquad \qquad \quad \ \ \\

4057: \qquad \quad + \, \mP^\eps_\flat \, \left 

4058: \lbrack \, ( d^* \cdot \Grade^\eps_\flat) \, d^* \, 

4059: \right \rbrack^* \, .

4060: \end{array} \right. $$

4061: And error terms 

4062: 

4063: \medskip

4064: 

4065: $ er^\eps_1 \, := \, \eps^{-\iota-\frac{3}{2}} \ P \ 

4066: \bar f^\eps_\flat \, , \qquad er^\eps_2 \, := \, 

4067: \eps^{-\iota-1} \ \mP^\eps_\flat \ f^{\eps *}_\flat \, . $

4068: 

4069: \medskip

4070:  

4071: \noindent{With these conventions, the expression $ d^\eps $ 

4072: is subjected to}

4073: \begin{equation} \label{interface}

4074: \ \left \{ \begin{array}{l}

4075: P \, \partial_t \bar d^\eps + \cL^\eps_{11} \, \bar d^\eps 

4076: + \cL^\eps_{12} \, d^{\eps *} + A^\eps_{11} \, \bar d^\eps \\ 

4077: \qquad \quad + \, \eps^{\iota - 1} \ Q^\eps_1 + er^\eps_1

4078: \, = \, \nu \ P \ \Delta_x \bar d^\eps \, , \\

4079: \mP^\eps_\flat \, \partial_t d^{\eps *} + \cL^\eps_{21} \, 

4080: \bar d^\eps + \cL^\eps_{22} \, d^{\eps *} + A^\eps_{21} \, 

4081: \bar d^\eps +  A^\eps_{22} \, d^{\eps *} \qquad \qquad 

4082: \qquad \quad \ \\ 

4083: \qquad \quad + \, \eps^{\iota - 1} \ Q^\eps_2 + er^\eps_2

4084: \, = \, \nu \ \mP^\eps_\flat \ \Delta \, d^{\eps *} \, .

4085: \end{array} \right.

4086: \end{equation}

4087: 

4088: \bigskip

4089: 

4090: \noindent{Energy estimates are obtained at the level 

4091: of (\ref{interface}).} Below, we just sketch the related 

4092: arguments which are classical. 

4093: 

4094: \bigskip

4095: 

4096: \noindent{$ \bullet $ {\bf $ L^2 - \, $estimates for the 

4097: linear problem.}} The linearized equations of Euler 

4098: equations along the approximate solution $ u^\eps_\flat $ 

4099: are obtained by removing $ Q^\eps_1 $ and $ Q^\eps_2 $ 

4100: from (\ref{interface}). It yields a system which, at 

4101: first sight, involves coefficients which are singular 

4102: in $ \eps $. In fact, this is not the case. Let us 

4103: explain why.

4104: 

4105: \medskip

4106: 

4107: \noindent{This is clear for $ \cL^\eps_{11} $, $ \cL^\eps_{21} $

4108: and $ A^\eps_{11} $.}

4109: 

4110: \medskip

4111: 

4112: \noindent{Since $ u^{\eps*}_\flat = \bigcirc( \eps^{\frac{1}{l}} ) $, 

4113: this is also true for $ \cL^\eps_{12} $ and $ A^\eps_{21} $.}

4114: 

4115: \medskip

4116: 

4117: \noindent{The contributions which in $ \cL^\eps_{22} $

4118: have $ \eps^{-1} $ in factor give no trouble since}

4119: 

4120: \medskip

4121: 

4122: $ \part_t \varphi^\eps_\flat + \bar u^\eps_\flat \cdot \nabla 

4123: \varphi^\eps_\flat = \bigcirc ( \eps^{\frac{N}{2}} ) = \bigcirc (

4124: \eps^{d+4} ) \, , \qquad u^{\eps *}_\flat \cdot \nabla \varphi^\eps_\flat 

4125: = v^{\eps *}_\flat = \bigcirc ( \eps^{1 + \frac{1}{l}} ) \, . $

4126: 

4127: \medskip

4128: 

4129: \noindent{Now, look at $ A^\eps_{22} $.} Recall that $ d^{\eps *} =

4130: \mP^\eps_\flat \, d^{\eps *} $ which means that

4131: 

4132: \medskip

4133: 

4134: $ \eps^{-1} \ d^{\eps *} \cdot \nabla \varphi^\eps_\flat \, = \, 

4135: - \, \Div \, d^{\eps *} \, . $

4136: 

4137: \medskip

4138: 

4139: \noindent{Therefore}

4140: 

4141: \medskip

4142: 

4143: $ \eps^{-1} \ \mP^\eps_\flat \, \left \lbrack \, ( d^{\eps *} 

4144: \cdot \Grade^\eps_\flat) \, u^{\eps*}_\flat \, \right \rbrack^* \, 

4145: = \, T^\eps (t,x,\nabla) \, d^{\eps *} \, , $

4146: 

4147: \medskip

4148: 

4149: \noindent{where $ T^\eps $ is some differential operator of 

4150: order $ 1 $ with bounded coefficients.} 

4151: 

4152: \medskip

4153: 

4154: \noindent{Observe that these manipulations and the blow up procedure 

4155: induce a {\it loss} of hyperbolicity. When $ \nu = 0 $, this is the 

4156: source of hidden instabilities. When $ \nu \geq \nu_N > 0 $ with 

4157: $ \nu_N $ large enough, this can be compensated by the viscosity. 

4158: This is the key to $ L^2 - \, $estimates. 

4159: 

4160: \bigskip

4161: 

4162: \noindent{$ \bullet $ {\bf The non linear problem and higher 

4163: order estimates.}} Let $ \sigma $ be the smaller integer such

4164: that $ \sigma \geq \frac{d+3}{2} $. If the life span $ T_\eps $

4165: of the exact solution $ u^\eps $ is finite, we must have

4166: $$ \lim_{t \, \longrightarrow \, T_\eps} \ \ \parallel u^\eps 

4167: (t,\cdot) \parallel_{H^\sigma} \, = \, + \infty \, . \qquad 

4168: \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad $$

4169: Thus, the Theorem \ref{ciprin} is a consequence of the following 

4170: majoration

4171: 

4172: \medskip 

4173: 

4174: $ \sup \ \bigl \{ \, \parallel u^\eps (t,\cdot) \parallel_{

4175: H^\sigma} \, ; \ t \in [0 , \min \, (T_\eps,T) ] \, \bigr \} 

4176: \, \leq \, C \, < \, \infty \, . $

4177: 

4178: \medskip

4179: 

4180: \noindent{Consider the set}

4181: 

4182: \medskip

4183: 

4184: $ \cZ_\eps \, := \, \bigl \{ \, \md_{0,\eps} \, , \, \cdots \, , 

4185: \, \md_{d,\eps} \, , \, \partial_\theta \, \bigr \} \, . $

4186: 

4187: \medskip

4188: 

4189: \noindent{Extract the operators}

4190: 

4191: \medskip

4192: 

4193: $ \cZ^k_\eps \, := \, \cZ_1 \circ \, \cdots \, \circ \cZ_k \, , 

4194: \qquad \cZ_j \in \cZ_\eps \, , \qquad k \leq \sigma \, . $

4195: 

4196: \medskip

4197: 

4198: \noindent{It suffices to show that}

4199: $$ \ \max_{\, 0 \leq k \leq \sigma} \ \ \sup \ \bigl \{ \, 

4200: \parallel \eps^{-k} \ \cZ^k_\eps \ u^\eps (t,\cdot) 

4201: \parallel_{L^2} \, ; \ t \in [0 , \min \, (T_\eps,T) ] \, 

4202: \bigr \} \, \leq \, C \, < \, \infty \, . \ $$

4203: Pick some $ \cZ^k_\eps $ with $ k \leq \sigma $. Apply 

4204: $ \cZ^k_\eps $ on the left of (\ref{interface}). Use 

4205: the point i) of Lemma \ref{projeler} to pass through 

4206: $ \mP^\eps_\flat $. Then, observe that the commutator 

4207: of two vector fields in $ \cZ^\eps $ is a linear 

4208: combination of elements of $ \cZ^\eps $ with 

4209: coefficients in $ C^\infty $. Thus, we get an 

4210: equation on $ \cZ^k_\eps \, d^{\eps *} $. 

4211: 

4212: \medskip

4213: 

4214: \noindent{The linear part is managed as in the preceding

4215: paragraph.} Take $ \iota = 1 $. The contributions due to 

4216: $ Q^\eps_1 $ and $ Q^\eps_2 $ are controled by way of the 

4217: a priori estimate and the viscosity. The condition on $ N $ 

4218: is to make sure that 

4219: 

4220: \medskip

4221: 

4222: $ \frac{N}{2} - \iota - \frac{3}{2} - \sigma \, \geq \, 0 \, . $

4223: 

4224: \medskip

4225: 

4226: \noindent{Thereby, the contributions brought by the error 

4227: terms $ er^\eps_1 $ and $ er^\eps_2 $ remain bounded in 

4228: the procedure.}

4229: 

4230: \bigskip

4231: 

4232: \bibliographystyle{alpha}

4233: 

4234: \begin{thebibliography}{10}

4235: 

4236: \bibitem[1]{B} C. Bardos, {\it What use for the mathematical theory of the Navier-Stokes equations},

4237: Mathematical fluid mechanics, 1-25, Adv. Math. Fluid Mech.,

4238: Birkh\"auser, Basel (2001).

4239: 

4240: \bibitem[2]{BKM} J. T. Beale, T. Kato, A. Majda, {\it Remarks on the

4241: breakdown of smooth solutions for the 3-D Euler equations},

4242: Commun. Math. Phys., P. 61-66 (1994).

4243: 

4244: \bibitem[3]{C} T. Chacon Rebollo, {\it Oscillations due to the transport of microstructures},

4245: SIAM J. Appl. Math., Vol 48, No 5, Octobre 1988.

4246: 

4247: \bibitem[4]{Chem} J.-Y. Chemin, {\it Perfect incompressible fluids}. Translated from the 1995 French 

4248: original by Isabelle Gallagher and Dragos Iftimie. Oxford Lecture Series in Mathematics and its 

4249: Applications, 14. The Clarendon Press, Oxford University Press, New York, 1998. x+187 pp.

4250: 

4251: \bibitem[5]{CDGG} J-Y. Chemin, B. Desjardins, I. Gallagher, E. Grenier, {\it Fluids with

4252: anisotropic viscosity}, Mathematical Modelling and Numerical Analysis, M2AN, Vol. 34,

4253: 2, 315--335 (2000).

4254: 

4255: \bibitem[6]{Che} C. Cheverry, {\it Propagation of Oscillations in Real

4256: Vanishing Viscosity Limit}, To appear in  Commun. Math. Phys.

4257: 

4258: \bibitem[7]{CGM} C. Cheverry, O. Gu\`es, G. M\'etivier, {\it Oscillations fortes sur un champ

4259: lin\'eairement d\'eg\'en\'er\'e}, Annales Scientifiques de l'ENS, 5 (2003).

4260: 

4261: \bibitem[8]{CGM1} C. Cheverry, O. Gu\`es, G. M\'etivier, {\it Large

4262: amplitude high frequency waves for quasilinear hyperbolic systems},

4263: submitted in Advances in Differential equations.

4264: 

4265: \bibitem[9]{CF} P. Constantin, C. Fefferman, {\it Direction of 

4266: vorticity and the problem of global regularity for the 

4267: Navier-Stokes equations}, Indiana Univ. Math. Journal, 42, No. 3,

4268: 775--789 (1993).

4269: 

4270: \bibitem[10]{CG} A. Corli, O. Gu\`es,

4271: {\it Stratified solutions for systems of conservation laws}, Trans. Amer. Math. Soc. 353, 

4272: no. 6, 2459--2486 (2001).

4273: 

4274: \bibitem[11]{D} R.-J. DiPerna, A.-J. Majda, {\it Oscillations and Concentrations in Weak Solutions 

4275: of the Incompressible Fluid Equations}, Commun. Math. Phys. 108, 667--689 (1987).

4276: 

4277: \bibitem[12]{E} W. E, {\it Homogenization of linear and non linear transport equations}, Comm. Pure 

4278: Appl. Math.  45,  no. 3, 301--326 (1992).

4279: 

4280: \bibitem[13]{E2} W. E, {\it Propagation of oscillations in the solutions of $

4281: 1-D $ compressible fluid equations}, Com. P. D. E. 17 (3/4),

4282: 347--370 (1992).

4283: 

4284: \bibitem[14]{FMRT} C. Foias, P.-P. Manley, R. Rosa, R. Temam, {\it Cascade of energy in turbulent flows}, 

4285: C. R. Acad. Sci. Paris, t. 332, Série I, p. 509-514 (2001).

4286: 

4287: \bibitem[15]{FSV} S. Friedlandler, W. Strauss, M. Vishik, {\it Nonlinear

4288:  instability in an ideal fluid},

4289: Ann. Inst. Henri Poincar\'e, Vol. 14, no. 2 (1997), 187--209.

4290: 

4291: \bibitem[16]{Ge} P. G\'erard, {\it Microlocal defect measures}, Comm. in Partial Differential Equations, 

4292: 16, 1761--1794 (1991). 

4293: 

4294: \bibitem[17]{Gr} E, Grenier, {\it On the nonlinear instability of Euler and

4295: Prandtl equations}, Comm.

4296: Pure Appl. Math. 53 (2000), no. 9, 1067--1091.

4297: 

4298: \bibitem[18]{G} O. Gu\`es, {\it D{\'e}veloppement asymptotique de solutions exactes de syst{\`e}mes 

4299: hyperboliques quasilin{\'e}aires}, Asymptotic Anal. 6, no. 3, 241--269 (1993). 

4300: 

4301: \bibitem[19]{G2} O. Gu\`es, {\it Ondes multidimensionnelles $ \eps-$stratifi\'ees et oscillations}, 

4302: Duke Math. J. 68, no. 3, 401-446 (1992).

4303: 

4304: \bibitem[20]{JMR1} J-L Joly, G. M\'etivier, J. Rauch, {\it Transparent non linear geometric optics and 

4305: Maxwell-Bloch equations}, Journal of Differential Equations, 166,

4306: 175--250 (2000).

4307: 

4308: \bibitem[21]{La} D. Lannes, {\it Secular growth for symmetric hyperbolic systems}, 

4309: Journal of Differential Equations, 90, 466--503 (2003).

4310: 

4311: \bibitem[22]{L} M. Lesieur, {\it Turbulence in fluids}, Fluid Mechanics and its Applications, 

4312: Kluwer Academic Publishers Group, Dordrecht, 515 pp. ISBN: 0-7923-4415-4

4313: (1997).

4314:  

4315: \bibitem[23]{Li} P.-L. Lions, {\it Mathematical Topics in fluid

4316: Mechanics}, Volume 1, Incompressible Models, Oxford Science Publications.

4317: 

4318: \bibitem[24]{MPP} D. W. Mclaughlin, G. C. Papanicolaou, O. R. Pironneau, {\it Convection of

4319: microstructure and related problems}, SIAM J. Appl. Math., Vol 45, No 5, Octobre 1985.

4320: 

4321: \bibitem[25]{Sc} S. Schochet, {\it Fast singular limits of hyperbolic PDEs}, J. Dif. Eq. 114, 

4322: 476--512 (1994).

4323: 

4324: \bibitem[26]{Se} D. Serre, {\it Oscillations nonlin\'eaires de haute fr\'equence ; dim $\geq 2$}, 

4325: In Marino A. and Murthy M. K. V., editors, Nonlinear variational problems and partial differential 

4326: equations, volume 320 of Pitman Res. notes in Math., pages 245-294, London, 1995. Longman.

4327: 

4328: \bibitem[27]{Si} T. Sideris, {\it Formation of Singularities in Compressible Fluids}, 

4329: Commun. in Math. Phys. 101, 475--485 (1985). 

4330: 

4331: 

4332: \end{thebibliography}

4333: 

4334: \end{document}

4335: 

4336: