math0511564/e12.tex
1: \documentclass[11pt]{amsart}
2: \usepackage{amsmath,amssymb,vmargin,enumerate}
3: 
4: \pagestyle{headings}
5: 
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%85
7: 
8: \addtolength{\headheight}{2.5pt}
9: \setlength{\textwidth}{6in}
10: \setlength{\textheight}{8.5in}
11: 
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: 
14: 
15: \title{Entropy boundary layers}
16: \author{Franck SUEUR}
17: \date{\today}
18: 
19: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
20: 
21: \providecommand{\diver}{\mathrm{div}}
22: \newcommand{\eps}{\varepsilon}
23: \newcommand{\X}{\textbf{X}}
24: \newcommand{\K}{\textbf{K}}
25: \renewcommand{\L}{\textbf{L}}
26: \renewcommand{\S}{\textbf{S}}
27: \providecommand{\B}{\textbf{B}}
28: \newcommand{\U}{\textbf{U}}
29: \newcommand{\V}{\textbf{V}}
30: \newcommand{\W}{\textbf{W}}
31: \renewcommand{\u}{\textbf{u}}
32: \providecommand{\Phi}{\textbf{\Phi}}
33: \newcommand{\noi}{\noindent}
34: \newcommand{\un}{\underline}
35: \newcommand{\til}{\tilde}
36: \newcommand{\rond}{\mathring}
37: \newcommand{\M}{\mathcal{M}}
38: \renewcommand{\H}{\mathcal{H}}
39: \providecommand{\Hdroit}{\mathbb{H}}
40: \renewcommand{\P}{\mathcal{P}}
41: \providecommand{\Pb}{\mathbb{P}}
42: \providecommand{\E}{\mathcal{E}}
43: \providecommand{\F}{\mathcal{F}}
44: \providecommand{\A}{\mathcal{A}}
45: \providecommand{\R}{\mathbb{R}}
46: \providecommand{\N}{\mathbb{N}}
47: \providecommand{\Nrond}{\mathcal{N}}
48: \providecommand{\tend}{\rightarrow}
49: \providecommand{\dev}{development }
50: \providecommand{\Dev}{Development }
51: \providecommand{\sub}{substitution }
52: \providecommand{\reg}{regular }
53: \providecommand{\Th}{theorem }
54: \providecommand{\prob}{problem }
55: \providecommand{\ver}{verify }
56: \providecommand{\sec}{read }
57: \providecommand{\cons}{consider }
58: \providecommand{\hyp}{hyperbolic }
59: \providecommand{\lin}{linear}
60: \providecommand{\sym}{symmetric }
61: \providecommand{\prol}{prolongement }
62: \providecommand{\si}{  iteration sheme }
63: \providecommand{\esti}{estimate}
64: \providecommand{\sol}{solution }
65: \providecommand{\coeff}{coefficient }
66: \providecommand{\Bdroit}{\mathbb{B} }
67: \providecommand{\Brond}{\mathcal{B} }
68: \providecommand{\Adroit}{\mathbb{A} }
69: \providecommand{\Arond}{\mathcal{A} }
70: \providecommand{\Z}{\mathcal{Z}}
71: \providecommand{\carac}{characteristic}
72: 
73: \newtheorem{theo}{Theorem}[section]
74: \newtheorem{coro}[theo]{Corollary}
75: \newtheorem{lem}[theo]{Lemma}
76: \newtheorem{souslem}[theo]{Sublemma}
77: \newtheorem{prop}[theo]{Proposition}
78: \newtheorem{rem}[theo]{Remark}
79: \newtheorem{hypo}[theo]{Assumption}
80: \newtheorem{hypodefi}[theo]{Assumption-Definition}
81: \newtheorem{defi}[theo]{Definition}
82: \newtheorem{propr}[theo]{Property}
83: \newtheorem{step}{Step}
84: \newtheorem{step2}{Step}
85: 
86: \begin{document}
87: 
88: 
89: \begin{abstract}
90: 
91: We consider the Euler system of compressible and entropic gaz dynamics in a bounded open domain of $\R^d$ with 
92:  wall boundary condition. 
93: We prove the existence and the stability of families of solutions which correspond to a ground state 
94: plus a large entropy boundary layer.
95: The ground state is a solution of the Euler system which satisfies some explicit additional conditions on the boundary.
96: These conditions are used in a reduction of the system.
97: We construct BKW expansions at all order. 
98: The profile problems are linear thanks to a transparency property.
99: We prove the stability of these expansions by proving $\eps$-conormal estimates
100:  for a characteristic boundary value problem.
101:  
102: \end{abstract}
103: 
104: \maketitle
105: 
106: 
107: \section{Introduction}
108: 
109: 
110:   We consider the  \textbf{Euler system of compressible and entropic gaz dynamics}. 
111: We respectively denote by  $\mathtt{s}$, $\mathtt{v} $  and  $\mathtt{p}$ the  entropy, the speed and the pressure. 
112:  The volumic density, noted $\mathtt{\rho}$ , is a function of  $\mathtt{p}$ and $\mathtt{s}$.  
113:  We assume that  $\mathtt{\rho}(\mathtt{p},\mathtt{s}) > 0$ for all $\mathtt{p}$ and  $\mathtt{s}$. 
114: We also introduce the function $\alpha (\mathtt{p},\mathtt{s}) := 
115: (\mathtt{\rho} (\mathtt{p} ,\mathtt{s}))^{-1} \,  \mathtt{\rho}'_\mathtt{p} (\mathtt{p},\mathtt{s}) $.
116:   We assume that $\alpha (\mathtt{p},\mathtt{s}) > 0$ for all $\mathtt{p}$ and  $\mathtt{s}$.
117:   We denote by $t$ the time variable and by $x=(x_1 , ...,x_d )$ the space variable. 
118:   We denote by $\nabla$ the gradient with respect to $x$.
119:    The Euler system is:
120:    %
121:  \begin{eqnarray*}
122:   \X_{\mathtt{v}}  \,  \mathtt{v} + \mathtt{\rho}^{-1} \,  \nabla  \mathtt{p} =0 ,
123: \quad \X_{\mathtt{v}} \,  \mathtt{p} +  \alpha^{-1} \, \mathrm{div} \  \mathtt{v}  =0 \quad \X_{\mathtt{v}} \,  \mathtt{s}=0 ,
124: \end{eqnarray*}
125:   %
126:   where  $\X_{\mathtt{v}} \, $ is the particle derivative $\X_{\mathtt{v}} \,  := \partial_t + \mathtt{v} .\nabla $.
127: The previous system  is a nonconservative form of a system of conservation laws. 
128: Moreover, this system is  hyperbolic symmetrizable. It has three characteristic fields.
129: One of them is linearly degenerate (cf. \textit{section $\ref{setting}$}). 
130:  We consider these equations in a bounded open domain  $\Omega \subset \R^d$ lying on one side
131: of its $C^\infty$  boundary $\Gamma$. 
132: More precisely, since we will need an equation of
133: the boundary $\Gamma $, we fix once for all a function
134: $\varphi \in C^{\infty}(\R^d, \R)$ and we assume that
135: $\Omega = \{ \varphi > 0 \}$, $ \Gamma = \{ \varphi = 0 \}$ 
136: and  $| \nabla \varphi (x) | = 1 $ in an open neighborhood $\mathcal{V}_\Gamma $ 
137: of $\Gamma $\footnote{ Hence for $x \in \Omega \cap \mathcal{V}_\Gamma$: 
138: $\varphi(x) = dist(x,\Gamma )$.}. 
139: We consider the natural boundary condition $ \mathtt{v}. \mathfrak{n} =0$, where $\mathfrak{n}$ is
140: the unit outward normal to $\Gamma$. 
141: For $T>0$, the \textbf{boundary value problem} reads:
142: %
143: \begin{eqnarray}
144: \label{aller}
145: \left\{
146: \begin{array}{cc}
147: \left.
148: \begin{array}{l}
149:   \X_{\mathtt{v}} \,  \mathtt{v} + \mathtt{\rho}^{-1} \,  \nabla  \mathtt{p} =0 
150: \\  \X_{\mathtt{v}} \,  \mathtt{p} +  \alpha^{-1} \, \mathrm{div} \ \mathtt{v}  =0
151: \\ \X_{\mathtt{v}} \,  \mathtt{s}=0
152: \end{array}
153: \right\}
154: \quad &\mathrm{where} \, (t,x) \in (0,T) \times \Omega
155: \\  \mathtt{v}. \mathfrak{n} = 0 \quad &\mathrm{where} \, (t,x) \in (0,T) \times \Gamma
156: \end{array}
157: \right.
158: \end{eqnarray}
159: %
160: The boundary $\Gamma$ is  characteristic for the linearly degenerate field. 
161: We introduce the tangential velocity $\mathtt{v}_\mathfrak{t} $ and 
162: the  normal velocity $ \mathtt{v}_\mathfrak{n}$. 
163: Thus, we have $ \mathtt{v} = \mathtt{v}_\mathfrak{t} + \mathtt{v}_\mathfrak{n} $. 
164: The choice of the set of thermodynamic variables $\mathtt{v}$, $\mathtt{p}$ and  $\mathtt{s}$ is particularly well adapted 
165: to boundary problem (cf. \cite{moi3}).
166: The existence of \textbf{local regular solutions} of $(\ref{aller})$ is given in \cite{cho} and \cite{Gu90}.
167: If $\mathcal{O}$ is an open subset of $\R^d$, we denote by $H^\infty (\mathcal{O})$ the set of $u \in L^2 (\mathcal{O})$ 
168: such that all the derivatives of $u$ are in $L^2 (\mathcal{O})$. 
169: From now on, we will assume that a real $T_0 > 0$ and 
170: a solution $u^0 :=(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) \in H^\infty ((0,T_0) \times \Omega)$ of $(\ref{aller})$ are given.
171: 
172: 
173: An interesting question about the Euler system is the study of the convergence of a more acute model: 
174: the Navier-Stokes system, which includes a viscosity term,  when the amplitude of the viscosity goes to $0$.  
175: The difficulty is linked to the existence of a \textbf{boundary layer} i.e. of a rapid variation 
176: of the solutions of the viscous model
177:  near the boundary. 
178: There is a sensitivity on the type of boundary conditions imposed for the viscous model. 
179: The most delicate is the homogeneous Dirichlet condition. 
180: In this case, there are in general some large characteristic boundary layers
181:  of large amplitude\footnote{For other boundary conditions the amplitude of the boundary layers 
182:  can be weaker or the boundary layers can be noncharacteristic.
183: Then the problem is simpler (see \cite{fg}, \cite{cmr}, \cite{moi3}, \cite{Temam}).}.
184:  In one space  dimension, a simple case is the isentropic one since there is no boundary layer 
185:  and the solutions of the Navier-Stokes
186: system are regular perturbations of the solutions of the Euler system. 
187: For the entropic Navier-Stokes  equations, 
188: an answer was given by F.Rousset in
189: \cite{Rousset} using boundary layers analysis. 
190: In several space dimensions, the analysis is quite more complicated even for the incompressible Navier-Stokes equations.
191: There is a huge literature about the tangential velocity boundary layers which appear (see, for example, 
192: the papers of Z.Xin and T.Yanagisawa \cite{Xin}, M.Sammartino and R.E.Caflisch
193: \cite{Samm1}, \cite{Samm2}, E.Grenier \cite{Gr1}, \cite{Gr2},...).
194: An attempt of analysis involves Prandtl equations
195: (see the surveys of W.E \cite{E} and E.Grenier \cite{Grenier}). 
196: 
197: Here we do not consider the  Navier-Stokes equations but only the Euler system.
198: The idea to investigate first the stability of boundary layers type solutions for the Euler equations is a classical approach in
199: fluid mechanics (see the books of P.G.Dranzin and W.H.Reid \cite{Drazin}, of C.Marchioro and M.Pulvirenti \cite{Marchioro},
200: S.Schochet \cite{Sch}. 
201: This idea was followed  more recently by E.Grenier for the study of velocity boundary layers  \cite{Gr1}, \cite{Gr2}.
202: Such a strategy is also possible for entropy boundary layers since they are characteristic, like the velocity ones.  
203: Thus our goal in this paper is to study  \textbf{entropy boundary layers} for the Euler system.
204: As far as we know there was no mathematical study of entropy boundary layer in \textbf{several space dimensions}. 
205: 
206: \section{Overview of the results}
207: 
208: 
209: We introduce the space $\Nrond (T) :=  H^\infty  ( (0,T) \times  \Omega , \mathcal{S}(\R_+)$ 
210: where we  denote by $\mathcal{S} (\R_+ )$ the Schwartz space of $C^\infty$ rapidly decreasing functions. 
211:  Thus a function $U (t,x,X) \in  \Nrond (T) $ is $C^\infty$ rapidly decreasing with respect to $X$.
212:  Let us begin to look naively for solutions $ u^\eps :=( \mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^\eps )_{0 < \eps \leqslant 1}$
213:  of $(\ref{aller})$ with $\mathtt{s}^\eps$ of the form 
214:  %
215:   \begin{eqnarray}
216:  \label{elec} 
217:  \mathtt{s}^\eps (t,x) :=  \mathtt{s}^0  (t,x) + \tilde{\mathtt{S}}^0 (t,x,\frac{\varphi(x)}{\eps})
218:  \end{eqnarray}
219:  %
220:  where the function $\tilde{\mathtt{S}}^0 $ is in $\Nrond (T_0)$ and
221: the function $u^0 := (\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) $ is the ground state  given above.
222: Replacing in the third equation of the system $(\ref{aller})$ leads to the following linear transport equation
223:  that  necessarily the function $\tilde{\mathtt{S}}^0 $ verifies:
224:  %
225:  \begin{eqnarray}
226:  \label{cle} 
227:   (\partial_t  + \mathtt{v}^0 .\nabla
228:  +  \frac{\mathtt{v}^0_{\mathfrak{n}}}{\varphi(x)}  .X \partial_X ) \tilde{\mathtt{S}}^0 =
229: 0 \quad \mathrm{where} \, (t,x,X) \in (0,T_0) \times \Omega  \times \R_+ .
230:  \end{eqnarray}
231:  %
232:  Thanks to the boundary condition (see $(\ref{aller})$), the function $\mathtt{v}^0_{\mathfrak{n}} \, /\varphi(x)$ is 
233:  $C^\infty$\footnote{Notice that for this transport equation, the boundary $\Gamma \times \{0\}$ of the domain $\Omega \times \R_+ $ is totally
234:  characteristic. As a consequence,  none boundary condition needs to be prescribed for $\tilde{\mathtt{S}}^0 $.}.
235: In general these necessary conditions are not sufficient to insure 
236: that the functions $( \mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^\eps )_{0 <\eps \leqslant 1}$ are solutions of $(\ref{aller})$.
237: Indeed, because the functions $\mathtt{\rho}$ and $\alpha$ depend on $\mathtt{p}$ and $\mathtt{s}$, the two first equations of
238: $(\ref{aller})$ are not satisfied.
239: However if in addition we assume that the ground state $(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) $ satisfies 
240: %
241: \begin{eqnarray}
242: \label{pro}
243: \X_{\mathtt{v}^0} \,  \,  \mathtt{v}^0 =0 , \quad \mathrm{div}_x \ \mathtt{v}^0 =0 , \quad \nabla_{t,x} \, \mathtt{p}^0 =0 \quad \mathrm{where} \,
244: (t,x) \in (0,T_0) \times \Omega,
245: \end{eqnarray}
246: %
247: then it is easy to check that the functions $(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^\eps )_{0 <\eps \leqslant 1}$ 
248: are solutions of $(\ref{aller})$.
249:  \textit{In section $\ref{ground}$}, we will prove the existence of ground states 
250:  $(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) \in H^\infty$ solutions of
251:  $(\ref{aller})$ and verifying the conditions $(\ref{pro})$.
252: 
253:  Our goal is to relax the conditions $(\ref{pro})$ into   \textbf{conditions localized on the boundary $\Gamma$}.
254:   In fact the conditions $(\ref{pro})$ were first introduced in a paper of C.Cheverry, O.Gu\`es and G.M\'etivier  \cite{cgm2}
255:    where the existence and the stability of large amplitude high frequency
256:   entropy waves are shown.  
257:   Here we look for entropy  boundary layers which are local singularities. 
258:   It seems rather natural that local conditions are sufficient to deal with  boundary layers.
259:   We reach our goal and claim the following theorem.
260:   
261:   \begin{theo}
262:   \label{int}
263:   
264:   Assume that a solution  $u^0 :=(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) \in H^\infty ((0,T_0) \times \Omega)$ of
265:   $(\ref{aller})$ verifying  
266:   %
267:   \begin{eqnarray}
268:   \label{cond}
269: \X_{\mathtt{v}^0} \,  \mathtt{v}^0 =0 \ \mathrm{and} \ \X_{\mathtt{v}^0} \,  \mathtt{p}^0 =0
270:  \quad  \mathrm{for} \ (t,x) \in (0,T_0) \times \Gamma
271:   \end{eqnarray}
272:   %
273:    and a solution $ \tilde{\mathtt{S}}^0 \in \Nrond (T_0)$ of  $(\ref{cle})$ are given.
274:   For $0 < \eps \leqslant 1$, we denote by  $ u^\eps $ the function 
275:   $ u^\eps :=( \mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^\eps )$ where
276:   $\mathtt{s}^\eps$ is given by $(\ref{elec})$.
277:   Then there exists a family of solutions  $(\tilde{u}^\eps )_{0 < \eps \leqslant 1}$ in $H^\infty ((0,T_0) \times \Omega) $
278:   of $(\ref{aller})$ such that  $\tilde{u}^\eps - u^\eps$ tends to $0$ in $H^1 ((0,T_0) \times \Omega)$ when $\eps \rightarrow 0^+ $.
279:   
280:    \end{theo}
281:   
282:   In fact Theorem $\ref{int}$ is a corollary of more acute results involving \textbf{WKB} (Wentzel-Kramers-Brillouin) expansions.
283:  We will prove the \textbf{existence and stability} of families of solutions 
284:   $( \mathtt{v}^\eps , \mathtt{p}^\eps ,\mathtt{s}^\eps)_{0 < \eps \leqslant 1}$
285: of the Euler system with a large amplitude
286: entropy boundary layer i.e.  of the form 
287:  %
288: \begin{eqnarray}
289: \label{pre}  \quad
290:    \left\{
291:    \begin{array}{l}
292:   \mathtt{v}_\mathfrak{t}^\eps  (t,x) := \mathtt{v}_\mathfrak{t}^0  (t,x) + \eps
293: (\mathtt{V}_\mathfrak{t}  (t,x)  
294: + \tilde{\mathtt{V}}_\mathfrak{t} (t,x,\frac{\varphi(x)}{\eps})) + O( \eps^2 ) ,
295: \vspace{0.2cm} \\  \mathtt{v}_\mathfrak{n}^\eps (t,x) := \mathtt{v}_\mathfrak{n}^0  (t,x)+ \eps \mathtt{V}_\mathfrak{n} (t,x)  
296: + O( \eps^2 ) ,
297: \vspace{0.2cm} \\  \mathtt{p}^\eps  (t,x):= \mathtt{p}^0 (t,x) + \eps \mathtt{P} (t,x)  + O( \eps^2 ) ,
298: \vspace{0.2cm} \\  \mathtt{s}^\eps (t,x) := \mathtt{s}^0 (t,x)  +  \tilde{\mathtt{S}}^0 (t,x,\frac{\varphi(x)}{\eps})  + O(\eps ) ,
299: \end{array}
300:   \right.
301: \end{eqnarray}
302:  %
303: where $u^0 :=(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 )$ is a solution of $(\ref{aller})$  verifying the conditions $(\ref{cond})$.
304:  This analysis is inspired by \cite{cgm2}, where the propagation of large amplitude high frequency entropy waves is shown 
305:  for ground states $u^0 $ solutions of $(\ref{aller})$  
306:  verifying the conditions $(\ref{pro})$
307:  (in the first equation of section $2.3$ of \cite{cgm2}, read $ \nabla_{t,x} \mathtt{p} =0$ instead of $\nabla_{x} \mathtt{p}=0$).
308:  In our analysis the condition on the particle derivative is localized on the boundary. 
309:  Considering directly the Euler system, we use implicitly the  general structure conditions of \cite{cgm2}.
310:  In particular, the choice of the set of thermodynamic variables $\mathtt{v}$, $\mathtt{p}$ and  $\mathtt{s}$ is a key point. 
311: 
312:  An important feature of the expansion $(\ref{pre})$ is that there are some boundary layers not only on the entropy 
313:  but also on the
314: other components, ponderated by some $\eps$.
315: More accurately, boundary layer appears on tangential velocity with an amplitude $\eps$ whereas boundary layer
316: appears on the normal velocity and the pressure with an amplitude $\eps^2$. 
317:  The conditions $(\ref{cond})$ on the ground state play a main role in the fact that the large boundary layer keeps polarized on
318:  the entropy. 
319:  \textit{In section $\ref{redbull}$}, we reduce the system, thanks to  a change of unknown singular with respect to
320:  $\eps$. 
321:  This \textbf{reduction} is inspired by \cite{cgm2}. 
322:  The conditions $(\ref{cond})$ on the ground state will be used at this step.
323:  Because the localized conditions $(\ref{cond})$ are weaker than the conditions $(\ref{pro})$ of \cite{cgm2}, our reduction is much
324:  more delicate than in \cite{cgm2}. We now describe
325: in more detail the rest of the contents of the paper. 
326:  \\
327:  \\ \textit{In section $\ref{secform}$}, we look for \textbf{formal WKB solutions} of the problem  $(\ref{aller})$. 
328: This means that we construct WKB expansions of infinite order.
329: Let us precise this.
330: We introduce the profile space  
331: %
332: \begin{eqnarray}
333: \label{defp}  \P (T):= \{ \mathcal{U} \in L^2 ((0,T) \times \Omega \times \R_+  )  \ \mathrm{such} \ \mathrm{that} \  \mathrm{there} \ 
334: \mathrm{exist} \ \un{ \mathcal{U}} \in H^\infty ((0,T) \times \Omega ) \ \mathrm{and}
335: \\  \nonumber \ \tilde{ \mathcal{U}} 
336: \in  \Nrond (T) \   \mathrm{such} \ \mathrm{that} \ \mathrm{for} \ \mathrm{all} \ (t,x,X)
337:  \in (0,T) \times \Omega \times \R_+ ,
338:   \  \mathcal{U}  (t,x, X)= \un{ \mathcal{U}} (t,x) 
339: + \tilde{ \mathcal{U}} (t,x,  X)   \}. 
340: \end{eqnarray}
341:  %
342:   The function  $\un{ \mathcal{U}}$ is the regular part and  $\tilde{ \mathcal{U}}$ is
343:    the characteristic boundary layer term. 
344:    We will  split $\mathcal{U} \in \P(T)$ into $\mathcal{U} =(\mathtt{V},\mathtt{P} ,\mathtt{S})$ and $\mathtt{V}$ 
345:  into $\mathtt{V} := \mathtt{V}_\mathfrak{t} + \mathtt{V}_\mathfrak{n}$ where $\mathtt{V}_\mathfrak{n} :=  (\mathtt{V}  .
346:  \mathfrak{n})  \mathfrak{n}$.
347:  The function  $\mathtt{V}$ (respectively  $\mathtt{P}$) takes its values in $\R^{d}$ (resp. $\R$).
348:  The function  $\mathtt{V}_\mathfrak{t} $ (respectively  $\mathtt{V}_\mathfrak{n}$) takes its values in $\R^{d-1}$ (resp. $\R$).
349: By abuse of notations, we will say that $\mathtt{V}$, $\mathtt{V}_\mathfrak{t} $,  $\mathtt{V}_\mathfrak{n} $ 
350: and $\mathtt{P}$ are in $\P(T)$ even if they do not take values in $\R^{d+2}$. 
351: 
352:    We look for formal solutions  $(u^\eps )_\eps $ of $(\ref{aller})$ of the form
353: %
354: \begin{eqnarray}
355: \label{sep}
356:  u^\eps (t,x) = \sum_{n \geqslant 0} \eps^n \, \mathcal{U}^n (t,x, \frac{\varphi(x) }{\eps})
357: \end{eqnarray}
358: %
359:  where each  $\mathcal{U}^n$ belongs to  $\P (T) $ and $\un{\mathcal{U}}^0 = u^0 $. 
360:  Let us explain what is meant by \textit{formal} \textit{solutions}.
361:  Plugging the expansion $(\ref{sep})$ into the system, using Taylor expansions and ordering the terms in powers of $\eps$, 
362:   we get a formal expansion in power series of $\eps$:
363: %
364: \begin{eqnarray*}
365: \sum_{n \geqslant -1} \eps^n \,  \Phi^n (t,x, \frac{\varphi(x) }{\eps})
366: \end{eqnarray*}
367: %
368:  where the  $(\Phi^n )_{n \geqslant -1}$ are in  $\P (T) $.
369:  We say that $(u^\eps )_\eps $ is a formal solution when all the resulting $\Phi^n $ are identically zero.
370:  Theorem $\ref{try}$ will sum up the main results of  section $\ref{secform}$.
371:  It  states that the system has formal solutions of the form
372:  %
373:  \begin{eqnarray}
374:  \label{pre2}  \quad
375:    \left\{
376:    \begin{array}{l}
377:   \mathtt{v}_\mathfrak{t}^\eps (t,x) = \mathtt{v}_\mathfrak{t}^0 (t,x) 
378:   + \displaystyle{\sum_{j \geqslant 1}} \eps^j \,  \mathtt{V}_\mathfrak{t}^{j-1} (t,x, \frac{\varphi(x) }{\eps}) ,
379: \vspace{0.2cm}  \\ \mathtt{v}_\mathfrak{n}^\eps (t,x) =  \mathtt{v}_\mathfrak{n}^0 (t,x) 
380:   + \eps \,  \mathtt{V}_\mathfrak{n}^0 (t,x) 
381:   + \displaystyle{\sum_{j \geqslant 2}} \eps^j \,  \mathtt{V}_\mathfrak{n}^{j-1} (t,x, \frac{\varphi(x)}{\eps}) ,
382: \vspace{0.2cm}  \\ \mathtt{p}^\eps (t,x) =  \mathtt{p}^0 (t,x) + \eps \, \mathtt{P}^0 (t,x) 
383:   +  \displaystyle{\sum_{j \geqslant 2}} \eps^j \,  \mathtt{P}^{j-1} (t,x, \frac{\varphi(x)}{\eps}) ,
384:   \vspace{0.2cm}  \\ \mathtt{s}^\eps (t,x) = \displaystyle{\sum_{j \geqslant 0}} \eps^j \,  \mathtt{S}^{j} (t,x,
385:   \frac{\varphi(x)}{\eps}) ,
386:     \end{array}
387: \right.
388:     \end{eqnarray}
389: %
390:   and that we can prescribe arbitrary initial values to the $(\mathtt{S}^j |_{t=0})_{j \in \N } $  
391:   and to the $(\mathtt{V}_\mathfrak{t}^j |_{t=0})_{j \in \N } $. 
392:    The profiles $\mathtt{V}_\mathfrak{t}^0$, $\mathtt{V}_\mathfrak{n}^0$ and $\mathtt{P}^0$ involved in $(\ref{pre2})$ are the
393:   profiles  $\mathtt{V}_\mathfrak{t}$, $\mathtt{V}_\mathfrak{n}$ and $\mathtt{P}$ involved in $(\ref{pre})$. 
394:   In $(\ref{pre2})$, we use the index $0$ to match with notations of section  $\ref{secform}$.
395:   In $(\ref{pre})$, we did not write the index in order to avoid heavy notations. 
396:    
397:   The two first equations of $(\ref{aller})$ involve the entropy $\mathtt{s}$, through the functions $\rho$ et $\alpha$.
398:   A priori the large entropy boundary layer could contaminate the 
399:   velocity\footnote{which can create a destabilizing effect  (see \cite{Gr1}, \cite{Gr2})}.
400:   We prove that since we consider some ground states  $u^0 := (\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 )$ solutions of 
401:  $(\ref{aller})$  verifying the conditions $(\ref{cond})$, there is no contamination at
402:  order $\eps^0$: when looking at the expansion $(\ref{pre2})$, we notice that there is no large pressure boundary layer and
403:  no large  velocity boundary layer.
404:  To use  the conditions $(\ref{cond})$, we reduce the system (cf. section $\ref{redbull}$). 
405:  This step is difficult.
406:  Let us briefly mention here some key points of our strategy.
407:  For sake of clarity, we  begin with the first equation of $(\ref{aller})$ only.
408:  We look for solutions $(u^\eps )_\eps $ 
409:    where $\mathtt{v}^\eps$ and $\mathtt{p}^\eps$ are of the form
410:      $\mathtt{v}^\eps:= \mathtt{v}^0 + \eps \mathtt{V}^\eps $, 
411:      $\mathtt{p}^\eps := \mathtt{p}^0 + \eps \mathtt{P}^\eps $.
412:  We split the particle derivative $\X_{\mathtt{v}^\eps} \, \mathtt{v}^\eps$ into
413:     %
414:  \begin{eqnarray*}
415:    \X_{\mathtt{v}^\eps} \, \mathtt{v}^\eps =  \X_{\mathtt{v}^0} \, \mathtt{v}^0 
416:    + \eps  \X_{\mathtt{v}^\eps} \, \mathtt{V}^\eps
417:    +  \eps \mathtt{V}^\eps . \nabla  \mathtt{v}^0 .
418:    \end{eqnarray*}
419: %%%
420:   Thanks to the conditions $(\ref{cond})$, there exists a function $\phi \in C^\infty ((0,T_0 ) \times \Omega )$ 
421:   such that for all  $(t,x) \in (0,T_0 ) \times \Omega $, 
422:    $(\X_{\mathtt{v}^0} \, \mathtt{v}^0 )(t,x)= \varphi(x)\, \phi(t,x) $.
423:    Remark that if $\tilde{ \mathcal{U}} \in   \Nrond (T_0 ) $, then 
424:    %
425:  \begin{eqnarray*}
426:   \X_{\mathtt{v}^0} \, \mathtt{v}^0 \, . \tilde{\mathcal{U}} (t,x, \frac{\varphi(x)}{\eps}) =
427:   \eps (\phi(t,x) X \tilde{ \mathcal{U}} (t,x,X))|_{X=\frac{\varphi(x)}{\eps}} 
428:   \quad \mathrm{and} \quad X \tilde{ \mathcal{U}} \in  \Nrond (T_0 ).
429:   \end{eqnarray*}
430: %%%
431: To use this remark, we develop the term $\rho(\mathtt{p}^\eps , \mathtt{s}^\eps )$. 
432: Thus, we write  $\mathtt{s}^\eps$ under the form  
433:  %
434:  \begin{eqnarray}
435:  \label{SSS}
436:  \mathtt{s}^\eps (t,x) = \mathtt{s}^0 (t,x) + \tilde{\mathtt{S}}^0  (t,x ,\frac{\varphi(x)}{\eps})  
437:    + \eps \mathtt{S}^{\eps,\flat} (t,x),
438:  \end{eqnarray}
439: %
440: with $\tilde{\mathtt{S}}^0  \in  \Nrond (T_0 ) $.
441: We get for the term $\rho( \mathtt{p}^\eps , \mathtt{s}^\eps )$ the following expansion:
442:  %
443:  \begin{eqnarray*}
444: \rho( \mathtt{p}^\eps , \mathtt{s}^\eps ) &=&  \rho( \mathtt{p}^0 , \mathtt{s}^0 )
445: +  (\rho( \mathtt{p}^0 ,  \mathtt{s}^0 + \tilde{\mathtt{S}}^0 ) -  \rho( \mathtt{p}^0 , \mathtt{s}^0 ))
446: +  \eps D \rho( \mathtt{p}^0 , \mathtt{s}^0 ). ( \mathtt{P}^\eps , \mathtt{S}^{\eps,\flat} )
447: + O(\eps^2 ) .
448: \end{eqnarray*}
449: %%%
450: Because  $u^0$ satisfies the equation  $(\ref{aller})$, we get the equation:
451: %
452:  \begin{eqnarray*}
453:  \rho( \mathtt{p}^\eps , \mathtt{s}^\eps ) .\X_{\mathtt{v}^\eps } \,  \mathtt{V}^\eps + \nabla   \mathtt{P}^\eps   
454:    + \mathfrak{P}_\mathtt{v} = O(\eps ) \quad
455:    \mathrm{with}  \quad
456:   \mathfrak{P}_\mathtt{v} := \mathfrak{P}_\mathtt{v}^a + \mathfrak{P}^b_\mathtt{v} + \mathfrak{P}^c_\mathtt{v}
457:  \\ \nonumber  
458:   \mathrm{where}  \quad 
459:   \left\{
460: \begin{array}{l}
461:  \mathfrak{P}_\mathtt{v}^a :=
462:   \rho( \mathtt{p}^0 , \mathtt{s}^\eps ). \mathtt{V}^\eps . \nabla \mathtt{v}^0 , 
463: \vspace{0.2cm}\\  \mathfrak{P}^b_\mathtt{v} := 
464: \eps^{-1} \, (\rho( \mathtt{p}^0 ,  \mathtt{s}^0 + \tilde{\mathtt{S}}^0 ) -  \rho( \mathtt{p}^0 , \mathtt{s}^0
465: )).\X_{\mathtt{v}^0} \, \mathtt{v}^0 , 
466: \vspace{0.2cm}\\  \mathfrak{P}^c_\mathtt{v} :=
467:   D \rho( \mathtt{p}^0 , \mathtt{s}^0 ). ( \mathtt{P}^\eps , \mathtt{S}^{\eps,\flat} ) .\X_{\mathtt{v}^0} \,
468: \mathtt{v}^0 .
469: \end{array}
470: \right.
471:  \end{eqnarray*}
472: %
473: With the previous remark, we get
474: %
475:  \begin{eqnarray*}
476:  \mathfrak{P}^b_\mathtt{V} =  \{ \phi X (\rho( \mathtt{p}^0 ,  \mathtt{s}^0 + \tilde{\mathtt{S}}^0 ) 
477:  -  \rho( \mathtt{p}^0 , \mathtt{s}^0 )) \}|_{X=\frac{\varphi(x)}{\eps}} .
478:   \end{eqnarray*}
479: %%%
480: Proceed in the same way for the second equation of $(\ref{aller})$, we get the equation:
481: %
482:  \begin{eqnarray*}
483:   \alpha ( \mathtt{p}^\eps , \mathtt{s}^\eps ) .\X_{\mathtt{v}^\eps } \,  \mathtt{P}^\eps + \mathrm{div} \,  \mathtt{V}^\eps 
484: + \mathfrak{P}_\mathtt{p}  = O(\eps ) \quad
485:   \mathrm{with}  \quad 
486:   \mathfrak{P}_\mathtt{p} := \mathfrak{P}^a_\mathtt{p} + \mathfrak{P}^b_\mathtt{p} + \mathfrak{P}^c_\mathtt{p} 
487:   \\ \nonumber  
488:   \mathrm{where}  \quad 
489:   \left\{
490: \begin{array}{l}
491: \mathfrak{P}^a_\mathtt{p} :=  \alpha ( \mathtt{p}^0 , \mathtt{s}^\eps ). \mathtt{V}^\eps . \nabla \mathtt{p}^0 ,
492: \vspace{0.2cm}\\ \mathfrak{P}^b_\mathtt{p} := \eps^{-1} \, ( \alpha  ( \mathtt{p}^0 ,  \mathtt{s}^0 + \tilde{\mathtt{S}}^0 ) 
493: - \alpha   ( \mathtt{p}^0 , \mathtt{s}^0
494: )).\X_{\mathtt{v}^0} \, \mathtt{p}^0
495: \vspace{0.2cm}\\  \qquad  =    \{ \phi X (\alpha  ( \mathtt{p}^0 ,  \mathtt{s}^0 + \tilde{\mathtt{S}}^0 ) 
496:  - \alpha  ( \mathtt{p}^0 , \mathtt{s}^0 )) \}|_{X=\frac{\varphi(x)}{\eps}} ,
497: \vspace{0.2cm}\\ \mathfrak{P}^c_\mathtt{p} := 
498: D \alpha (\mathtt{p}^0 , \mathtt{s}^0 ). (\mathtt{P}^\eps ,\mathtt{S}^{\eps,\flat}) .\X_{\mathtt{v}^0} \,
499: \mathtt{p}^0 .
500: \end{array}
501: \right.
502:  \end{eqnarray*}
503: %
504: In other words, the unknown $U^\eps := (\mathtt{V}^\eps ,\mathtt{P}^\eps , \mathtt{s}^\eps )$ verify the Euler system $(\ref{aller})$
505: except  the perturbation terms  $\mathfrak{P}_\mathtt{V}$, $  \mathfrak{P}_\mathtt{P}$, and 
506:  $O(\eps )$.
507: We see that the terms $\mathfrak{P}_\mathtt{v}^a$, $\mathfrak{P}_\mathtt{v}^b$, $\mathfrak{P}_\mathtt{p}^a$ and 
508: $\mathfrak{P}_\mathtt{p}^b$ do not have any singular factor with respect to $\eps$.
509: This is a consequence of  $(\ref{cond})$.
510: These terms are expressed in function of the unknown $U^\eps$, of the ground state $u^0$  and of the boundary layer $\tilde{\mathtt{S}}^0$.
511: The terms  $\mathfrak{P}^c_\mathtt{v}$ and $\mathfrak{P}^c_\mathtt{p}$, them, involve $\mathtt{S}^{\eps,\flat}$.
512: If we try to eliminate $\mathtt{S}^{\eps,\flat}$ via $(\ref{SSS})$, we involve  the unknown $U^\eps$ in a singular way  
513: via the term $\eps^{-1} \, \mathtt{s}^\eps$. 
514: We overcome this difficulty  in Lemma  $\ref{psg}$  using that the terms  $\X_{\mathtt{v}^0}
515: \, \mathtt{p}^0$ and  $\X_{\mathtt{v}^0} \, \mathtt{v}^0$ are respectively in factor of  $\mathfrak{P}^c_\mathtt{v}$ and
516: $\mathfrak{P}^c_\mathtt{p}$.
517:  Moreover the terms  $\mathfrak{P}_\mathtt{V}$ and $  \mathfrak{P}_\mathtt{P}$ are affine with respect to
518:   $ (\mathtt{V}^\eps ,\mathtt{P}^\eps )$. 
519:   
520:    The profile equations are linear, thanks to some original \textbf{transparency} properties of the Euler system. 
521: On one hand, the entropy boundary layer profile $\tilde{\mathtt{S}}^0$ verifies a transport equation
522:  which is linear with respect to the entropy (cf. equation $(\ref{clem})$). 
523: On the other hand, the amplitude of the boundary layer on the tangential velocity is weak (of order $\eps$) and the boundary is
524: characteristic for a linearly degenerate field. 
525: Thanks to this, the tangential velocity  boundary layer  profile $\tilde{\mathtt{V}}_\mathfrak{t}$ satisfies a linear equation,
526:  without Burgers-like nonlinearity
527: (cf. equation $(\ref{poli})$).
528: This is a transparency phenomenon analogous to the one observed in \cite{moi3}.
529: A interesting point is that such transparency phenomena does not occur for large amplitude high frequency entropy waves (see
530: Theorem $3.9$ of  \cite{cgm2}).
531: \\
532: \\  \textit{In section $\ref{stab}$}  we are interested in the \textbf{existence} (cf. Theorem $\ref{di}$) and
533:  the \textbf{propagation} (cf. Theorem $\ref{propa}$) of \textbf{exact solutions}
534:  of $(\ref{aller})$ asymptotic to approximate solutions obtained by truncating formal solutions 
535:  constructed in section $\ref{secform}$.
536:  Theorem $\ref{int}$ given in the introduction is a consequence of Theorem $\ref{try}$ and Theorem  $\ref{di}$. 
537:   After a reduction (cf. Prop. $\ref{Reduction2}$, subsection $\ref{Reduction}$), 
538:  we will face a singular perturbation problem because of boundary layers which corresponds to variations
539: in $\frac{\varphi(x)}{\eps}$. 
540:  More precisely we deal with
541:   a family of quasi-linear symmetric hyperbolic boundary value problem.
542:  As for the originating Euler problem, the boundary is conservative and characteristic of constant multiplicity.
543:  To tackle this characteristic problem  we get inspired by the paper \cite{Gu90} of O.Gu\`es which uses 
544:  the notion of conormal regularity and  the spaces 
545: %
546: \begin{eqnarray*}
547: E^m (T) := \{   u  \in L^2  (T)  /  \quad 
548: \sum_{0 \leqslant 2k+|\alpha| \leqslant m}  || \partial_\mathfrak{n}^k  \, Z^\alpha \, u ||_{L^2 ( T )}  < \infty  \} ,
549: \end{eqnarray*}
550:  %
551:  with $\alpha := (\alpha_0 ,..., \alpha_d ) \in \N^{d+1}$, $|\alpha|:= \alpha_0 +...+ \alpha_d $, $Z^\alpha := Z_0^{\alpha_0} ...
552: Z_d^{\alpha_d} $ where $( Z_0,...,Z_d) $ generates the  algebra of $C^\infty$ tangent vector to $\Gamma$.
553: To simplify, we  denote $L^2 (T) :=  L^2 ((0,T) \times \Omega ) $.
554:  For these spaces  one normal derivative corresponds to two conormal derivatives.
555:  We adapt the method of \cite{Gu90} by substituting the derivative $\eps \partial_\mathfrak{n} $ 
556:  to the derivative $\partial_\mathfrak{n} $ 
557: in order to obtain uniform estimates and will use the following subsets of $L^2 (T)^{]0,1]}$:
558: %
559: \begin{eqnarray*}
560: \mathbf{E}^m (T) := \{   (u^\eps)_\eps   \in L^2 (T ) /  \quad 
561:  \mathrm{sup}_{0 < \eps \leqslant 1}  
562:  \sum_{0 \leqslant 2k+|\alpha| \leqslant m}  ||  (\eps \partial_\mathfrak{n} )^k \, Z^\alpha \, u^\eps ||_{L^2 ( T )}  < \infty  \} .
563: \end{eqnarray*}
564: %
565: This idea to  use some derivatives with $\eps$ in factor for some  singular perturbation problems is natural and was also used in
566: the papers of  \cite{Gu93},  \cite{Gu92} with the $\eps$-stratified notion, \cite{cgm2} with
567:  the $\boldsymbol{\eps}$\textbf{-conormal} notion. 
568: Here, this idea is applied to (characteristic) boundary value problem and anisotropic Sobolev spaces.
569:  
570:  
571:   At first look, this system we obtained is singular with respect to $\eps$ but a trick allows to overcome this false singularity
572: (see subsection $\ref{Reduction}$).
573:  We will use a family of iterative schemes. 
574: Thus we will supply in subsection $\ref{Linesti}$ linear estimates which are the core the proof. 
575: We will successively perform $L^2$ estimates, conormal estimates and normal estimates.
576: A main difficulty lies in the way to deal with commutators (cf. Proposition $\ref{lapin}$).
577: This strategy yields exact solutions till $T_0$.
578: The proof of Theorem  $\ref{di}$ needs carefulness about the existence of compatible initial data.      
579: Subsection  $\ref{bobo}$ is devoted to this question.
580:   
581:  It is possible to obtain $L^\infty$ estimates, in spite of the fact that $d \geqslant 1$. 
582: We refer to  papers \cite{m2}, \cite{sonique} of G.M\'etivier, paper \cite{RR} of J.Rauch and M.Reed and paper \cite{Gu90}.
583: This idea is still relevant when adapting $\eps$-conormal regularity to characteristic boundary value problem.
584: Therefore we can weaken the regularity of the solution and prove a propagation result for some solutions admitting only
585: one normal derivative in $L^2$. 
586: We introduce the following subsets of $L^2 (T)^{]0,1]}$:
587: %
588: \begin{eqnarray*}
589: \mathbf{A}^m (T) := \{   (u^\eps)_\eps   /  \quad 
590:  \mathrm{sup}_{0 < \eps \leqslant 1}  
591:  \sum_{0 \leqslant |\alpha| \leqslant m}  ||  Z^\alpha  \, u^\eps ||_{L^2 ( T)} 
592:  +  \sum_{0 \leqslant  |\alpha| \leqslant m-2}  ||  \eps \partial_\mathfrak{n}  \, Z^\alpha  \, u^\eps ||_{L^2 ( T)} < \infty  \} .
593: \end{eqnarray*}
594: %
595: We will also use some norms built on $L^\infty$. 
596: Because the boundary is characteristic, we will need not only the Lipschitz norms but higher order $L^\infty$ control, 
597: as O. Gu\`es in \cite{Gu90} and G. M\'etivier in \cite{sonique}.
598: We will denote by $L^\infty (T) $ the space     $ L^\infty (T)  :=  L^\infty ((0,T) \times \Omega ) $.
599:  We will introduce the norms
600: %
601: \begin{eqnarray*}
602:  ||u||^*_{\eps,T}  &:=& \sum_{0 \leqslant |\alpha| \leqslant 2} ||Z^\alpha \, u||_{L^\infty (T)}
603:  +  \sum_{0 \leqslant |\alpha| \leqslant 1} ||Z^\alpha  \, \eps \partial_\mathfrak{n}  \, u||_{L^\infty (T)} ,
604: \end{eqnarray*}
605: %
606: and the following subsets of $L^\infty (T)^{]0,1]}$:
607: %
608: \begin{eqnarray*}
609: \mathbf{\Lambda}^m (T) := \{  (u^\eps)_\eps    /  \quad 
610:  \mathrm{sup}_{0 < \eps \leqslant 1} \, ||u^\eps ||^*_{\eps,T} < \infty  \} . 
611: \end{eqnarray*}
612: %
613: Theorem $\ref{propa2}$ states a propagation result in the spaces $\mathbf{A}^m (T) \cap \mathbf{\Lambda}^m (T) $.  
614:   
615:  One quality of our method is that we need approximate solutions with only a few profiles.
616: The minimum number of profiles required is linked to the lost of a factor $\eps^{\frac{1}{2}}$ in a Sobolev embedding Lemma (Lemma
617: $\ref{sobo}$). 
618: Let us explain one motivation to minimize the number of profiles needed.
619:  In this paper, we consider a ground state $u^0$ in $H^\infty ((0,T_0) \times \Omega )$  and formal solutions with $H^\infty$
620: regularity. It could also be possible to extend to ground states of high but finite regularity. 
621: 
622:  \section{Forthcoming}
623:  \label{coming}
624: 
625: We plan to show in a further work that for more general ground states $u^0$ which are solutions of $(\ref{aller})$,
626: which verify the condition $\X_{\mathtt{v}^0} \,  \mathtt{v}^0 =0$ for $(t,x) \in (0,T_0) \times \Gamma$
627:  but which do not verify the
628: condition $\X_{\mathtt{v}^0} \,  \mathtt{p}^0 =0$ for $(t,x) \in (0,T_0) \times \Gamma$, 
629: it is still possible to construct, in small time, some nontrivial formal solutions of $(\ref{aller})$ but
630: of the more general form
631:  %
632:  \begin{eqnarray*}
633:     \left\{
634:    \begin{array}{l}
635:   \mathtt{v}_\mathfrak{t}^\eps (t,x) = 
636:    \displaystyle{\sum_{j \geqslant 0}} \eps^j \,  \mathtt{V}_\mathfrak{t}^{j} (t,x, \frac{\varphi(x) }{\eps}) ,
637: \vspace{0.2cm}  \\ \mathtt{v}_\mathfrak{n}^\eps (t,x) =  \mathtt{v}_\mathfrak{n}^0 (t,x) 
638:   + \displaystyle{\sum_{j \geqslant 1}} \eps^j \,  \mathtt{V}_\mathfrak{n}^{j} (t,x, \frac{\varphi(x)}{\eps}) ,
639: \vspace{0.2cm}  \\ \mathtt{p}^\eps (t,x) =  \mathtt{p}^0 (t,x) + \eps \mathtt{p}^1 (t,x) 
640:   +  \displaystyle{\sum_{j \geqslant 2}} \eps^j \,  \mathtt{P}^{j} (t,x, \frac{\varphi(x)}{\eps}) ,
641:   \vspace{0.2cm}  \\ \mathtt{s}^\eps (t,x) = \displaystyle{\sum_{j \geqslant 0}} \eps^j \,  \mathtt{S}^{j} (t,x,
642:   \frac{\varphi(x)}{\eps}) .
643:     \end{array}
644: \right.
645:     \end{eqnarray*}
646: %
647: Moreover we will show  that we can prescribe arbitrary initial values for the $(\mathtt{V}_\mathfrak{t}^{j} )_{j \in \N}$ 
648: and the $(\mathtt{S}^{j} )_{j \in \N}$.
649:   However these formal solutions can be unstable.
650:   
651:   
652: \section{Setting of the notations}
653: \label{setting}
654: 
655: To simplify and avoid heavy notations, we will consider from now on that the domain $\Omega$ 
656: is the half-space $ \Omega := \{ x \in \R^d / \quad x_d \geqslant 0 \}$.
657:  This assumption does not change the mathematical analysis of the problem. 
658:  We fix a notation. 
659: If $A$ is a $d+2$ by $d+2$ square matrix, we denote by $A^\star$ the $d+1$ by $d+1$ extracted square matrix which
660: contains the $d+1$ first rows of the $d+1$ first lines.
661: With $u=(\mathtt{v}, \mathtt{p},\mathtt{s})$, the Euler system is of the form
662:  %
663:  \begin{eqnarray}
664:  \label{po}
665:  \X_\mathtt{v} \, u + \M (u,\partial_x ) u =0 
666:  \end{eqnarray}
667: %
668:  with
669:  %
670:  \begin{eqnarray*}
671: \M (u,\xi) := 
672:  \begin{bmatrix}
673:  \M^\star (u,\xi) &0 
674:  \\ 0 & 0 
675:   \end{bmatrix}
676:   ,\quad 
677:   \M^\star (u,\xi) := 
678:  \begin{bmatrix}
679:   0 &  \mathtt{\rho}^{-1} \ \xi
680: \\  \mathtt{\alpha}^{-1} \  {}^{t} \xi & 0 
681: \end{bmatrix},
682: \end{eqnarray*}
683:  %
684: We recall that we assume that  $\alpha (\mathtt{p},\mathtt{s}) > 0$ for all $\mathtt{p}$ and  $\mathtt{s}$. 
685: We  denote by
686: %
687: \begin{eqnarray*}
688: S(u) :=
689: \begin{bmatrix}
690:  S^\star (u) & 0
691: \\   0 &  1
692:   \end{bmatrix}
693:    ,\quad 
694:  S^\star (u) :=
695: \begin{bmatrix}
696:      \mathtt{\rho} I_d & 0 
697: \\    0  & \alpha 
698:  \end{bmatrix} .
699: \end{eqnarray*}
700: %
701: Multiply Eq. $(\ref{po})$ by  the  matrix $S(u)$
702: to get the equation 
703: %
704: \begin{eqnarray}
705:  \label{po2}
706: \mathfrak{L}(u,\partial) u =0 
707:  \end{eqnarray}
708: %
709: where  $\mathfrak{L}(u,\partial) := S(u)  \X_\mathtt{v}  + \L (\partial_x )$ and  for all $\xi \in \R^d$, 
710: %
711: \begin{eqnarray*}
712: \L(\xi) :=  S(u) \M (u,\xi) =
713: \begin{bmatrix}
714:  \L^{\star} (\xi) &0 
715:  \\ 0 & 0 
716: \end{bmatrix} 
717:  ,\quad 
718:  \L^{\star} (\xi) :=
719: \begin{bmatrix}
720:    0 &  \xi
721: \\    {}^{t} \xi  & 0 
722:   \end{bmatrix}.
723: \end{eqnarray*}
724: %
725: We also introduce the operator  $\mathfrak{L}^{\star} (u,\partial) := S^{\star}(u)  \X_\mathtt{v}  + \L^{\star} (\partial_x )$.
726: The matrix $S(u)$ is symmetric  definite nonnegative and depends in a $C^\infty $ way of  $u$.
727: The system $(\ref{po2})$ is therefore symmetric hyperbolic.
728: For this system, the boundary conditions $\mathtt{v}. \mathfrak{n} = 0$  are conservative.
729: We denote by $v:=  (\mathtt{v}, \mathtt{p})$ and by   $ w:=\mathtt{s}$.
730: We introduce for all $\xi \in \R^d - \{ 0 \}$ the subspace $ \mathbb{F} (u,\xi) := \ker \M(u,\xi)$ of $ \R^N $.
731: We denote by $(e_1,...,e_{d+2})$ the canonical basis of $\R^{d+2}$.
732: Notice that $ \mathbb{F} (u,\xi) = \xi^{\bot} \oplus \R e_{d+2} $ and that
733: $\mathbf{\lambda} (u,\xi):= \mathtt{v}. \xi$ is a linearly degenerate eigenvalue with constant multiplicity $d> 0 $ i.e.
734: %
735: \begin{eqnarray*}
736: \left.
737: \begin{array}{c}
738: r. \nabla_u \mathbf{\lambda} (u,\xi) = 0 ,\quad \mathrm{for} \  \mathrm{all} \ r \in \mathbb{F} (u,\xi) ,
739:   
740: \\  \dim \mathbb{F} (u,\xi) = d ,
741: \end{array}
742: \right\}
743: \quad \mathrm{for} \  \mathrm{all} \  (u,\xi) \in \R^{d+2} \times (\R^d - \{ 0 \}) .
744: \end{eqnarray*}
745:  %
746: Notice that $\{ v=0 \} \subset \ker \M(u,\xi)$, for all  $ \xi \in \R^d - \{ 0 \}$.
747: We  denote by $P_0$ the  orthogonal projector on $\ker \L_d$ i.e.  
748: %
749: \begin{eqnarray*}
750: P_0 := 
751: \begin{bmatrix}
752: P_0^{\star} & 0
753: \\ 0 & 1 
754: \end{bmatrix},
755: \end{eqnarray*}
756: %
757: where $P_0^{\star}$ is a $(d+1) \times (d+1) $ matrix 
758: %
759: \begin{eqnarray*}
760: P_0^{\star} := 
761: \begin{bmatrix}
762:  I_{d-1} & 0 & 0
763: \\ 0 & 0 & 0
764: \\ 0 & 0 & 0
765: \end{bmatrix}.
766: \end{eqnarray*}
767: %
768: We will often split $\mathcal{U} \in \P(T)$ into $\mathcal{U} =( \mathcal{V},  \mathcal{W})$.
769: The function  $\mathcal{V}$ (respectively  $\mathcal{W}$) takes its values in $\R^{d+1}$ (resp. $\R$).
770:  We will sometimes split $\mathcal{V}$ into  $\mathcal{V}=(\mathtt{V},\mathtt{P})$ and $\mathtt{V}$ 
771:  into $\mathtt{V} := (\mathtt{V}_\mathfrak{t} , \mathtt{V}_d)$. 
772:  The function  $\mathtt{V}$ (respectively  $\mathtt{P}$) takes its values in $\R^{d}$ (resp. $\R$).
773:  The function  $\mathtt{V}_\mathfrak{t} $ (respectively  $\mathtt{V}_d$) takes its values in $\R^{d-1}$ (resp. $\R$).
774: By abuse of notations, we will say that $\mathcal{V}$, $\mathcal{W}$, $\mathtt{V}$, $\mathtt{V}_\mathfrak{t} $,  $\mathtt{V}_d$ 
775: and $\mathtt{P}$ are in $\P(T)$ even if they do not take values in $\R^{d+2}$. 
776: 
777: \section{Overdetermined ground states} 
778: \label{ground}
779: 
780: In this section, we  prove the existence of ground states $(\mathtt{v}^0 , \mathtt{p}^0 ,\mathtt{s}^0 ) \in H^\infty$ solutions of
781:  $(\ref{aller})$ and verifying the conditions $(\ref{pro})$.
782:  
783: \begin{theo}
784: \label{tt}
785: Given $h \in C^1 (\R^d ,\R^d )$ such that $h'(x)$ is nilpotent in a neighborhood of $0$ and such that $h_d (x) = 0$ when $x_d =0$,
786: there exists a local $C^1$ solution $\mathtt{v}$ of the initial boundary value problem:
787: %
788: \begin{eqnarray*}
789: \partial_t \mathtt{v} + (\mathtt{v} . \nabla_x ) \mathtt{v} = 0 , \quad \mathrm{div}_x \mathtt{v} =0 ,\quad \mathtt{v}_d |_{x_d =0 }
790: =0 ,\quad \mathtt{v} |_{t =0 } = h .
791: \end{eqnarray*}
792: %
793: \end{theo}
794: %
795: \begin{proof}
796: %
797: Local existence of $C^1$ solution of multidimensional Burgers equation can be achieved with characteristic method. 
798: The classical relation $\mathtt{v} (t,x+t h(x)) = h(x)$ holds in a neighborhood of $0$. 
799: Because $h_d (x) = 0$ when $x_d =0$, we get $\mathtt{v}_d  (t,x) = 0$ for $x_d =0$.
800: The divergence free relation is a consequence of the nilpotence of $h'(x)$ as proved in Theorem $2.6$ of \cite{cgm2}.
801: %
802: \end{proof}
803: %
804: %
805: Assume that $d=2$ for a moment and let us give some examples of initial velocity $h$
806:  which satisfy the assumptions of Theorem $\ref{tt}$.
807: If $h$ satisfies $h_1 (x)= x_2$ and $h_2 (x) =0$ in a neighborhood  of $0$, then $h$ is convenient. 
808: We now detail a more general process to get convenient $h$.
809: Let $F$ be a function in $ C^1 (\R_+ ,\R )$ such that $F(0)=0$ and $F'(x) > 0$ for all $x \in \R_+ $.
810: Let $a_{init} \in C^1 (\R_+ , \R)$. 
811: There exists a local solution $a \in C^1 (\R \times \R_+ ,\R )$ of the scalar initial boundary value problem  
812: %
813: \begin{eqnarray*}
814: \partial_1 a + \partial_2 a =0 , \quad a|_{x_2 =0} =0  , \quad a|_{x_1 =0} = a_{init} .
815: \end{eqnarray*}
816: %
817: Consider $h(x) := (a(x),F \circ a (x))$. 
818: Then $h$ satisfies the assumptions of Theorem $\ref{tt}$.
819: 
820:  Of course, because conditions $(\ref{pro})$ imply conditions  $(\ref{cond})$, what precedes proves the existence of $T_0 > 0$ and
821: of a ground state $u^0 := (v^0 , w^0 ) \in H^\infty ((0,T_0 ) \times \Omega )$ solution of  $(\ref{aller})$ and verifying the
822: conditions $(\ref{cond})$.
823: From now on, we will assume that such a ground state is given.
824: 
825: \section{Reduction of the system} 
826: \label{redbull}
827: 
828: We look for solutions of  $(\ref{aller})$  of the  form
829: %
830: \begin{eqnarray*}
831: u^\eps =(v^\eps := v^0 + \eps V^\eps , w^\eps := W^\eps ).
832: \end{eqnarray*}
833: %
834: We are going to realize a singular change of unknown, by looking for an equation for $U^\eps := (V^\eps , W^\eps )$. 
835: 
836: \begin{prop}
837: \label{Re1}
838: 
839: The function $u^\eps $ is solution of the equation $(\ref{po})$ if and only if the function $U^\eps$ verify the equation 
840: %
841:  \begin{eqnarray}
842:  \label{Re2}
843: \mathfrak{L}(u^\eps ,\partial)  U^\eps +  K^\eps  =0 ,
844:  \end{eqnarray}
845: %
846:  where
847:   %
848:  \begin{eqnarray}
849:  \label{K1} 
850:   K^\eps  := \frac{1}{\eps} 
851:  \begin{bmatrix}   
852:  K_1 ( v^0 , \partial v^0  , u^\eps) 
853:  \\  0  
854:  \end{bmatrix} \quad  \mathrm{with} \  
855:  K_1 (v^0 , \partial v^0 , u^\eps) =  \mathfrak{L}^\star (u^\eps ,\partial ) v^0 .
856:   \end{eqnarray}
857: 
858: \end{prop}
859:  %
860: \begin{proof}
861:  %
862: The equation $(\ref{po})$ is equivalent to the two following equations:
863: %
864: \begin{eqnarray}
865:  \label{Re3}
866:  \X_{\mathtt{v}^\eps} \,   v^\eps + \M^\star ( u^\eps , \partial_x ) v^\eps =0 ,
867: \quad  \X_{\mathtt{v}^\eps} \,   w^\eps  =0 ,
868: \end{eqnarray} 
869: %
870: By dividing  the first equation by $\eps$, it's also  equivalent to 
871: %
872: \begin{eqnarray*}
873:  \X_{\mathtt{v}^\eps} \,   V^\eps + \M^\star ( u^\eps , \partial_x ) V^\eps 
874:  + \frac{1}{\eps} (  \X_{\mathtt{v}^\eps} \,   v^0 + \M^\star ( u^\eps , \partial_x ) v^0 ) = 0 ,
875: \quad  \X_{\mathtt{v}^\eps} \,   W^\eps =0 .
876: \end{eqnarray*}
877: %
878:  We get:
879:  %
880: \begin{eqnarray}
881: \label{pet0}
882:  \X_{ \mathtt{v}^\eps} U^\eps +
883:   \mathcal{M} U^\eps + 
884:  \begin{bmatrix} 
885:  \frac{1}{\eps} (\X_{ \mathtt{v}^\eps} v^0 +   \mathcal{M}^\star ( u^\eps , \partial_x) v^0 )
886:  \\ 0 
887:  \end{bmatrix}
888:  =0  .
889: \end{eqnarray}
890:  %
891:  We multiply $(\ref{pet0})$ to the left by the matrix  $S (u^\eps )$ to end the proof. 
892:   %
893: \end{proof}
894: %
895: 
896: The underlying idea of the previous proposition is that because $v^0$ is given as a ground state, $U^\eps$ is the real unknown.
897: For each $\eps$, we obtain that $U^\eps$ satisfies a hyperbolic system. However in order to achieve an asymptotic analysis for
898: $\eps \rightarrow 0^+$, we can be \textit{a priori} worried about the singular factor $\eps^{-1}$ within $K^\eps$.
899: Let us recall the way \cite{cgm2} deals with this term, expliciting their calculus for this particular case of the Euler system.
900: First because $\X_{ \mathtt{v}^\eps} v^0 = \X_{\mathtt{v}^0} \,  v^0 + \eps \mathtt{V}^\eps . \nabla v^0 $, we get 
901: %
902: \begin{eqnarray*} 
903: K^\eps := \begin{bmatrix}
904: \frac{1}{\eps} S^\star (u^\eps ) \X_{\mathtt{v}^0} \,  v^0  + \frac{1}{\eps} \L^\star (\partial_x ) v^0 
905: +  S^\star (u^\eps )  \mathtt{V}^\eps . \nabla v^0 
906:  \\ 0 
907:  \end{bmatrix}.
908:  \end{eqnarray*}
909:  %
910:  Remember that \cite{cgm2} use the condition  $(\ref{pro})$ on the ground state $u^0$
911:   which are stronger than the condition $(\ref{cond})$
912:   we use in the present paper.
913:  The condition  $(\ref{pro})$ reads  
914: $ \X_{\mathtt{v}^0} \,  v^0 = \L^\star (\partial_x ) v^0 = 0$ and 
915:   %
916:   \begin{eqnarray*}
917:      K^\eps = \begin{bmatrix} S^\star (u^\eps )  \mathtt{V}^\eps . \nabla 
918: v^0 \\ 0 
919:  \end{bmatrix}
920:  \end{eqnarray*}
921:  %
922:   does not contain any singular factor $\eps^{-1}$ anymore.
923: The following proposition will show that under condition $(\ref{cond})$ the term $K^\eps$ does not contain 
924: any singular factor $\eps^{-1}$ too.
925: In order to help the reader, we give some hints about our strategy.
926: We will take into account that we search $W^\eps$ of the form 
927: %
928: \begin{eqnarray*} 
929: W^{\eps} (t,x)  = \mathcal{W}^0 (t,x,\frac{x_d}{\eps}) + \eps W^{\eps,\flat} (t,x)
930: \end{eqnarray*}
931:  %
932:  with $\mathcal{W}^0
933: \in \P (T )$ (cf. $(\ref{defp})$ for the definition) and $ \underline{\mathcal{W}}^0 =w^0$. 
934: Under the conditions $(\ref{pro})$, we have only expanded $v^\eps$.
935: Because we assumed that $\X_{ \mathtt{v}^0} v^0 =  0$, the term  $S^\star (u^\eps ) \X_{\mathtt{v}^0} \,  v^0 $ vanished.
936: Under the localized condition $(\ref{cond})$, the idea is tricker.
937: We will expand $W^\eps$ too, into $W^\eps = w^0 + \tilde{\mathcal{W}}^0 +\eps W^{\eps,\flat} $ within $S^\star (u^\eps )$.
938: Imagine in a first time that $\tilde{\mathcal{W}}^0$  is identically zero.
939: Then we obtain by a Taylor first order expansion that
940: %
941: \begin{eqnarray*} 
942: K^\eps =
943:  \begin{bmatrix} 
944:  \frac{1}{\eps} \mathfrak{L}(u^0,\partial)  v^0  + \mathrm{terms} \ \mathrm{of} \  \mathrm{order} \ \eps^0 
945:   \\ 0 
946:  \end{bmatrix} .
947:   \end{eqnarray*}
948:  %
949:  Because the ground state $u^0$ satisfies $(\ref{aller})$, the first term in the right member is equal to zero and $K^\eps$ does
950:  not appear singular with respect to  $\eps$ anymore.
951:  Of course we want to deal with some nonvanishing function $\tilde{\mathcal{W}}^0$. 
952: A key difference with paper \cite{cgm2} is that here $\tilde{\mathcal{W}}^0$ denotes a boundary layer and we will see in the
953:  following proposition that its singular contribution within $K^\eps$ contains the trace of $\X_{ \mathtt{v}^0} v^0 $ 
954:  on the boundary $\Gamma := \{ x_d =0 \}$ as a factor.
955: Therefore the localized conditions $(\ref{cond})$ allow to conclude.
956: In prevision of the following sections, we will be careful with the way the term $K^\eps$ depends of $V^\eps$ and
957: $W^{\eps,\flat}$.   
958:  
959:  
960: In order to avoid heavy notations, we will omit the two first arguments and write $ K^\eps_1 (u^\eps)$ 
961: instead of  $ K^\eps_1  (v^0 , \partial v^0 , u^\eps)$.
962: Furthermore, we want now to use the special form of the solutions $u^\eps$ we are looking for. 
963: Thus we will write $ K^\eps_1  (v^\eps , w^\eps)$.
964: We also introduce some notations.
965: A Taylor expansion proves that there exist two $C^\infty$ functions $v^{0,\flat}$ and $w^{0,\flat}$
966:  such that $ v^0 = \rond{v}^0 + x_d \, v^{0,\flat} $ 
967: and $ w^0 = \rond{w}^0 +  x_d \,  w^{0,\flat} $, where $\rond{u}$ is the trace of $u$ on $x_d =0$. 
968: 
969: \begin{prop} 
970: \label{sev}
971: 
972: Assume that $W^{\eps}$ is of the form 
973: %
974: \begin{eqnarray} 
975: \label{qz}
976: W^{\eps} (t,x)  = \mathcal{W}^0 (t,x,\frac{x_d}{\eps}) + \eps W^{\eps,\flat} (t,x) .
977: \end{eqnarray}
978: %
979:  where $ \mathcal{W}^0 \in \P(T) $
980:  with  $ \un{\mathcal{W}}^0 = w^0$.
981: Then there is a  $C^\infty $ matrix  $K_1^{\flat} $ such that  
982: %
983: \begin{eqnarray} 
984:  \label{singul}   K_{1} ( u^\eps ) = 
985:  \eps K_{1}^{\flat} (\eps, x_d ,\rond{v}^0 , v^{0,\flat} , V^\eps , \eps V^\eps , \rond{w}^0 , w^{0,\flat} ,  \tilde{\mathcal{W}}^0
986:  ,\frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\eps,\flat} , \eps W^{\eps,\flat})
987:  \end{eqnarray}
988: %
989:  where $K_{1}^{\flat}$ is affine with respect to its fifth argument $V^\eps$ and affine with respect to its eleventh argument
990:  $W^{\eps,\flat}$ with $\X_{\mathtt{v}^0} \,  v^0$ as factor in the leading coefficient.  In  $(\ref{singul})$, to avoid heavy
991:  notations we denote
992:  $\tilde{\mathcal{W}}^0$ instead of $\tilde{\mathcal{W}}^0  (t,x,\frac{x_d}{\eps})$.
993: 
994: \end{prop}
995: %
996: \begin{proof} 
997: %
998: %
999: We begin giving a technical lemma. We will use it twice.
1000: 
1001: \begin{lem}
1002: \label{tech}
1003: 
1004: There exist some $d+1$  square matrices $K^\flat_{1,1} (u_1 , u_2 )$  and some  $K^\flat_{1,2} (u_1
1005: , u_2 ) \in \R^{d+1}$ both $C^\infty$ with respect to  to their arguments
1006: $u_1 := (v_1 , w_1 )$,  $u_2 := (v_2 , w_2 ) \in \R^{d+1} \times \R   $ such that $K^\flat_{1,2}$ 
1007: has  $\X_{\mathtt{v}_1} v^0$ as factor and that 
1008: %
1009: \begin{eqnarray} 
1010:  \label{zou} \quad K_1 (v_1 + v_2 , w_1 + w_2 ) = K_1 (u_1) +  K_{1,1}^{\flat} (u_1 , u_2 ) .v_2
1011:  + w_2 \,  K_{1,2}^{\flat} ( u_1 , u_2 ).
1012:  \end{eqnarray}
1013: %
1014: \end{lem}
1015: 
1016: \begin{proof} 
1017: %
1018: We will proceed in two steps.  
1019: 
1020:  \begin{enumerate}[$1$.]
1021: 
1022: \item  A Taylor first order expansion yields the existence of matrices $S_{1}^{\star,\flat} (u_1 , u_2 )$, 
1023: $S_{2}^{\star,\flat} (u_1 , u_2)$
1024:  such that for all $(u_1 ,u_2)$, 
1025: %
1026: \begin{eqnarray}
1027: \label{tuok} \quad  S^{\star} (v_1 +v_2 , w_1 + w_2) = S^{\star} (u_1 ) + v_2 . S_{1}^{\star,\flat} (u_1 , u_2) 
1028: +  w_2 . S_{2}^{\star,\flat} (u_1 ,u_2)  .
1029:  \end{eqnarray}
1030: %
1031: Here $S_{1}^{\star,\flat} (u_1 , u_2)$ is a $(d+1)^2 \times (d+1)$ matrix and 
1032: $S_{2}^{\star,\flat} ( u_1 , u_2)$  is a $d+1 $ square matrix.
1033: We write in a block form the matrix 
1034: %
1035: \begin{eqnarray*}
1036: S_{1}^{\star,\flat} := 
1037: \begin{bmatrix}
1038: S_{1,1}^{\star,\flat}
1039: \\ .
1040: \\ .
1041: \\ .
1042: \\ S_{1,d+1}^{\star,\flat}
1043: \end{bmatrix} ,
1044:  \end{eqnarray*}
1045: %
1046: where the $(S_{1,j}^{\star,\flat} )_{j=1,...,d+1}$ are some $d+1$ square matrices. 
1047: In Eq. $(\ref{tuok})$, 
1048: the notation $ v. S_{1}^{\star, \flat}$ stands for the $d+1$ square matrix
1049:  $v . S_{1}^{\star,\flat} := v_1 S_{1,1}^{\star,\flat}  + ... +  v_{d+1} S_{1,d+1}^{\star,\flat} $.
1050:   
1051:   \item We introduce, for all $(u_1 , u_2)$, the $d+1$ by $d+1$ matrix $K_{1,1}^{\flat} ( u_1 ,u_2) $ 
1052:  such that for all $z=(\mathtt{z},z_{d+1}) \in \R^{d} \times \R$,
1053: %
1054: \begin{eqnarray*}
1055:  K_{1,1}^{\flat} (u_1 , u_2).z &:=& z. S_{1}^{\star,\flat} ( u_1 , u_2) . \X_{ \mathtt{v}_1 + \mathtt{v}_2   }  \, v^0 
1056: + ( S^{\star} ( u_1 ) + w_2 . S_{2}^{\star,\flat} ). \mathtt{z}. \nabla v^0 ,
1057:  \end{eqnarray*}
1058: %
1059: and the vector
1060: %
1061: \begin{eqnarray*}
1062: K_{1,2}^{\flat} ( u_1 ,u_2) :=   S_{2}^{\star,\flat} (u_1 , u_2) . \X_{\mathtt{v}_1 } \,  v^0 \in \R^{d+1} .
1063: \end{eqnarray*}
1064: %
1065:  As by definition,  for all $(v ,w ) \in \R^d \times \R$, 
1066:  %
1067:  \begin{eqnarray*}
1068:   K_1 (v ,w)  :=  S^\star (v ,w ) \X_{\mathtt{v}} \,  v^0 + \L^\star (\partial_x ) v^0 , 
1069:   \end{eqnarray*}
1070: %
1071:    we get for all $(u_1 , u_2)$, 
1072:  %
1073:  \begin{eqnarray*}
1074:   K_1 (v_1 + v_2 , w_1 + w_2) - K_1 (v_1  , w_1 ) =  
1075:   S^\star (v_1 + v_2 ,w_1 + w_2  ) \X_{\mathtt{v}_1 + \mathtt{v}_2}  \, v^0 
1076:   \\ - S^\star (v_1 ,w_1) \X_{\mathtt{v}_1}  \, v^0 .
1077:   \end{eqnarray*}
1078: %
1079:  Thanks to Eq. $(\ref{tuok})$ and because $\X_{\mathtt{v}_1 + \mathtt{v}_2   } v^0 = \X_{\mathtt{v}_1 } v^0 + \mathtt{v}_2 .\nabla
1080:  v^0 $,  we get $(\ref{zou})$.
1081: %
1082: \end{enumerate}
1083: \end{proof}
1084: 
1085: This technical lemma given, we will proceed in three steps. 
1086: 
1087: \begin{step}[First order expansion]
1088: 
1089: We are going to prove that there exists a $C^\infty $ function $ K_1^{\flat,1}$, affine
1090: with respect to its third variable and affine
1091: with respect to its sixth variable with $\X_{\mathtt{v}^0} \,  v^0$ as factor in the leading coefficient such that 
1092:  %
1093: \begin{eqnarray} 
1094:  \label{zoub}
1095:   K_1 ( u^\eps ) =  K_1 (v^0  , \mathcal{W}^0 )
1096:    + \eps  K_1^{\flat,1} (\eps , v^0 , V^\eps , \eps V^\eps , \mathcal{W}^0 ,  W^{\eps,\flat} ,\eps W^{\eps,\flat} ).
1097:  \end{eqnarray}
1098: %
1099: \end{step}
1100: 
1101: We use a first time Lemma $\ref{tech}$ and apply $(\ref{zou})$ to 
1102: %
1103: \begin{eqnarray*}
1104: (v_1 , w_1 ,v_2 , w_2)  =(v^0 ,\mathcal{W}^0 ,\eps V^\eps ,\eps W^{\eps,\flat}), 
1105:  \end{eqnarray*}
1106: %
1107: 
1108: we get $(\ref{zoub})$ with 
1109: %
1110: \begin{eqnarray*}
1111:  K_1^{\flat,1} (\eps , v^0 , V^\eps , \eps V^\eps , \mathcal{W}^0 ,  W^{\eps,\flat}, \eps W^{\eps,\flat} ) &:=&
1112:  V^\eps  K_{1,1}^{\flat}   (v^0  , \mathcal{W}^0 , \eps V^\eps , \eps W^{\eps,\flat})
1113: \\  &&+ W^{\eps,\flat}  K_{1,2}^{\flat}   (v^0  ,  \mathcal{W}^0 ,\eps V^\eps ,\eps W^{\eps,\flat}). 
1114:  \end{eqnarray*}
1115: %
1116: It is clear that the function $K_1^{\flat,1}$ is affine
1117: with respect to its third variable and affine
1118: with respect to its sixth variable with $\X_{\mathtt{v}^0} \,  v^0$ as factor in the leading coefficient.
1119: 
1120: \begin{step}[Do $x_d =0$]
1121: 
1122:   We are going to prove that there exists a $C^\infty $ function $K_1^{\flat,1}$, 
1123:  affine with respect to its fifth argument and affine with respect to its eleventh argument with $\X_{\mathtt{v}^0} \, 
1124:  v^0$ as factor in the leading coefficient, such that       
1125: %
1126: \begin{eqnarray} 
1127:  \label{sing}   K_{1} ( u^\eps ) &= &  K_{1} (u^0 )
1128:   + (  K_{1} (\rond{v}^0   ,\rond{w}^0 + \tilde{\mathcal{W}}^0) - K_{1} (\rond{v}^0  , \rond{w}^0 ))
1129:  \nonumber \\  & +& \eps   K_{1}^{\flat} (\eps, x_d ,\rond{v}^0 , v^{0,\flat} , V^\eps , \eps V^\eps , \rond{w}^0 , 
1130:  w^{0,\flat} ,  \tilde{\mathcal{W}}^0 , \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\eps,\flat} , \eps W^{\eps,\flat}) .
1131:  \end{eqnarray}
1132:  %
1133: \end{step} 
1134: 
1135:  Thus we have to do the analysis of  $K_1 (v^0 ,\mathcal{W}^0)$. 
1136: We write naively 
1137: %
1138: \begin{eqnarray}
1139: \label{N}
1140:   K_1 (v^0  , \mathcal{W}^0 ) =  K_1 (u^0  ) + (  K_1 (v^0  , \mathcal{W}^0) -  K_1 (u^0 )) .
1141:  \end{eqnarray}
1142: %
1143: and do an estimate for the error 
1144: by substituting 
1145: %
1146: \begin{eqnarray*}
1147:  K_1 (\rond{v}^0  , \rond{w}^0 + \tilde{\mathcal{W}}^0 ) -  K_1 (\rond{v}^0 ,\rond{w}^0 ) \quad
1148: \mathrm{to} \  K_1 (v^0  , \mathcal{W}^0) -  K_1 ( u^0 )  .
1149:  \end{eqnarray*}
1150: %
1151: %
1152: \begin{lem}
1153: 
1154: There exists a  $C^\infty $ function $ K_1^{\flat,2}$ such that 
1155: %
1156: \begin{eqnarray}
1157: \label{zoub2}
1158:   K_1 (v^0  , \mathcal{W}^0) -  K_1 (u^0  ) &=& K_1 (\rond{v}^0 , \rond{w}^0 + \tilde{\mathcal{W}}^0 ) 
1159:   - K_1 (\rond{v}^0  , \rond{w}^0 ) 
1160:   \\ \nonumber && + \eps K_1^{\flat,2} (x_d ,\rond{v}^0 , v^{0,\flat} , \rond{w}^0  , w^{0,\flat}  ,\tilde{\mathcal{W}}^0 ,
1161:   \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ).
1162:  \end{eqnarray}
1163:  %
1164:  \end{lem}
1165:  %
1166:  \begin{proof}
1167:   %
1168:  We will proceed in three steps.
1169:    \begin{enumerate}[$1$.]
1170:  
1171:  \item We use use  Lemma $\ref{tech}$ again and apply respectively $(\ref{zou})$ to 
1172:  %
1173: \begin{eqnarray*} 
1174: (v_1 , w_1 , v_2 , w_2) &=& ( \rond{v}^0 ,  \rond{w}^0 + \tilde{\mathcal{W}}^0  ,  x_d  v^{0,\flat} ,   x_d  w^{0,\flat} ) 
1175: \quad \mathrm{and} \  \mathrm{to} 
1176: \\   (v_1 , w_1 , v_2 , w_2)  &=& ( \rond{v}^0 ,  \rond{w}^0  ,  x_d  v^{0,\flat} ,   x_d  w^{0,\flat} )
1177:  \end{eqnarray*}
1178: %
1179:  we get 
1180: %
1181: \begin{eqnarray} 
1182: \label{mlp} \qquad \left\{
1183: \begin{array}{ccc}
1184:   K_1 (v^0  , \mathcal{W}^0) &=&  K_1 (\rond{v}^0  , \rond{w}^0 + \tilde{\mathcal{W}}^0) 
1185:   + x_d   K_{1}^{a} (x_d ,\rond{v}^0 , v^{0,\flat} ,  \rond{w}^0 ,  w^{0,\flat} ,  \tilde{\mathcal{W}}^0  ) ,
1186:   \\   K_1 (v^0  , w^0 )  &=&  K_1 (\rond{v}^0  , \rond{w}^0 ) 
1187:   + x_d   K_{1}^{b} (x_d ,\rond{v}^0 ,  v^{0,\flat} ,  \rond{w}^0 ,  w^{0,\flat}  ),
1188:   \end{array}
1189:   \right.
1190:  \end{eqnarray}
1191: %
1192: with
1193: %
1194: \begin{eqnarray*} 
1195:  K_{1}^{a} (x_d , \rond{v}^0 ,  v^{0,\flat} ,  \rond{w}^0 , w^{0,\flat}  ,  \tilde{\mathcal{W}}^0  )
1196:     &:=&  v^{0,\flat}  K_{1,1}^{\flat}   ( \rond{v}^0  ,  \rond{w}^0  + \tilde{\mathcal{W}}^0  , x_d v^{0,\flat},  x_d  w^{0,\flat} ) 
1197:  \\ &+&  w^{0,\flat}  K_{1,2}^{\flat}   (  \rond{v}^0 ,  \rond{w}^0 + \tilde{\mathcal{W}}^0 , x_d   v^{0,\flat} ,  x_d   w^{0,\flat} ),
1198: \\  K_{1}^{b} (x_d , \rond{v}^0 , v^{0,\flat}  ,   \rond{w}^0  , w^{0,\flat} )
1199:     &:=&   v^{0,\flat}  K_{1,1}^{\flat}  (\rond{v}^0 , \rond{w}^0 , x_d v^{0,\flat} , x_d  w^{0,\flat}) 
1200:  \\ &+& w^{0,\flat}  K_{1,2}^{\flat}   ( \rond{v}^0  , \rond{w}^0 , x_d  v^{0,\flat},  x_d  w^{0,\flat} ).
1201:  \end{eqnarray*}
1202:  %
1203:  and we denote by $ K_1^{\flat,r,2} :=  K_{1}^{a} -  K_{1}^{b}$.
1204:  
1205:   \item By a first order Taylor expansion, we obtain that there exist some $(d+1) \times (d+1)$ matrices $G_{1,1}^{\flat}$
1206:   and some $G_{1,2}^{\flat} \in \R^{d+1}$, $C^\infty$ with respect to their arguments
1207:    %
1208:    \begin{eqnarray*} 
1209:   u_1 :=  (v_1 , w_1 ) \ ,  u_2 := (v_2 , w_2 ) \in  \R^{d+1} \times \R \ ,  w_3 \in \R  
1210:   \end{eqnarray*}
1211:  %
1212:  
1213: such that
1214:  
1215:  %
1216: \begin{eqnarray*}
1217: K_{1,1}^{\flat} (v_1 ,w_1 + w_3 , u_2 ) - K_{1,1}^{\flat} (u_1 , u_2 ) = 
1218:  w_3 . G_{1,1}^{\flat} (u_1 ,u_2 , w_3),
1219: \\ K_{1,2}^{\flat} (v_1 ,w_1 + w_3 , u_2 ) - K_{1,2}^{\flat} (u_1 , u_2 ) = 
1220:  w_3 .G_{1,2}^{\flat}  (u_1 ,u_2 , w_3).
1221: \end{eqnarray*}
1222:  
1223:   \item Thanks the two previous points, we obtain $(\ref{zoub2})$ with
1224:  %
1225:  \begin{eqnarray*}
1226:  K_1^{\flat,2} (x_d ,\rond{v}^0 , v^{0,\flat} , \rond{w}^0  , w^{0,\flat}  ,\tilde{\mathcal{W}}^0 ,
1227:   \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ) = 
1228: \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 \{  v^{0,\flat} . G_{1,1}^{\flat} (\rond{v}^0 ,\rond{w}^0 , x_d v^{0,\flat} ,x_d w^{0,\flat} ,\tilde{\mathcal{W}}^0) 
1229: \\ + w^{0,\flat}   G_{1,2}^{\flat}   (\rond{v}^0  ,\rond{w}^0 , x_d  v^{0,\flat}  ,  x_d  w^{0,\flat}  , \tilde{\mathcal{W}}^0 )\} .
1230:  \end{eqnarray*}
1231: %
1232: Because $G_{1,1}^{\flat}$ and $G_{1,2}^{\flat}$ are $C^\infty$, the function $ K_1^{\flat,2}$ is also $C^\infty$.
1233: 
1234: 
1235: \end{enumerate}
1236:  %
1237:  %
1238: \end{proof}
1239: 
1240:  We denote by 
1241: %
1242: \begin{eqnarray*}
1243:  K_1^{\flat}   (\eps, x_d ,\rond{v}^0 , v^{0,\flat} , V^\eps , \eps V^\eps , \rond{w}^0 , w^{0,\flat}   , \tilde{\mathcal{W}}^0 
1244:  ,\frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\eps,\flat} ,\eps W^{\eps,\flat} ) := 
1245:  \\  K_1^{\flat,2} (x_d ,\rond{v}^0 , v^{0,\flat}  , \rond{w}^0   , w^{0,\flat}   , \tilde{\mathcal{W}}^0 , \frac{x_d}{\eps} \tilde{\mathcal{W}}^0  ) 
1246:  +  K_1^{\flat,1} (\eps , v^0 , V^\eps , \eps V^\eps , \mathcal{W}^0 ,  W^{\eps,\flat} ,\eps W^{\eps,\flat} )  
1247:   \end{eqnarray*}
1248:   %
1249:  Because  the first term does not involve $V^\eps$ and $W^{\eps,\flat}$ and
1250:   the function $ K_1^{\flat,1}$ is affine with respect to its third variable $V^\eps$ and affine
1251: with respect to its sixth variable $W^{\eps,\flat}$ with $\X_{\mathtt{v}^0} \,  v^0$ as factor in the leading coefficient, 
1252:  the function  $ K_1^{\flat} $ is  affine with respect to its fifth variable $V^\eps$ and affine with respect to its eleventh argument 
1253:  $W^{\eps,\flat}$  with
1254:  $\X_{\mathtt{v}^0} \,  v^0$ as factor in the leading coefficient.
1255:  Note that  as the profile $ \tilde{W}_0$ is rapidly decreasing in $X$, 
1256:  the  profile $X \tilde{W}_0 $  is also rapidly decreasing. 
1257:  Combine $(\ref{zoub})$, $(\ref{N})$ and $(\ref{zoub2})$ to find  $(\ref{sing})$.
1258: 
1259: \begin{step}[Use the ground state properties]
1260: 
1261: We are going to prove that the terms  $ K_1 (u^0  )$ and  $K_1 (\rond{v}^0  , \rond{w}^0 + \tilde{\mathcal{W}}^0 ) -
1262: K_1 (\rond{u}^0  )$ are equal to $0$.
1263: 
1264: \end{step}
1265: 
1266: First because the ground state $u^0$ satisfies $(\ref{aller})$,  
1267: %
1268: \begin{eqnarray*}
1269:  K_1 (u^0  )  =   S^\star  (u^0  )  (\X_{\mathtt{v}^0} \,    + \M^\star (u^0 ,\partial_x ))   v^0 =0 .
1270: \end{eqnarray*}
1271: %
1272: Moreover, referring to  $(\ref{K1})$ and thanks to $(\ref{cond})$, we obtain:
1273: %
1274: \begin{eqnarray*}
1275:  K_1 (\rond{v}^0 , \rond{w}^0 + \tilde{\mathcal{W}}^0) - K_1 (\rond{u}^0 )
1276: = (  S^\star   (v^0 , \rond{w}^0  + \tilde{\mathcal{W}}^0) -   S^\star ( v^0 , \rond{w}^0 ) ) \X_{ \rond{\mathtt{v}}^{0}}  \rond{v}^{0} 
1277: = 0  .
1278: \end{eqnarray*}
1279: %
1280: \end{proof} 
1281: 
1282: 
1283: \section{WKB expansions}
1284: \label{secform}
1285: 
1286: \subsection{Formal solutions}
1287: 
1288: We look for formal solutions  $(u^\eps )_\eps $ of $(\ref{aller})$ of the  form $ u^\eps =(v^0 + \eps V^\eps , W^\eps )$ 
1289: where  $U^\eps := (V^\eps , W^\eps )$ is an expansion  
1290: %
1291: \begin{eqnarray}
1292: \label{pk}
1293: \sum_{n \geqslant 0} \eps^n \, \mathcal{U}^n (t,x, \frac{x_d}{\eps})
1294: \end{eqnarray}
1295: %
1296:  where each  $ \mathcal{U}^n $ belongs to  $\P (T) $. 
1297:  We rewrite  $ \mathcal{U}^n $ as  $ \mathcal{U}^n = ( \mathcal{V}^n ,  \mathcal{W}^n )$ 
1298:  and we suppose  $  \un{\mathcal{W}}^0 := w^0$. 
1299:  Let us explain what is meant by \textit{formal} \textit{solutions}.
1300:  Plugging the expansion $(\ref{pk})$ into the system, using Taylor expansions and ordering the terms in powers of $\eps$, 
1301:   we get a formal expansion in power series of $\eps$:
1302: %
1303: \begin{eqnarray}
1304: \label{resu}
1305: \sum_{n \geqslant -1} \eps^n \, \Phi^n (t,x, \frac{x_d}{\eps})
1306: \end{eqnarray}
1307: %
1308:  where the  $(\Phi^n )_{n \geqslant -1}$ are in  $\P (T) $.
1309:  We say that $(u^\eps )_\eps $ is a formal solution when all the resulting $\Phi^n $ are identically zero.
1310:  The following theorem states that the system has formal solutions and that we can prescribe arbitrary initial values to the $(P_0
1311:  \, \mathcal{U}^j |_{t=0} )_{j \in \N } $. 
1312:   We introduce    $\Nrond_{init}  :=  H^\infty  (\R^d_+  , \mathcal{S}(\R_+))$ 
1313: and the profile space  
1314: %
1315: \begin{eqnarray*}
1316:  \P_{init} := \{ \mathcal{U} \in L^2 (\R^d_+ \times \R_+ ) \ \mathrm{such} \  \mathrm{that} \
1317:  \mathrm{there} \  \mathrm{exist} \ \un{ \mathcal{U}} \in H^\infty (\R^d_+ ), \ \tilde{ \mathcal{U}} \in  \Nrond_{init} 
1318: \\ \mathrm{such} \  \mathrm{that} \ \mathrm{for} \  \mathrm{all} \
1319:   (x,X) \in \R^d_+ \times \R_+ ,\quad  \mathcal{U}  (x,X) = \un{ \mathcal{U}} (x) + \tilde{ \mathcal{U}} (x,X)  \}.
1320:  \end{eqnarray*}
1321: %
1322:  
1323: \begin{theo}
1324: \label{try}
1325: 
1326: Assume that some profiles $(\mathcal{U}^j_{init} )_{j \in \N }$ such that $ (Id - P_0) \mathcal{U}^j_{init} =0$  and
1327: $\mathcal{W}^0_{init} = w^0 |_{t=0}$ are given, 
1328: then there exists a formal solution $(u^\eps) $  of $(\ref{aller})$ on $(0,T_0 )$ with some profiles  $(\mathcal{U}^j )_{j \in \N }$
1329: in $\P (T_0 )$ such that $\un{\mathcal{W}}^0 := w^0 $ and that for all $j \in \N$, $P_0 \, \mathcal{U}^j |_{t=0}= \mathcal{U}^j_{init} $.
1330: Moreover the profile $\mathcal{U}^0$ is polarized in the sense
1331:  that $(Id - P_0) \tilde{\mathcal{U}} =0$ for all $(t,x) \in (0,T_0 ) \times \Omega$.
1332: 
1333: \end{theo}
1334: 
1335: Notice that Theorem $\ref{try}$ gives the existence of a formal solution till $T_0$.
1336: At first sight, this may seems strange because Theorem $\ref{try}$ deals with large boundary layers and in some similar settings
1337: large boundary layers have in general a nonlinear behavior (see for example paper \cite{MZ} of G.M\'etivier and K.Zumbrun about large
1338: viscous boundary layers for noncharacteristic nonlinear hyperbolic problems and paper \cite{Gr1} of E.Grenier about large velocity
1339: boundary layers for the Euler equations). 
1340: It is also different from the result on entropy waves of C.Cheverry, O.Gu\`es and G.M\'etivier  \cite{cgm2}. 
1341: In \cite{cgm2}, Theorem $3.4$ gives the local existence of formal solutions with large amplitude oscillations on the entropy.
1342: More precisely Theorem $3.4$ of \cite{cgm2} gives the existence of the profile $\mathcal{U}^0$ for small time and we see on the
1343: system $(3.36)$ of \cite{cgm2} that this is not because of a lack of analysis but because of an actual nonlinear effect.
1344: At the opposite in the proof of Theorem $\ref{try}$ we will construct a formal solution thanks to linear profile problems.
1345: This shows that a system can reveals additional transparency properties when looking at boundary layers instead of high frequency
1346: oscillations.
1347: The transparency property of \cite{moi3} supports this 
1348: remark.\footnote{This phenomenon can also be observed when looking at boundary layers of "strong amplitude", following the terminology of
1349: \cite{cgm1}.
1350: Instead of nonlinear modulation equation as in \cite{cgm2}, we obtain linear profile equation.}
1351: 
1352: Since the profile $\mathcal{U}^0$ is polarized, we get $\tilde{\mathtt{V}}_\mathfrak{n}^0 = \tilde{\mathtt{P}}^0 = 0$ (see section
1353: $\ref{setting}$ for the definition of $P_0$) as we have
1354: mentioned it in the introduction. 
1355: This means that the amplitude of normal velocity and pressure boundary layers is weaker than
1356: the amplitude of tangential boundary layer. 
1357: The fact that the profile  $\tilde{\mathtt{V}}_\mathfrak{n}^0 $ vanishes is a consequence of the conditions $(\ref{cond})$ and more
1358: precisely of the fact that $\X_{\mathtt{v}^0} \,  \mathtt{p}^0 =0$
1359: for $(t,x) \in (0,T_0) \times \Gamma$. Without this last condition, it is still possible to construct some nontrivial formal
1360: solutions of the form $(\ref{pk})$ but with a nonvanishing profile $\tilde{\mathtt{V}}_\mathfrak{n}^0 $ (cf. section $\ref{coming}$). 
1361: However such formal solutions can be unstable. 
1362: We do not consider such formal solutions in this paper.
1363: \\
1364: \\ Next subsection is devoted to the proof of Theorem $\ref{try}$.
1365: We outline here some key tools of the proof.
1366: First we use the reduction of the system of section $3$.
1367: We use in a essential way Proposition $\ref{sev}$.
1368: For example because $K_1$ contains a factor $\eps$, the term $K^\eps$ is not singular with respect to $\eps$.
1369: This is crucial because we look for formal solutions without boundary layers for $v^\eps$ but only with large boundary layers for
1370: $V^\eps$.
1371: In order to find the expansion  $(\ref{resu})$, we use several Taylor expansions with respect to $\eps$ and to $x_d$. 
1372: The underlying idea is to obtain an expansion $(\ref{resu})$ inside which the profiles  $(\mathcal{U}^j )_{j \in \N }$ appear at the
1373: highest possible order. 
1374: We notably use the fact that $ \mathtt{v}_d^0$ and $\X_{\mathtt{v}^0} \, v^0$  vanish on the boundary and the fact that $K_1^\flat$ is affine with respect
1375: to its eleventh argument with $\X_{\mathtt{v}^0} \, 
1376:  v^0$ as factor in the leading coefficient (cf. Proposition $\ref{sev}$ and  Lemma  $\ref{psg}$).     
1377: In order to find some  profiles  $(\mathcal{U}^j )_{j \in \N }$ for which the profiles  $(\Phi^j )_{j \in \N }$ 
1378: of the resulting expansion  $(\ref{resu})$ are identically zero, 
1379: we follow the strategy of \cite{Gu95}, \cite{moi}, \cite{moi2}, \cite{moi3} and exhibit a
1380: sequence of profile problem. 
1381: Theorem $\ref{casca}$ gives the existence of profiles $(\mathcal{U}^j )_{j \in \N }$ solving this sequence of problem.
1382: Theorem $\ref{try}$ will be proved as a consequence of theorem $\ref{casca}$.
1383: One key point in the proof of Theorem $\ref{casca}$ is that profile problems are linear, thanks to transparency properties.
1384: One manifestation of this property is that $K_1^\flat$ is affine with respect
1385: to its fifth argument. The profiles problems are linearized Euler equations for the regular part of the profiles and linear totally
1386: characteristic transport equations for the boundary layer part of the profiles.
1387: 
1388: 
1389: \subsection{Proof of Theorem $\ref{try}$}
1390: 
1391: We will proceed in two steps. 
1392: First we explicit the expansion $(\ref{resu})$.
1393:  Then we look for profiles  $(\mathcal{U}^j )_{j \in \N }$ for which the profiles  $(\Phi^j )_{j \in \N }$ 
1394: of the resulting expansion  $(\ref{resu})$ are identically zero solving a sequence of profile problems.
1395: 
1396: \subsubsection{Expliciting the expansion $(\ref{resu})$} 
1397: 
1398: We introduce the operator $\H^0$ which maps a function $U \in H^\infty ((0,T) \times \R^d_+ )$ to the function 
1399: $\H^0 (U, W^{\flat} ,  \tilde{\mathcal{W}}^0 ,  \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , \partial ) U \in H^\infty ((0,T) \times
1400: \R^d_+ )$ where 
1401:  %
1402:  \begin{eqnarray*}
1403:   \H^0 (U, W^{\flat} ,  \tilde{\mathcal{W}}^0 ,  \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , \partial ) U:= 
1404:   \mathfrak{L} (v^0 ,W, \partial ) U
1405:  +  \begin{bmatrix}
1406:  K^0 (V, \tilde{\mathcal{W}}^0 , \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ,    W^{\flat} )  
1407:   \\ 0
1408:  \end{bmatrix}
1409:   \end{eqnarray*}
1410: %
1411:   with 
1412:  %
1413:  \begin{eqnarray*}
1414:  K^0 (V, \tilde{\mathcal{W}}_0 , \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\flat} ) := 
1415:  K_{1}^{\flat} ( 0, x_d ,\rond{v}^0 , v^{0,\flat} , V , 0 , \rond{w}^0 , \rond{w}^{0,\flat}  ,  \tilde{\mathcal{W}}^0 , \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ,  
1416:  W^{\flat} , 0) .
1417:  \end{eqnarray*}
1418: %
1419:  Remember that the term $ K_{1}^{\flat}  $ is defined in Proposition $\ref{sev}$.
1420:  As in $(\ref{singul})$, we denote
1421:  $\tilde{\mathcal{W}}^0$ instead of $\tilde{\mathcal{W}}^0  (t,x,\frac{x_d}{\eps})$.
1422:  Thanks to Prop $\ref{sev}$, the function $K^0 (V, \tilde{W} , X \tilde{W} ,W)$ is affine with respect
1423:   to its first argument $V$ and affine with
1424:  respect to its fourth argument $W$ with $\X_{\mathtt{v}^0} \,   v^0$ as factor in the leading coefficient. 
1425: 
1426: \begin{prop}
1427: 
1428: \begin{enumerate}[$(i)$]
1429: 
1430: \item We obtain 
1431: %
1432:  \begin{eqnarray*}
1433: \un{\Phi}^{-1} =0,  \ \un{\Phi}^{0} = \H^0  (\un{  \mathcal{U}}^0 ,0,0 , \un{\mathcal{W}}^1 ,\partial )  \un{ \mathcal{U}^0 }, 
1434:   \end{eqnarray*}
1435: %
1436:  and for $j \geqslant 1$,
1437: %
1438:  \begin{eqnarray*}
1439: \un{\Phi}^{j} =  \H^0 (\un{  \mathcal{U}}^0 , 0,0,\un{ \mathcal{W}}^1 , \partial ) \un{  \mathcal{U}}^j -  \un{\mathcal{Q}}^j .
1440: \end{eqnarray*}
1441: %
1442: The terms $\un{\mathcal{Q}}^j$ depend on  $(t,x)$, on the profiles $( \un{\mathcal{U}}^k )_{k  <j}$ and on their derivatives and affinely
1443: on the profiles  $\un{\mathcal{U}}^{j}$ and $\un{\mathcal{W}}^{j+1}$. 
1444: In particular, $ \un{\mathcal{Q}}^1_w = -  \un{\mathtt{V}}^0 . \nabla \un{\mathcal{W}}^0 $.
1445: 
1446: 
1447: \item We obtain 
1448:  %                                                                                                                                                                                                                                                                                                                        
1449: \begin{eqnarray*}
1450:  \tilde{\Phi}^{-1} &=& \L_d \, \partial_X \tilde{\mathcal{U}}^0
1451: \\  \tilde{\Phi}^{0} &=& \L_d \, \partial_X \tilde{\mathcal{U}}^1 
1452:  + \H' (\mathcal{U}^0  , \partial ,  \partial_X ) \tilde{ \mathcal{U}}^0 ,
1453: \\  \tilde{\Phi}^{j} &=& \L_d \, \partial_X \tilde{\mathcal{U}}^{j+1} 
1454: +  \H' ( \mathcal{U}^0  ,  \partial ,  \partial_X ) \tilde{ \mathcal{U}}^j  
1455:  -  \tilde{ \mathcal{Q}}^j \quad \mathrm{for} \ j \geqslant 1
1456: \end{eqnarray*}
1457:  %
1458: where
1459: 
1460: \begin{enumerate}[$\bullet$]
1461: 
1462: \item  the operator $ \H'  $ 
1463:  maps a function $\mathcal{U} \in  \P(T) $ to
1464:    $ \H' (\mathcal{U} ,\partial , \partial_X ) \tilde{\mathcal{U}} \in  \mathcal{N}(T) $ where
1465:   %
1466: \begin{eqnarray*}
1467:  \H' (\mathcal{U} ,\partial , \partial_X ) \tilde{\mathcal{U}} :=
1468:  (X  \mathtt{v}^{0,\flat}_d +  \rond{\underline{\mathtt{V}}}_d +  \tilde{\mathtt{V}}_d ) S (v^0 ,\mathcal{W} )  \partial_X  \tilde{\mathcal{U}} 
1469:  + \mathfrak{L} (v^0 , \mathcal{W} ,\partial)  \tilde{\mathcal{U} }
1470: \\  + 
1471: (S (v^0 ,\mathcal{W} ) - S (v^0 , \un{ \mathcal{W} })) \X_{\mathtt{v}^0}  \, \un{ \mathcal{U}}
1472: + 
1473: \begin{bmatrix}
1474:   K^0_I (  \mathcal{V} , \tilde{\mathcal{W}} , X \tilde{\mathcal{W}} )
1475: \\ 0
1476: \end{bmatrix} ,
1477:  \end{eqnarray*}
1478:  %
1479: where $K^0_I ( V , W , X W )$ is a $C^\infty$ function affine with respect to
1480: its first argument $V$ such that   $K^0_I (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 ) \in 
1481: \mathcal{N}(T) $.
1482:  
1483: 
1484: \item the terms $ \tilde{ \mathcal{Q}}^j $  depend of  $(t,x)$, of the profiles $( \mathcal{U}^k )_{k  <j}$, 
1485: of their traces  and of their derivatives and
1486: affinely of the profiles $\mathcal{U}^j $ and   $ \un{\mathcal{W}}^{j+1} $.
1487: 
1488: \end{enumerate}
1489: 
1490: \end{enumerate}
1491: \end{prop}
1492:  %
1493: \begin{proof}[Proof of $(i)$]
1494:  %
1495:  %
1496: We introduce the operator $\H^\eps$ by the formula
1497: %
1498: \begin{eqnarray*}
1499:  \H^\eps (U,\partial ) U:=  \mathfrak{L}( v^0 + \eps V ,W,\partial)   U + K^\eps .
1500: \end{eqnarray*}
1501: %
1502: Remember that the term $K^\eps$ is defined in Proposition $\ref{Re1}$.
1503: Assume that  $W $ is of the form $ W^{\eps} = \mathcal{W}^0 +\eps W^{\flat} $ where $ \mathcal{W}^0 \in \P $ 
1504: with  $\un{\mathcal{W}}^0 = w^0$.
1505: By a first order Taylor expansion in respect to $\eps$, using Proposition $\ref{sev}$, we see that  $\H^\eps $ can be put under the form:
1506: %
1507: \begin{eqnarray*}
1508:  \H^\eps (U,\partial ) U = \H^0 (U, W^{\flat} , \tilde{\mathcal{W}}^0 ,  \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ,  \partial ) U
1509:  + \sum_{j=1}^d B_{j}^{\eps} (v^0 , U) \eps \partial_j U + \eps \mathbf{M}^\eps (t,x,U ). 
1510:  \end{eqnarray*}
1511: %
1512:  where the $B_{j}^{\eps}$ and $\mathbf{M}^\eps$ are $(d+2) \times (d+2) $ matrices depending in a $C^\infty$ way on their arguments
1513: including $\eps$ up to $\eps=0$. They are  of the form 
1514: %
1515:  \begin{eqnarray*}
1516:  B_{j}^{\eps} := 
1517: \begin{bmatrix}
1518:  B_{j}^{\star,\eps} & 0
1519: \\ 0 & \mathtt{V}_j
1520:  \end{bmatrix}
1521:  , \quad
1522:  \mathbf{M}^\eps :=
1523:  \begin{bmatrix}
1524:   \mathbf{M}^{\star,\eps} & 0
1525: \\ 0 & 0
1526:  \end{bmatrix}.
1527:  \end{eqnarray*}
1528: %
1529:  Then we easily complete the proof of $(i)$.
1530: % 
1531: \end{proof}
1532: %
1533: \begin{proof}[Proof of $(ii)$]
1534: %
1535:  We now look at the  $\tilde{\Phi}^{j}$.
1536: Because the terms $\tilde{\mathcal{U}}^{j} (t,x,\frac{x_d}{\eps})$ contain a factor $\eps^{-1}$ in their argument, the normal derivative $\partial_d$ plays a
1537: crucial role. 
1538: Therefore, we define  
1539: %
1540: \begin{eqnarray}
1541: \label{fd}
1542: \X'_\mathtt{v} :=\X_\mathtt{v} - \mathtt{v}_d \partial_d \  \mathrm{and} \  \L'   (u, \partial )= \L  ( \partial) + S(u) \X'_\mathtt{v} .
1543: \end{eqnarray}
1544: %
1545: Write $U := (V,W)$ to find that 
1546: %
1547: \begin{eqnarray}
1548:  \L'   (u, \partial ) U = 
1549:  \begin{bmatrix}
1550:   \L^{\star,'}   (u, \partial ) V 
1551: \\ \X'_\mathtt{v}  W
1552: \end{bmatrix}
1553: \quad \mathrm{where} \
1554:  \L^{\star,'}   (u, \partial ) := \L^\star  ( \partial) + S^\star (u) \X'_\mathtt{v} .
1555: \end{eqnarray}
1556: %
1557: Then
1558: %
1559: \begin{eqnarray*}
1560:  \H^\eps (U,\partial ) U  &=&  (\mathtt{v}^0_d + \eps \mathtt{V}_d ) S^\eps (v^0 + \eps V ,W)  \partial_d U 
1561:  + \L' (v^0 + \eps V ,W ,\partial) U   + K^\eps  ,
1562: \\ \H^0 (U, W^{\flat}, \tilde{\mathcal{W}}^0 ,  \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 ,  \partial ) U   &=&  
1563: \mathtt{v}^0_d \, \S (v^0 ,W)  \partial_d U + \L' (v^0 , W,\partial) U  
1564:  + K^0   (V, \tilde{\mathcal{W}}^0 ,  \frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\flat}). 
1565: \end{eqnarray*}
1566: %
1567: A key point when calculating the  $\tilde{\Phi}^{j}$ is that when a boundary layer profile  is in factor, 
1568: we can do formally $"x_d =0"$. The
1569: underlying idea is that for a $C^\infty$ scalar function $\Phi (x_d )$ and $\tilde{\mathcal{U}} \in  \mathcal{N}(T)$
1570: using a Taylor expansion yields the existence of a $C^\infty$ scalar function $\Phi^\flat (x_d )$
1571:  such that $\Phi (x_d ) = \Phi (0) + x_d \Phi^\flat (x_d ) $. Then
1572: %
1573: \begin{equation*}
1574: \Phi (x_d ) \tilde{\mathcal{U}} (t,x,\frac{x_d}{\eps}) = \Phi (0) \tilde{\mathcal{U}} (t,x,\frac{x_d}{\eps}) 
1575: + \eps \Phi^\flat (x_d ) (X \tilde{\mathcal{U}} (t,x,X)) |_{X=\frac{x_d}{\eps}} .
1576: \end{equation*}
1577: %
1578: Because $X  \tilde{\mathcal{U}} (t,x,X) $ still is in $\mathcal{N}(T)$, the second term can be put with the term of
1579: order $\eps$. 
1580: If $\Phi (0 )=0$, then you obtain that $\tilde{\mathcal{U}}$ does not appear in the resulting expansion at the order $0$.
1581: We are going to apply successively this idea to the terms  
1582: %
1583: \begin{eqnarray*}
1584:  \mathtt{v}^0_d \, S (v_0  ,W) \partial_d  \mathcal{U},  \ B_{d}^{\eps} (v_0 , \mathcal{U}) \eps \partial_d \mathcal{U} \ 
1585: \mathrm{and} \ K^0 (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,X \tilde{\mathcal{W}}^0 ,\mathcal{W}^1) . 
1586: \end{eqnarray*}
1587: %
1588: \begin{enumerate}[$\bullet$]
1589: 
1590: \item  Because  $\rond{\mathtt{v}}^0_{d} =0$, we have  $\mathtt{v}^0_d  = x_d  \mathtt{v}^{0,\flat}_d$.
1591: Applying $ \mathtt{v}^0_d \, S (v^0  ,W) \partial_d$ to $ \tilde{\mathcal{U}}(t,x,\frac{x_d}{\eps})$ where 
1592:   $\tilde{\mathcal{U}} \in \mathcal{N}(T) $ 
1593: yields the term 
1594: %
1595: \begin{eqnarray*}
1596: (\mathtt{v}^{0,\flat}_d \, S^0 (v^0  ,W) X \partial_X \tilde{\mathcal{U}}
1597:  + \mathtt{v}^{0} \,  S^0 (v^0  ,W) \partial_d \tilde{\mathcal{U}} )|_{X=\frac{x_d}{\eps}} .
1598:  \end{eqnarray*}
1599: %
1600: \item  We now look at the term  $B_{d}^{\eps} (v^0 ,\mathcal{U} )$.
1601: 
1602: \begin{lem}
1603: 
1604: There exist some matrices $B_{d}^{\eps,\flat}$ $C^\infty$ with respect to its arguments, including $\eps$ up to $0$, such that for
1605: all $ \mathcal{U} \in \P(T)$, 
1606: %
1607: \begin{eqnarray}
1608: \label{frg}
1609: B_{d}^{\eps} (v^0 ,\mathcal{U} )  = \mathtt{V}_d \, S (v^0 ,\mathcal{W} ) 
1610: + \eps B_{d}^{\eps,\flat} (x_d , X, \rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} ).
1611: \end{eqnarray}
1612: %
1613: Moreover, the matrices  $B_{d}^{\eps,\flat}$ are affine with respect to $X$.
1614: 
1615: \end{lem}
1616: %
1617: \begin{proof}
1618: %
1619: We will proceed in three steps.
1620: \\
1621: \begin{enumerate}[$1$.]
1622: 
1623: \item  First  by a Taylor expansion there exist some matrices $B_{d}^{\eps,\flat,1}$, $C^\infty$ with respect to its arguments, 
1624: including $\eps$ up to $0$, such that for all $\mathcal{U} \in \P(T)$, 
1625: %
1626: \begin{eqnarray*}
1627: B_{d}^{\eps} (v^0 , \mathcal{U}) = B_{d}^{0} (v^0 , \mathcal{U}) 
1628: + \eps B_{d}^{\eps,\flat} (v^0 , \mathcal{U}  ),
1629: \end{eqnarray*}
1630: %
1631:  where
1632: %
1633: \begin{eqnarray*}
1634:  B_{d}^{0} (v^0 ,  \mathcal{U}) :=  \mathtt{V}_d \, S (v^0 ,\mathcal{W} ) + \mathtt{v}^0_d \, DS (v^0 ,\mathcal{W} ).( \mathcal{V},0) .
1635: \end{eqnarray*}
1636: %
1637: \item  By a Taylor expansion, there exist some matrices $B_{d}^{0,\flat,1}$ and $B_{d}^{0,\flat,2}$,
1638:  $C^\infty$ with respect to their arguments 
1639:  %
1640:  \begin{eqnarray*}
1641:  (v_1 , \mathcal{U}_1 ,v_2 , \mathcal{U}_2 ) \in \R^{d+1} \times \R^{d+2} \times \R^{d+1} \times \R^{d+2} ,
1642:  \end{eqnarray*}
1643: %
1644:  such that for all $(v_1 , \mathcal{U}_1 ,v_2 ,\mathcal{U}_2 )$, 
1645:  %
1646:  \begin{eqnarray}
1647:  \label{tyh}
1648: \quad  B_{d}^{0} (v_1 + v_2 , \mathcal{U}_1 + \mathcal{U}_2  ) =
1649:  B_{d}^{0} ( v_1 , \mathcal{U}_1  ) + v_2 . B_{d}^{0,\flat,1} (v_1 ,\mathcal{U}_1 ,v_2 ,\mathcal{U}_2 )
1650: \\  \nonumber + U_2 . B_{d}^{0,\flat,2} (v_1 ,\mathcal{U}_1 ,v_2 ,\mathcal{U}_2 ) .
1651:  \end{eqnarray}
1652: %
1653:  We write  $ \mathtt{v}^0 = \rond{\mathtt{v}}^0 + x_d \, \mathtt{v}^{0,\flat}$ and for all  $\mathcal{U} \in \P (T)$, 
1654:  $\un{\mathcal{U}} = \rond{\un{\mathcal{U}}} + x_d \, \un{\mathcal{U}}^\flat $.
1655:  We apply  $(\ref{tyh})$ with 
1656:  %
1657:  \begin{eqnarray*}
1658:  (v_1 ,\mathcal{U}_1 ,v_2 ,\mathcal{U}_2  )= (\rond{v}^0 , \rond{\un{\mathcal{U}}}+ \tilde{\mathcal{U}} , x_d \, v^{0,\flat} ,  x_d \, \un{\mathcal{U}}^\flat ) ,
1659:  \end{eqnarray*}
1660: %
1661:    and obtain for all $\mathcal{U} \in \P (T)$,
1662: %
1663: \begin{eqnarray*}
1664:  B_d^0 (v^0 , \mathcal{U}) = B_d^0 (\rond{v}^0 ,\rond{\un{\mathcal{U}}} + \tilde{\mathcal{U}}) 
1665:  +  x_d \, B_{d}^{0,\flat,2} (x_d ,\rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} ),
1666: \end{eqnarray*}
1667: %
1668: with 
1669: %
1670: \begin{eqnarray*}
1671:  B_{d}^{0, \flat,2} (x_d ,\rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{U}^\flat ,\tilde{\mathcal{U}} )  &:= &
1672:  v^{0,\flat}  . B_{d,1}^{0,\flat} (\rond{v}^0 ,\rond{\un{\mathcal{U}}} + \tilde{\mathcal{U}},  x_d \, v^{0,\flat} ,  x_d \, \un{U}^\flat )
1673: \\ && + \un{U}^\flat . B_{d,2}^{0,\flat} (\rond{v}^0 ,\rond{\un{\mathcal{U}}} + \tilde{\mathcal{U}},  x_d \, v^{0,\flat} ,  x_d \, \un{U}^\flat ) .
1674:  \end{eqnarray*}
1675: %
1676: \item   Moreover, for all $\mathcal{U} \in \P  (T) $, as $ \rond{\mathtt{v}}^0_d  =0$, 
1677:  we have $ B_d^0 (\rond{v}^0 ,\mathcal{U}) =  \mathtt{V}_d S^0 (v^0 ,W)$.
1678: As a consequence, we get $(\ref{frg})$ with 
1679: %
1680: \begin{eqnarray}
1681: \label{NN}
1682:  B_{d}^{\eps,\flat} (x_d , X, \rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{U}^\flat ,\tilde{\mathcal{U}} ):= B_{d}^{\eps,\flat} (v^0 , \mathcal{U}  )
1683:  + X  B_{d}^{0, \flat,2} (x_d ,\rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} ) 
1684: \ \end{eqnarray}
1685: %
1686: which is affine with respect to $X$.
1687: 
1688: \end{enumerate}
1689: \end{proof}
1690: 
1691: Write $\mathtt{V}_d = \rond{\underline{\mathtt{V}}}_d + x_d \mathtt{\underline{V}}_d^{\flat}$ and define
1692: %
1693: \begin{eqnarray*}
1694:  B_{d}^{\eps,\flat,2} (x_d , X, \rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} )
1695: =  B_{d}^{\eps,\flat} (x_d , X, \rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} )
1696: +  \underline{\mathtt{V}}_d^{\flat} S (v^0 , \mathcal{W}) X  
1697: \end{eqnarray*}
1698: %
1699: to see that by applying $B_{d}^{\eps} (v^0 ,\mathcal{U}) \eps \partial_d$ to $\tilde{\mathcal{U}}(t,x,\frac{x_d}{\eps})$ where 
1700:   $\tilde{\mathcal{U}} \in \mathcal{N}(T) $,
1701:  we get the term 
1702: %
1703: \begin{eqnarray*}
1704: ( (\rond{\underline{\mathtt{V}}}_d +  \tilde{\mathtt{V}}_d )  S (v^0 , \mathcal{W}) \partial_X \tilde{\mathcal{U}} 
1705: + \eps \{B_{d}^{\eps,\flat,2} (x_d , X, \rond{v}^0 , v^{0,\flat} ,\rond{\un{\mathcal{U}}} , \un{\mathcal{U}}^\flat ,\tilde{\mathcal{U}} ).\partial_X 
1706: +  B_{d}^{\eps} (v^0 ,\mathcal{U}) \partial_d  \} \tilde{\mathcal{U}} ) |_{X=\frac{x_d}{\eps}}  .
1707: \end{eqnarray*}
1708: %
1709: Moreover, since the matrices $B_{d}^{\eps,\flat}$ are affine with respect to $X$ (see $(\ref{NN})$), 
1710: the two terms of the sum are in $\mathcal{N}(T)$.
1711: \\
1712: \item At first sight, the term 
1713: %
1714:  \begin{eqnarray*}
1715:  T := K^0 (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0, \mathcal{W}^1 )
1716:   -  K^0 ( \un{\mathcal{V}}^0 ,0 , 0,  \un{\mathcal{W}}^1 )
1717:  \end{eqnarray*}
1718: %
1719:   must appear in 
1720: $\tilde{\Phi}^{0}$ but we are going to redevelop this term.
1721: 
1722: \begin{lem}
1723: \label{psg}
1724: 
1725: There exist two  $C^\infty$ functions $K^0_I$ and $K^0_{II}$  affine with respect to their  first argument $\mathcal{V}^0$ such that 
1726:  %
1727:  \begin{eqnarray}
1728:  \label{ty}
1729: T  &:=& 
1730:   K^0_I (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0  , X \tilde{\mathcal{W}}^0 ) 
1731:  + \eps K^0_{II} (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 , \tilde{\mathcal{W}}^1 , X \tilde{\mathcal{W}}^1). 
1732: \end{eqnarray}
1733: %
1734: \end{lem}
1735: 
1736: Moreover, the functions $K^0_I (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0  , X \tilde{\mathcal{W}}^0 )$ and $ K^0_{II} (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X
1737: \tilde{\mathcal{W}}^0 , \tilde{\mathcal{W}}^1 , X \tilde{\mathcal{W}}^1)$ are in $\mathcal{N}(T)$.
1738: %
1739: %
1740: \begin{proof}
1741: %
1742: We will proceed in six steps.
1743: \\
1744: \begin{enumerate}[$1$.]
1745: \item
1746: We begin to split the term $T$ into three parts:  $T= T_1 +  T_2 +  T_3$ where
1747:  %
1748:  \begin{eqnarray*}
1749:  T_1 &:=& K^0 (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0, \mathcal{W}^1 ) - K^0 (\mathcal{V}^0
1750:  ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0, \un{\mathcal{W}}^1 ),
1751: \\  T_2 &:=& K^0 (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0, \un{\mathcal{W}}^1 ) 
1752: -  K^0 (\mathcal{V}^0 ,0 , 0, \un{\mathcal{W}}^1 ),
1753: \\  T_3 &:=&  K^0 (\mathcal{V}^0 ,0 , 0, \un{\mathcal{W}}^1 ) - K^0 ( \un{\mathcal{V}}^0 ,0 , 0, \un{\mathcal{W}}^1 ).
1754: \end{eqnarray*}
1755: %
1756: \item  Because the term $ K^0 (V,\tilde{\mathcal{W}} ,  X \tilde{\mathcal{W}} , W)$ is affine with respect to its 
1757:  first argument $V$ and affine  with respect to its fourth argument $W$ with $\X_{\mathtt{v}^0} \,   v^0$ as factor 
1758:  in the leading coefficient, there exist some $C^\infty$ functions
1759:  $K^0_1$ and  $K^0_2$  affine with respect to their first argument such that for all $(V,\tilde{\mathcal{W}},W) \in \R^{d+1} \times
1760:  \R\times \R$,
1761:   %
1762:  \begin{eqnarray}
1763:  \label{L}
1764:  K^0 (V ,\tilde{\mathcal{W}} , X \tilde{\mathcal{W}} , W ) &:=& K^0_1 (V ,\tilde{\mathcal{W}}, X \tilde{\mathcal{W}}  )
1765:   + \X_{\mathtt{v}^0} \,   v^0 . K^0_2 (V ,\tilde{\mathcal{W}}, X \tilde{\mathcal{W}} ) .W .
1766:    \end{eqnarray}
1767:  %
1768:  \item To deal with the term $T_1$, we use twice the equality $(\ref{L})$, first with
1769:   $(V ,\tilde{\mathcal{W}}, W)=( \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , \mathcal{W}^1 )$ and then with 
1770:  $(V ,\tilde{\mathcal{W}}, W)=( \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  \un{\mathcal{W}}^1 )$ to find 
1771:   %
1772:  \begin{eqnarray}
1773:   \label{La}
1774:  T_1 = \X_{\mathtt{v}^0} \,   v^0 .  K^0_2 (V ,\tilde{\mathcal{W}}, X \tilde{\mathcal{W}} ) .\tilde{\mathcal{W}}^1
1775:  \end{eqnarray}
1776: %
1777:   \item To deal with the term $T_2$, we use twice the equality $(\ref{L})$, first with
1778:   $(V ,\tilde{\mathcal{W}}, W)= ( \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  \un{\mathcal{W}}^1 )$ and then with 
1779:  $(V ,\tilde{\mathcal{W}}, W)=( \mathcal{V}^0 ,0 , \un{ \mathcal{W}}^1 )$  to find 
1780:   %
1781:  \begin{eqnarray}
1782:   \label{Lb}
1783:     T_2 &=&  K^0_1 ( \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  X \tilde{\mathcal{W}}^0 ) -  K^0_1 ( \mathcal{V}^0 ,0,0)  
1784: \\ \nonumber &&+  \X_{\mathtt{v}^0} \,   v^0 . ( K^0_2 (  \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 ) 
1785:    -   K^0_2 (  \mathcal{V}^0 , 0, 0 ))
1786:    .\un{\mathcal{W}}^1 .
1787:  \end{eqnarray}
1788: %
1789:  \item To deal with the term $T_3$,  we use twice the equality $(\ref{L})$, first with
1790:   $(V ,\tilde{\mathcal{W}}, W)= (\mathcal{V}^0 ,0 ,  \un{\mathcal{W}}^1 )$ and then with 
1791:  $(V ,\tilde{\mathcal{W}}, W)=( \un{\mathcal{V}}^0 ,0 ,  \un{\mathcal{W}}^1 )$
1792:   %
1793:  \begin{eqnarray}
1794:   \label{Lc}
1795:   T_3 =  K^0_1 ( \mathcal{V}^0 ,0 , 0 ) -  K^0_1 (  \un{\mathcal{V}}^0 ,0,0) + 
1796:  \X_{\mathtt{v}^0} \,   v^0 . ( K^0_2 (  \mathcal{V}^0 ,0 ,0) -   K^0_2 ( \un{\mathcal{V}}^0 , 0,0 ))  .\un{\mathcal{W}}^1 .
1797:   \end{eqnarray}
1798: %
1799:    \item
1800:  Using $(\ref{La})$, $(\ref{Lb})$, $(\ref{Lc})$, we get $(\ref{ty})$ with
1801:   %
1802:  \begin{eqnarray*}
1803:    K^0_I (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0 ,  X \tilde{\mathcal{W}}^0 ) &:=&  K^0_1 ( \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  X \tilde{\mathcal{W}}^0 ) - 
1804:     K^0_1 (  \un{\mathcal{V}}^0 ,0,0)
1805:  \\  K^0_{II} (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 , \un{\mathcal{W}}^1 , X \tilde{\mathcal{W}}^1) &:=&
1806:     \frac{ \X_{\mathtt{v}^0} \,   v^0 }{x_d} X \{  K^0_2 (  \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  X \tilde{\mathcal{W}}^0 ). \tilde{\mathcal{W}}^1
1807:  \\ &&+ ( K^0_2 (  \mathcal{V}^0 ,\tilde{\mathcal{W}}^0 ,  X\tilde{\mathcal{W}}^0 ) - K^0_2 (  \un{\mathcal{V}}^0 ,0,0)).
1808:  \un{\mathcal{W}}^1 \}.
1809:  \end{eqnarray*}
1810: %
1811: 
1812:  We have used that $x_d = \eps X$.
1813:  Because the term $\X_{\mathtt{v}^0} \,  v^0$ vanishes on the boundary, it contains a factor $x_d$. 
1814: Therefore the functions $ K^0_I$ and $K^0_{II}$
1815:  are $C^\infty$, affine with respect to their first argument. 
1816:  Moreover, the functions 
1817:   %
1818:  \begin{eqnarray*}
1819:  K^0_I (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0 ,  X \tilde{\mathcal{W}}^0 ) \quad \mathrm{and} \quad   K^0_{II} (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X
1820: \tilde{\mathcal{W}}^0 , \tilde{\mathcal{W}}^1 , X \tilde{\mathcal{W}}^1)
1821:  \end{eqnarray*}
1822: %
1823:  are in $\mathcal{N}(T)$.
1824: 
1825:  \end{enumerate}
1826: \end{proof}
1827: 
1828: As a consequence, the function  $K^0_{II} (\mathcal{V}^0 ,\tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 , \un{\mathcal{W}}^1 , X
1829: \tilde{\mathcal{W}}^1)$ can be put in $\tilde{\Phi}_{1}$.
1830: Therefore the boundary layer profile $\mathcal{W}^1$ does not appear
1831: in the contribution of the term  $T$  at the order $0$.
1832: \\
1833: \end{enumerate}
1834: %
1835:  Finally,  we get  the result and complete the proof of $(i)$.                                                                                                                            
1836:  %
1837: \end{proof}
1838: 
1839: Write $ \mathcal{U} := (\mathcal{V} ,\mathcal{W} )$ to find that 
1840: %
1841: \begin{eqnarray}
1842: \label{spil}
1843:  \H' (X, \mathcal{U}  ,\partial , \partial_X ) \tilde{\mathcal{U}} :=
1844:  \begin{bmatrix}
1845:   \H^{\star,'} ( \mathcal{U}   ,\partial , \partial_X )  \tilde{\mathcal{V}}
1846:  \\  (X  \mathtt{v}^{0,\flat}_d + \rond{\un{\mathtt{V}}}_d +  \tilde{\mathtt{V}}_d  ) \partial_X  \tilde{\mathcal{W}} 
1847:  + \X_{\mathtt{v}^0}  \, \tilde{\mathcal{W}}
1848:  \end{bmatrix}.
1849:  \end{eqnarray}
1850:  %
1851:  where
1852:   %
1853:   \begin{eqnarray*}
1854:  \H^{\star,'} (X,\mathcal{U} ,\partial , \partial_X )  \tilde{\mathcal{V}} :=
1855:  (X \mathtt{v}^{0,\flat}_d + \rond{\un{\mathtt{V}}}_d +  \tilde{\mathtt{V}}_d  ) S^\star (v^0 , \mathcal{W})  \partial_X \tilde{\mathcal{V}} 
1856: +  \mathfrak{L}^{\star} (v^0 , \mathcal{W} ,\partial)  \tilde{\mathcal{V}} 
1857: \\ + (S^\star  (v^0 ,\mathcal{W} ) - S^\star  (v^0 , \un{ \mathcal{W} })) \X_{\mathtt{v}^0}  \, \un{ \mathcal{V}}
1858:  +  K^0_I (  \mathcal{V} , \tilde{\mathcal{W}} , X  \tilde{\mathcal{W}}).
1859:  \end{eqnarray*}
1860:  
1861:  \subsubsection{A sequence of profile problems}
1862:  \label{tabl}
1863: 
1864: We split, for $j \geqslant 0$,  $ \un{\Phi}^{j}$  into
1865:    %
1866:   \begin{eqnarray*}
1867:   \un{\Phi}^{j} := (  \un{\Phi}^{j}_{v} ,  \un{\Phi}^{j}_{w} ) \in \R^{d+1} \times  \R .
1868:   \end{eqnarray*}
1869:  %
1870:    We define the problem 
1871:  %
1872: \begin{eqnarray*}
1873: (\mathcal{S}^j (T)): \quad
1874: \left\{
1875: \begin{array}{ccc}
1876: &(Id -P_0 ) \til{\Phi}^{j-1} = P_0   \, \til{\Phi}^{j} = \un{\Phi}^{j}_{v} =  \un{\Phi}^{j+1}_{w}   =0
1877:    \quad &\mathrm{where} \,  (t,x)  \in   (0,T) \times \Omega
1878: \\    &\mathtt{V}^j_d =  0   \quad &\mathrm{where} \,  (t,x)   \in  (0,T) \times  \Gamma
1879: \end{array}
1880: \right.
1881: \end{eqnarray*}
1882:  %
1883: We illustrate our  strategy with the following table
1884: \\
1885: \\ \begin{tabular}{ccccc}
1886:                     &  $(\mathcal{S}^{0})$                              & $(\mathcal{S}^{1})$                              & $(\mathcal{S}^{2})$           &   
1887: 		                 ...
1888: \\                  &                                         &                                        &                            
1889:         &  
1890: \\    $\Phi^{-1}$   & $(Id - P_0 )    \tilde{\Phi}^{-1} $     &                                        &                            
1891:         & 
1892: \\   $\Phi^{0}$     & $ P_0  \, \tilde{\Phi}^0 ,\un{\Phi}^{0}_{v} $ & $(Id - P_0 ) \tilde{\Phi}^0$        &                            
1893:         & 
1894: \\    $\Phi_{1 }$   &   $ \un{\Phi}^{1}_{w} $                   & $ P_0  \, \tilde{\Phi}^1 ,\un{\Phi}^{1} $  &  $(Id - P_0 )   
1895: \tilde{\Phi}^1 $   & 
1896: \\ $\Phi^{2 }$      &                                         &    $ \un{\Phi}^{2}_{w} $                 &   $ P_0  \, \tilde{\Phi}^2
1897: ,\un{\Phi}^{2}_{v} $   &                        
1898: \\   $\Phi^{3 }$ &         &           &  $ \un{\Phi}^{3}_{w} $  &                    
1899: \\ ... & & & &
1900: \end{tabular}
1901: \\
1902: \\ Each element of the left column contain  the  sum of the  corresponding line. When we solve successively 
1903: the  problems $(\mathcal{S}^j (T))$ the unknown profiles are for all $j \geqslant 0$, $\mathcal{V}^j $, $ \un{\mathcal{W}}^{j+1} $
1904: and $\tilde{\mathcal{W}}^{j}$ when solving the  problem  $(\mathcal{S}^j (T))$. 
1905:   Notice that $ \un{\Phi}^{0}_{w} =  \X_{\mathtt{v}^0} \,  w^0  =0 $ because the ground state $(v^0 ,\mathcal{W}^0 )$ 
1906:  is solution of the Euler system.  
1907:   Our goal in this section is to prove the following theorem:
1908: 
1909:  \begin{theo}
1910:  \label{casca}
1911:  
1912:  There exist profiles $(\mathcal{U}^j )_{j \geqslant 0}$ in $\mathcal{P} (T_0 )$  which verify $\un{\mathcal{W}}^0 :=
1913:  w^0 $, $(\mathcal{S}^j (T_0))_{j \geqslant 0} $ and
1914:   $ P_0 \mathcal{U}^j |_{t=0} =  \mathcal{U}^j_{init}$ for all $j \geqslant 0$.
1915:  Moreover the profile $\mathcal{U}^0$ is polarized in the sense
1916:  that $(Id - P_0) \tilde{\mathcal{U}} =0$ for all $(t,x) \in (0,T_0 ) \times \Omega$.
1917: 
1918: \end{theo}
1919: 
1920: 
1921: Theorem $\ref{try}$ is a consequence of theorem $\ref{casca}$.
1922:  %
1923: \begin{proof}
1924:  %
1925: We will proceed in two steps first studying the problem $(\mathcal{S}^0 (T_0))$ then for all $ j \geqslant 1$, 
1926: the problems  $(\mathcal{S}^j (T_0) )$.
1927: The problem $(\mathcal{S}^0 (T))$ splits into two sub-problems 
1928:  %
1929: \begin{eqnarray}
1930: \label{1P}
1931: \left\{
1932: \begin{array}{ccc}
1933: & \un{\Phi}^{0}_{v} = \un{\Phi}^{1}_{w} = 0  &\quad \mathrm{where} \,  (t,x)  \in  (0,T) \times  \Omega
1934: \\  & \un{\mathtt{v}}^0_d =  0              &\quad \mathrm{where} \,  (t,x)   \in   (0,T) \times \Gamma
1935: \end{array}
1936: \right.
1937: \end{eqnarray}
1938:  %
1939: and   
1940:  %
1941: \begin{eqnarray}
1942: \label{2P}
1943: (Id - P_0 ) \tilde{\Phi}^{-1} =  P_0  \, \tilde{\Phi}^0 =0  \quad \mathrm{where} \,  (t,x)  \in   (0,T) \times \Omega
1944: \end{eqnarray}
1945:  %
1946: Let us begin to look at $(\ref{1P})$: we have  
1947:  %
1948: \begin{eqnarray*}
1949:  \un{\Phi}^{0}_{v} &=&  \mathfrak{L}^\star (u^0 ,\partial)    \un{\mathcal{V}}^0 +  K^0  ( \un{\mathcal{V}}^0 ,0 ,0, \un{\mathcal{W}}^1 ) ,
1950:  \\  \un{\Phi}^{1}_{w} &=&   \X_{\mathtt{v}^0} \,  \un{\mathcal{W}}^1 +  \mathtt{\un{V}}^0 . \nabla \un{\mathcal{W}}^0 .
1951: \end{eqnarray*}
1952:  %
1953: In  $(\ref{1P})$, the determinations of $\un{\mathcal{V}}^0$ and $\un{\mathcal{W}}^1$ are coupled. 
1954: The existence of solutions is given by the following lemma.
1955: 
1956:  \begin{lem}
1957: \label{noncon}
1958: 
1959: There exist some profiles $ (\un{\mathcal{V}}^0 , \un{\mathcal{W}}^1)$ 
1960: in $H^\infty ( (0,T_0 ) \times \Omega)$ solution of:
1961:  %
1962: \begin{eqnarray}
1963: \label{elair} \qquad 
1964: \left\{
1965: \begin{array}{cc}
1966: \mathfrak{L} (u^0 ,\partial) 
1967:     \begin{bmatrix}
1968:      \un{\mathcal{V}}^0
1969:       \\ \un{\mathcal{W}}^1
1970:    \end{bmatrix}
1971:    +  
1972:     \begin{bmatrix}
1973:    K^0 ( \un{\mathcal{V}}^0  ,0 ,0, \un{\mathcal{W}}^1 ) 
1974:     \\ \un{\mathtt{V}}^0 . \nabla \un{\mathcal{W}}^0 
1975:    \end{bmatrix}
1976:     = 0  
1977:       &\quad \mathrm{where} \,  (t,x) \in      (0,T_0 ) \times \Omega
1978: \\  \underline{\mathtt{V}}^0_d =0  &\quad \mathrm{where} \,  (t,x) \in  (0,T_0 ) \times \Gamma
1979: \end{array}
1980: \right.
1981: \end{eqnarray}
1982:  %
1983:  such that 
1984:   %
1985:  \begin{eqnarray*}
1986:   P_0^\star \, \un{\mathcal{V}}^0 |_{t=0} =  \un{\mathcal{V}}^0_{init},  \ 
1987:  \un{\mathcal{W}}^1 |_{t=0} = \un{\mathcal{W}}^1_{init} .
1988:  \end{eqnarray*}
1989:   %
1990:  \end{lem}
1991:  %
1992: \begin{proof}
1993:  %
1994: We will proceed in three steps.
1995: 
1996: \begin{enumerate}[$(i)$]
1997: 
1998: \item 
1999: It is important to see that these equations are linear because the function  $K^0 $ is affine 
2000: with respect to its first and third argument.
2001: 
2002: \item 
2003: Some compatibility conditions on the corner $\{t=x_d =0 \}$ are necessary to the existence of solutions in $H^\infty$.
2004: The existence of initial compatible initial data $(\hat{\un{\mathcal{V}}}^0_{init} , \hat{\un{\mathcal{W}}}^1_{init} )$ are given in 
2005: Proposition $3.10$ of  \cite{moi}. Moreover Proposition $3.10$ of \cite{moi} insures that we can prescribe arbitrarily  $P_0 \, 
2006: \hat{\un{\mathcal{V}}}^0_{init} $ and $ \hat{\un{\mathcal{W}}}^1_{init} $. It is therefore possible to choose $P_0 \, 
2007: \hat{\un{\mathcal{V}}}^0_{init} := \un{\mathcal{V}}^0_{init}$ and $\hat{\un{\mathcal{W}}}^1_{init} = \un{\mathcal{W}}^1_{init}$.
2008: 
2009: \item
2010: Then  Th\'eor\`eme $3.10$ of  \cite{moi} yields the existence of a solution $(\un{\mathcal{V}}^0 , \un{\mathcal{W}}^1)$ in $H^\infty
2011: ((0,T_0 ) \times \Omega)$.
2012: 
2013: \end{enumerate}
2014: \end{proof}
2015:  %
2016:  %
2017: \begin{rem}
2018: 
2019: Referring to Proposition $\ref{sev}$, we can  compute the term  $ K^0 ( \un{\mathcal{V}}^0  ,0 ,0, \un{\mathcal{W}}^1 ) $ and see
2020: that the Prob. $(\ref{elair})$ can be restated as a boundary value problem for the linearized Euler system. 
2021: To see this by another way, plug $v^\eps = v^0 + \eps \un{\mathcal{V}}^0$ and 
2022: $w^\eps = w^0 + \eps \un{\mathcal{W}}^0$ in $(\ref{po})$.
2023: 
2024: \end{rem}
2025: 
2026:  Let us attack  $(\ref{2P})$. 
2027:  To solve it we use the following result:
2028: 
2029:  \begin{lem}
2030:  \label{lem2}
2031:  
2032:  There is a unique  profile  $\til{U}^0$  in a function of  $\Nrond (T_0)$ solution of
2033:   %
2034: \begin{eqnarray}
2035: \label{lem2a}
2036: &(Id-P_0 )  \til{\mathcal{U}}^0 =0 \quad &\mathrm{where} \,  (t,x,X) \in (0,T_0) \times \Omega_ \times \R_+ ,
2037: \\ \label{lem2b}
2038: & P_0  \,  \H' (X ,v_0  , \mathcal{W}^0 , \partial , \partial_X ) \til{\mathcal{U}}^0 = 0 
2039:   \quad &\mathrm{where} \,  (t,x,X) \in  (0,T_0) \times \Omega  \times \R_+ ,
2040:   \\ \label{lem2c}
2041: &  P_0  \, \til{\mathcal{U}}^0 = \til{\mathcal{U}}^0_{init}  \quad &\mathrm{where} \,  t=0 .
2042: \end{eqnarray}
2043:   %
2044: \end{lem}
2045:  %
2046: \begin{proof}
2047:  %
2048: If we solve directly the problem, because the system is \`a-priori nonlinear, we can only claim the local existence of a solution $
2049: \til{\mathcal{U}}^0$. A more acute analysis is possible using transparency properties.
2050: In order to exploit them, we are going to proceed in two steps first looking for $ \til{\mathcal{W}}^0$ 
2051: then looking  for the tangential velocity $\mathtt{V}_\mathfrak{t}$.
2052: 
2053: \begin{enumerate}[$(i)$]
2054: 
2055: \item Referring to $(\ref{spil})$, and thanks to the equations $(\ref{lem2a})$ and $(\ref{lem2b})$,  
2056:  we find  for $\til{\mathcal{W}}^0$ the equation
2057:  %
2058:  \begin{eqnarray*}
2059: (X  \mathtt{v}^{0,\flat}_d + \rond{\un{\mathtt{V}}}_d +  \tilde{\mathtt{V}}_d ) \partial_X  \,  \tilde{\mathcal{W}}^0 
2060:  + \X_{\mathtt{v}^0}  \,  \tilde{\mathcal{W}}^0 =0 . 
2061: \end{eqnarray*}
2062:  %
2063: Thanks to $(\ref{lem2a})$ and  $(\ref{elair})$,  we have 
2064: %
2065:  \begin{eqnarray}
2066:  \label{prune}
2067:  \til{\mathtt{V}}^0_d =0 \quad \mathrm{and} \  \rond{\un{\mathtt{V}}}_d^0 =0 .
2068:  \end{eqnarray}
2069:  %
2070:  Therefore we get
2071:  %
2072:  \begin{eqnarray}
2073:  \label{clem}  \quad
2074:  \mathtt{v}^{0,\flat}_d  \, X \partial_X  \tilde{\mathcal{W}}^0 
2075:  + \X_{\mathtt{v}^0}  \,  \tilde{\mathcal{W}}^0 =0 . 
2076: \end{eqnarray}
2077:  %
2078: This is a linear transport equation for which the  boundary  is totally characteristic.
2079: As a consequence, there is none compatibility condition at the corner $\{ t= x_d =0 \}$. 
2080: Therefore there exists one (and only one) solution $\til{\mathcal{W}}^0$ in $\P(T_0 )$ of $(\ref{clem})$ such that 
2081: $\til{\mathcal{W}}^0 |_{t=0} = \til{\mathcal{W}}^0_{init} $.
2082: 
2083: \item 
2084: We now look for $P_0^\star \, \til{\mathcal{V}}^0$ i.e. for $\tilde{\mathtt{V}}^0_\mathfrak{t}$.
2085:  Referring to $(\ref{spil})$, using once more $(\ref{prune})$, we get 
2086:  %
2087:  \begin{eqnarray}
2088:   \label{M1}
2089:  \H^{\star,'} (X ,\mathcal{U}  ,\partial , \partial_X ) \tilde{\mathcal{V}}  = 
2090:  \mathtt{v}^{0,\flat}_d \, S^\star (v^0 , \mathcal{W} )  X \partial_X \tilde{\mathcal{V}}  +  
2091:  \mathtt{v}^{0}_d  \,S^\star  (v^0 , \mathcal{W})  \partial_d  \tilde{\mathcal{V}}
2092:  + \L^{\star,'} (v^0 , \mathcal{W} ,\partial)  \tilde{\mathcal{V}} 
2093: \\ \nonumber +(S^\star  (v^0 ,\mathcal{W}^0 ) - S^\star  (v^0 , \un{ \mathcal{W} })) \X_{\mathtt{v}^0} \, \un{ \mathcal{U}}
2094:  +   K^0_I (  \mathcal{V} , \tilde{\mathcal{W}} , X \tilde{\mathcal{W}} ) .
2095:  \end{eqnarray}
2096:  %
2097:  Referring to section $(\ref{setting})$,  we get 
2098:  %
2099:  \begin{eqnarray}
2100:  \label{M3}
2101:   P_0^{\star}  \, S^\star  (v^0 , \un{ \mathcal{W} }) P_0^{\star} = \rho( \mathtt{p}^0 , \mathcal{W}^0 ) Id_{d-1}
2102:  \quad \mathrm{and} \  P_0^{\star}  \,  \L^{\star} P_0^{\star} = 0, 
2103:   \end{eqnarray}
2104:    %
2105: We denote by $K^0_{I,\mathfrak{t}}$ the $d-1$ first components of $K^0_{I}$.
2106:  Thanks to  $(\ref{M1})$ and $(\ref{M3})$, we deduce from the equation $(\ref{lem2b})$ the following equation for the tangential
2107:  velocity $\tilde{\mathtt{V}}^0_\mathfrak{t}$:
2108:   %
2109:  \begin{eqnarray}
2110:  \label{poli} \quad
2111: \\ \nonumber  \rho( \mathtt{p}^0 ,\mathcal{W}^0)  (\mathtt{v}^{0,\flat}_d X \partial_X 
2112:  + \X_{\mathtt{v}^0})  \, \tilde{\mathtt{V}}^0_\mathfrak{t}
2113:   + K^0_{I,\mathfrak{t}} (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 ) 
2114:  =  (  \rho (\mathtt{p}^0 ,\mathcal{W}^0 ) 
2115: -  \rho ( \mathtt{p}^0 , \un{ \mathcal{W}}^0)) \X_{\mathtt{v}^0}  \, \un{ \mathtt{V}}^0_\mathfrak{t}  .
2116:  \end{eqnarray}
2117:  %
2118:  We face a totally characteristic initial boundary value problem and once more time, there is none compatibility conditions. 
2119:  The system is first order symmetric hyperbolic.
2120: Because the function $K^0_{I,\mathfrak{t}}$ is affine with respect to its first argument, this equation $(\ref{poli})$ is linear.
2121: Therefore there exists one (and only one) $\tilde{\mathcal{V}}_0 $ solution of  the equation $(\ref{poli})$ 
2122: on  $(0,T_0) \times \Omega  \times \R_+ $ with $P_0^\star \, \tilde{\mathcal{V}}^0 |_{t=0} = \tilde{\mathcal{V}}^0_{init} $.
2123: 
2124: \end{enumerate}
2125: %
2126: \end{proof}
2127: 
2128: \begin{rem}
2129: 
2130: Referring to Proposition $(\ref{sev})$, we can compute 
2131:  %
2132:  \begin{eqnarray*}
2133: K^0_{I,\mathfrak{t}} (  \mathcal{V}^0 , \tilde{\mathcal{W}}^0 , X \tilde{\mathcal{W}}^0 ) :=  
2134: (  \rho (\mathtt{p}^0 ,\mathcal{W}^0 ) -  \rho ( \mathtt{p}^0 , \un{ \mathcal{W}}^0))
2135:  \frac{\X_{\mathtt{v}^0}  \,  \mathtt{v}^0_\mathfrak{t} }{x_d } X 
2136: \\ + (  \rho (\mathtt{p}^0 ,\mathcal{W}^0 ) \mathtt{V}^0 - \rho (\mathtt{p}^0 , \un{\mathcal{W}}^0 ) \un{\mathtt{V}}^0 ).  \nabla 
2137: \mathtt{v}^0_\mathfrak{t} .
2138:  \end{eqnarray*}
2139:  %
2140: 
2141: \end{rem}
2142: 
2143:  For all $ j \geqslant 1$, the problem  $(\mathcal{S}^j (T_0) )$ splits into several problems. 
2144: First, we solve:
2145: %
2146: \begin{eqnarray}
2147: \label{depol}
2148: (Id-P_0 )  \tilde{\Phi}^{j-1} =0 \quad \mathrm{where} \, (t,x,X) \in  (0,T_0) \times \Omega \times \R_+
2149: \end{eqnarray}
2150: %
2151: then 
2152: %
2153: \begin{eqnarray}
2154: \label{pol}  
2155:  P_0  \,  \tilde{\Phi}^{j} =0   \quad \mathrm{where} \, (t,x,X) \in  (0,T_0) \times  \Omega \times \R_+
2156: \end{eqnarray}
2157: %
2158: and 
2159: %
2160:  \begin{eqnarray}
2161: \label{rft}
2162: \left\{
2163: \begin{array}{ccc}
2164: &\un{\Phi}^{j}_{v}  =   \un{\Phi}^{j+1}_{w}  =   0 \quad &\mathrm{where} \,  (t,x) \in  (0,T_0) \times  \Omega
2165: \\    &\un{\mathtt{V}}^j_d  =0   \quad &\mathrm{where} \,  (t,x) \in  (0,T_0) \times \Gamma
2166: \end{array} 
2167: \right. 
2168: \end{eqnarray}
2169: %
2170: To do so, we will proceed in three steps.
2171: 
2172: \begin{enumerate}[$(i)$]
2173: 
2174: \item 
2175: We solve  $(\ref{depol})$ and defining $(Id-P_0) \til{\mathcal{U}}^j$ as the unique solution in  $\Nrond (T)$ of
2176: %
2177: \begin{eqnarray*}
2178: \L_d   \, \partial_X (Id-P_0) \tilde{\mathcal{U}}^{j}= 
2179: - (Id-P_0 ) ( \H' (X, \mathcal{U}^0 ,   \partial ,  \partial_X ) \tilde{ \mathcal{U}}^{j-1} 
2180:  -  \tilde{ \mathcal{Q}}^{j-1} ).
2181: \end{eqnarray*}
2182: %
2183: We stress that $\partial_X $ is an automorphism of  $\Nrond (T)$.
2184: Notice that we cannot prescribe arbitrarily $(Id-P_0) \tilde{\mathcal{U}}^{j}$ at $t=0$.
2185: 
2186: \item
2187: The problem  $(\ref{rft})$ is equivalent to a linearized Euler problem as in Lemma $\ref{noncon}$ for
2188: $(\un{\mathcal{V}}^{j},\un{\mathcal{W}}^{j+1})$.
2189: We obtain easily the existence of profiles $ \un{\mathcal{V}}^j$, $\un{\mathcal{W}}^{j+1}$ 
2190:  in $H^\infty ((0,T_0) \times \Omega)$ solution of this problem with 
2191:   %
2192:   \begin{eqnarray*}
2193:    P_0^\star  \,  \un{  \mathcal{V}}^j |_{t=0} =  \un{\mathcal{V}}^j_{init} , \  
2194:  \un{  \mathcal{W}}^{j+1} |_{t=0} = \un{  \mathcal{W}}^{j+1}_{init} .
2195:  \end{eqnarray*}    
2196: %
2197: \item We  define $P_0 \, \til{\mathcal{U}^j}$ as the unique solution in  $\Nrond (T_0)$ of 
2198: the totally characteristic linear hyperbolic initial boundary value:
2199: %
2200: \begin{eqnarray*}
2201: P_0  \, \H' (X, \mathcal{U}^0 , \partial, \partial_X )  P_0  \,   \tilde{ \mathcal{U}}^j 
2202:  =  P_0 ( \tilde{ \mathcal{Q}}^{j}   
2203:  - (Id-P_0 ) \H' (X, \mathcal{U}^0 ,  \partial , \partial_X ) \tilde{ \mathcal{Q}}^j )
2204: \end{eqnarray*}    
2205: %
2206: with $ P_0  \, \tilde{ \mathcal{U}}^j |_{t=0} =  \tilde{ \mathcal{U}}^j_{init}$.
2207: Thus $(\ref{pol})$ is proved. 
2208: 
2209: \end{enumerate}
2210: \end{proof}
2211: 
2212: 
2213: \section{Stability}
2214: \label{stab}
2215: 
2216: In this section, we are interested in the existence and the propagation of exact solutions of $(\ref{aller})$ asymptotic to the
2217: formal solutions constructed in the previous section. 
2218: Thus we assume that formal solutions $(u^\eps_{formal})_\eps $ of
2219: $(\ref{aller})$ on $(0,T_0 ) \times \Omega $ of the  form 
2220: %
2221: \begin{eqnarray*}
2222:  u^\eps_{formal} =(v^0 + \eps V^\eps_{formal} , W^\eps_{formal} ) \quad
2223: \mathrm{where} \   U^\eps_{formal} := (V^\eps_{formal} , W^\eps_{formal} )
2224: \end{eqnarray*}    
2225: %
2226:  is an expansion  
2227: %
2228: \begin{eqnarray*}
2229: \sum_{n \geqslant 0} \eps^n \, \mathcal{U}^n (t,x, \frac{x_d}{\eps})
2230: \ \mathrm{with} \ 
2231: \mathcal{U}^n = (\mathcal{V}^n , \mathcal{W}^n ) \in \P (T_0 ), \  \underline{\mathcal{W}}^0 = w^0 
2232: \end{eqnarray*}    
2233: %
2234: are given.
2235: We obtain approximate solutions $u^\eps_a = ( v^\eps_a =   v^0 + \eps V^\eps_a , W^\eps_a )$ 
2236: of the system $(\ref{aller})$, choosing for $n \in \N$, 
2237: %
2238: \begin{eqnarray*}
2239: V^\eps_a  (t,x) &:=& \sum_{k=0}^n \eps^{k} \, \mathcal{V}^k (t,x, \frac{x_d}{\eps}),
2240: \\ W^\eps_a  (t,x) &:=& \sum_{k=0}^n \eps^{k} \, \mathcal{W}^k (t,x, \frac{x_d}{\eps}) + \eps^{n+1} \, \un{\mathcal{W}}^{n+1} (t,x) .
2241: \end{eqnarray*}    
2242: %
2243: We  denote $Z_0 := \partial_t$, $Z_i := \partial_i$ for $1 \leq i \leq n-1$ and $Z_d := h(x_d ) \partial_d$, where $h$ is a
2244: bounded and $C^\infty$ function on $\R$ such that $h(x_d) \neq 0$ for $x_d  \neq 0$, $h(x_d) =x_d$ when $0\leqslant x_d \leqslant 1$
2245: and $h(x_d) = 1$ when $ x_d \geqslant 2$. 
2246: The family $(Z_i )_{0 \leqslant i \leqslant d}$ generates the algebra of $C^\infty$ tangent vector fields to $\Gamma$.
2247: For all $l \in \N$, we  denote by $Z^l$ the collection of the operators of the form  $Z_0^{\alpha_0} ... Z_d^{\alpha_d}$ where
2248: $\alpha_0 ,... , \alpha_d $ are in $\N$ and satisfy $|\alpha|:= \alpha_0 +... + \alpha_d =l$. 
2249: To simplify, we introduce the notations 
2250: %
2251: \begin{eqnarray*}
2252: L^2 (T) :=  L^2 ((0,T) \times \R_+^d) \quad \mathrm{and} \quad  L^\infty (T)  :=  L^\infty ((0,T) \times \R_+^d ).
2253: \end{eqnarray*}
2254: %
2255:  The family $(U^\eps_a )_{\eps \in ]0,1]}$ verifies the following estimates:
2256: %
2257: \begin{eqnarray}
2258: \label{linf}
2259: \mathrm{sup}_{0 < \eps \leqslant 1} || (\eps \partial_d)^k \, Z^l \, U^\eps_a ||_{L^\infty (T_0)  } < \infty .
2260: \end{eqnarray}
2261: %
2262: For $m \in \N$ and $T >0$, we denote by $\mathbf{E}^m (T)$ the set 
2263: %
2264: \begin{eqnarray*}
2265: \mathbf{E}^m (T) := \{   (u^\eps)_\eps   \in L^2 ( T ) /  \quad 
2266:  \mathrm{sup}_{0 < \eps \leqslant 1}  
2267:  \sum_{0 \leqslant 2k+l \leqslant m}  ||  (\eps \partial_d )^k \,  Z^l \,  u ||_{L^2 ( T)}  < \infty  \} .
2268: \end{eqnarray*}
2269: %
2270: To be clear about our notations, let us stress that in the previous inequality, the sum  
2271: %
2272: \begin{eqnarray*}
2273: \sum_{0 \leqslant 2k+l \leqslant m}  ||  (\eps \partial_d )^k  \, Z^l \,  u ||_{L^2 ( T)}
2274: \end{eqnarray*}
2275: %
2276:  stands for
2277: %
2278: \begin{eqnarray*}
2279: \sum_{0 \leqslant 2k+|\alpha| \leqslant m}  ||  (\eps \partial_d )^k  \, Z^\alpha \,  u ||_{L^2 ( T)}
2280: \end{eqnarray*}
2281: %
2282: Let us explain why we use the set $\mathbf{E}^m (T)$. 
2283: We begin with a brief review about smooth solutions of characteristic hyperbolic initial boundary value problem.
2284: We referred to the work of O.Gu\`es \cite{Gu90}.\footnote{An analysis in $L^2$ was achieved by J.Rauch \cite{R}.}
2285: First the boundary matrix of the system $(\ref{po2})$ is $A_d (u) := S(u) \mathtt{v}_d + \mathbf{L}_d $.
2286: When $x_d =0$, $\mathtt{v}_d =0$ and the rank of $A_d (u) = \mathbf{L}_d $ is constant.
2287: This suggests to use an extra-derivative namely $x_d \partial_d$ to the tangential derivatives $\partial_t, ...,\partial_{d-1}$.
2288: This yields the notion of conormal regularity.
2289: Then, handle normal derivatives for characteristic problem needs carefulness.
2290: An example by \cite{mo} shows that in general there are not $H^s$ estimates (unlike the noncharacteristic case). 
2291: We can extirpate $A_d \partial_d u$ from the equation but because the matrix $A_d$ is not invertible, 
2292: this does not provide estimates
2293: for the components in the kernel of $A_d$.
2294: However these components satisfy a transport equation (cf. \cite{rare}, \cite{mefr}, \cite{Gu90})
2295: with a source term which contains two conormal derivatives of all the
2296: components.   
2297: An iteration yields the idea to use one normal derivative for two conormal derivatives.
2298: Thus in \cite{Gu90}, O.Gu\`es uses the spaces
2299: %
2300: \begin{eqnarray*}
2301: E^m (T) := \{   u  \in L^2 ( T ) /  \quad 
2302:  \sum_{0 \leqslant 2k+l \leqslant m}  || \partial_d^k  \, Z^l  \, u ||_{L^2 ( T )}  < \infty  \} .
2303: \end{eqnarray*}
2304: %
2305: Here we face a singular perturbation problem. 
2306: More precisely, we look at boundary layers which corresponds to variations
2307: in $\frac{x_d}{\eps}$. 
2308: That is why we introduce some more adapted sets $\mathbf{E}^m (T)$ with the derivatives $\eps \partial_d $ instead of $\partial_d $.
2309: This idea of using some derivatives with $\eps$ in factor for some singular perturbations problems was used in
2310: \cite{Gu93},
2311: \cite{Gu92}, \cite{cgm2},... 
2312: Here, this idea is applied to anisotropic Sobolev spaces.
2313: Let us mention one technical point.
2314: In \cite{Gu90}, O.Gu\`es uses a reduction of the system.
2315: For the Euler system, this corresponds to choose the thermodynamic variables $p$, $\mathtt{v}$, $s$.
2316: 
2317: Looking at the table of subsection $\ref{tabl}$,  we see that  
2318: %
2319: \begin{eqnarray}
2320: \label{E}
2321: \H^\eps (U^\eps_a , \partial) U^\eps_a = \eps^{n+ \frac{1}{2}} 
2322:  \begin{bmatrix}
2323:  R^\eps_v
2324: \\ \eps R^\eps_W
2325: \end{bmatrix} 
2326: \quad \mathrm{for} \ \mathrm{all} \ (t,x) \in (0,T_0) \times \Omega ,
2327: \end{eqnarray}  
2328:   %
2329: with
2330: %
2331:  \begin{eqnarray*}
2332:  R^\eps_v &:=& (Id - P_0^* ) \tilde{\Phi}^n_v  + \eps \un{\Phi}^{n+1}_v + ...
2333: \\ R^\eps_W &:=&  \tilde{\Phi}^n_w + \eps \un{\Phi}^{n+2}_w + ...
2334: \end{eqnarray*}  
2335:  %
2336:   Notice that if $\tilde{\Phi} \in \mathcal{N}(T)$ then the family $(\Phi^\eps )_{\eps}$ defined by
2337:   %
2338:   \begin{eqnarray*}
2339:  \Phi^\eps (t,x) := \eps^{-\frac{1}{2}} \,  \tilde{\Phi} (t,x,\frac{x_d}{\eps}) \quad \mathrm{for} \ \mathrm{all} \ (t,x) \in (0,T) \times \Omega ,
2340:  \end{eqnarray*}  
2341:   %
2342:  is in  $\mathbf{E}^m (T) $, for all $m \in \N$.
2343:  Thus the family $( \eps^{-\frac{1}{2}} \,  R^\eps )_{\eps}$ defined by
2344:   %
2345:  \begin{eqnarray*}
2346:  R^\eps (t,x) :=  \begin{bmatrix}
2347: \eps R^\eps_v
2348: \\ R^\eps_{W}
2349: \end{bmatrix}
2350: \quad \mathrm{for} \ \mathrm{all} \ (t,x) \in (0,T) \times \Omega ,
2351: \end{eqnarray*}  
2352:   %
2353:  is in  $\in \mathbf{E}^m (T) $, for all $m \in \N$. 
2354:  
2355:  The system of equations $(\ref{E})$ is equivalent to the two equations 
2356:  %
2357:   \begin{eqnarray}
2358:   \label{E1} \quad &\mathfrak{L}^\star (u^\eps_a ,\partial)  V^\eps_a 
2359:   + \frac{1}{\eps} K_1 (v^0 , \partial v^0 , u^\eps_a )
2360: = \eps^{n+ \frac{1}{2}} R^\eps_v 
2361:  \quad &\mathrm{for} \ \mathrm{all} \ (t,x) \in (0,T_0) \times \Omega,
2362:   \\ \label{E2}
2363:  & \X_{\mathtt{v}^\eps_a } W^\eps_a = \eps^{n+ \frac{3}{2}} R^\eps_W \quad &\mathrm{for} \ \mathrm{all} \ (t,x) \in (0,T_0) \times \Omega.
2364:   \end{eqnarray}  
2365: %
2366:  Multiplying the equation $(\ref{E1})$ by $\eps$, using that $u^0$ satisfies the equation 
2367:  %
2368:  \begin{eqnarray*}
2369:  \mathfrak{L} (u^0 ,\partial)  (\partial_x ) v^0 =0 ,
2370:  \end{eqnarray*}  
2371: %
2372:  we obtain that the family $(u^\eps_a )_\eps $ satisfy
2373: %
2374: \begin{eqnarray*}
2375: \mathfrak{L} (u^\eps_a ,\partial) u^\eps_a 
2376: =
2377: \eps^{n+ \frac{1}{2}}
2378: \begin{bmatrix}
2379: \eps R^\eps_v
2380: \\ \eps R^\eps_{W}                                                                                                                   
2381:                                                       
2382: \end{bmatrix} \quad \mathrm{when} \ x_d > 0 ,
2383: \\ \mathtt{v}^\eps_{a,d} = 0  \quad \mathrm{when} \ x_d = 0 .
2384: \end{eqnarray*}    
2385: %%
2386: %
2387: We look for solutions $(u^\eps )_\eps \in H^{\infty} $ of the problem $(\ref{aller})$ of the form 
2388: %
2389: \begin{eqnarray}
2390: \label{fofo}
2391: u^\eps := (v^\eps, W^\eps ) \ \mathrm{with} \ v^\eps = v^\eps_a + \eps^{M+1} \, 
2392: V^\eps_R , \ W^\eps = W^\eps_a + \eps^{M} \, W^\eps_R .
2393: \end{eqnarray}    
2394: %
2395:  We denote $U^\eps_R := (V^\eps_R  , W^\eps_R )$.
2396:  We begin with a result of propagation.
2397: 
2398: \begin{theo}
2399: \label{propa}
2400: 
2401: Let $m >\frac{d}{2}+5 $, $n \geqslant 1$, $M \in ]\frac{1}{2}, n+ \frac{1}{2} [$ and  $T \in ]0,T_0[$. 
2402: We assume that we have a family of exact solutions $(u^\eps )_\eps \in H^{\infty} ((0,T) \times \Omega )$ of the
2403: problem $(\ref{aller})$ of the form $(\ref{fofo})$ where the family $(U^\eps_R )_{\eps}$ is in the set $\mathbf{E}^m (T)$.
2404: Then there is $\eps_0 \in ]0,1]$ such that for all $\eps \in ]0,\eps_0 ]$, the solution $u^\eps$ can be extended in a solution in
2405: $H^\infty ((0,T_0 ) \times \Omega )$ of the problem $(\ref{aller})$ and is of the form $(\ref{fofo})$ where we have extended
2406: $(U^\eps_R )_{\eps}$ in a family of $\mathbf{E}^m (T_0)$.
2407: 
2408: \end{theo}
2409: 
2410: The assumption $n+ \frac{1}{2} > M$ insures that the $u^\eps_a$ are accurate enough approximate solutions and the assumption $M >
2411: \frac{1}{2}$ that the $u^\eps_a$ are close to $u^\eps$. 
2412: Thanks to these assumptions,  Theorem $\ref{propa}$ claims that it is possible to extend the $u^\eps$, for $\eps \in ]0,\eps_0 ]$,
2413: till $T_0$ i.e. till the ground state exists.
2414: Notice that we can extend the result of Theorem $\ref{propa}$ to the more general case $n \in \N$, $M \geqslant \frac{1}{2}$.
2415: To do so, consider  $\hat{n} \geqslant M - \frac{1}{2}$ and 
2416: approximate solutions of order  $\hat{n}$:
2417: %
2418: \begin{eqnarray*}
2419: \hat{u}^\eps_a = ( \hat{v}^\eps_a , \hat{w}^\eps_a ) =  ( v^0 + \eps \hat{V}^\eps_a , \hat{W}^\eps_a )
2420: \end{eqnarray*}    
2421: %
2422: of the system $(\ref{aller})$, choosing
2423: %
2424: \begin{eqnarray*}
2425: \hat{V}^\eps_a  (t,x) &:=&  \sum_{k=0}^n \eps^{k} \, \mathcal{V}^k (t,x, \frac{x_d}{\eps}),
2426: \\ \hat{W}^\eps_a  (t,x) &:=& \sum_{k=0}^n \eps^{k} \, \mathcal{V}^k (t,x, \frac{x_d}{\eps}) + \eps^{n+1} \, \un{\mathcal{W}}^{n+1} (t,x) .
2427: \end{eqnarray*} 
2428: %
2429: We define the family  $(\hat{U}^\eps_R )_{\eps}$ by 
2430: %
2431: \begin{eqnarray*}
2432: \hat{U}^\eps_R := U^\eps_R + \eps^{-M} \, (U^\eps - \hat{U}^\eps_a ) \in  \mathbf{E}^m (T) .
2433: \end{eqnarray*} 
2434: %
2435: %
2436: On $(0,T)$, we get $U^\eps =\hat{U}^\eps_a + \eps^{M} \hat{U}^\eps_R $. 
2437: Applying Theorem $\ref{propa}$, we obtain some extensions of the $(\hat{U}^\eps_R )_{\eps}$ in a family of $\mathbf{E}^m (T_0)$. 
2438: Then we extend $( U^\eps_R  )_{\eps}$ in a family of $\mathbf{E}^m (T_0)$ , setting 
2439: %
2440: \begin{eqnarray*}
2441: U^\eps_R = \hat{U}^\eps_R  - \eps^{-M} \, (U^\eps - \hat{U}^\eps_a ).
2442: \end{eqnarray*} 
2443: %
2444: %
2445: In this paper, we consider a ground state $u^0$ in $H^\infty ((0,T_0) \times \Omega )$  and formal solutions with $H^\infty$
2446: regularity. It could also be possible to extend to ground states of high but finite regularity. 
2447: 
2448: We also give a result of existence.
2449: 
2450: \begin{theo}
2451:  \label{di}
2452: 
2453: Let $m >\frac{d}{2}+4 $, $n \geqslant 1$, $M \in ]\frac{1}{2}, n+ \frac{1}{2} [$.
2454: There exists $\eps_0 \in ]0,1]$ such that for all $\eps \in ]0,\eps_0 ]$, 
2455: there exists a solution $u^\eps \in H^\infty ((0,T_0 ) \times \Omega )$ of
2456: $(\ref{aller})$ of the form $(\ref{fofo})$ where $(U^\eps_R)_{\eps}$ is in $\mathbf{E}^m (T_0)$.
2457: 
2458: \end{theo}
2459: 
2460: It is important to understand that Theorem $\ref{propa}$ and $\ref{di}$ yields exact solutions till $T_0$. 
2461: Our method is based on the sets $\mathbf{E}^m$ and on some estimates uniform with respect to $\eps$. 
2462: Theorem $\ref{int}$ given in the introduction is a consequence of Theorem $\ref{try}$ and Theorem  $\ref{di}$. 
2463: 
2464: Because the problem $(\ref{aller})$ comes from a system of conservation laws and $\mathbf{\lambda}(u,\xi)$ 
2465: is an eigenvalue with constant
2466: multiplicity (cf. section $2$), it is possible to obtain $L^\infty$ estimates, even for $d \geqslant 1$. 
2467: We refer to  papers \cite{m2}, \cite{sonique} of G.M\'etivier, paper \cite{RR} of J.Rauch and M.Reed and paper \cite{Gu90}.
2468: Therefore we can weaken the regularity of the solution and prove a propagation result for some solutions admitting only
2469: one normal derivative in $L^2$. 
2470: We introduce the sets
2471: %
2472: %
2473: \begin{eqnarray*}
2474: \mathbf{A}^m (T) := \{   (u^\eps)_\eps   \in L^2 ( T) /  \quad 
2475:  \mathrm{sup}_{0 < \eps \leqslant 1}  
2476:  \sum_{0 \leqslant l \leqslant m}  ||  Z^l  \, u ||_{L^2 ( T )} 
2477:  +  \sum_{0 \leqslant l \leqslant m-2}  ||  \eps \partial_d   \, Z^l  \, u ||_{L^2 ( T )} < \infty  \} .
2478: \end{eqnarray*}
2479: %
2480: We will also use some norms built on $L^\infty$.
2481: We denote by $Z^\eps $ the collection of the derivatives  $Z_0 $,...,$Z_d$ and $ \eps \partial_d $. 
2482: Because the boundary is characteristic, we will need not only the Lipschitz norms but higher order $L^\infty$ control, 
2483: as O.Gu\`es in
2484: \cite{Gu90} and G.M\'etivier in \cite{sonique}.
2485: We denote by $L^\infty (T)$ the space  $L^\infty (T) =L^\infty ((0,T) \times \R_+^d )$.
2486: We introduce the norms
2487: %
2488: \begin{eqnarray*}
2489: ||u||_{k,T }  &:=& \sum_{0 \leqslant k \leqslant m} ||  Z^k u ||_{L^\infty (T )} ,
2490: \\ ||u||^*_{\eps,T}  &:=& ||u||_{0,T} + ||Z^\eps  u||_{1,T} ,
2491: \\  ||u||_{\eps,Lip,T}  &:=&  ||u||_{0,T} +   ||Z^\eps u ||_{0,T} .
2492: \end{eqnarray*}
2493: %
2494: Remember that, by abuse of notation, we denote for example  $||Z^\eps u ||_{0,T}$
2495:  for 
2496:  %
2497: \begin{eqnarray*}
2498:  \sum_{0 \leqslant i  \leqslant d} ||Z_i u ||_{0,T} + ||  \eps \partial_d u ||_{0,T} .
2499:  \end{eqnarray*}
2500: %
2501: We introduce the sets
2502: %
2503: \begin{eqnarray*}
2504: \mathbf{\Lambda}^m (T) := \{  (u^\eps)_\eps   \in L^\infty ( T)  /  \quad 
2505:  \mathrm{sup}_{0 < \eps \leqslant 1} \, ||u||^*_{\eps,T} < \infty  \} . 
2506: \end{eqnarray*}
2507: %
2508: It is also possible to tackle the limit case $M = \frac{1}{2} $, but we can prove the propagation only till $T_1 \in ]T,T_0]$.
2509: We incorporate this limit case in the following Theorem.
2510: 
2511: \begin{theo}
2512: \label{propa2}
2513: 
2514: Let $m > \frac{d}{2}+5$,  $M \geqslant \frac{1}{2} $, $T \in ]0,T_0[$. 
2515: We assume that we have a family of exact solutions $(u^\eps )_\eps $ of the
2516: problem $(\ref{aller})$ on $(0,T)$ of the form $(\ref{fofo})$ 
2517: where $(U^\eps_R)_{\eps}$ is in $\mathbf{A}^m (T) \cap \mathbf{\Lambda}^m (T) $.
2518: Then there exists $T_1 \in ]T,T_0]$ and there is $\eps_0 \in ]0,1]$ such that for all $\eps \in ]0,\eps_0 ]$,
2519:  the solution $u^\eps$ can be extend in a solution on
2520: $(0,T_1 )$ of the problem $(\ref{aller})$ and is of the form $(\ref{fofo})$ where we have extended
2521: $(U^\eps_R)_{\eps}$ in a function of $\mathbf{A}^m (T_1) \cap \mathbf{\Lambda}^m (T_1)$.
2522: Moreover if $M> \frac{1}{2} $, we can take $T_1 = T_0$.  
2523: 
2524: \end{theo}
2525: 
2526: It could be also possible to treat the limit case $M = \frac{1}{2} $ with the sets $\mathbf{E}^m (T_0)$ and incorporate a result of
2527: propagation till $T_1 \in ]T,T_0]$ in Theorem $\ref{propa}$. We did not do so for sake of clarity
2528: The rest of this section is devoted to the proof of Theorem $\ref{propa}$ and Theorem  $\ref{di}$.
2529: The proof of Theorem  $\ref{di}$ needs carefulness about the existence of compatible initial data.      
2530: Subsection  $\ref{bobo}$ is devoted to this question.
2531: In subsection $\ref{Reduction}$ we perform a reduction in a problem for $(U^\eps_R )_{\eps}$ which are the real unknown.
2532: We obtain (cf. Prop. $\ref{Reduction2}$) that $U^\eps_R$ satisfies a quasi-linear symmetric hyperbolic boundary value problem.
2533: As for the originating Euler problem, the boundary is conservative and characteristic of constant multiplicity.
2534: At first look, this system is singular with respect to  $\eps$ because of a factor $\eps^{-1}$ in the equation of $V^\eps_R$.
2535: However a further analysis reveals that in fact the singular term contains $\frac{x_d}{\eps} W^\eps_R $ as factor.
2536: This will be a key point in order to surmount the apparent singularity. 
2537: In order obtain existence of  $(U^\eps_R )_\eps$ till $T_0$, we will use a family of iterative schemes. 
2538: Thus we will supply in subsection $\ref{Linesti}$ linear estimates which are the core the proof. 
2539: We will successively perform $L^2$ estimates, conormal estimates and normal estimates.
2540: Several difficulties occur and are melt.
2541: First, the boundary is characteristic. 
2542: As we have explained it above, to tackle this problem we get inspired by the paper of O.Gu\`es \cite{Gu90}. 
2543: We adapt the method of O.Gu\`es substituting the derivative $\eps \partial_d $ to the derivative $\partial_d $ 
2544: in order to obtain uniform estimates. 
2545: Moreover, we use estimates of $\frac{x_d}{\eps} W^\eps_R $ in order to surmount the apparent singularity.
2546: In subsection $\ref{Itsch}$, 
2547: 
2548: The proof of Theorem $\ref{propa2}$ is not detailed as it obtained from Th\'eor\`eme $3$ of \cite{Gu90} in the same way Theorem
2549: $\ref{propa}$ and Theorem  $\ref{di}$ are inspired from Th\'eor\`eme $1$ and $2$ of \cite{Gu90}.
2550: The discussion about $M-\frac{1}{2}$ appears in the proof in subsection $\ref{Itsch}$ when we use the Sobolev embedding Lemma 
2551: $\ref{sobo}$. Some minor modifications at this step allow to tackle the limit case $M =\frac{1}{2}$.
2552: 
2553: 
2554: \subsection{Reduction}
2555: \label{Reduction}
2556: 
2557: Because the $(U^\eps_R )_\eps$ are the real unknown, we begin with a reduction.
2558: We will denote
2559: %
2560: \begin{eqnarray*}
2561: W^{\eps,\flat}_a  (t,x) := \sum_{k=1}^n \eps^{k-1} \, \mathcal{W}^k (t,x, \frac{x_d}{\eps}) 
2562: + \eps^{n} \, \un{\mathcal{W}}^{n+1} (t,x) .
2563: \end{eqnarray*}
2564: %
2565: %
2566: \begin{prop}
2567: \label{Reduction2}
2568: 
2569: There are some $(d+1) \times (d+1)$ matrices $J^\alpha$, 
2570:  some functions $J^{\beta,1}$, $J^{\beta,2}$
2571:  with values in $\R^{d+1}$, such that for all $\eps \in ]0,1]$,
2572:  a function $u^\eps = (v^\eps, W^\eps )$ of the form $(\ref{fofo})$ verifies $(\ref{aller})$ 
2573:  if and only if $U^\eps_R := (V^\eps_R  , W^\eps_R )$ verifies 
2574: %
2575: \begin{eqnarray}
2576: \label{poi}
2577:  \\ \nonumber (\mathfrak{L}^* (u^\eps ,\partial) + J^\alpha ).V^\eps_R
2578: +   W^\eps_R .(J^{\beta,1} +  \frac{x_d}{\eps} .J^{\beta,2}) = - \eps^{n+\frac{1}{2}-M} R^\eps_v \quad &\mathrm{where} \, (t,x) \in (0,T) \times \Omega
2579: \\ \label{poi2} \X_{\mathtt{v}^\eps }  \, W^\eps_R + \eps \mathtt{V}^\eps_R . \nabla W^\eps_a = -\eps^{n+\frac{3}{2}-M}  R^\eps_W 
2580: &\quad \mathrm{where} \, (t,x) \in (0,T) \times \Omega
2581: \\ \label{poi3} \mathtt{V}^\eps_{R,d} = 0  &\quad \mathrm{where} \, (t,x) \in (0,T) \times \Gamma .
2582: \end{eqnarray}
2583: %
2584: Moreover, the matrices $J^\alpha$ and the functions  $J^{\beta,1}$, $J^{\beta,2}$ depend in a $C^\infty$ way of 
2585: %
2586: \begin{eqnarray*}
2587: (\eps,V^\eps_a , Z_\eps V^\eps_a ,W^{\eps,\flat}_a, \eps^{M} \, U^\eps_R ). 
2588: \end{eqnarray*}
2589: %
2590: %
2591: \end{prop}
2592: %
2593: \begin{proof}
2594: %
2595: We only sketch the if part.
2596: The converse is left to the reader.
2597: Thus we assume that $(u^\eps)$ satisfies $(\ref{aller})$. 
2598: According to Proposition $\ref{Re1}$, the function $U^\eps$ verifies $(\ref{Re2})$ 
2599: which is equivalent to the two coupled equations $(\ref{Re3})$. 
2600: \\ 
2601: \\ We begin to show $(\ref{poi})$.
2602: Because the family of functions $(u^\eps )_\eps$ is of the form $(\ref{fofo})$, for each $\epsilon$, 
2603: the function $W^\eps$ is of the form $(\ref{qz})$ where $ \mathcal{W}^0 \in \P(T)$
2604:  is such that $\un{\mathcal{W}}^0 = w^0$ and $ W^{\eps,\flat} =  W^{\eps,\flat}_a + \eps^{M-1}  \, W^{\eps}_R $.
2605: Thus we can use Proposition $\ref{sev}$. 
2606: We will denote 
2607: %
2608: \begin{eqnarray*}
2609: \label{rr}
2610: K_1^\flat (V^\eps, \eps V^\eps , W^{\eps,\flat} ,\eps W^{\eps,\flat}) 
2611: \end{eqnarray*}
2612: %
2613: instead of
2614: %
2615: \begin{eqnarray*}
2616: K_{1}^{\flat} (\eps, x_d ,\rond{v}^0 , v^{0,\flat} , V^\eps , \eps V^\eps , \rond{w}^0 , w^{0,\flat} ,  \tilde{\mathcal{W}}^0
2617:  ,\frac{x_d}{\eps} \tilde{\mathcal{W}}^0 , W^{\eps,\flat} , \eps W^{\eps,\flat}) 
2618:  \end{eqnarray*}
2619: %
2620: %
2621: because these arguments are the ones which
2622: are important at this step of the analysis.
2623: Let us introduce our strategy. 
2624: We would like to write $(\ref{rr})$
2625: as a perturbation of 
2626: %
2627: \begin{eqnarray*}
2628: K_1^\flat (V^\eps_a , \eps V^\eps_a , W^{\eps,\flat}_a ,\eps W^{\eps,\flat}_a )
2629: \end{eqnarray*}
2630: %
2631: %
2632:  of order $\eps^M$. 
2633: But because $W^{\eps,\flat} := W^{\eps,\flat}_a + \eps^{M-1}  \, W^{\eps}_R $ a naive Taylor expansion fails to give the desired
2634: result.
2635: The trick lies in the way $K_1^\flat $ depends of $W^{\eps,\flat}$ (cf. Proposition $\ref{sev}$).
2636: This allows to factorize by $\eps^M$ and $ \frac{x_d}{\eps}$.
2637: 
2638: \begin{lem}
2639: \label{Redlem1}
2640: 
2641: There are  some $(d+1) \times (d+1)$ matrices $K_1^{\flat,\alpha}$ and  some $C^\infty$ 
2642: functions $K_1^{\flat,\beta,1}$, $K_1^{\flat,\beta,2}$  with values in $\R^{d+1}$ such that 
2643: %
2644: \begin{eqnarray}
2645: \label{KK}
2646: \quad K_1^\flat (V^\eps, \eps V^\eps , W^{\eps,\flat} ,\eps W^{\eps,\flat}) &=&
2647:  K_1^\flat (V^\eps_a , \eps V^\eps_a  , W^{\eps,\flat}_a  ,\eps W^{\eps,\flat}_a )
2648: \\ \nonumber &&+ \eps^{M} \,   K_1^{\flat,\alpha} (\eps, V^\eps_a ,W^{\eps,\flat}_a , \eps^{M}  \, U^\eps_R) .V^\eps_R
2649: \\ \nonumber &&+ \eps^{M} \,  W^\eps_R . K_1^{\flat,\beta,1} (\eps, V^\eps_a ,W^{\eps,\flat}_a , \eps^{M}  \, U^\eps_R) 
2650: \\ \nonumber &&+ \eps^{M} \, \frac{x_d}{\eps} W^\eps_R . K_1^{\flat,\beta,2} (\eps, V^\eps_a , W^{\eps,\flat}_a ). 
2651: \end{eqnarray}
2652: %
2653: \end{lem}
2654: %
2655: \begin{proof}
2656: %
2657: We will proceed in three steps. 
2658: 
2659: \begin{enumerate}
2660: 
2661: \item \label{pl1} By a first order Taylor development, there exist some $C^\infty$ $(d+1) \times (d+1)$ matrices $K_1^{\flat,\alpha}$ 
2662: and some  $C^\infty$ functions $K_1^{\flat,\beta,1}$ with values in $\R^{d+1}$ such
2663: that 
2664: %
2665: \begin{eqnarray*}
2666: K_1^\flat (V^\eps, \eps V^\eps , W^{\eps,\flat} ,\eps W^{\eps,\flat}) &=&
2667:  K_1^\flat (V^\eps_a , \eps V^\eps_a  , W^{\eps,\flat}  ,\eps W^{\eps,\flat}_a )
2668: \\ &&+ \eps^{M} \, K_1^{\flat,\alpha} (\eps, V^\eps_a ,W^{\eps,\flat}_a , \eps^{M}  \, U^\eps_R) .V^\eps_R 
2669: \\ &&+ \eps^{M} \, W^\eps_R . K_1^{\flat,\beta,1} (\eps, V^\eps_a  , W^{\eps,\flat}_a , \eps^{M}  \, U^\eps_R). 
2670: \end{eqnarray*}
2671: %
2672: \item \label{pl2} According to Proposition $\ref{sev}$, the function 
2673: %
2674: \begin{eqnarray*}
2675: K_1^\flat (V^\eps_a , \eps V^\eps_a  , W^{\eps,\flat} ,\eps
2676: W^{\eps,\flat}_a)
2677: \end{eqnarray*}
2678: %
2679:  is affine with respect to $  W^{\eps,\flat} $ with $\X_{\mathrm{v}^0} v^0$ as a factor in the leading term.
2680: This means that  there is a $C^\infty$ function $K_1^{\flat,\eta}$ with values in $\R^{d+1}$ such that
2681: %
2682: \begin{eqnarray*}
2683: K_1^\flat (V^\eps_a , \eps V^\eps_a  , W^{\eps,\flat} , \eps W^{\eps,\flat}_a) &=&
2684: K_1^\flat (V^\eps_a , \eps V^\eps_a  , W^{\eps,\flat}_a , \eps W^{\eps,\flat}_a) 
2685: \\ &&+ \eps^{M-1} \, \X_{\mathrm{v}^0}  \, v^0  .W^\eps_R . K_1^{\flat,\eta}
2686: (\eps,V^\eps_a  ,  W^{\eps,\flat}_a ) .
2687: \end{eqnarray*}
2688: %
2689: \item Thanks to the conditions $(\ref{cond})$, we get $ \X_{\mathrm{v}^0} v^0 =0$ when $x_d =0$.
2690: We define
2691: %
2692: \begin{eqnarray*}
2693:  K_1^{\flat,\beta,2} (\eps,V^\eps_a , W^{\eps,\flat}_a ) :=  \frac{1}{x_d} \X_{\mathrm{v}^0}  \, v^0 .K_1^{\flat,\eta}
2694: (\eps,V^\eps_a  ,  W^{\eps,\flat}_a ) .
2695: \end{eqnarray*}
2696: %
2697: The function $K_1^{\flat,\beta,2}$ is  $C^\infty$ and thanks to $(\ref{pl1})$ and $(\ref{pl2})$ 
2698: we get $(\ref{KK})$.
2699: 
2700: 
2701: \end{enumerate}
2702: %
2703: \end{proof}
2704: 
2705: We now look at the term $S^* (u^\eps ) \X_{\mathtt{v}^\eps} \,  V^\eps $.
2706: 
2707: \begin{lem}
2708: \label{Redlem2}
2709: 
2710: There are some $(d+1) \times (d+1)$ matrices $S^{*,\alpha}$ and some functions  $S^{*,\beta}$ with values in
2711: $\R^{d+1}$ such that 
2712: %
2713: \begin{eqnarray*}
2714: S^* (u^\eps ) \X_{\mathrm{v}^\eps} V^\eps = 
2715: S^* (u^\eps_a ) \X_{\mathrm{v}^\eps_a} V^\eps_a + \eps^{M} \, S^* (u^\eps )  \X_{\mathrm{v}^\eps}  \, V^\eps_R
2716:             + \eps^{M} S^{*,\alpha} . V^\eps_R + \eps^{M} \, W^\eps_R . S^{*,\beta} .
2717: \end{eqnarray*}
2718: %
2719: Moreover, the matrices $S^{*,\alpha}$ and $S^{*,\beta}$ depend in a $C^\infty$ way of 
2720: %
2721: \begin{eqnarray}
2722: \label{RR}
2723: (\eps, V^\eps_a , Z_\eps  V^\eps_a ,  \eps^{M} \, U^\eps_R )  .
2724:  \end{eqnarray}
2725: %
2726: \end{lem}
2727: %
2728: \begin{proof}
2729: %
2730: We proceed in three steps.
2731: 
2732: \begin{enumerate}
2733: 
2734: \item We apply  $(\ref{zou})$  with $(v_1 , w_1 ,v_2 , w_2 ) = (u^\eps_a , \eps^{M} \,  ( \eps V^\eps_R ,  W^\eps_R  ))$ and obtain
2735: %
2736: \begin{eqnarray}
2737: \label{tyj}
2738: S^* (u^\eps ) = S^* (u^\eps_a ) &+& \eps^{M+1} \, V^\eps_R . S^{*,\flat}_1 (u^\eps_a , \eps^{M} \,  ( \eps V^\eps_R ,  W^\eps_R  )) 
2739: \\    \nonumber                          &+&  \eps^{M} \, W^\eps_R . S^{*,\flat}_2 (u^\eps_a , \eps^{M} \, ( \eps V^\eps_R ,  W^\eps_R  )) .
2740: \end{eqnarray}
2741: %
2742: %
2743: \item We get 
2744: %
2745: \begin{eqnarray*}
2746:  \X_{\mathrm{v}^\eps} V^\eps = \X_{\mathrm{v}^\eps_a} V^\eps_a + \eps^{M} \, \X_{\mathrm{v}^\eps} V^\eps_R 
2747:  + \eps^{M} \, \mathtt{V}^\eps_R .(\eps \nabla  V^\eps_a  ).
2748: \end{eqnarray*}
2749: %
2750: \item We introduce the $(d+1) \times (d+1)$ matrices $S^{*,\alpha}$ such that for all $V \in \R^{d+1}$,
2751: %
2752: \begin{eqnarray*}
2753: S^{*,\alpha} V := \eps V . S^{*,\flat}_1 . ( \X_{\mathrm{v}^\eps_a}  \, V^\eps_a  
2754: +  \eps^{M} \, \mathtt{V}^\eps_R . (\eps \nabla  V^\eps_a  ). 
2755: \end{eqnarray*}
2756: %
2757: %
2758: Notice that $\X_{\mathrm{v}^\eps_a} V^\eps_a $ is not singular and can be expressed thanks to $V^\eps_a $, $Z_\eps V^\eps_a$.
2759: Thus the matrices $S^{*,\alpha} $ are $C^\infty$ with respect to $(\ref{RR})$.
2760: 
2761: We also introduce the functions 
2762: %
2763:  \begin{eqnarray*}
2764: S^{*,\beta}  :=  S^{*,\flat}_2 . (\X_{\mathrm{v}^\eps_a}  \, V^\eps_a  +  \eps^{M} \, \mathtt{V}^\eps_R. (\eps \nabla  V^\eps_a  ))
2765: \end{eqnarray*}
2766: %
2767: which take values in $\R^{d+1}$ and are $C^\infty$ with respect to $(\ref{RR})$.
2768: Thanks to $1$ and $2$, we get $(\ref{tyj})$. 
2769: 
2770: \end{enumerate}
2771: %
2772: \end{proof}
2773: 
2774: We define  the matrices $(d+1) \times (d+1)$ matrices $J^\alpha$
2775: %
2776: \begin{eqnarray*}
2777: J^\alpha (\eps,V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a , \eps^{M} \, U^\eps_R ) &:=& 
2778:  K_1^{\flat,\alpha} (\eps, V^\eps_a ,W^{\eps,\flat}_a , \eps^{M} \, U^\eps_R) 
2779: \\  &&+ S^{*,\alpha} (\eps, V^\eps_a , Z_\eps  V^\eps_a , \eps^{M} \, U^\eps_R  ) ,
2780: \end{eqnarray*}
2781: %
2782: and, for $i \in \{1,2\}$, the $\R^{d+1}$-valued function
2783: %
2784: \begin{eqnarray*}
2785:  J^{\beta,i} (\eps,V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a , \eps^{M} \, U^\eps_R ) &:=&
2786:   K_1^{\flat,\beta,i} (\eps,V^\eps_a ,  W^{\eps,\flat}_a , \eps^{M} \, U^\eps_R ) 
2787: \\ &&+ S^{*,\beta} (\eps, V^\eps_a , Z_\eps  V^\eps_a , \eps^{M} \, U^\eps_R ) .
2788:  \end{eqnarray*}
2789:  %
2790: Thanks to Lemma $\ref{Redlem1}$ and $\ref{Redlem2}$ and Equation $(\ref{E1})$, we show that $V^\eps_R$  satisfies $(\ref{poi})$.
2791: \\
2792: \\ As
2793: %
2794: \begin{eqnarray*}
2795:  \X_{\mathrm{v}^\eps}  \, W^\eps = \X_{\mathrm{v}^\eps_a}  \, W^\eps_a + \eps^{M}  \, \X_{\mathrm{v}^\eps}  \, W^\eps_R 
2796:  + \eps^{M}  \, \mathtt{V}^\eps_R .(\eps \nabla  W^\eps_a  ),
2797: \end{eqnarray*}
2798:  %
2799:  and using Equation $(\ref{E2})$,  we get $(\ref{poi2})$.
2800: We easily show $(\ref{poi3})$ and complete the proof.
2801:  %
2802: \end{proof}
2803:  %
2804: In the formulas $(\ref{poi})-(\ref{poi2})-(\ref{poi3})$, there is a singular factor $\eps^{-1}$ 
2805: which appears in the term $\frac{x_d}{\eps} W^\eps_R J^{\beta,2}$. 
2806: One idea would be to try to obtain estimates for $W^\eps_R$ ponderated by some $\eps$. 
2807: The difficulty lies in the fact that the equation $(\ref{poi3})$ of $W^\eps_R$ involves $V^\eps_R$ 
2808:  in return by the term $\eps  \mathtt{V}^\eps_R  .\nabla  W^\eps_a $.
2809: Fortunately, because of the special form of $\nabla  W^\eps_a $, we will see that it is possible to find good estimates of
2810: $\frac{x_d}{\eps} W^\eps_R$. This allows to overcome the false singularity.
2811: 
2812: 
2813: \subsection{Initial data}
2814: \label{bobo}
2815: 
2816: In order to obtain smooth solutions of $(\ref{poi})-(\ref{poi2})-(\ref{poi3})$, some compatibility conditions for the initial data
2817: $(U^\eps_{R,init})_{\eps \in ]0,1]}$ are necessary. 
2818: At the order $0$, this reads 
2819: %
2820: \begin{eqnarray*}
2821: \mathcal{R}^{0,\eps} : \  \mathtt{V}^\eps_{R,init,d} |_{x_d =0}= 0.
2822: \end{eqnarray*}
2823: %
2824: We now explain what are the compatibility conditions at order $j \geqslant 1$. 
2825: From the equation $(\ref{poi})$, we can extirpate $\partial_t \mathtt{V}^\eps_{R,init,d} $ in function of spatial derivatives and so by
2826: restriction its trace $\partial_t \mathtt{V}^\eps_{R,init,d} |_{ t=x_d =0 }$ on the corner $\{  t=x_d =0 \}$.
2827: More precisely, there exists a $C^\infty$ function $\H_1$ such that 
2828: %
2829: \begin{eqnarray*}
2830: \partial_t \mathtt{V}^\eps |_{ t=x_d =0 } = \H_1 ((\partial_x^l U^\eps_{R,init} |_{x_d =0 } )_{l \leqslant 1} ) - \partial_t
2831: R^\eps_{\mathtt{V}_d}  |_{t=x_d =0} .
2832: \end{eqnarray*}
2833: %
2834: in fact, the function $\H_1$ also depends in a $C^\infty$ way of $t,x,v^0$ and its derivatives, the profiles $\mathcal{U}^j$ and
2835: their $\eps$-conormal derivatives.
2836: We purposely not specify these arguments for sake of clarity.
2837: by iteration, we can also express the time derivatives $\partial_t^j \mathtt{V}^\eps_{R,d}$ for $j \geqslant 2$ by the equation
2838: $(\ref{poi})$. Therefore there exist some $C^\infty$ functions $\H^j$ such that 
2839: %
2840: \begin{eqnarray*}
2841: \partial_t^j \mathtt{V}^\eps_R |_{ t=x_d =0 } = \H_j ((\partial_x^l U^\eps_{R,init} |_{x_d =0 } )_{l \leqslant j} ) - \partial_t^j
2842: R^\eps_{\mathtt{V}_d}  |_{t=x_d =0} .
2843: \end{eqnarray*}
2844: %
2845: For $j \geqslant 1$, we define the $j$th order compatibility condition
2846: %
2847: \begin{eqnarray*}
2848: (\mathcal{R}^{j,\eps}): \ \H_j ((\partial_x^l U^\eps_{R,init} |_{x_d =0 } )_{l \leqslant j} ) = \partial_t^j
2849: R^\eps_{\mathtt{V}_d}  |_{t=x_d =0} .
2850: \end{eqnarray*}
2851: %
2852: Thus for $\eps \in ]0,1]$, if for $T >0$, $U^\eps_{R} \in H^\infty ( (0,T) \times \Omega) $ satisfies the equation
2853: $(\ref{poi})-(\ref{poi2})-(\ref{poi3})$ and $U^\eps_{R} |_{t=0} = U^\eps_{R,init}$, then $U^\eps_{R} $  satisfies the compatibility
2854: condition $(\mathcal{R}^{j,\eps})_{j \in \N}$. 
2855: Next proposition will show that there exist some families $(U^\eps_{R,init})_{\eps \in ]0,1]}$ which satisfy the  compatibility
2856: condition $(\mathcal{R}^{j,\eps})_{j \in \N}$ and some estimates uniform with respect to $\eps$.
2857: Moreover we can prescribe arbitrarily the components $(\mathtt{V}^\eps_{R,init,y} , \mathtt{S}^\eps_{R,init})_{\eps}$ 
2858: among the families
2859: which satisfy convenient uniform estimates.
2860: 
2861: \begin{prop}
2862: \label{exidi}
2863: 
2864: Let $(\mathtt{V}^\eps_{R,init,y} , \mathtt{S}^\eps_{R,init})_{\eps}$ be a family bounded in $H^\infty (\Omega)$. 
2865: Then there exists a family
2866: $(\mathtt{V}^\eps_{R,init,d} , \mathtt{P}^\eps_{init})_{\eps}$ bounded in $H^\infty (\Omega)$ such that for all $\eps \in ]0,1]$, for all $j\in
2867: \N$, the compatibility condition $\mathcal{R}^{\eps,j}$ is verified.
2868: 
2869: \end{prop}
2870: 
2871: The underlying reason to the possibility  to choose arbitrarily the tangential velocity and the entropy is that these components are
2872: the characteristic ones. Indeed we will see in the proof below the crucial role played by the normal derivatives.
2873: We need for uniform estimates in order to find a family of solutions $(U^\eps_R )_\eps$ in $E^\infty ((0,T_0) \times \Omega )$.
2874: We succeed in this goal obtaining in Proposition $\ref{exidi}$ a family $(U^\eps_{R,init})_\eps$ which even do not contain
2875: singularities in $\frac{x_d}{\eps}$.
2876: We refer to \cite{moi}, \cite{moi2} for other examples of existence of initial data compatible at all order with uniform existence in
2877: a setting of boundary layer theory.
2878: We will use in the proof of Proposition $\ref{exidi}$, as in the references above, a Borel lemma.
2879: %
2880: \begin{proof}
2881: %
2882: We will proceed in three steps.
2883: %
2884: \begin{enumerate}
2885: %
2886: \item We begin to analyze more accurately the compatibility conditions $ \mathcal{R}^{\eps,j}$ and more particularly 
2887: the way the functions $(\H_j )_{j \in \N^*}$ involve the normal derivatives of $\mathtt{v}^\eps_{R,init,d}$ and $p^\eps_{init}$.
2888: Indeed there are some $C^\infty$ functions $(\tilde{\H}_j )_{j \in \N^*}$
2889: such that for all $j \in \N^*$,
2890: %
2891: \begin{eqnarray*}
2892: \H^{2j} ((\partial_x^l U^\eps_{R,init} |_{x_d =0 } )_{l \leqslant 2j} ) = 
2893: \tilde{\H}^{2j} ((\partial_x^\alpha U^\eps_{R,init} |_{x_d =0 } )_{\alpha \in I_j} ) + (\frac{1}{\alpha^\eps})^j \partial_d^{2j}
2894: \mathtt{v}^\eps_{R,init,d} ,
2895: \\  \H^{2j+1} ((\partial_x^l U^\eps_{R,init} |_{x_d =0 } )_{l \leqslant 2j+1} ) = 
2896: \tilde{\H}^{2j} ((\partial_x^\alpha U^\eps_{R,init} |_{x_d =0 } )_{\alpha \in I'_j} ) + (\frac{1}{\rho^\eps})^j \partial_d^{2j}
2897: p^\eps_{R,init} .
2898: \end{eqnarray*}
2899: %
2900: \item For all $j \in \N$, there exists a family of functions $( \rond{\mathtt{V}}^{\eps,j}_{R,init,d} , \rond{P}^{\eps,j}_{R,init}
2901: )_\eps$ bounded in $H^\infty ((0,T_0) \times \Gamma )$ such that for all $j \in \N^*$
2902: %
2903: \begin{eqnarray*}
2904: \tilde{\H}^{2j} ((\partial_x^{\alpha'} \rond{U}^{\eps,\alpha_d}_{R,init}  )_{\alpha \in I_j} )
2905:  + (\frac{1}{\alpha^\eps})^j  \rond{\mathtt{V}}^{\eps,2j}_{R,init,d} =0,
2906: \\  \tilde{\H}^{2j+1} ((\partial_x^{\alpha'} \rond{U}^{\eps,\alpha_d}_{R,init}  )_{\alpha \in I'_j} ) 
2907: + (\frac{1}{\rho^\eps})^j  \rond{\mathtt{P}}^{\eps,2j+1}_{R,init} =0.
2908: \end{eqnarray*}
2909: %
2910: \item We use a Lemma by E.Borel in order to end the proof (cf. \cite{moi}).
2911: %
2912: %
2913: \end{enumerate}
2914: \end{proof}
2915: %
2916: \subsection{Linear estimate}
2917: \label{Linesti}
2918: 
2919: We begin to look at the following linear problem:
2920: %
2921: \begin{eqnarray}
2922:  \label{AL} \qquad (\mathfrak{L}^\star ( \underline{u}^\eps ,\partial)
2923: + \underline{J}^\alpha ).V^\eps_R +  W^\eps_R .(\underline{J}^{\beta,1} + 
2924: \frac{x_d}{\eps} \underline{J}^{\beta,2})
2925:  = - R^\eps_v
2926:  \quad \mathrm{when} \ x_d >0 ,
2927: \\ \label{BL} \X_{\underline{\mathtt{v}}^\eps } W^\eps_R + \eps \mathtt{V}^\eps_R . \nabla W^\eps_a = -\eps R^\eps_W 
2928: \quad \mathrm{when} \ x_d > 0,
2929: \\ \label{CL} \mathtt{V}^\eps_{R,d} = 0 \quad \mathrm{when} \ x_d =0,
2930: \end{eqnarray}
2931: %
2932: where $\underline{J}^\alpha $ denotes the $(d+1) \times (d+1)$ matrix 
2933: %
2934: \begin{eqnarray*}
2935: \underline{J}^\alpha &:=& J^\alpha (\eps, V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a, \eps^{M} \, \underline{U}^\eps_R  ),
2936: \end{eqnarray*}
2937: %
2938: and for $i \in \{1,2\}$,  $\underline{J}^{\beta,i} $ denotes the $\R^{d+1}$-valued function
2939: %
2940: \begin{eqnarray*}
2941: \underline{J}^{\beta,i} &:=& J^{\beta,i} (\eps, V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a, \eps^M \, \underline{U}^\eps_R  ) .
2942: \end{eqnarray*}
2943: %
2944: %
2945:   The family $(\underline{u}^\eps )_{\eps} := (\underline{v}^\eps , \underline{W}^\eps )_{\eps} $ is a given family of the form 
2946:   %
2947:   \begin{eqnarray}
2948:   \label{linform}
2949:     \underline{v}^\eps := v^\eps_a + \eps^{M+1} \,  \underline{V}^\eps_R ,  
2950:    \quad \underline{W}^\eps  := W^\eps_a + \eps^{M} \,  \underline{W}^\eps_R 
2951:    \end{eqnarray}
2952: %
2953:    with
2954:    %
2955:     \begin{eqnarray*}
2956:    \underline{U}^\eps_R := (\underline{V}^\eps_R , \underline{W}^\eps_R ) \in \mathbf{E}^m (T_0  ). 
2957:   \end{eqnarray*}
2958: %
2959:   We introduce the classic spaces of conormal distributions 
2960:   %
2961:    \begin{eqnarray*}
2962:   H^{0,m} (T) := \{ u \in L^2 (T ) / \ 0 \leqslant k  \leqslant m, \ Z^k u \in L^2 ( T ) \} .
2963:    \end{eqnarray*}
2964: %
2965:   These spaces will be endowed by the following weighted norms:
2966: %
2967: \begin{eqnarray}
2968: \label{defi}
2969: |u|_{m,\lambda,T} &:=& \sum_{0 \leqslant k \leqslant m} \lambda^{m-k}  \, ||e^{-\lambda T} Z^k u ||_{L^2 (T)} .
2970: \end{eqnarray}
2971: %
2972: Because we face a characteristic boundary problem with functions in $\frac{x_d}{\eps}$, we will also use the following  norms:
2973: %
2974: \begin{eqnarray}
2975: \label{defiN}
2976:  |u|^\mathcal{N}_{\eps, m,\lambda,T} &:=& |u|_{m,\lambda,T}  +  |(\eps \partial_d )u|_{m,\lambda,T}
2977: \\ \label{defiE}  |u|^\mathbf{E}_{\eps, m,\lambda,T} &:=& 
2978: \sum_{0 \leqslant 2k+l \leqslant m} \lambda^{m-2k-l}  \, |Z^l (\eps \partial_d )^k |_{0,\lambda,T}
2979: \end{eqnarray}
2980: %
2981: %
2982: A link between the $L^2$-type norm: $|.|^\mathbf{E}_{\eps,m,\lambda,T}$ and the $L^\infty$-type norm: $||.||^*_{\eps,T}$ is given by
2983: the following Sobolev Embedding lemma:
2984: 
2985: \begin{lem}[Sobolev Embedding]
2986: \label{sobo}
2987: 
2988: Let  $m > \frac{n}{2}+5$. There is  $c >0$ such that for all $T \in [0,T_0]$, 
2989:   for all $u \in C_0^\infty ((0,T) \times \Omega )$,
2990: %
2991: \begin{eqnarray*}
2992: \sqrt{\eps}  ||u||^*_{\eps,T} \leqslant c T e^{\lambda T }  \, |u|^\mathbf{E}_{\eps,m,\lambda,T} .
2993: \end{eqnarray*}
2994: 
2995: \end{lem} 
2996: %
2997: \begin{proof}
2998: %
2999:  Let us introduce the family of functions $(\tilde{u}^\eps )_{\eps \in ]0,1]}$ 
3000: by 
3001: %
3002: \begin{eqnarray*}
3003: \tilde{u}^\eps (t,x ):= u(t,y,\eps x_d ) \quad \mathrm{for} \ \mathrm{all} \ (\eps,t,x )  \in ]0,1] \times [0,T_1]  \times 
3004: \Omega. 
3005: \end{eqnarray*}
3006: %
3007: %
3008: We introduce $ ||u||^*_{T}  := ||u||_{0,T} + ||  u||_{1,T}$.
3009: We get
3010: %
3011: \begin{eqnarray*}
3012: || \tilde{u}^\eps ||^*_T := ||u ||^*_{\eps,T} 
3013:  \ \mathrm{and} \   || \tilde{u}^\eps ||^E_{m,\lambda,T} := \eps^{-\frac{1}{2}} \, ||u ||^\mathbf{E}_{\eps,m,\lambda, T} .
3014:  \end{eqnarray*}
3015: %
3016:  %
3017:   However Lemma $II.1.1$ of \cite{Gu90} shows that there exists $c >0$ such that for all $u \in E^m$, 
3018:   %
3019:   \begin{eqnarray*}
3020: ||u||^*_{T} \leq cTe^{\lambda T } |u|^E_{m,\lambda,T}  .
3021: \end{eqnarray*}
3022: %
3023: Using this for the family $(\tilde{u}^\eps)_\eps $ leads to Lemma $\ref{sobo}$.
3024:  %
3025: \end{proof}
3026:  %
3027: We will use the following version of the Gagliardo-Nirenberg estimates (cf. \cite{Gu90}).
3028:  %
3029: \begin{lem}[Gagliardo-Nirenberg]
3030: \label{Ga1}
3031:  %
3032: There exists $C>0$ such that for all $T \in ]0,T_0]$, for all $u \in L^\infty (T) \cap H^{0,m} (T)$, 
3033: for all $l$ and $m$ such that $l  \leqslant k  \leqslant m$ and for all real $\lambda \geqslant 1$
3034: %
3035: \begin{eqnarray*}
3036: \lambda^{k-l} |e^{-\frac{2}{p}\lambda t}  \, Z^l u |_{L^p ((0,T) \times \Omega )} \leqslant C ||u||^{1-\frac{k}{m}}_{0,T}
3037: |u|^{\frac{k}{m}}_{m,\lambda,T} ,
3038: \end{eqnarray*}
3039: %
3040: where $p:= \frac{2m}{k}$.
3041:  %
3042: \end{lem} 
3043:  %
3044: We will also use 
3045:  %
3046: \begin{lem}[Gagliardo-Nirenberg]
3047: \label{Ga2}
3048:  %
3049: Let $m \in \N$ be even.
3050: There exists $C>0$ such that for all $T \in ]0,T_0]$, for all $u \in L^\infty (T) \cap E (T)
3051: $, for all $l$ and $m$ such that $l  \leqslant k  \leqslant m$, for all $\eps \in ]0,1]$ and for all real $\lambda \geqslant 1$,
3052: %
3053: \begin{eqnarray*}
3054: \lambda^{k-l} |e^{-\frac{2}{p}\lambda t} (\eps \partial_d )^k Z^l u |_{L^p ((0,T) \times \Omega )} \leqslant 
3055: C ||u||^{1-\frac{k}{m}}_{0,T} (|u|^\mathbf{E}_{\eps,m,\lambda,T})^{\frac{k}{m}},
3056: \end{eqnarray*}
3057: %
3058: where $p:= \frac{2m}{k}$.
3059:  %
3060: \end{lem} 
3061: %
3062: \begin{proof}
3063: %
3064: It is a special case of $(Ap-II-3)$ given in \cite{Gu90}, p.$643$.
3065: %
3066: \end{proof}
3067: %
3068: Lemma $\ref{Ga1}$ and $\ref{Ga2}$ imply the following Moser's type inequalities.
3069:  %
3070:  %
3071: \begin{lem}[Moser Inequalities]
3072: \label{Moser}
3073:  %
3074: Let $F \in C^\infty  $ such that   $F(0) =0$, and  $\mathfrak{R} >0$. There exists a real  $l$ such that 
3075:  for all $T \in [0,T_0]$, if  
3076:  %
3077:  \begin{eqnarray*}
3078:  g \in L^\infty (T ) \cap  H^{0,m} (T) 
3079:  \quad (\mathrm{resp}. \ L^\infty (T ) \cap E^m (T ) \ \mathrm{and} \ m \  \mathrm{even} ),
3080:  \end{eqnarray*}
3081: %
3082:    verifies  $||g||_{0,T} \leqslant  \mathfrak{R}$,
3083:   we have, for all  $\lambda  \geqslant 1$, for all  $\eps \in ]0,1]$, we have: 
3084: %
3085: \begin{eqnarray*}
3086: | F(g) |_{m,\lambda,T} \leqslant l | g |_{m,\lambda,T} ,\quad (\mathrm{resp}. \ | F(g) |^\mathbf{E}_{\eps, m,\lambda,T} 
3087: \leqslant l | g |^\mathbf{E}_{\eps, m,\lambda,T} ).
3088: \end{eqnarray*}
3089: %
3090: \end{lem} 
3091:  %
3092: \begin{lem}
3093: \label{Moser2}
3094:  %
3095: For all  $\underline{m} \in \N$, there exists a real $c$ such that for all  $T$ verifying  $0 \leqslant T \leqslant T_0$, 
3096: and  $\alpha, \beta \in L^\infty (T )$, we have:
3097:  %
3098: \begin{enumerate}[$(1)$]
3099:  %
3100: \item \label{mosera}
3101: If  $\alpha, \beta \in H^{0,m} (T)$ and if  $j,k,l \in \N$ are such that 
3102:  $0 \leqslant j+l \leqslant k \leqslant \underline{m} $:
3103: %
3104: \begin{eqnarray*}
3105: \lambda^{\underline{m}-k} |Z^j \alpha \ Z^l \beta |_{\eps,0,\lambda,T} 
3106: \leqslant c( ||\alpha ||_{0,T} \ |\beta |_{ \underline{m},\lambda,T} +  |\alpha |_{\underline{m},\lambda,T} \ ||\beta ||_{ 0,T })
3107: \\ |\alpha \ \beta |_{\underline{m},\lambda,T} \leqslant c ( ||\alpha ||_{0,T} \  | \beta |_{ \underline{m},\lambda,T}  
3108:  +  |\alpha |_ {\underline{m} ,\lambda,T} \   ||\beta ||_{ 0,T })
3109: \end{eqnarray*}
3110: %
3111: \item If  $m$ is even, if $\alpha$, $\beta \in \mathbf{E}^m (T) ,$ and $ k,k',l,l',p  \in \N$ are 
3112:  such  that  $2(k+k')+l+l' \leqslant p \leqslant \underline{m}$:
3113: %
3114: \begin{eqnarray*}
3115: \lambda^{\underline{m}-p} |(\partial_n^k Z^l \alpha )(\partial_n^{k'} Z^{l'} \alpha ) |_{0,\lambda,T}
3116:  &\leqslant& c (|| \alpha ||_{0,\lambda,T} \  |\beta|^\mathbf{E}_{\eps,\underline{m},\lambda,T} 
3117:  + | \alpha |^\mathbf{E}_{\eps,\underline{m},\lambda,T} \  || \beta ||_{0,\lambda,T} ) ,
3118: \\ | \alpha \  \beta |^\mathbf{E}_{\eps,\underline{m},\lambda,T} 
3119: &\leqslant& c ( || \alpha  ||_{0,T}  \ |\beta|^\mathbf{E}_{\eps,\underline{m},\lambda,T}
3120:  + | \alpha |^\mathbf{E}_{\eps,\underline{m},\lambda,T} \  || \beta ||_{0,T} ).
3121: \end{eqnarray*}
3122: %
3123: \end{enumerate}
3124:  %
3125: \end{lem}
3126:  %
3127: We combine Lemma $\ref{Moser}$ and Lemma $\ref{Moser2}$ to find the following corollary.
3128:  %
3129:  %
3130: \begin{coro}
3131:  %
3132: Let  $F \in C^\infty (\R^k ,\R )$ such that  $F(0)=0$ and  $\mathfrak{R} >0$. 
3133:  There exists  $C>0$ such that, for all  $T \in [0,T_0 ]$, 
3134: for all  function $g \in Lip((0,T) \times \Omega ) \cap \mathcal{N} (T )$, $g: \Omega_T \rightarrow \R^k $
3135:  which verify  $||g||_{Lip(T)} \leqslant \mathfrak{R}$ then  $F(g) \in  \mathcal{N} (\Omega_T )$ and we have:
3136: %
3137: \begin{eqnarray*}
3138: |F(g)|^\mathcal{N}_{\eps,m,\lambda,T} \leqslant C |g|^\mathcal{N}_{\eps,m,\lambda,T} .
3139: \end{eqnarray*}
3140: %
3141: \end{coro}
3142: %
3143: \begin{proof}
3144: %
3145: Referring to the definition of $|.|^\mathcal{N}_{\eps,m,\lambda,T}$ (see $(\ref{defiN})$),
3146:  we get
3147: %
3148: \begin{eqnarray*}
3149: |F(g)|^\mathcal{N}_{\eps,m,\lambda,T} &:=& |F(g)|_{m,\lambda,T} + | \eps  \partial_d F(g)|_{m-2,\lambda,T} .
3150: \end{eqnarray*}
3151: %
3152: We use Lemma $\ref{Moser}$ to bound $|F(g)|_{m,\lambda,T}$.
3153: Referring to the definition of $|.|_{m,\lambda,T}$ (see  $(\ref{defi})$),
3154:  we get
3155: %
3156: \begin{eqnarray*}
3157:  | \eps  \partial_d F(g)|_{m-2,\lambda,T}    & =&  
3158:   \sum_{0 \leqslant k \leqslant m-2} \lambda^{m-2-k} ||e^{-\lambda T}  \, Z^k \eps \partial_d F(g)||_{L^2 (T)} 
3159: \\     & =&  \sum_{0 \leqslant k= k_1 + k_2  \leqslant m-2} \lambda^{m-2-k} \mathfrak{N}_{k_1 , k_2 } 
3160: \end{eqnarray*}
3161: %
3162: where 
3163: %
3164: \begin{eqnarray*}
3165:  \mathfrak{N}_{k_1 , k_2 } := ||e^{-\lambda T}   \, Z^{k_1}  D_g F(g) .  Z^{k_1} \eps \partial_d g ||_{L^2 (T )} .
3166: \end{eqnarray*}
3167:  %
3168: Then we use Lemma $\ref{Moser2}$ $(\ref{mosera})$. 
3169: %
3170: \end{proof}
3171: %
3172: We will prove in subsection $\ref{L2esti}$ and $\ref{hoe}$ the following linear $\eps$-uniform estimates 
3173: which will be useful in subsection  $\ref{Itsch}$. 
3174:  %
3175: \begin{theo}
3176: \label{pok}
3177:  %
3178: Let $m$ be an even integer larger than $2$. Let $\mathfrak{R}$ be a strictly positive real. 
3179: Then there is $\lambda_0 > 1$ such that if $\eps^M ||
3180: \underline{U}_R^\eps ||^*_{\eps,T_0} \leqslant \mathfrak{R}$ then 
3181: for all $\lambda \geqslant \lambda_0$, for all $\eps \in ]0,\eps_0]$,
3182: %
3183: \begin{eqnarray*}
3184: |U_R^\eps |^\mathbf{E}_{\eps,m,\lambda,T_0 } \leqslant
3185:  \lambda^{-1} \lambda_0 .( 1+ \eps^M  \, || U_R^\eps ||^*_{\eps,T_0} . | \underline{U}_R^\eps |^\mathbf{E}_{\eps,m,\lambda,T_0}  ).
3186: \end{eqnarray*}
3187: %
3188: \end{theo}
3189: 
3190: We will proceed in several steps. 
3191: In subsection $\ref{L2esti}$, we will perform $L^2$ estimates proceeding in three steps, estimating $W^\eps_R$ then $\frac{x_d}{\eps}
3192: W^\eps_R$ and finally $V^\eps_R$. 
3193: higher order estimates are more delicate to obtain.
3194: In subsection $\ref{hoe}$, we will give preliminary results with some commutators estimates (cf. Proposition $\ref{lapin}$ and with
3195: some  estimates of $\mathtt{v}^\eps_d . \partial_d U^\eps_R $ (cf. Lemma $\ref{carotte}$). We will begin with conormal estimates
3196: (subsection $\ref{conorm}$) then looking for normal estimates (subsection $\ref{norm}$). 
3197: As in the $L^2$ estimate, we will look for estimates about $W^\eps_R $, $\frac{x_d}{\eps}
3198: W^\eps_R$ and finally $V^\eps_R$.
3199: Furthermore, in the normal estimate, we will distinguish the estimates about tangential velocity $\mathtt{v}^\eps_{R,y}$ which
3200: corresponds to characteristic components and the estimates about normal velocity $\mathtt{v}^\eps_{R,d}$ which
3201: corresponds to a noncharacteristic component. 
3202:  %
3203:  %
3204: \subsection{$L^2$ estimate}
3205: \label{L2esti}
3206:  %
3207: We will proceed in three steps estimating $W^\eps_R$ then $\frac{x_d}{\eps}
3208: W^\eps_R$ and finally $V^\eps_R$.  
3209: According to section $5$, $\nabla W^\eps_a$ is of the form $\beta_1 (\eps,t,x) +
3210: \eps^{-1} \beta_2 (\eps,t,x,\frac{x_d}{\eps})$ where $\beta_1$ and $\beta_2$ are
3211: $C^\infty_b$ with respect to  their arguments including $\eps$ up to $0$. 
3212: Moreover  $\beta_2$ is rapidly decreasing  with respect to  its last argument.
3213:  %
3214:  %
3215: \subsubsection{Estimate of $W^\eps_R$}
3216: \label{estiW}
3217:  %
3218: The aim of this subsection is to prove 
3219:  %
3220: \begin{prop}
3221:  %
3222: Let $\mathfrak{R} >0$. There exists $\lambda_0 > 1$ such that if $\eps^M \, || \underline{V}^\eps_R ||^*_{\eps,T_0} \leqslant \mathfrak{R}$
3223:  then for all $\lambda \geqslant \lambda_0$, 
3224:  %
3225:  \begin{eqnarray*}
3226:  | W^\eps_R |_{0,\lambda,T_0} \leqslant \frac{\lambda}{\lambda_0} (1+ | V^\eps_R |_{0,\lambda,T_0}).
3227:  \end{eqnarray*}
3228: %
3229: %
3230: \end{prop}
3231: %
3232: \begin{proof}
3233: %
3234: We rewrite the equation $(\ref{BL})$ as  $\X_{\underline{\mathtt{v}^\eps}} W^\eps_R =f^\eps_I$ 
3235: where $ f^\eps_I :=  (\eps \beta_1 + \beta_2 ) V^\eps_R -\eps R^\eps_W $. 
3236: By a classic $L^2$ estimate, we obtain that for $\mathfrak{R} >0$, there exists $\lambda_0 \geqslant 1$ such that if
3237: $||\underline{\mathtt{v}}^\eps ||_{Lip} \leqslant  \mathfrak{R}$ then
3238:  for all $\lambda \geqslant \lambda_0$, 
3239:  %
3240:  \begin{eqnarray*}
3241:  | W^\eps_R |_{0,\lambda,T_0} \leqslant \frac{\lambda_0}{\lambda}
3242: |f^\eps_I  |_{0,\lambda,T_0}.
3243:  \end{eqnarray*}
3244: %
3245: Using $(\ref{linform})$
3246: and estimating $| f^\eps_I  |_{0,\lambda,T_0}$ we
3247: end the proof.
3248: %
3249: \end{proof}
3250: %
3251: \subsubsection{Estimate of $\frac{x_d}{\eps} W^\eps_R$}
3252: \label{estiXW}
3253:  %
3254: We will exploit the special form of the equation $(\ref{BL})$ thanks to a second estimate 
3255: which concerns $\frac{x_d}{\eps} W^\eps_R$.
3256:  %
3257: \begin{prop}
3258:  %
3259: Let $\mathfrak{R}  >0$. 
3260: There exists $\lambda_0 > 1$ such that if $\eps^M || \underline{V}^\eps ||_{\eps,T_0}^* \leqslant \mathfrak{R}$ 
3261: then for all $\lambda \geqslant \lambda_0$, 
3262: $ |\frac{x_d}{\eps} W^\eps_R |_{0,\lambda,T_0} \leqslant \frac{\lambda}{\lambda_0} (1+ | V^\eps_R |_{0,\lambda,T_0})$.
3263:  %
3264: \end{prop}
3265: %
3266: \begin{proof}
3267: %
3268: We will proceed in two steps.
3269: First we will look for an equation for $\frac{x_d}{\eps} W^\eps_R$ and then we will use a $L^2$
3270: estimate for this equation.
3271:  %
3272: \begin{enumerate}[$1$.]  
3273:  %
3274: \item We begin to calculate 
3275: %
3276: \begin{eqnarray}
3277: \label{aq}
3278: \X_{\underline{\mathtt{v}}^\eps}  \, ( \frac{x_d}{\eps} W^\eps_R ) = \frac{x_d}{\eps} \X_{\underline{\mathtt{v}}^\eps}  \, W^\eps_R +
3279: \underline{\mathtt{v}}_d^\eps . \frac{1}{\eps} W^\eps_R .
3280: \end{eqnarray}
3281: %
3282: Because $\underline{\mathtt{v}}_d^\eps =0 $ when $x_d =0$, there exist some $C^\infty$ functions $
3283: \underline{\mathtt{v}}^{\eps,\flat}$ such that $\underline{\mathtt{v}}_d^\eps = x_d  \underline{\mathtt{v}}^{\eps,\flat}$.
3284: Using the equation $(\ref{BL})$, we deduce from the equation $(\ref{aq})$ that $\frac{x_d}{\eps} W^\eps_R $ verifies the equation 
3285: %
3286: \begin{eqnarray}
3287: \label{aq2}
3288: \X_{\underline{\mathtt{v}}^\eps}  \, ( \frac{x_d}{\eps} W^\eps_R ) -  \underline{\mathtt{v}}^{\eps,\flat} ( \frac{x_d}{\eps} W^\eps_R ) 
3289: = f^\eps_{II} ,
3290: \end{eqnarray}
3291: %
3292: where 
3293: %
3294: \begin{eqnarray}
3295: f^\eps_{II} &:=& \frac{x_d}{\eps} (- \eps \mathtt{V}^\eps_R . \nabla W^\eps_a -  \eps R^\eps_W ) ,
3296: \\          &:=& -x_d \beta_1 V^\eps_R - \frac{x_d}{\eps} \beta_2 V^\eps_R - x_d R^\eps_W  .
3297: \end{eqnarray}
3298: %
3299: \item By a classic $L^2$ estimate, we obtain that for $\mathfrak{R} >0$, there exists $ \lambda_0 \geqslant 1$ such that if 
3300: %
3301: \begin{eqnarray*}
3302: ||\underline{\mathtt{v}}^{\eps} ||_{Lip ((0,T_0 ) \times \Omega )} 
3303: + || \underline{\mathtt{v}}_d^{\eps} ||_{L^\infty (T_0  )}  \leqslant R
3304: \end{eqnarray*}
3305: %
3306:  then for all $\lambda  \geqslant  \lambda_0$ ,
3307: %
3308: \begin{eqnarray*}
3309: | \frac{x_d}{\eps} W^\eps_R |_{0,\lambda,T_0} \leqslant \frac{\lambda_0}{\lambda} | f^\eps_{II} |_{0,\lambda,T_0} .
3310: \end{eqnarray*}
3311: %
3312:   %  
3313: Because of the form of $\underline{\mathtt{v}}^\eps $ (see $(\ref{linform})$), the family $(||
3314: \underline{\mathtt{v}}^\eps ||_{Lip} + || \underline{\mathtt{v}}^{\eps,\flat} ||_{L^\infty} )_\eps $ 
3315: is bounded when the family $(\eps^{M} || \underline{V}^\eps_R  ||^*_{\eps,T_0}  )_\eps$ is bounded. 
3316: Moreover because $\beta_2$ is rapidly decreasing with respect to its last argument,  
3317: the family $(\frac{x_d}{\eps} \beta_2 (\eps,t,x,\frac{x_d}{\eps}))_\eps $ is
3318: bounded in $L^\infty$. This ends the proof.
3319:  %
3320: \end{enumerate}
3321:  %
3322: \end{proof}
3323:  %
3324: \subsubsection{Estimate of $ V^\eps_R$}
3325: \label{estiV}
3326:  %
3327: We will prove the following result:
3328:  %
3329: \begin{lem}
3330:  %
3331: Let $\mathfrak{R} >0$. There exists $ \lambda_0 > 0 $ such that 
3332: if $\eps^{M} || \underline{U}^\eps_R ||_{\eps,T_0}^* \leqslant  \mathfrak{R}$ 
3333: then for all $\lambda \geqslant \lambda_0$, $ | V^\eps_R |_{0,\lambda,T_0} \leqslant  \frac{ \lambda_0}{\lambda}$.
3334:  %
3335: \end{lem}
3336:  %
3337: \begin{proof}
3338: %
3339: For each $\eps \in ]0,1]$, the function $V^\eps_R$ is solution of the following boundary value problem: 
3340: %
3341: \begin{eqnarray}
3342: \label{Sx}
3343: \left\{
3344: \begin{array}{ccc}
3345: &(  \mathfrak{L}^\star (\underline{u}^\eps  ,\partial) + \underline{J}^\alpha  ).V^\eps_R =
3346: f^\eps_{III} \quad &\mathrm{when} \ x_d > 0 ,
3347: \\ &\mathtt{V}^\eps_{R,d} = 0  &\quad \mathrm{when} \ x_d = 0 ,
3348: \end{array}
3349: \right.
3350: \end{eqnarray}
3351: %
3352: where 
3353: %
3354: \begin{eqnarray*}
3355: f^\eps_{III}  :=  - W^\eps_R \underline{J}^{\beta,1}  - \frac{x_d}{\eps}  W^\eps_R \underline{J}^{\beta,2} - R^\eps_W .
3356: \end{eqnarray*}
3357: %
3358:  %
3359: It is a symmetric hyperbolic system for which the boundary is conservative and characteristic of constant multiplicity. 
3360: Thanks to \cite{R}, the following $L^2$ estimate holds: for $\mathfrak{R}> 0$, there exists $\lambda > 0$, such that if  
3361: %
3362: \begin{eqnarray*}
3363: ||\underline{\mathtt{v}}^\eps ||_{Lip} + || \underline{J}^\alpha ||_{L^\infty} 
3364: + || \underline{J}^\beta ||_{L^\infty} \leqslant \mathfrak{R}
3365: \end{eqnarray*}
3366: %
3367: then for all $\lambda \geqslant \lambda_0$, 
3368: %
3369: \begin{eqnarray*}
3370: |V^\eps_R |_{0,\lambda,T_0} \leqslant  \frac{\lambda}{\lambda_0} |f^\eps_{III} |_{0,\lambda,T_0}  .
3371: \end{eqnarray*}
3372: %
3373: Using the form of $\underline{\mathtt{v}}^\eps $, $\underline{J}^\alpha$, $\underline{J}^\beta$ and 
3374: estimating  $|f^\eps_{III} |_{0,\lambda,T_0}$ thanks to subsection $\ref{estiW}$ and $\ref{estiXW}$ yields the conclusion.
3375:  %
3376: \end{proof}
3377:  %
3378:  %
3379: \subsection{Higher order estimates} 
3380: \label{hoe}
3381:  %
3382: We look for estimates uniform with respect to $0<\eps \leqslant 1$ of 
3383: %
3384: \begin{eqnarray*}
3385: \lambda^{m-2k-l} |(\eps \partial_d )^k Z^l U^\eps_R |_{0,\lambda,T_0 } \quad \mathrm{for} \  2k+l \leqslant m .
3386: \end{eqnarray*}
3387: %
3388: In this subsection, we give crucial preliminary technical results. 
3389: We group them in two kinds.
3390: In a first time, we look for commutators estimates. 
3391: In a second time, we look for estimates which will be useful when estimating source terms.
3392:  %
3393: \subsubsection{Commutators estimates} 
3394:  %
3395: We begin with the estimates of the commutators $[ \underline{J} , (\eps \partial_d )^k Z^l ] \varphi$.
3396: Because we will proceed in two steps, estimating first conormal estimates then normal estimates, we give specific estimates
3397: of the commutators with the derivatives $Z^l$ for $l \leqslant m$ (i.e. in the limit case $k=0$).
3398:  %
3399: \begin{prop}
3400: \label{pomme}
3401:  %
3402: Let $m$ be an integer. 
3403: There is an increasing function $C:\R_+ \rightarrow \R_+^*$ such that for all $\eps \in ]0,1]$, 
3404: if $\eps^M  \, || \underline{U}^\eps_R ||_{0,T}
3405: \leqslant  \mathfrak{R}$ then for all $\varphi \in C^\infty_0 ((0,T) \times \overline{\Omega} )$,
3406:  for all $k,l$ such that $2k+l \leqslant m$, for all $\lambda
3407: \geqslant \lambda_0$,
3408:  %
3409: \begin{enumerate}[$a)$]
3410:  %
3411: \item \label{pomme1}
3412: if $m$ is even, 
3413: %
3414: \begin{eqnarray*}
3415: \lambda^{m-2k-l} |[ \underline{J} (\underline{u}^\eps ) , (\eps \partial_d )^k  Z^l ] \varphi|_{0,\lambda,T} 
3416: \leqslant C(\mathfrak{R}) ( |\varphi |^\mathbf{E}_{\eps, m,\lambda,T} 
3417: + \eps^M || \varphi ||_{0,T} .|\underline{U}^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} ) , 
3418: \end{eqnarray*}
3419: %
3420:  %
3421:  \item \label{pomme2}
3422:  for any $m$, 
3423: %
3424: \begin{eqnarray*}
3425: \lambda^{m-l} |[ \underline{J}  (\underline{u}^\eps ) ,  Z^l ] \varphi |_{0,\lambda,T}
3426:  \leqslant C(\mathfrak{R}) ( |\varphi |_{m,\lambda,T} 
3427: + \eps^M || \varphi ||_{0,T} .|\underline{U}^\eps_R |_{m,\lambda,T} ) , 
3428: \end{eqnarray*}
3429: %
3430: \end{enumerate}
3431:  %
3432: where $\underline{J}$ stands for $\underline{J}^\alpha $, $\underline{J}^{\beta,1}$ or $\underline{J}^{\beta,2}$.
3433:  %
3434: \end{prop}
3435: %
3436: \begin{proof}[Proof of $(\ref{pomme2})$]
3437: %
3438: We introduce a notation: if $l \in \N, l \neq 0$ and $\phi$ a vector-valued function which components are
3439: denoted  $\phi_i$, we denote $Z^{<l>} \phi$ the collection of the terms
3440: of the form $Z^{l_1}  \phi_{i_1} ... Z^{l_r}  \phi_{i_r} $
3441:  where $1 \leqslant  r  \leqslant l , \ l_1 + ... l_r = l $
3442: and $l_i \geqslant 1$ and if $l =0$, $Z^{<l>} \phi =0$.
3443: The commutator $[ \underline{J} ,  Z^l ] \varphi$ is a linear combination of terms of the form 
3444: %
3445: \begin{eqnarray*}
3446: \Phi (\eps, V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a , \eps^M \, \underline{U}^\eps_R ). \mathfrak{P}
3447: \end{eqnarray*}
3448: %
3449:  with 
3450:  %
3451:  \begin{eqnarray*}
3452:  \mathfrak{P} := Z^{<l_1>} ( V_{a}^{\eps} ) . Z^{<l_2>} ( Z_\eps  V_{a}^{\eps} ) . Z^{<l_3>} (  W^{\eps,\flat}_a  ) .
3453: Z^{<l_4>} (\eps^M \, \underline{U}^\eps_R ).  Z^{l_5}  \phi ,
3454: \end{eqnarray*}
3455:  %
3456: and  $l_1 + l_2 + l_3 + l_4 + l_5= l $.
3457:  Thanks to estimates $(\ref{linf})$ and Lemma $\ref{Moser}$, we get
3458:  %
3459:  \begin{eqnarray*}
3460: \lambda^{m-l} |[ \underline{J}  (\underline{u}^\eps ) ,  Z^l ] \varphi |_{0,\lambda,T}
3461:  \leqslant C(\mathfrak{R}) \lambda^{m-l} 
3462:  | Z^{<l_4>} (\eps^M \, \underline{U}^\eps_R ).  Z^{l_5}  \phi |_{m,\lambda,T}  .
3463:  \end{eqnarray*}
3464: %
3465: We end the proof thanks to Lemma $\ref{Moser2}$ $(1)$.
3466: %
3467: \end{proof}
3468: %
3469: \begin{proof}[Proof of $(\ref{pomme1})$]
3470: %
3471: We introduce a notation: if $l \in \N, l \neq 0$ and $\phi$ a vector-valued function which components are
3472: denoted  $\phi_i$, we denote $(\eps \partial_d)^{<k>} \phi$ the collection of the terms
3473: of the form $ (\eps \partial_d)^{k_1}  \phi_{i_1} ... (\eps \partial_d)^{k_r}  \phi_{i_r} $
3474:  where $1 \leqslant  r  \leqslant l , \ k_1 + ... + k_r = l $
3475: and $k_i \geqslant 1$ and if $l =0$, $(\eps \partial_d)^{<k>} \phi =0$.
3476: The commutator $[ \underline{J} ,  (\eps \partial_d )^k Z^l ] \varphi$ is a linear combination of terms of the form 
3477: %
3478: \begin{eqnarray*}
3479: \Phi (\eps, V^\eps_a , Z_\eps V^\eps_a , W^{\eps,\flat}_a , \eps^M \, \underline{U}^\eps_R ). \tilde{\mathfrak{P}}
3480: \end{eqnarray*}
3481: %
3482: with 
3483:  %
3484:  \begin{eqnarray*}
3485: \tilde{\mathfrak{P}} := 
3486: (\eps \partial_d)^{<k_1>} Z^{<l_1>}  ( V_{a}^{\eps} ) . (\eps \partial_d)^{<k_2>} Z^{<l_2>} ( Z_\eps  V_{a}^{\eps} ) 
3487:  (\eps \partial_d)^{<k_3>} Z^{<l_3>} (  W^{\eps,\flat}_a  ) .
3488: \\ .(\eps \partial_d)^{<k_4>}  Z^{<l_4>}   (\eps^M \, \underline{U}^\eps_R ).  (\eps \partial_d)^{<k_5>} Z^{l_5} \underline{J} \phi ,
3489: \end{eqnarray*}
3490: %
3491:  with $l_1 + l_2 + l_3 + l_4 + l_5= l $, $k_1 + k_2 + k_3 + k_4 + k_5 = k $.
3492:  Thanks to estimates $(\ref{linf})$ and Lemma $(\ref{Moser})$, we get
3493:  %
3494:  \begin{eqnarray*}
3495: \lambda^{m-2k-l} |[ \underline{J}  (\underline{u}^\eps ) ,  Z^l ] \varphi |_{0,\lambda,T}
3496:  \leqslant C(\mathfrak{R}) \lambda^{m-l} 
3497:  | Z^{<l_4>} (\eps^M \, \underline{U}^\eps_R ).  Z^{l_5} \underline{J} \phi |_{m,\lambda,T}  .
3498:  \end{eqnarray*}
3499: %
3500: We end the proof thanks to Lemma $(\ref{Moser2})$ $(2)$.
3501:  %
3502: \end{proof}
3503:  %
3504:   %
3505:  A main point would be to estimate some commutators involving $S(\underline{u}^\eps )
3506: X_{\underline{\mathtt{v}}^\eps}$.   
3507: We will estimate separately $S(\underline{u}^\eps ) \underline{\mathtt{v}}^\eps_d \partial_d$ and the one $S(\underline{u}^\eps )
3508: X'_{\underline{\mathtt{v}}^\eps}$ (see $(\ref{fd})$ for the definition of $X'_{\underline{\mathtt{v}}^\eps}$).
3509: Moreover because we will proceed in two steps, estimating first conormal estimates then normal estimates, we give specific estimates
3510: of the commutators with the derivatives $Z^l$ for $l \leqslant m$ (i.e. in the limit case $k=0$).
3511:  %
3512: \begin{prop}
3513: \label{lapin}
3514:  %
3515: Let $m$ be an integer. 
3516: There is an increasing function $C:\R_+ \rightarrow \R_+^*$ such that for all $\eps \in ]0,1]$, 
3517: if $\eps^M  \, || \underline{U}^\eps_R ||^*_{\eps,T}
3518: \leqslant  \mathfrak{R}$ then for all $\varphi \in C^\infty_0 ((0,T) \times \overline{\Omega} )$,
3519:  for all $k,l$ such that $2k+l \leqslant m$, for all $\lambda
3520: \geqslant \lambda_0$,
3521:  %
3522: \begin{enumerate}[$a)$]
3523:  %
3524: \item if $m$ is even, 
3525: %
3526: \begin{eqnarray*}
3527: \lambda^{m-2k-l} |[S(\underline{u}^\eps ).\underline{\mathtt{v}}^\eps_d \partial_d , (\eps \partial_d )^k  Z^l ] \varphi
3528: |_{0,\lambda,T} \leqslant C(\mathfrak{R}) ( |\varphi |^\mathbf{E}_{\eps, m,\lambda,T} 
3529: + \eps^M || \varphi ||^*_{\eps,T} .|\underline{U}^\eps_R |^\mathbf{E}_{\eps,
3530: m,\lambda,T} ) , 
3531: \end{eqnarray*}
3532: %
3533: \item for any $m$, 
3534: %
3535: \begin{eqnarray*}
3536: \lambda^{m-l}  |[S(\underline{u}^\eps ).\underline{\mathtt{v}}^\eps_d \partial_d ,  Z^l ] \varphi |_{0,\lambda,T} \leqslant 
3537: C(\mathfrak{R}) (|\varphi|^\mathcal{N}_{\eps, m,\lambda,T} 
3538: + \eps^M || \varphi ||^*_{\eps,T} .|\underline{U}^\eps_R |^\mathcal{N}_{\eps,m,\lambda,T}),
3539: \end{eqnarray*}
3540: %
3541: \item if $m$ is even, 
3542: %
3543: \begin{eqnarray*}
3544: \lambda^{m-2k-l} |[S(\underline{u}^\eps ). \X'_{\underline{\mathtt{v}}^\eps} , (\eps \partial_d )^k  Z^l ] \varphi
3545: |_{0,\lambda,T} \leqslant C(\mathfrak{R})  ( |\varphi |^\mathbf{E}_{\eps, m,\lambda,T} 
3546: + \eps^M ||\varphi||_{1,T} .|\underline{U}^\eps_R |^\mathbf{E}_{\eps,
3547: m,\lambda,T} ) , 
3548: \end{eqnarray*}
3549: %
3550: \item for any $m$, 
3551: %
3552: \begin{eqnarray*}
3553: \lambda^{m-l}  |[S(\underline{u}^\eps ). \X'_{\underline{\mathtt{v}}^\eps} ,  Z^l ] \varphi |_{0,\lambda,T} \leqslant 
3554:   C(\mathfrak{R})  (|\varphi|_{m,\lambda,T} 
3555:   + \eps^M || \varphi ||_{1,T} .|\underline{U}^\eps_R |_{m,\lambda,T}).
3556: \end{eqnarray*}
3557: %
3558: \end{enumerate}
3559:  %
3560: \end{prop}
3561:  %
3562: We begin with the proof of $c)$ and $d)$ which are simpler.
3563: Because $X'_{\underline{\mathtt{v}}} := \partial_t + \sum_{i=1}^{d-1} \mathtt{v}_i \partial_i $, the estimates  $c)$ and $d)$ are
3564: consequences of the following lemma.
3565:  %
3566: \begin{lem}
3567:  %
3568: Let $m$ be an integer.
3569: Let $F$ be a $C^\infty$ function.
3570: There is an increasing function $C:\R_+ \rightarrow \R_+^*$ such that for all $\eps \in ]0,1]$, 
3571: if $\eps^M  \, || \underline{U}^\eps_R ||^*_{\eps,T}
3572: \leqslant  \mathfrak{R}$ then for all $\varphi \in C^\infty_0 ((0,T) \times \overline{\Omega} )$,
3573:  for all $k,l$ such that $2k+l \leqslant m$, for all $\lambda
3574: \geqslant \lambda_0$,
3575:  %
3576: \begin{enumerate}[$(i)$]
3577:  %
3578: \item \label{re1} if $m$ is even, 
3579: %
3580: \begin{eqnarray*}
3581: \lambda^{m-2k-l} |[ F(\underline{u}^\eps ). Z , (\eps \partial_d )^k  Z^l ] \varphi|_{0,\lambda,T} \leqslant 
3582: C(\mathfrak{R})  ( |\varphi |^\mathbf{E}_{\eps, m,\lambda,T}  
3583: + \eps^M ||\varphi||_{1,T} .|\underline{U}^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} ) , 
3584: \end{eqnarray*}
3585: %
3586: \item \label{re2} for any $m$, 
3587: %
3588: \begin{eqnarray*}
3589: \lambda^{m-l}  |[ F(\underline{u}^\eps ). Z ,  Z^l ] \varphi |_{0,\lambda,T} \leqslant 
3590:   C(\mathfrak{R})  (|\varphi|_{m,\lambda,T} 
3591:   + \eps^M || \varphi ||_{1,T} .|\underline{U}^\eps_R |_{m,\lambda,T}).
3592: \end{eqnarray*}
3593: %
3594: \end{enumerate}
3595:  %
3596: \end{lem}
3597:  %
3598: \begin{proof}[Proof of $(\ref{re2})$]
3599: %
3600: The commutator $[ F(\underline{u}^\eps ). Z ,  Z^l ] \varphi $ is a linear combination of terms of the form
3601: %
3602: \begin{eqnarray}
3603: \label{bb}
3604: Z^{l_1} F(\underline{u}^\eps ). Z^{l_2} \varphi \quad \mathrm{when} \ l_1 + l_2 = l +1, \ l_1 \geqslant 1, \ l_2  \geqslant 1.
3605: \end{eqnarray}
3606: %
3607: We introduce $l_1' := l_1 -1$, $l_2' := l'_2 -1$.
3608: Thus the term $(\ref{bb}$) reads 
3609: %
3610: \begin{eqnarray*}
3611: Z^{l'_1} (Z F(\underline{u}^\eps )). Z^{l'_2} (Z \varphi ).
3612: \end{eqnarray*}
3613: %
3614: We apply Lemma $\ref{Moser2}$ $(1)$ with 
3615: %
3616: \begin{eqnarray*}
3617: \alpha := Z F(\underline{u}^\eps ) \ ,  \beta:= Z \varphi \ , \underline{m} := m-1   
3618: \end{eqnarray*}
3619: %
3620: and obtain the estimate
3621: %
3622: \begin{eqnarray*}
3623: \lambda^{m-l}  |[ F(\underline{u}^\eps ). Z ,  Z^l ] \varphi |_{0,\lambda,T} \leqslant 
3624: C ( ||F(\underline{u}^\eps ) ||_{1,T} .  |\varphi |_{m,\lambda,T}
3625:   +  |F(\underline{u}^\eps ) |_{m,\lambda,T} . || \varphi  ||_{1,T} ).
3626: \end{eqnarray*}
3627: %
3628: We use that $(u^\eps )_\eps$ is of the form $(\ref{linform})$ and apply Lemma $\ref{Moser}$ to complete the proof.
3629: %
3630: \end{proof}
3631:  %
3632: \begin{proof}[Proof of $(\ref{re1})$]
3633: %
3634: We assume that $k \geqslant 1$. 
3635: The case $k=0$ corresponds to $(\ref{re2})$.
3636: The commutator $[ F(\underline{u}^\eps ). Z , (\eps \partial_d )^k  Z^l ] \varphi$ is a linear combination of terms of the form
3637: %
3638: \begin{eqnarray}
3639: \label{bb1}
3640: &(\eps \partial_d )^k  Z^{l_1} F(\underline{u}^\eps ). Z^{l_2} \varphi
3641:  \quad &\mathrm{with} \ l_1 + l_2 = l +1, \ l_1 \geqslant 1, \ l_2  \geqslant 1,
3642: \\ \label{bb2}
3643: &(\eps \partial_d )^{k_1}  F(\underline{u}^\eps ). (\eps \partial_d )^{k_2} Z^{l+1} \varphi 
3644: \quad &\mathrm{with} \ k_1 + k_2 = k, \ k_1  \geqslant 1 .
3645: \end{eqnarray}
3646: %
3647: %
3648: \begin{enumerate}[$\bullet$]
3649:  %
3650: \item We begin to deal with the term $(\ref{bb1})$.
3651: We introduce $k' := k -1$, $l_2' := l'_2 -1$.
3652: We commute $\eps \partial_d$ with $Z^{l_1}$.
3653: Thus it suffices to deal  with terms of the form
3654: %
3655: \begin{eqnarray*}
3656:  (\eps \partial_d )^{k'}  Z^{l_1} (\eps \partial_d  F(\underline{u}^\eps )). Z^{l'_2} (Z \varphi ) .
3657: \end{eqnarray*}
3658: %
3659: %
3660: We apply Lemma $\ref{Moser2}$ $(1)$ with 
3661: %
3662: \begin{eqnarray}
3663: \label{jus}
3664: \alpha :=  \eps \partial_d F(\underline{u}^\eps ) \ ,  \beta:= Z \varphi \ , \underline{m} := m-2 .   
3665: \end{eqnarray}
3666: %
3667: We obtain the estimate
3668: %
3669: \begin{eqnarray*}
3670: \lambda^{m-l}  |[ F(\underline{u}^\eps ). Z ,  Z^l ] \varphi |_{0,\lambda,T} \leqslant 
3671: C ( ||F(\underline{u}^\eps ) ||_{\eps,Lip,T} .  |\varphi |^\mathbf{E}_{m,\lambda,T}
3672:   +  |F(\underline{u}^\eps ) |^\mathbf{E}_{m,\lambda,T} . || \varphi  ||_{1,T} ).
3673: \end{eqnarray*}
3674: %
3675: We end the estimate as in the proof of $(\ref{re2})$.
3676:  %
3677: \item
3678: To deal with the term $(\ref{bb2}$) and complete the proof, we directly apply  Lemma $\ref{Moser2}$ $(1)$ to $(\ref{jus})$ and
3679: proceed in the same way.
3680:  %
3681: \end{enumerate}
3682: %
3683: \end{proof}
3684:  %
3685: We tackle the Proof of $b)$.
3686:  %
3687: \begin{proof}[Proof of $b)$]
3688: %
3689: The commutator $[S(\underline{u}^\eps).\underline{\mathtt{v}}^\eps_d \partial_d ,  Z^l ] \varphi $
3690: is a linear combination of some terms of the form 
3691: %
3692: \begin{eqnarray*}
3693:  Z^{l_1} S(\underline{u}^\eps) . Z^{l_2} \underline{\mathtt{v}}^\eps_d . Z^{l_3} \partial_d  \varphi , 
3694: \quad \mathrm{with} \  l_1 + l_2 + l_3  = l, \  l_1 + l_2 \geqslant 1 .
3695: \end{eqnarray*}
3696: %
3697: Because $\underline{\mathtt{v}}^\eps$ is of the form $\underline{\mathtt{v}}^\eps =  \mathtt{v}^0  + \eps
3698: \underline{\mathtt{V}}^\eps  $, it suffices to deal with the terms 
3699: %
3700: \begin{eqnarray}
3701: \label{ff1}  
3702: Z^{l_1} S(\underline{u}^\eps) . Z^{l_2} \mathtt{v}^0_d . Z^{l_3} \partial_d  \varphi ,
3703: \\ \label{ff2} 
3704: Z^{l_1} S(\underline{u}^\eps) . Z^{l_2} \underline{\mathtt{V}}^\eps_d . Z^{l_3} (\eps \partial_d ) \varphi .
3705: \end{eqnarray}
3706: %
3707: \begin{step2} 
3708: We deal with the terms of the form $(\ref{ff1})$.
3709: \end{step2}
3710:  %
3711:  %
3712: Using that $ \mathtt{v}^0_d = x_d \mathtt{v}^{0,\flat}_d$ and commuting $x_d$ with $Z^{l_3}$, 
3713: we have to deal with the terms of the form
3714: %
3715: \begin{eqnarray*}
3716: Z^{l_1} S(\underline{u}^\eps) . Z^{l_2} \mathtt{v}^0_d . Z^{l'_3} Z \varphi ,
3717: \quad \mathrm{with} \  l_3' \leqslant l_3 .
3718: \end{eqnarray*}
3719: %
3720: We use $L^\infty$ estimate of $\mathtt{v}^0$.
3721: Thus we have to control 
3722: %
3723: \begin{eqnarray}
3724: \label{mm}
3725: \lambda^{m-l} |Z^{l_1} S(\underline{u}^\eps) . Z^{l'_3} Z \varphi |_{0,\lambda,T} .
3726: \end{eqnarray}
3727: %
3728: We distinguish two cases:
3729:  %
3730: \begin{enumerate}[$\bullet$]
3731:  %
3732: \item if $l_1 \leqslant 1$, we control $Z^{l_1} S(\underline{u}^\eps)$ in $L^\infty$ norm by Lemma
3733: $\ref{Moser}$ and  the term $(\ref{mm})$ is bounded by $c. |\varphi |_{m,\lambda,T} $. 
3734: 
3735: \item if $l_1 \geqslant 2$, we apply Lemma $\ref{Moser2}$ with 
3736: %
3737: \begin{eqnarray*}
3738: \alpha := Z^2 S(\underline{u}^\eps) ,\ \beta :=  Z \varphi  ,\ \underline{m} := m-2.  
3739: \end{eqnarray*}
3740: %
3741: %
3742: \end{enumerate}
3743:  %
3744: \begin{step2} 
3745: We deal with the terms of the form $(\ref{ff2})$.
3746: \end{step2}
3747:  %
3748:   %
3749: We distinguish two cases:
3750:  %
3751: \begin{enumerate}[$\bullet$]
3752: 
3753: \item if $l_1 \leqslant 1$, we control $Z^{l_1} S(\underline{u}^\eps )$ in $L^\infty$ norm and then we
3754: deal with 
3755: %
3756: \begin{eqnarray*}
3757: \lambda^{m-l} |Z^{l_2} \underline{\mathtt{V}}^\eps_d .  Z^{l_3} (\eps \partial_d ) \varphi |_{0,\lambda,T} .
3758: \end{eqnarray*}
3759: %
3760: with $l_2 + l_3 \leqslant l$. We distinguish two sub cases:
3761:  %
3762: \begin{enumerate}[$(a)$]
3763:  %
3764: \item if $l_2 \geqslant 2$, we introduce $l_2' := l_2 -2$.
3765: Because $\underline{\mathtt{V}}^\eps_d = \underline{\mathtt{V}}^\eps_{a,d} + \eps^M \, \underline{\mathtt{V}}^\eps_{R,d}$,
3766: we study the terms 
3767: %
3768: \begin{eqnarray}
3769: \label{AAA1}
3770: \lambda^{m-l} | Z^{l'_2} (Z^2 \mathtt{V}^\eps_{a,d} ) . Z^{l_3} (\eps  \partial_d ) \varphi |_{0,\lambda,T} ,
3771: \\ \label{AAA2}
3772: \eps^M \lambda^{m-l} | Z^{l'_2} (Z^2 \underline{\mathtt{V}}^\eps_{R,d} ) . Z^{l_3} (\eps  \partial_d ) \varphi |_{0,\lambda,T} .
3773: \end{eqnarray}
3774: %
3775: We use the $L^\infty$ estimates of $(Z^2 \mathtt{V}^\eps_{a,d} )$ to control the term $(\ref{AAA1})$.
3776: To control the term $(\ref{AAA2})$, we apply Lemma $\ref{Moser2}$ with 
3777: %
3778: \begin{eqnarray*}
3779: \alpha := Z^2 \underline{\mathtt{V}}^\eps_{R,d} ,\ \beta :=  \eps \partial_d  \varphi  ,\ \underline{m} := m-2.  
3780: \end{eqnarray*}
3781: %
3782: %
3783: \item if $l_2 \leqslant 1$, we use the equality $\underline{\mathtt{V}}^\eps_{d} = x_d \underline{\mathtt{V}}^{\eps,\flat}_{d}$.
3784:  We commute the factor $x_d$ with the derivative $ Z^{l_3}$. 
3785:  Thus, to control the term $(\ref{AAA2})$, we study terms of the form
3786: %
3787: \begin{eqnarray*} 
3788: \lambda^{m-l} |  Z^{l_2} (\eps \underline{\mathtt{V}}^{\eps,\flat}_{d} ) .  Z^{l'_3} (x_d  \partial_d ) \varphi |_{0,\lambda,T} ,
3789: \end{eqnarray*}
3790: %
3791: with $l'_3 \leqslant l_3$.
3792: We control $Z^{l_2}  (\eps \underline{\mathtt{V}}^{\eps,\flat}_{d})$ in $L^\infty$ norm by 
3793: $ || \underline{\mathtt{V}}^\eps_{R,d} ||^*_{\eps,T}$. 
3794:  %
3795:  %
3796: \end{enumerate}
3797:  %
3798: \item if $l_1 \geqslant 2$, we apply twice  the Moser inequality of Lemma $\ref{Moser2}$ with  
3799: %
3800: \begin{eqnarray*} 
3801: \alpha := Z^2 S(\underline{u}^\eps ) ,\ \beta := \underline{\mathtt{V}}^{\eps}_{d} ,\ \gamma := \eps \partial_d \varphi ,\ \underline{m} := m-2. 
3802: \end{eqnarray*}
3803: %
3804: \end{enumerate}
3805: %
3806: \end{proof}
3807:  %
3808: We now give the proof of $a)$. 
3809:  %
3810: \begin{proof}[Proof of $a)$]
3811: %
3812: To explain in a clearer way our method, we will deal with the commutator
3813: %
3814: \begin{eqnarray}
3815: \label{jj}
3816: [\underline{\mathtt{v}}^\eps_d \partial_d , (\eps \partial_d )^k Z^l ] \varphi 
3817: \end{eqnarray}
3818: %
3819: instead of 
3820: %
3821: \begin{eqnarray*}
3822: [S(\underline{u}^\eps).\underline{\mathtt{v}}^\eps_d \partial_d , (\eps \partial_d )^k Z^l ] \varphi .
3823: \end{eqnarray*}
3824: %
3825: This avoids heavy notations and does not change the mathematical analysis.
3826: The commutator $(\ref{jj})$ is a linear combination of some
3827: terms of the form 
3828: %
3829: \begin{eqnarray}
3830: \label{tf1} &(\eps \partial_d )^k Z^{l_1} \underline{\mathtt{v}}^\eps_d . Z^{l_2} \partial_d  \varphi , 
3831:  \quad &\mathrm{with} \ \mathrm{l_1 \neq 0} \ , l_1 + l_2 = l,
3832: \\ \label{tf2} &(\eps \partial_d )^{k_1} \underline{\mathtt{v}}^\eps_d .  (\eps \partial_d )^{k_2} Z^{l} \partial_d  \varphi ,
3833:   \quad &\mathrm{with} \ \mathrm{k_1 \neq 0} \ ,  k_1 + k_2 = k.
3834: \end{eqnarray}
3835: %
3836: We look at the terms of the form $(\ref{tf1})$. 
3837: Because $\underline{\mathtt{v}}^\eps$ is of the form $\underline{\mathtt{v}}^\eps =  \mathtt{v}^0  + \eps
3838: \underline{\mathtt{V}}^\eps  $, it suffices to deal with the terms 
3839: %
3840: \begin{eqnarray}
3841: \label{rf1} & (\eps \partial_d )^k Z^{l_1} \mathtt{v}^0_d . Z^{l_2} \partial_d  \varphi , \quad  &\mathrm{with} \ \mathrm{l_1 \neq 0} \ , l_1 + l_2 = l,
3842: \\ \label{rf2} &(\eps \partial_d )^k Z^{l_1}  \underline{\mathtt{V}}^\eps_d .  Z^{l_2}  (\eps \partial_d ) \varphi 
3843:  \quad  &\mathrm{with} \ \mathrm{l_1 \neq 0} \ , l_1 + l_2 = l .
3844: \end{eqnarray}
3845: %
3846: We begin with the terms of the form $(\ref{rf1})$. 
3847: Using that $ \mathtt{v}^0_d = x_d \mathtt{v}^{0,\flat}_d$ and commuting $x_d$ with $Z^{l_2}$, 
3848: we have to deal with the terms of the form
3849: %
3850: \begin{eqnarray*}
3851: (\eps \partial_d )^k Z^{l_1} \mathtt{v}^{0,\flat}_d . Z^{l'_2} (\eps \partial) \varphi 
3852: \quad \mathrm{with} \ l_2' \leqslant l_2 .
3853: \end{eqnarray*}
3854: %
3855: Using $L^\infty$ estimate of $\mathtt{v}^0$ yields the result.
3856: We now have a look for the terms of the form $(\ref{rf2})$. 
3857: We distinguish three cases:
3858: 
3859: \begin{enumerate}[$\bullet$]
3860: 
3861: \item if $k \neq 0$ then we write $k=k'+1$ and we commute $\eps \partial_d$ with $Z^{l_1}$. 
3862: This yields some terms of the form 
3863: %
3864: \begin{eqnarray*}
3865: (\eps \partial_d )^{k'} Z^{l'}  (\eps \partial_d \underline{\mathtt{V}}^\eps_d ) Z^{l_2} \eps \partial_d \varphi 
3866: \end{eqnarray*}
3867: %
3868: with $l' \leqslant l_1$. 
3869: We use that $\underline{\mathtt{V}}^\eps_d = \underline{\mathtt{V}}^\eps_{a,d} + \eps^M \, \underline{\mathtt{V}}^\eps_{R,d}$ 
3870: and study the terms 
3871: %
3872: \begin{eqnarray}
3873: \label{Ai}
3874: \lambda^{m-2k-l} |   (\eps \partial_d )^{k'} Z^{l'}  (\eps \partial_d \underline{\mathtt{V}}^\eps_{a,d} ) Z^{l_2} \eps \partial_d
3875: \varphi  |_{0,\lambda,T} ,
3876: \\  \label{Bi}
3877: \lambda^{m-2k-l} | \eps^M \, (\eps \partial_d )^{k'} Z^{l'}  (\eps \partial_d \underline{\mathtt{V}}^\eps_{R,d} ) Z^{l_2} \eps
3878: \partial_d \varphi  |_{0,\lambda,T} .
3879: \end{eqnarray}
3880: %
3881: Using the $L^\infty$ estimates $(\ref{linf})$, we get
3882: $ (\ref{Ai}) \leqslant C  | \varphi |^\mathbf{E}_{\eps,m,\lambda,T} $.
3883: We apply Lemma $\ref{Moser2}$ $(2)$ to 
3884: %
3885: \begin{eqnarray*}
3886: \alpha := \eps \partial_d \mathtt{V}^\eps_{d,R} , \ \beta := \eps \partial_d  \varphi , \ \underline{m} =m-2
3887: \end{eqnarray*}
3888: %
3889:  to bound $(\ref{Bi})$ by 
3890: %
3891: \begin{eqnarray}
3892: \label{AM}
3893: C \eps^M (|\varphi|^\mathbf{E}_{\eps,m,\lambda,T} + ||\varphi ||_{\eps,Lip,T} .|\underline{U}^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T}). 
3894: \end{eqnarray}
3895: %
3896: %
3897: \item if $k=0$ and $l_1 \geqslant 2$ then with $l_1 := l_1' + 2$, the term $(\ref{rf2})$ can be rewritten 
3898: %
3899: \begin{eqnarray*}
3900: (\eps \partial_d )^{k} Z^{l'_1  }  (Z^{2} \underline{\mathtt{V}}^\eps_d ). Z^{l_2} (\eps  \partial_d ) \varphi .
3901: \end{eqnarray*}
3902: %
3903: We use that $\underline{\mathtt{V}}^\eps_d = \mathtt{V}^\eps_{a,d} + \eps^M \underline{\mathtt{V}}^\eps_{R,d}$ and study
3904: the terms 
3905: %
3906: \begin{eqnarray}
3907: \label{Aj}
3908: \lambda^{m-2k-l} |(\eps \partial_d )^{k} Z^{l'_1}  (Z^{2} \mathtt{V}^\eps_{a,d} ). Z^{l_2} (\eps  \partial_d ) \varphi |_{0,\lambda,T},
3909: \\  \label{Bj}
3910: \lambda^{m-2k-l} |\eps^M  (\eps \partial_d )^{k} Z^{l'_1}  (Z^{2} \underline{\mathtt{V}}^\eps_{R,d} ). Z^{l_2} (\eps  \partial_d )
3911: \varphi |_{0,\lambda,T}.
3912: \end{eqnarray}
3913: %
3914: We use the $L^\infty$ estimates $(\ref{linf})$ to bound the term  $(\ref{Aj})$. 
3915: We apply Lemma $\ref{Moser2}$ with 
3916: %
3917: \begin{eqnarray*}
3918: \alpha := Z^2 \underline{V}^\eps_{R,d} ,\ \beta := \eps \partial_d \varphi ,\ \underline{m} = m-2
3919: \end{eqnarray*}
3920: %
3921: %
3922:  to bound $(\ref{Bj})$ by $(\ref{AM})$.
3923:  %                 \\  
3924: \item if $k=0$ and $l_1 =1$, using that $\underline{\mathtt{V}}^\eps_d = x_d \underline{\mathtt{V}}^{\eps,\flat}_d $ 
3925: and commuting $x_d$ with
3926: $Z^{l_2}$ yield a sum of terms of the form 
3927: %
3928: \begin{eqnarray*}
3929: Z^{l_1} \eps \underline{\mathtt{V}}^{\eps,\flat}_d .Z^{l_2} (x_d  \partial_d \varphi )  .
3930: \end{eqnarray*}
3931: %
3932: We bound it by $|| \eps \underline{\mathtt{V}}^{\eps,\flat}_d  ||_{1,T} . |\varphi |_{m,\lambda,T}$ and conclude with $|| \eps
3933: \underline{\mathtt{V}}^{\eps,\flat}_d  ||_{1,T} \leqslant || \underline{\mathtt{V}}^{\eps}_d ||^*_{\eps,T} $. 
3934: We can deal with the terms $(\ref{tf2})$ with the same methods.
3935: %
3936: \end{enumerate}
3937:  For the proof of $b)$, proceed in the same way and substitute Lemma $\ref{Moser2}$ $(2)$ by Lemma $\ref{Moser2}$ $(1)$.
3938: %
3939: \end{proof}
3940:  %
3941: \subsubsection{Source term estimates}
3942: 
3943: We now give two lemmas which will be useful when estimating source terms.
3944: 
3945: \begin{lem}
3946: \label{poire}
3947: Let $m$ be an integer. 
3948: There is an increasing function $C:\R_+ \rightarrow \R_+^*$ such that  for all $\eps \in ]0,1]$, 
3949: if $\eps^M \, || \underline{U}^\eps_R ||_{0,T}
3950: \leqslant  \mathfrak{R}$ then for all $\varphi \in C^\infty_0 ((0,T) \times \overline{\Omega} )$,
3951:  for all $\lambda \geqslant \lambda_0$,
3952: 
3953: \begin{enumerate}[$a)$]
3954: 
3955: \item for any $m$, 
3956: %
3957: \begin{eqnarray*}
3958: |  \underline{J}. \varphi  |_{m,\lambda,T}
3959: \leqslant C(\mathfrak{R}) (|\varphi  |_{ m,\lambda,T} 
3960: + \eps^M ||\varphi  ||_{0,T} .|\underline{U}^\eps_R |_{m,\lambda,T}),
3961: \end{eqnarray*}
3962: %
3963: \item for $m$ even,
3964: %
3965: \begin{eqnarray*}
3966: |  \underline{J}. \varphi   |^\mathbf{E}_{\eps,m,\lambda,T}
3967: \leqslant  C(\mathfrak{R}) (  |\varphi |^\mathbf{E}_{\eps,m,\lambda,T} 
3968: +  \eps^M || \varphi ||_{0,T} . | \underline{U}^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} ).
3969: \end{eqnarray*}
3970: \end{enumerate}
3971:  %
3972: where $\underline{J}$ stands for $\underline{J}^\alpha $, $\underline{J}^{\beta,1}$ or $\underline{J}^{\beta,2}$.
3973:  %
3974: \end{lem}
3975: 
3976: 
3977: The proof of Lemma $\ref{poire}$ is straightforward and mainly lies on  Lemma $\ref{Moser2}$. 
3978: It is left to the reader.
3979: 
3980: \begin{lem}
3981: \label{carotte}
3982:  %
3983: Let $m$ be an integer. 
3984: There is an increasing function $C:\R_+ \rightarrow \R_+^*$ such that  for all $\eps \in ]0,1]$, 
3985: if $\eps^M \, || \underline{U}^\eps_R ||^*_{\eps,T}
3986: \leqslant  \mathfrak{R}$ then  for all $\lambda \geqslant \lambda_0$,
3987:  %
3988: \begin{enumerate}[$a)$]
3989:  %
3990: \item for any $m$, 
3991: %
3992: \begin{eqnarray*}
3993: | \underline{\mathtt{v}}^\eps_d \partial_d  U^\eps_R |_{m-1,\lambda,T}
3994: \leqslant C(\mathfrak{R}) (|U^\eps_R  |^\mathcal{N}_{\eps, m,\lambda,T} 
3995: + \eps^M || U^\eps_R  ||^*_{\eps,T} .|\underline{\mathtt{v}}^\eps_R |^\mathcal{N}_{\eps,m,\lambda,T}),
3996: \end{eqnarray*}
3997: %
3998: \item for $m$ even,
3999: %
4000: \begin{eqnarray*}
4001: | \underline{\mathtt{v}}^\eps_d \partial_d  U^\eps_R |^\mathbf{E}_{\eps,m-1,\lambda,T}
4002: \leqslant  C(\mathfrak{R}) (  |U^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} 
4003: +  \eps^M || U^\eps_R ||_{\eps,Lip} . | \underline{\mathtt{v}}^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} ).
4004: \end{eqnarray*}
4005: %
4006: \end{enumerate}
4007: \end{lem}
4008: 
4009: 
4010: \begin{proof}[Proof of $a)$]
4011: %
4012: We will proceed in four steps.
4013:  %
4014: \begin{enumerate}[$(i)$]
4015:  %
4016: \item 
4017:  %
4018: Because $ \underline{\mathtt{v}}^\eps_d =  \underline{\mathtt{v}}^0_d + \eps \underline{\mathtt{V}}^\eps_d $ (cf. $(\ref{linform})$)
4019:  and $\mathtt{v}^0_d = x_d \mathtt{v}^{0,\flat}_d$, we get  
4020: %
4021: \begin{eqnarray*}
4022: |\underline{\mathtt{v}}^\eps_d .\partial_d U^\eps_R |_{m-1,\lambda,T}
4023: \leqslant C | U^\eps_R |_{m,\lambda,T} 
4024: + |\underline{\mathtt{V}}^\eps_d .(\eps \partial_d ) U^\eps_R |_{m-1,\lambda,T} .
4025: \end{eqnarray*}
4026: %
4027: \item 
4028:  %
4029: In order to estimate $ |\underline{\mathtt{V}}^\eps_d .(\eps \partial_d ) U^\eps_R |_{m-1,\lambda,T}$, 
4030: we will control the terms of the form
4031: %
4032: \begin{eqnarray}
4033: \label{cordf}
4034: \lambda^{m-1-l} |(  Z^{l_1} \underline{\mathtt{V}}^\eps_d ). 
4035:  Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} ,
4036: \end{eqnarray}
4037: %
4038: where $l_1, l_2 \in \N, \ l_1 + l_2 =l \leqslant m-1 $.
4039: Because $\underline{V}^\eps = V^\eps_a + \eps^M \, \underline{V}^\eps_R$, 
4040: it suffices to deal with the terms of the form 
4041: %
4042: \begin{eqnarray}
4043: \label{cordfa}
4044: \lambda^{m-1-l} |{ Z^{l_1} \mathtt{V}^\eps_{a,d} }. 
4045:  Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} ,
4046: \\ \label{cordfb}
4047: \lambda^{m-1-l} |{  Z^{l_1} \underline{\mathtt{V}}^\eps_{R,d} }. 
4048: Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} .
4049: \end{eqnarray}
4050: %
4051: \item 
4052:  %
4053: We begin with the term $(\ref{cordfa})$. We distinguish two cases:
4054: \\
4055: \begin{enumerate}[$\bullet$]
4056:  %
4057: \item If $l_1 > 0$, then  $l_2 \leqslant m$ we use the $L^\infty$ estimates $(\ref{linf})$
4058:  and bound the term $(\ref{cordfa})$ by 
4059:  %
4060: \begin{eqnarray*}
4061:  c.| (\eps \partial_d ) U^\eps_R |_{m-2,\lambda,T}  \leqslant  c.| U^\eps_R |_{m,\lambda,T} .
4062: \end{eqnarray*}
4063: %
4064: %
4065: \item If $l_1 =0$, then we write $\mathtt{V}^\eps_{a,d}  = x_d . \underline{\mathtt{V}}^{\eps,\flat}_{a,d} $ 
4066: and commute $x_d$ with $ Z^{l_2}$ to find that $(\ref{cordfa})$ is a sum of terms of the form
4067: %
4068: \begin{eqnarray}
4069: \label{cordfa3}
4070: \lambda^{m-1-l} |\eps \underline{\mathtt{V}}^{\eps,\flat}_{a,d}  Z^{l'_2} U^\eps_R |_{0,\lambda,T} .
4071: \end{eqnarray}
4072: %
4073: with $l'_2 \leqslant l_2 + 1$. 
4074: Thus 
4075: %
4076: \begin{eqnarray*}
4077: (\ref{cordfa3}) \leqslant C || \eps \underline{\mathtt{V}}^{\eps,\flat}_{a,d} ||_\infty . | U^\eps_R |_{m,\lambda,T }
4078: \leqslant C  || \mathtt{V}^{\eps}_{a,d} ||_{\eps,T}^*  . | U^\eps_R |_{m,\lambda,T } . 
4079:  \end{eqnarray*}
4080: %
4081:     \end{enumerate}
4082:  %
4083: \item 
4084: 
4085: We now look at the term $(\ref{cordfb})$. We distinguish three cases:
4086: \\
4087: \begin{enumerate}[$\bullet$]
4088:  %
4089: \item If $l_1 \neq 0$, we apply Lemma $\ref{Moser2}$ $(1)$ to 
4090: %
4091: \begin{eqnarray*}
4092: \alpha := Z \underline{\mathtt{V}}^\eps_{R,d} , \
4093:  \beta := (\eps \partial_d ) U^\eps_R , \ \underline{m} := m-2. 
4094:  \end{eqnarray*}
4095: %
4096: The term $(\ref{cordfb})$ is bounded by 
4097: %
4098: \begin{eqnarray*}
4099: \label{cordf1}
4100:  c. \{ || \underline{\mathtt{V}}^\eps_{R,d} ||_{1,T} .| U^\eps_R |^\mathcal{N}_{\eps,m,\lambda,T} 
4101: + | \underline{\mathtt{V}}^\eps_{R,d} |_{m,\lambda,T} . || U^\eps_R ||_{\eps,Lip}  \}.
4102: \end{eqnarray*}
4103: %
4104: \item If $l_1 =0$, we write $\underline{\mathtt{V}}^\eps_{R,d}  = x_d . \underline{\mathtt{V}}^{\eps,\flat}_{R,d} $ 
4105: and commute $x_d$ with $ Z^{l_2}$ to find that $(\ref{cordf})$ is a sum of terms of the form
4106: %
4107: \begin{eqnarray}
4108: \label{cordf3}
4109: \lambda^{m-1-l} |\eps \underline{\mathtt{V}}^{\eps,\flat}_{R,d} Z^{l'_2} U^\eps_R |_{0,\lambda,T} .
4110: \end{eqnarray}
4111: %
4112: with  $l'_2 \leqslant l_2 + 1$. 
4113: Thus the term $(\ref{cordf3})$ is bounded by 
4114: $ || \eps \underline{\mathtt{V}}^{\eps,\flat}_{R,d} || _\infty . |  U^\eps_R |_{ m,\lambda,T }$.   
4115:    %
4116:     \end{enumerate}
4117: \end{enumerate}
4118: \end{proof}
4119:  %
4120: The proof of $a)$ follows the same path and is slightly more complicated.
4121:  %
4122:  %
4123: \begin{proof}[Proof of $b)$]
4124: %
4125: We will proceed in four steps.
4126:  %
4127: \begin{enumerate}[$(i)$]
4128:  %
4129: \item 
4130:  %
4131: Because $ \underline{\mathtt{v}}^\eps_d =  \underline{\mathtt{v}}^0_d + \eps \underline{\mathtt{V}}^\eps_d $ (cf. $(\ref{linform})$)
4132:  and $\mathtt{v}^0_d = x_d \mathtt{v}^{0,\flat}_d$, we get 
4133: %
4134: \begin{eqnarray*}
4135: |\underline{\mathtt{v}}^\eps_d .\partial_d U^\eps_R |^\mathbf{E}_{m-1,\lambda,T}
4136: \leqslant C | U^\eps_R |^\mathbf{E}_{m,\lambda,T} 
4137: + |\underline{\mathtt{V}}^\eps_d .(\eps \partial_d ) U^\eps_R |^\mathbf{E}_{m-1,\lambda,T} .
4138: \end{eqnarray*}
4139: %
4140: \item 
4141:  %
4142: In order to estimate $ |\underline{\mathtt{V}}^\eps_d .(\eps \partial_d ) U^\eps_R |^\mathbf{E}_{\eps,m-1,\lambda,T}$, 
4143: we will control the terms of the form
4144: %
4145: \begin{eqnarray}
4146: \label{rdf}
4147: \lambda^{m-1-2k-l}  \, |{ (\eps \partial_d )^{k_1} Z^{l_1} \underline{\mathtt{V}}^\eps_d }. 
4148: (\eps \partial_d )^{k_2} Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} ,
4149: \end{eqnarray}
4150: %
4151: where 
4152: %
4153: \begin{eqnarray*}
4154: k_i, l_i \in \N, \ k_1 + k_2 = k, \ l_1 + l_2 =l, \ 2k+l \leqslant m-1 .
4155: \end{eqnarray*}
4156: %
4157: %
4158: Because $\underline{V}^\eps = V^\eps_a + \eps^M \, \underline{V}^\eps_R$, 
4159: it suffices to deal with the terms of the form 
4160: %
4161: \begin{eqnarray}
4162: \label{rdfa}
4163: \lambda^{m-1-2k-l}  \, |{ (\eps \partial_d )^{k_1} Z^{l_1} \mathtt{V}^\eps_{a,d} }. 
4164: (\eps \partial_d )^{k_2} Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} ,
4165: \\ \label{rdfb}
4166: \lambda^{m-1-2k-l}  \, |{ (\eps \partial_d )^{k_1} Z^{l_1} \underline{\mathtt{V}}^\eps_{R,d} }. 
4167: (\eps \partial_d )^{k_2} Z^{l_2} (\eps \partial_d ) U^\eps_R |_{0,\lambda,T} .
4168: \end{eqnarray}
4169: %
4170: \item 
4171:  %
4172: We begin with the term $(\ref{rdfa})$. We distinguish two cases:
4173: \\
4174: \begin{enumerate}[$\bullet$]
4175:  %
4176: \item If $l_1 + k_1 > 0$, then  $2k_2 + l_2 \leqslant m$ we use the $L^\infty$ estimates $(\ref{linf})$
4177:  and bound the term $(\ref{rdfa})$ by 
4178:  %
4179: \begin{eqnarray*}
4180:  c.| (\eps \partial_d ) U^\eps_R |^\mathbf{E}_{m-2,\lambda,T}  \leqslant  c.| U^\eps_R |^\mathbf{E}_{m,\lambda,T} .
4181: \end{eqnarray*}
4182: %
4183: %
4184: \item If $l_1 = k_1=0$, then we write $\mathtt{V}^\eps_{a,d}  = x_d . \underline{\mathtt{V}}^{\eps,\flat}_{a,d} $ 
4185: and commute $x_d$ with $(\eps \partial_d )^{k_2} Z^{l_2}$ to find that $(\ref{rdfa})$ is a sum of terms of the form
4186: %
4187: \begin{eqnarray}
4188: \label{rdfa3}
4189: \lambda^{m-1-2k-l} \,  |\eps \underline{\mathtt{V}}^{\eps,\flat}_{a,d} (\eps \partial_d )^{k'_2} Z^{l'_2} U^\eps_R |_{0,\lambda,T} .
4190: \end{eqnarray}
4191: %
4192: with $k'_2 \leqslant k_2$, $l'_2 \leqslant l_2 + 1$. 
4193: Thus 
4194: %
4195: \begin{eqnarray*}
4196: (\ref{rdfa3}) \leqslant C || \eps \underline{\mathtt{V}}^{\eps,\flat}_{a,d} ||_\infty . |  U^\eps_R |^\mathbf{E}_{m,\lambda,T }
4197: \leqslant C  || \mathtt{V}^{\eps}_{a,d} ||_{\eps,T}^*  . | U^\eps_R |^\mathbf{E}_{m,\lambda,T } . 
4198:  \end{eqnarray*}
4199: %
4200:     \end{enumerate}
4201:  %
4202: \item 
4203:  %
4204: We now look at the term $(\ref{rdfb})$. We distinguish three cases:
4205: \\
4206: \begin{enumerate}[$\bullet$]
4207: 
4208: \item If $l_1 \neq 0$, we apply Lemma $\ref{Moser2}$ $(2)$ to 
4209: %
4210: \begin{eqnarray*}
4211: \alpha := Z \underline{\mathtt{V}}^\eps_{R,d} , \
4212:  \beta := (\eps \partial_d ) U^\eps_R , \ \underline{m} := m-2. 
4213:  \end{eqnarray*}
4214: %
4215: The term $(\ref{rdfb})$ is bounded by 
4216: %
4217: \begin{eqnarray*}
4218: \label{rdf1}
4219:  c. \{ || \underline{\mathtt{V}}^\eps_{R,d} ||_{1,T} .| U^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} 
4220: + | \underline{\mathtt{V}}^\eps_{R,d} |^\mathbf{E}_{\eps,m,\lambda,T} . || U^\eps_R ||_{\eps,Lip}  \}.
4221: \end{eqnarray*}
4222: %
4223: \item If $l_1 =0$ and $ k_1 \neq 0$, we apply Lemma $\ref{Moser2}$ $(2)$ to 
4224: %
4225: \begin{eqnarray*}
4226: \alpha := \eps \partial_d \underline{\mathtt{V}}^\eps_{R,d} , \  \beta := \eps \partial_d U^\eps_R , \  \underline{m} := m-2. 
4227: \end{eqnarray*}
4228: %
4229: %
4230: The term $(\ref{rdfb})$ is bounded by 
4231: %
4232: \begin{eqnarray*}
4233: \label{rdf2}
4234: \lambda^{-1} c. \{ || \underline{\mathtt{V}}^\eps_{R,d} ||_{\eps,Lip,T} .| U^\eps_R |^\mathbf{E}_{\eps,m,\lambda,T} 
4235: + | \underline{\mathtt{V}}^\eps_{R,d} |^\mathbf{E}_{\eps,m,\lambda,T} . || U^\eps_R ||_{\eps,Lip} \}.
4236: \end{eqnarray*}
4237: %
4238: \item If $l_1 = k_1=0$, we write $\underline{\mathtt{V}}^\eps_{R,d}  = x_d . \underline{\mathtt{V}}^{\eps,\flat}_{R,d} $ 
4239: and commute $x_d$ with $(\eps \partial_d )^{k_2} Z^{l_2}$ to find that $(\ref{rdf})$ is a sum of terms of the form
4240: %
4241: \begin{eqnarray}
4242: \label{rdf3}
4243: \lambda^{m-1-2k-l} |\eps \underline{\mathtt{V}}^{\eps,\flat}_{R,d} (\eps \partial_d )^{k'_2} Z^{l'_2} U^\eps_R |_{0,\lambda,T} .
4244: \end{eqnarray}
4245: %
4246: with $k'_2 \leqslant k_2$, $l'_2 \leqslant l_2 + 1$. 
4247: Thus  the term $(\ref{rdf3})$ is bounded by 
4248: $ || \eps \underline{\mathtt{V}}^{\eps,\flat}_{R,d} || _\infty . |  U^\eps_R |^\mathbf{E}_{m,\lambda,T }$.   
4249:    %
4250:     \end{enumerate}
4251: \end{enumerate}
4252: \end{proof}
4253:  %
4254: We now attack the conormal estimates of $U^\eps_R$.
4255:  %
4256: \subsection{Conormal estimates}
4257: \label{conorm}
4258:  %
4259: We look for estimates of $\lambda^{m-l} | Z^l U^\eps_R |_{0,\lambda,T_0}$.
4260: As for $L^2$ estimates, we will proceed in three steps estimating 
4261: %
4262: \begin{eqnarray*}
4263: \lambda^{m-l} | Z^l W^\eps_R |_{0,\lambda,T_0 }, \ \lambda^{m-l} | Z^l
4264: (\frac{x_d}{\eps} W^\eps_R )|_{0,\lambda,T_0 } \ \mathrm{and} \ \lambda^{m-l} | Z^l V^\eps_R |_{0,\lambda,T_0 } .
4265: \end{eqnarray*}
4266: %
4267: \subsubsection{Estimate of $W^\eps_R$}
4268: \label{coW}
4269:  %
4270: We apply the derivative $Z^l$ to the equation $\X_{\underline{v}^{\eps}} W^\eps_R = f^\eps_I $.
4271: We get $ \X_{\underline{v}^{\eps}} (Z^l W^\eps_R ) = \tilde{f}^\eps_I $, where $\tilde{f}^\eps_I  := [\X_{\underline{\mathtt{v}}^{\eps}}
4272: ,Z^l ] W^\eps_R + Z^l f^\eps_I $.
4273:  Then we estimate $\tilde{f}^\eps_I $ thanks to Lemma $\ref{Moser2}$ 
4274:   and use a classic $L^2$ estimate.
4275:   %
4276:  \subsubsection{Estimate of $\frac{x_d}{\eps}  W^\eps_R$}
4277: \label{coXW}
4278:  %
4279:  Because the control of the commutator $[ \underline{\mathtt{v}}^{\eps,\flat}_d \frac{x_d}{\eps} ,Z^l ] W^\eps_R $ seems difficult,
4280:  we write 
4281:  $\underline{\mathtt{v}}^{\eps,\flat}_d  \frac{x_d}{\eps} W^\eps_R  =
4282:   \underline{\mathtt{v}}^{0}_d \frac{x_d}{\eps} W^\eps_R + \underline{\mathtt{V}}_d^\eps. W^\eps_R  $. 
4283:  Thus   $\frac{x_d}{\eps}  W^\eps_R$ verifies the equation
4284:  %
4285:  \begin{eqnarray*}
4286: \label{coco}
4287:  \X_{\underline{v}^{\eps}} (\frac{x_d}{\eps}  W^\eps_R ) = 
4288:  \underline{\mathtt{v}}^{0,\flat}_d \frac{x_d}{\eps}  W^\eps_R +  \underline{\mathtt{V}}_d^\eps .
4289:  W^\eps_R + f^{\eps}_{II} .
4290: \end{eqnarray*}
4291: %
4292: We apply the derivative $Z^l$ to this equation and find for $Z^l (\frac{x_d}{\eps}  W^\eps_R )$ the equation
4293: %
4294: \begin{eqnarray*}
4295: ( \X_{\underline{\mathtt{v}}^{\eps}} -  \underline{\mathtt{v}}^{0,\flat}_d ) Z^l (\frac{x_d}{\eps}  W^\eps_R ) = \tilde{f}^{\eps}_{II} ,
4296: \end{eqnarray*}
4297: %
4298: where 
4299: %
4300: \begin{eqnarray*}
4301: \tilde{f}^{\eps}_{II} := [ \X_{\underline{\mathtt{v}}^{\eps}} -  \underline{\mathtt{v}}^{0,\flat}_d  , Z^l ] \frac{x_d}{\eps}  W^\eps_R 
4302: + Z^l ( \underline{\mathtt{V}}_d^\eps . W^\eps_R ) +  Z^l f^{\eps}_{II} .
4303: \end{eqnarray*}
4304: %
4305: %
4306: We estimate $[ \X_{\underline{\mathtt{v}}^{\eps}}  -  \underline{\mathtt{v}}^{0,\flat}_d ,  Z^l  ] (\frac{x_d}{\eps}  W^\eps_R )$ 
4307: thanks to Lemma $\ref{Moser2}$ 
4308: and $Z^l (  \underline{\mathtt{V}}_d^\eps . W^\eps_R )$ with Lemma $\ref{Moser2}$ and the previous subsection. 
4309: We finally use a classic $L^2$ estimate.
4310:  %
4311: \subsubsection{Estimate of $V^\eps_R$}
4312: \label{coV}
4313:  %
4314: We apply the derivative $Z^l$ to the boundary value problem $(\ref{Sx})$ and find that $Z^l V^\eps_R$ 
4315: is solution of the following boundary value problem:
4316: %
4317: \begin{eqnarray*}
4318: \left\{
4319: \begin{array}{ccc}
4320: &( \mathfrak{L}^\star (\underline{u}^\eps ,\partial)  + 
4321: \underline{J}^\alpha ).(Z^l V^\eps_R )=
4322: \tilde{f}^\eps_{III} \quad &\mathrm{when} \ x_d > 0 ,
4323: \\ &Z^l V^\eps_{R,d} = 0  \quad &\mathrm{when} \ x_d = 0 ,
4324: \end{array}
4325: \right.
4326: \end{eqnarray*}
4327:  %
4328:  where 
4329:  %
4330:  \begin{eqnarray*}
4331:   \tilde{f}^\eps_{III} := 
4332:  [S^\star (\underline{u}^\eps ) \X_{\underline{\mathtt{v}}^\eps}  + \L^\star (\partial_x ) + \underline{J}^\alpha , Z^l ] V^\eps_R +   Z^l
4333: f^\eps_{III}.
4334: \end{eqnarray*}
4335: %
4336: We estimate the commutator $[S^\star (\underline{u}^\eps ) \X_{\underline{\mathtt{v}}^\eps}  , Z^l ] V^\eps_R $ thanks to Proposition
4337: $\ref{lapin}$ and the commutator $[ \underline{J}^\alpha , Z^l ] V^\eps_R $ thanks to lemma  $\ref{Moser2}$. 
4338: We estimate $Z^l f^\eps_{III}$ thanks to the
4339: subsection $\ref{coW}$ and $\ref{coXW}$. 
4340: We now sketch the method to estimate $[\L^\star (\partial_x ), Z^l ] V^\eps_R $.
4341: First notice that $ [\L^\star (\partial_x ), Z^l ] V^\eps_R = [\L^\star_d \partial_d , Z^l ] V^\eps_R$ is a sum of terms of the form $Z^{l'}
4342: \L^\star_d \partial_d V^\eps_R $  with $l' \leqslant l$.
4343: Then we extirpate $\L^\star_d \partial_d V^\eps_R $  by the first equation of $(\ref{Sx})$.
4344: We can apply the previous methods for the resulting terms.
4345: We use Lemma $\ref{carotte}$ to handle the most delicate point i.e. the estimate of $| \underline{\mathtt{v}}_d^\eps . \partial_d U^\eps_R |_{m-1,\lambda,T_0}$.
4346: This supplies a good estimate of $[\L^\star (\partial_x ), Z^l ] V^\eps_R $.
4347: We conclude as in subsection $\ref{estiV}$ with a $L^2$ estimate.
4348:  %
4349:  %
4350: \subsection{Normal estimates}
4351: \label{norm}
4352: 
4353: We look for estimates of $\lambda^{m-2k-l} | (\eps \partial)^k Z^l U^\eps_R |_{0,\lambda,T}$, with $k>0$ and $2k+l \leqslant m$. 
4354: We begin to estimate for the noncharacteristic part: $\L_d^\star  V^\eps_{R}$.
4355: This corresponds to the pressure $p^\eps_R$ and $\mathtt{v}^\eps_{R,d}$. 
4356: Then we will estimate   $| W^\eps_{R} |^\mathbf{E}_{\eps,m,\lambda,T_0}$,   $\frac{x_d}{\eps}  | W^\eps_{R} |^\mathbf{E}_{\eps,m,\lambda,T_0}$ and finally
4357:  $| \mathtt{v}^\eps_{R,y} |^\mathbf{E}_{\eps,m,\lambda,T_0}$ where we denote by $\mathtt{v}_y$ the tangential velocity.
4358:  %
4359:  %
4360: \subsubsection{Normal estimates of $L_d^\star V^\eps_R$.}
4361: \label{norm1}
4362:  %
4363: We notice that 
4364: %
4365:  \begin{eqnarray*}
4366: | \L_d^\star \eps \partial_d V^\eps_R |^\mathbf{E}_{\eps,m-2,\lambda,T} \leqslant 
4367: \lambda^{-1} | \L_d^\star \eps \partial_d V^\eps_R |^\mathbf{E}_{\eps,m-1,\lambda,T}
4368: \end{eqnarray*}
4369:  %
4370: %
4371:  and extirpate $\L_d^\star \eps \partial_d V^\eps_R$  from the equation $(\ref{AL})$. 
4372: Among the resulting terms, the most delicate one to estimate is $| \underline{\mathtt{v}}^\eps_d . \partial_d 
4373: U^\eps_R|^\mathbf{E}_{\eps,m-1,\lambda,T_0} $ which is given by Lemma  $\ref{carotte}$.
4374: Notice that there is no problem with the term $\underline{J}^{\beta,2} \frac{x_d}{\eps} W^\eps_R $.
4375: 
4376: \subsubsection{Normal estimates of $W^\eps_R$ .}
4377: \label{norm2}
4378: 
4379: 
4380:  We apply the derivative $( \eps \partial_d )^k Z^l $ to the equation $\X_{\underline{\mathtt{v}}^{\eps}} W^\eps_R = f^\eps_I $.
4381:  We find that $( \eps \partial_d )^k Z^l  W^\eps_R $ verify the following equation: 
4382:  %
4383:  \begin{eqnarray*}
4384:  \X_{\underline{\mathtt{v}}^{\eps}}  (\eps \partial_d
4385:  )^k Z^l W^\eps_R = \hat{f}^\eps_I 
4386:   \end{eqnarray*}
4387:  %
4388:  %
4389:   where 
4390:  %
4391:   \begin{eqnarray*}
4392:  \hat{f}^\eps_I := [ \X_{\underline{\mathtt{v}}^{\eps}} , ( \eps \partial_d )^k Z^l ] W^\eps_R + ( \eps \partial_d )^k Z^l  f^\eps_I .
4393:   \end{eqnarray*}
4394:  %
4395:  We estimate the commutator $[ \X_{\underline{\mathtt{v}}^{\eps}} , ( \eps \partial_d )^k Z^l ] W^\eps_R $ thanks to Proposition $a)$ and
4396:  $c)$ and use a $L^2$ estimate.
4397:  
4398:  \subsubsection{Normal estimates of $ \frac{x_d}{\eps} W^\eps_R$ .}
4399: \label{norm3}
4400: 
4401:  We apply the derivative $( \eps \partial_d )^k Z^l $ to the equation $(\ref{coco})$ to find for $\hat{W}^\eps_R := ( \eps
4402:  \partial_d )^k Z^l ( \frac{x_d}{\eps} W^\eps_R )$ the equation $(\X_{\underline{\mathtt{v}}^\eps} - \mathtt{v}^{0,\flat}_d )
4403:  \hat{W}^\eps_R = \hat{f}^\eps_{III}$ where 
4404:  %
4405:  \begin{eqnarray*}
4406:  \hat{f}^\eps_{III} := [ \X_{\underline{\mathtt{v}}^\eps} - \mathtt{v}^{0,\flat}_d , ( \eps
4407:  \partial_d )^k Z^l ] \hat{W}^\eps_R + ( \eps \partial_d )^k Z^l ( \underline{\mathtt{V}}^\eps_d . W^\eps_R )
4408:  +  ( \eps \partial_d )^k Z^l f^\eps_{III} .
4409:  \end{eqnarray*}
4410:  %
4411:  We estimate the commutator $[ X_{\underline{\mathtt{v}}^\eps} , ( \eps \partial_d )^k Z^l ] \hat{W}^\eps_R$ thanks to 
4412:  Proposition $\ref{lapin}$, the second term of the right member thanks to Lemma $\ref{Moser2}$ and the previous subsection. 
4413:  We estimate $\hat{f}^\eps_{III}$ using that because $\beta_2$ is rapidly decreasing with respect to its last argument, 
4414:  all the derivatives of
4415:  $\frac{x_d}{\eps} \beta_2 (\eps,t,x,\frac{x_d}{\eps})$ are bounded in $L^\infty$ uniformly with respect to $\eps$ 
4416:  and $(f^\eps_{III} ) \in \mathbf{E}^m$.
4417:  We end with a $L^2$ estimate.
4418:  
4419:   \subsubsection{Normal estimates of $\mathtt{v}^\eps_{R,y} $}
4420:  \label{norm4}
4421: 
4422:  Looking at the first $d-1$ equations of $(\ref{AL})$, we find for $\mathtt{v}^\eps_{R,y}$ an equation of the form 
4423:  %
4424:  \begin{eqnarray*}
4425:  (\underline{\rho}^\eps .  \X_{\underline{\mathtt{v}}^\eps}  + \underline{J}_1^\alpha ) . \mathtt{v}^\eps_{R,y} =
4426:  f^\eps_{IV}
4427:   \end{eqnarray*}
4428:  %
4429:  where 
4430:  %
4431:  \begin{eqnarray*}
4432:  f^\eps_{IV} := - \underline{J}_2^\alpha . \mathtt{v}^\eps_{R,d} - \nabla_y p^\eps 
4433:  - \underline{J}^{\beta,1} \, W^\eps_R - \underline{J}^{\beta,2} \, \frac{x_d}{\eps}) W^\eps_R  
4434:  - R^\eps 
4435:   \end{eqnarray*}
4436:  %
4437:  and $\underline{J}_1^\alpha $, $\underline{J}_2^\alpha$ are some matrices extracted from $\underline{J}^\alpha$.
4438:  We apply the derivative $(\eps \partial_d )^k Z^l$ to this equation and find for 
4439:  $\hat{\mathtt{v}}^\eps_{R,y} := (\eps \partial_d )^k Z^l \mathtt{v}^\eps_{R,y}$ the equation 
4440:  %
4441:   \begin{eqnarray*}
4442:  ( \underline{\rho}^\eps  X_{\underline{\mathtt{v}}^\eps} + \underline{J}_1^\alpha ). \hat{\mathtt{v}}^\eps_{R,y} =
4443:  \hat{f}^\eps_{IV} 
4444:  \end{eqnarray*}
4445:  %
4446:   where 
4447:  %
4448:   \begin{eqnarray*}
4449:  \hat{f}^\eps_{IV} := [ \underline{\rho}^\eps  \X_{\underline{\mathtt{v}}^\eps} +
4450:  \underline{J}_1^\alpha  , (\eps \partial_d )^k  Z^l ] v^\eps_{R,y} + (\eps \partial_d )^k  Z^l f^\eps_{IV} .
4451:   \end{eqnarray*}
4452:  %
4453:  We estimate the commutator $[\underline{\rho}^\eps  \X_{\underline{\mathtt{v}}^\eps} , (\eps \partial_d )^k  Z^l ] v^\eps_{R,y}$
4454:  thanks to Proposition $\ref{lapin}$, the commutator $[\underline{J}_1^\alpha , (\eps \partial_d )^k  Z^l  ]  v^\eps_{R,y} $
4455:  thanks to Lemma $\ref{Moser2}$. 
4456:  When looking at the term $(\eps \partial_d )^k Z^l f^\eps_{IV}$, the contributions of $- \underline{J}_2^\alpha
4457:  .\underline{\mathtt{v}}^\eps_{R,d}$ and of $\underline{R}^\eps$ are easy to estimate.
4458:  We estimate the contribution of $\underline{J}^{\beta,1}  \, W^\eps_R$ and  $\underline{J}^{\beta,2} \,  \frac{x_d}{\eps} W^\eps_R  $ 
4459:   thanks to the previous subsections
4460:  ($\ref{norm2}$ and  $\ref{norm3}$) and to Lemma $\ref{Moser2}$. 
4461:  It remains to explain how to estimate 
4462:  %
4463:   \begin{eqnarray*}
4464:  \lambda^{m-2k-l}  \, |(\eps \partial_d )^k Z^l \nabla_y p^\eps |_{0,\lambda,T_0}. 
4465:   \end{eqnarray*}
4466:  %
4467:  Commuting $\eps \partial_d $ with $Z^l \nabla_y$, we are lead to estimate the terms 
4468:  %
4469:  \begin{eqnarray*}
4470:  \lambda^{m-2k-l}  \, | (\eps \partial_d )^{k'} Z^{l'} \nabla_y p^\eps |_{0,\lambda,T_0},
4471:  \end{eqnarray*}
4472:  %
4473:   with $k' \leqslant k-1$, $l' \leqslant
4474:  l-1$. All these terms are bounded by $|\eps \partial_d p^\eps |^\mathbf{E}_{\eps,m-1,\lambda,T_0}$ which is estimated in subsection
4475:  $\ref{norm2}$. 
4476:  
4477:  \subsection{An iterative scheme}
4478: \label{Itsch}
4479:  
4480:  We define an iterative scheme $(U_R^{\eps,\nu} )_{\eps,\nu }$ by 
4481:  %
4482:  \begin{eqnarray*}
4483:  ( \mathfrak{L}^\star (u^{\eps,\nu} ,\partial) + J^{\alpha,\nu} ) V_R^{\eps,\nu+1} 
4484:  + W^\eps_R  ( \underline{J}^{\beta,1,\nu} + \frac{x_d}{\eps}  .\underline{J}^{\beta,2,\nu}) = R^\eps_v 
4485:   \quad \mathrm{when} \ x_d >0,
4486:  \\ \X_{\mathtt{v}^{\eps,\nu}} W_R^{\eps,\nu+1} + \eps V_R^{\eps,\nu+1} .\nabla W_a^{\eps} = - \eps R^\eps_W \quad \mathrm{when}  \ x_d >0,
4487:  \\ \mathtt{V}_{R,d}^{\eps,\nu+1} =0 \quad \mathrm{when}  \ x_d =0 , 
4488:  \end{eqnarray*}
4489:  %
4490:  when $ J^{\alpha,\nu}$ denotes the $(d+1) \times (d+1)$ matrix 
4491:  %
4492:  \begin{eqnarray*}
4493:  J^{\alpha,\nu} := J^{\alpha} ( \eps, V_a^{\eps} , Z_\eps V_a^{\eps} , W_a^{\eps,\flat} , \eps^M \, U_R^{\eps,\nu} ) ,
4494:   \end{eqnarray*}
4495:  %
4496:  for $i \in \{1,2\}$, $J^{\beta,i,\nu}$ denotes the $\R^{d+1}$-valued functions 
4497:  %
4498:   \begin{eqnarray*}
4499:   J^{\beta,i,\nu} := J^{\beta,i} (\eps, V_a^{\eps} , Z_\eps V_a^{\eps} , W_a^{\eps,\flat}, \eps^M \, U_R^{\eps,\nu} ) 
4500:    \end{eqnarray*}
4501:   %
4502:    and
4503:   %
4504:    \begin{eqnarray*}
4505:  u^{\eps,\nu} := u^0 + \eps U^{\eps,\nu} , \quad U^{\eps,\nu} := U^{\eps}_a + \eps^M U_R^{\eps,\nu}.
4506:   \end{eqnarray*}
4507:  %
4508:    Because this step is now very classic in BKW theory (cf. \cite{Gu93}, \cite{Gu92}, \cite{Gu95}, \cite{moi}, \cite{moi2},
4509:  \cite{moi3}...), we only sketch as a preview how to deduce from the Sobolev embedding lemma $\ref{sobo}$ and from the linear estimates of
4510: Theorem $\ref{pok}$  the uniform boundedness of $ (U_R^{\eps,\nu})_{\eps,\nu} $. 
4511:  In order to do so, we fix two strictly positive real $h$ and $R$ , $\lambda$ and $\eps_0$ such that 
4512:  %
4513:   \begin{eqnarray}
4514:   \label{R}
4515:  \lambda^{-1} \, \lambda_0 \leqslant \mathrm{min} (\frac{h}{2},\frac{1}{2}) ,
4516:   \\ \eps_0^{M-\frac{1}{2}} \, c T_0 e^{\lambda T_0 } \leqslant  \frac{1}{2h} .
4517:   \end{eqnarray}
4518:  %
4519:  \begin{prop}
4520:   
4521:  If 
4522:  %
4523:  \begin{eqnarray*}
4524:  \eps^M || U^{\eps,\nu}_R ||^*_{\eps,T_0} \leqslant \mathfrak{R} \ \mathrm{and}  \ || U^{\eps,\nu}_R ||^\mathbf{E}_{\eps,m,\lambda,T_0}
4525:  \leqslant h ,
4526:   \end{eqnarray*}
4527:  %
4528:  then  
4529:  %
4530:  \begin{eqnarray*}
4531:  \eps^M || U^{\eps,\nu+1}_R ||^*_{\eps,T_0} \leqslant  \mathfrak{R} \ \mathrm{and}  \ || U^{\eps,\nu+1}_R ||^\mathbf{E}_{\eps,m,\lambda,T_0} \leqslant h .
4532:    \end{eqnarray*}
4533:  %
4534:  \end{prop}
4535:  
4536:   \begin{proof}
4537:    %
4538:    We will proceed in three steps.
4539:    
4540:    \begin{enumerate}
4541:     %
4542:  \item We begin applying the Sobolev embedding lemma $\ref{sobo}$:
4543:   %
4544:   \begin{eqnarray*}
4545: \sqrt{\eps}  || U^{\eps,\nu+1}_R  ||^*_{\eps,T_0} \leqslant c T_0 e^{\lambda T_0 } \,  | U^{\eps,\nu+1}_R  |^E_{\eps,m,\lambda,T_0} .
4546: \end{eqnarray*}
4547: %
4548:   Thanks to $(\ref{R})$, we obtain
4549:   %
4550:   \begin{eqnarray*}
4551:   \eps^M \  || U^{\eps,\nu+1}_R  ||^*_{\eps,T_0} \leqslant \frac{1}{2h} | U^{\eps,\nu+1}_R  |^E_{\eps,m,\lambda,T_0} .
4552:   \end{eqnarray*}
4553:   %
4554:  \item 
4555:   We apply Theorem $\ref{pok}$ and find
4556:   %
4557: \begin{eqnarray*}
4558: |U_R^{\eps,\nu+1} |^\mathbf{E}_{\eps,m,\lambda,T_0} \leqslant
4559:  \lambda^{-1} \lambda_0 .( 1+ \eps^M || U_R^{\eps,\nu+1} ||^*_{\eps,T_0} . | U_R^{\eps,\nu} |^\mathbf{E}_{\eps,m,\lambda,T_0}  ).
4560: \end{eqnarray*}
4561: %
4562:  Thanks to $(\ref{R})$, we obtain
4563:   %
4564:   \begin{eqnarray*}
4565:   |U_R^{\eps,\nu+1} |^\mathbf{E}_{\eps,m,\lambda,T_0} \leqslant h.
4566:   \end{eqnarray*}
4567:   %
4568:  \item 
4569:   Thanks to the Sobolev embedding lemma $\ref{sobo}$,  we obtain $ \eps^M || U^{\eps,\nu+1}_R ||^*_{\eps,T_0} \leqslant  \mathfrak{R}$.
4570:    %
4571:    %
4572: \end{enumerate}
4573:  %
4574:   \end{proof}
4575:    %
4576:   \textit{I thank Olivier Gu\`es for suggesting me this subject; in September $2001$.}
4577:  
4578: \nocite{*}
4579: \bibliographystyle{plain}
4580: \bibliography{a}
4581: 
4582: 
4583: $2000$ MSC: 76N20 Boundary-layer theory, 35Q35 Other equations arising in fluid mechanics, 35L50 Boundary value problems for
4584: hyperbolic systems of first-order PDE.
4585: \\
4586: \\ Franck SUEUR
4587: \\ \address{Laboratoire d'Analyse, de Topologie et de Probabilit\'e
4588: \\ Centre de Math\'ematiques et d'Informatique 
4589: \\ 39, rue F. Joliot Curie 
4590: \\ 13453 Marseille Cedex 13 }
4591: \\  \email{fsueur@cmi.univ-mrs.fr}
4592: \\ \urladdr{http://www.cmi.univ-mrs.fr/~fsueur/}
4593: 
4594:  \end{document} 
4595: