1:
2: \section{First anisotropic estimates}\label{s2}
3:
4: If $M$ is a generic CR submanifold of a complex manifold $X$, the
5: tangent bundle to $M$ admits a maximal complex subbundle
6: called the complex tangent bundle to $M$ and denoted by $HM$. It is
7: related to the CR structure of $M$ by $HM=TM\cap (H_{0,1}M+\ol H_{0,1}M)$.
8:
9:
10: Let us define now some anisotropic H\"older spaces of forms which
11: specify the complex tangent directions:
12:
13: - $\ac^\alpha(\ol D\cap M)$, $0<\alpha<1$, is the
14: set of continuous functions on $\ol D\cap M$ which are in
15: $\cc^{\alpha/2}(\ol D\cap M)$.
16:
17: - $\ac^{1+\alpha}(\ol D\cap M)$, $0<\alpha<1$, is the
18: set of functions $f$ such that $f\in\cc^{(1+\alpha)/2}(\ol D\cap M)$ and
19: $X_\cb f\in\cc^{\alpha/2}(\ol D\cap M)$, for all
20: complex tangent vector fields $X_\cb$ to $M$. Set
21: \begin{equation}\label{holderanis}
22: \|f\|_{A\alpha}=\|f\|_{(1+\alpha)/2}+\sup_{\|X_\cb\|\leq 1}\|X_\cb f\|_{\alpha/2}
23: \end{equation}
24:
25: - $\ac^{p+\alpha}(\ol D\cap M)$, $p\geq 2$, $0<\alpha<1$, is the
26: set of functions $f$ of class $\cc^{[p/2]}$ such that
27: $Xf\in\ac^{p-2+\alpha}(\ol D\cap M)$,
28: for all tangent vector fields $X$ to $M$ and
29: $X_\cb f\in\ac^{p-1+\alpha}(\ol D\cap M)$, for all
30: complex tangent vector fields $X_\cb$ to $M$.
31:
32: The anisotropic H\"older space of forms $\ac^{p+\alpha}_*(\ol D\cap
33: M)$, $p\geq 0$, $0<\alpha<1$, is then the space of continuous forms on
34: $\ol D\cap M$, whose
35: coefficients are in $\ac^{p+\alpha}(\ol D\cap M)$.
36:
37: This section is devoted to the study of the continuity properties
38: with respect to
39: these spaces of
40: the operators $T_r$
41: and $A_r$, $1\leq r\leq q$ and $n-k-q+1\leq r\leq n-k$, defined in
42: the previous section when $M$
43: is $q$-concave.
44:
45: We first prove some estimates for the operators defined by the local
46: kernels from the previous section. For this we may assume that
47: $M$ is embedded in $\cb^n$.
48: To clarify the exposition, we introduce a new kind of operators.
49:
50: \begin{defin}
51: Let $m\in\nb$, $\epsilon\in\{-1,0,1\}$, $\delta>0$ and $I\in\ic'(k)$. An
52: \emph{operator of type} $(m,\epsilon,\delta)$ is, by definition, a map
53: $$E~:~\cc^0_{n,*}(\Gamma_I)\to\cc^\infty_{n,*}(\Gamma_I)$$
54: such that there exist
55: \begin{itemize}
56: \item an integer $\kappa\geq 0$
57: \item a differential form $\widehat E(z,\zeta,\lambda)$ of type
58: $O_{|I|+1-2n+2\kappa+m+\epsilon}$ on $\Gamma_I\times
59: D^*_I\times\Delta_{I*}$ such that for all $f\in\cc^0_{n,*}(\Gamma_I)$
60: $$Ef(\zeta)=\int_{(z,\lambda)\in\Gamma_I\times\Delta_{I*}}
61: \wt f(z)\wedge\frac{\widehat
62: E(z,\zeta,\lambda)\wedge\Theta(z)}{(\Phi+\delta)^{\kappa+m}(z,\zeta,\lambda)},$$
63: where$\wt f\in \cc^0_{0,*}(\Gamma_I)$ is the form with
64: $$f(z)=\wt f(z)\wedge dz_1\wedge\dots\wedge dz_n,$$
65: and $\Theta=1$, if $m=0$, and either
66: $\Theta=\pa\rho_{i_1}\wedge\dots\wedge\pa\rho_{i_m}$ or
67: $\Theta=\opa\rho_{i_1}\wedge\dots\wedge\pa\rho_{i_m}$, if $m\geq 1$
68: with $i_1,\dots,i_m\in I$.
69: \end{itemize}
70: \end{defin}
71:
72: We consider the operators $\wt C_{I*}$, $I\in\ic'(k)$, defined by
73: \begin{equation}\label{operateurC}
74: \wt C_{I*} f(\zeta)=\int_{z\in D\cap M}f(z)\wedge
75: C_{I*}(z,\zeta)=\int_{(z,\lambda)\in D\cap
76: M\times\Delta_{I*}}f(z)\wedge K_{I*}(z,\zeta,\lambda)
77: \end{equation}
78: for $f\in\cc^1_{n,r}(D)$, $0\leq r\leq n-k$.
79: Using Stokes formula we get
80: \begin{equation}
81: \wt C_{I*} f(\zeta)=\int_{(z,\lambda)\in\Gamma_I\times\Delta_{I*}}\opa
82: f(z)\wedge
83: K_{I*}(z,\zeta,\lambda)+(-1)^{n+r}\int_{(z,\lambda)\in\Gamma_I\times\Delta_{I*}}f(z)\wedge
84: \opa_z K_{I*}(z,\zeta,\lambda).
85: \end{equation}
86:
87: By (\ref{singRC}) and ((\ref{singdbarRC}) the kernel $\wt C_{I*}$ is
88: the sum of an operator of type
89: $(m,1,0)$ and of an operator of type $(m,0,0)$, $0\leq m\leq k$, and
90: if $X_\cb$ denotes a complex tangent vector field to $M$, then by
91: (\ref{phitan}) the kernels
92: $X^\zeta_\cb\wt C_{I*}$ and $X^z_\cb\wt C_{I*}$ are the sum of an
93: operator of type $(m,0,0)$ and of an operator of type $(m,-1,0)$,
94: $0\leq m\leq k$.
95:
96: To prove the anisotropic estimates we need the following spaces
97: and norms on differential forms~:
98:
99: - $\bc^\beta_*(\Gamma_I)$, $\beta\geq 0$, is the space of forms
100: $f\in\cc^0_*(\ol\Gamma_I\setminus M)$ such that, for some constant $C>0$,
101: $$\|f(z)\|\leq C[{\rm dist}(z,M)]^{-\beta}, \quad z\in\Gamma_I,$$ where
102: ${\rm dist}(z,M)$ is the Euclidean distance between $z$ and $M$. Set
103: \begin{equation}\label{croissance}
104: \|f\|_{\beta}=\sup_{z\in \Gamma_I}\big(\|f(z)\|[{\rm
105: dist}(z,M)]^{\beta}\big)
106: \end{equation}
107: for $\beta\geq 0$ and $f\in\bc^\beta_*(\Gamma_I)$.
108:
109: It follows from the characterization of H\"older continuous functions
110: by mean of the Poisson integral that, if $f\in\cc^\alpha_*(\ol D\cap M)$,
111: $0<\alpha<1$, there exists for each $I\in\ic'(k)$ a continuous form
112: $f_I$ on $\ol\Gamma_I$ such that ${f_I}_{|_{M}}=f$ and
113: $X_If_I\in\bc^{1-\alpha}_*(\Gamma_I)$, for all vector
114: fields $X_I$ tangent to $\Gamma_I$.
115:
116: \begin{lem}\label{estimbeta}
117: If $f\in \bc^\beta_*(\Gamma_I)$, $0\leq \beta<1$, has compact
118: support in $D$ and if $E_I$ is an operator of type
119: $(m,\epsilon,\delta)$, then there exists a constant $C>0$,
120: independent of $\delta$, such that for $\zeta\in D^*_I$
121: \begin{align*}
122: |E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
123: dist}(\zeta,M)+\delta]^{1/2-\beta},\quad {\rm if}~ \epsilon=0\\
124: |E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
125: dist}(\zeta,M)+\delta]^{-\beta},\quad {\rm if}~ \epsilon=-1\\
126: |E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
127: dist}(\zeta,M)+\delta]^{1-\beta},\quad {\rm if}~ \epsilon=1
128: \end{align*}
129: \end{lem}
130: \begin{proof}
131: It follows from \cite{LaLefourier} and \cite{BaLa} that, after
132: integration in $\lambda$, an operator of type $(m,\epsilon,\delta)$ is
133: controled, since $n\geq 3$, by
134: \begin{equation*}
135: \int_{\Gamma_I}\frac{|f_I|~|\sigma\wedge d\rho_I{\wedge_{\nu=1}^s}dt_\nu|}
136: {(|\rho_I|+d+\delta+|\zeta-z|^2)\Pi_{\nu=1}^{s}(|t_\nu|+d+\delta+|\zeta-z|^2)^{1+1/s}
137: |\zeta-z|^{2n-k-s-3-\epsilon}},
138: \end{equation*}
139: with $1\leq s\leq k$,
140: where $\rho_I$ is the function defined by
141: $\rho_I(z)=\rho_{i_1}(z)=\dots=\rho_{i_k}(z)$ for $z\in\Gamma_I$,
142: $d=d(\zeta)={\rm dist}(\zeta,M)$,
143: $\sigma$ a monomial in $dz_1,\dots,dz_n,d\ol z_1,\dots,d\ol z_n$, $t_\nu= {\rm
144: Im}~\Phi(z,\zeta,\lambda^\nu)$ and $dt_\nu= d_z{\rm
145: Im}~\Phi(z,\zeta,\lambda^\nu)$ with $\lambda^1,\dots,\lambda^{k+1}$
146: some points in $\Delta_{I*}$ which are
147: linearly independent in $\rb^{k+1}$,
148:
149: We denote all the constants by $C$. Then since $d(z)\leq
150: C|\rho_I(z)|$ for $z\in\Gamma_I$, we get
151: \begin{align*}
152: \|E_I &f(\zeta)\|\leq\\
153: & C
154: \|f\|_{\beta}\int_{\Gamma_I}\frac{|\rho_I|^{-\beta} ~|\sigma\wedge
155: d\rho_I{\wedge_{\nu=1}^s}dt_\nu|}
156: {(|\rho_I|+d+\delta+|\zeta-z|^2)\Pi_{\nu=1}^{s}(|t_\nu|+d+\delta+|\zeta-z|^2)^{1+1/s}
157: |\zeta-z|^{2n-k-s-3-\epsilon}}
158: \end{align*}
159: for all $f\in \bc^\beta_*(\Gamma_I)$ and $\zeta\in D^*_I$. We can
160: use $\rho_I$ as a coordinate in $\Gamma_I$ and since $M$ is generic
161: the $t_\nu$'s can also be used as coordinates, so we obtain that
162: \begin{align*}
163: \|E_I &f(\zeta)\|\leq \\
164: &C
165: \|f\|_{\beta}\int_{y\in\rb^{2n-k+1}\atop |y|<c}\frac{|y_1|^{-\beta}
166: ~dy_1\wedge dy_2\wedge\dots
167: \wedge dy_{2n-k+1}}
168: {(|y_1|+d+\delta+|\zeta-z|^2)\Pi_{\nu=1}^{s}(|y_\nu|+d+\delta+|y|^2)^{1+1/s}
169: |y|^{2n-k-s-3-\epsilon}}.
170: \end{align*}
171: Since $\beta<1$, we can integrate with respect to $y_1$, which gives
172: \begin{equation*}
173: \|E_I f(\zeta)\|\leq C
174: \|f\|_{\beta}\int_{y\in\rb^{2n-k}\atop |y|<c}\frac{dy_2\wedge\dots
175: \wedge dy_{2n-k+1}}
176: {(d+\delta+|y|^2)^{\beta}\Pi_{\nu=1}^{s}(|y_\nu|+d+\delta+|y|^2)^{1+1/s}
177: |y|^{2n-k-s-3-\epsilon}}.
178: \end{equation*}
179: Then integrating with respect to $y_2,\dots,y_s$ and using spherical coordinates,
180: we have
181: \begin{equation*}
182: \|E_I f(\zeta)\|\leq C
183: \|f\|_{\beta}\int_0^C\frac{dr}
184: {(d+\delta+r^2)^{\beta+1} r^{-2-\epsilon}},
185: \end{equation*}
186: which proves the lemma following the values of $\epsilon$.
187: \end{proof}
188:
189: In the same way, following again \cite{LaLefourier}, we can prove
190: estimates for the gradient of $E_If$, when $E_I$ is an
191: operator of type $(m,\epsilon,\delta)$ and $f\in \bc^\beta_*(\Gamma_I)$,
192: $0\leq \beta<1$.
193:
194: \begin{lem}\label{estimbeta-gradient}
195: If $f\in \bc^\beta_*(\Gamma_I)$, $0\leq \beta<1$, has compact support
196: in $D$, $E_I$ is an
197: operator of type $(m,\epsilon,\delta)$ and if $\nabla_\zeta$ is one of the
198: operators $\frac{\pa}{\pa\zeta_1},\dots,\frac{\pa}{\pa\zeta_n}$ or
199: $\frac{\pa}{\pa\ol\zeta_1},\dots,\frac{\pa}{\pa\ol\zeta_n}$, then
200: there exists a constant $C>0$ independent of $\delta$ such
201: that for $\zeta\in D^*_I$
202: \begin{align*}
203: |\nabla_\zeta E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
204: dist}(\zeta,M)+\delta]^{-1/2-\beta},\quad {\rm if}~ \epsilon=0\\
205: |\nabla_\zeta E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
206: dist}(\zeta,M)+\delta]^{-1-\beta},\quad {\rm if}~ \epsilon=-1\\
207: |\nabla_\zeta E_I f(\zeta)|&\leq C \|f\|_{\beta}[{\rm
208: dist}(\zeta,M)+\delta]^{-\beta},\quad {\rm if}~ \epsilon=1
209: \end{align*}
210: \end{lem}
211:
212: Using the classical Hardy-Littlewood lemma, we deduce from Lemma
213: \ref{estimbeta}
214: and Lemma \ref{estimbeta-gradient}, the following estimates for the
215: operators $\wt C_{I*}$ and $X^\zeta_\cb\wt C_{I*}$
216:
217: \begin{prop}\label{estimci}
218: Let $f\in \bc^\beta_*(\Gamma_I)$, $0\leq \beta<1$, be a form with
219: compact support in $D$, then
220: \begin{align*}
221: &\wt C_{I*}f\in\cc^{1/2-\beta}_*(D^*_I),\quad {\rm if}~ 0\leq\beta<1/2\\
222: &\wt C_{I*}f\in\bc^{\beta-1/2}_*(D^*_I),\quad {\rm if}~ 1/2\leq\beta<1
223: \end{align*}
224: and
225: \begin{equation*}
226: X^\zeta_\cb\wt C_{I*}f\in \bc^\beta_*(D^*_I).
227: \end{equation*}
228: \end{prop}
229:
230: Let us recall Lemma 5.3 and 5.5 from \cite{BaLa}
231:
232: \begin{lem}
233: There exists $Y_1,\dots,Y_k$ tangential vector fields to $M$ such that
234: for all $\zeta\in U_0$ and all $1\leq i,j\leq k$
235: $$Y_i^\zeta\Phi_j(\zeta,\zeta)=\delta_{ij},$$
236: where $\delta_{ij}$ denotes the Kronecker index.
237: \end{lem}
238:
239: If $X^z$ is a vector field on $D$ in the variable $z$, we denote by
240: $X^\zeta$ the same vector field in the variable $\zeta$.
241:
242: \begin{lem}\label{derivation}
243: Let us fix $\delta>0$, if $\widehat E_\delta$ is the kernel of an
244: operator of type $(m,\epsilon,\delta)$, there exists $\eta>o$ such
245: that for $(z,\zeta)\in \ol D_I\times\ol D_I^*$ with $|z-\zeta|<\eta$
246: and $z\neq \zeta$ we have
247: \begin{equation}\label{eq-derivation}
248: X^z\widehat E_\delta=-X^\zeta\widehat
249: E_\delta+\frac{(X^z+X^\zeta)\Phi_{I*}}{Y^\zeta_\lambda\Phi_{I*}}Y^\zeta_\lambda
250: \widehat E_\delta+\widehat G_\delta,
251: \end{equation}
252: where $\widehat G_\delta$ is a finite sum of kernels associated to
253: operators of type $(m,\epsilon,\delta)$.
254: \end{lem}
255:
256: As regularity is a local problem, let us fix $\zeta_0$ in $D\cap M$
257: and choose a function $\chi$ with compact support such that
258: $\chi(z)=1$, if $|z-\zeta_0|\leq \eta/4$, and $\chi(z)=0$, if
259: $|z-\zeta_0|\geq \eta/2$, where $\eta$ is given by Lemma
260: \ref{derivation}.
261:
262: If $f$ is a continuous form on $D\cap M$ and $f_I$ a continuous
263: extension of $f$ to $\Gamma_I$, we write
264: \begin{equation*}
265: \wt C_{I*} f(\zeta)=\wt C_{I*} (\chi f)(\zeta)+\wt C_{I*} ((1-\chi)f)(\zeta)
266: \end{equation*}
267: As the singularity of the kernel associated to the operator $\wt
268: C_{I*}$ is concentrated on the diagonal, the regularity of the second
269: term, for $\zeta\in M$ with $|\zeta-\zeta_0|\leq \eta/4$, is directly
270: given by the regularity of this kernel, which is of class $\ci$ in
271: the first variable $z$ and of
272: class $\cc^{l-2}$ in the second variable $\zeta$ when $M$ is of class $\cc^l$.
273: \begin{thm}\label{estimlocale}
274: Assume $M$ is of class $\cc^l$.
275: The operator $\wt R_{M}=\sum_{I\in\ic'(k)} {\rm sgn}(I)
276: \wt C_{I*}$ is a bounded linear operator from $\ac^{p+\alpha}_*(\ol D\cap
277: M)$ into $\ac^{p+1+\alpha}_*(\ol D\cap
278: M)$ for $0\leq p\leq 2l-7$ and $0<\alpha<1$.
279: \end{thm}
280:
281: \begin{proof}
282: Let us first consider the case $p=0$. Let $f\in\ac^{\alpha}_*(\ol D\cap
283: M)$, i.e. $f\in\cc^{\alpha/2}_*(\ol D\cap M)$, then there exists, for each
284: $I\in\ic'(k)$, a continuous form
285: $f_I$ on $\ol\Gamma_I$ such that ${f_I}_{|_{M}}=f$ and
286: $X_If_I\in\bc^{1-\alpha/2}_*(\Gamma_I)$, for all vector
287: fields $X_I$ tangent to $\Gamma_I$. From the previous remark we have
288: only to study $\wt C_{I*} (\chi f)(\zeta)$.
289: Let us denote by $\wt C_{I*}^\delta$ the operator $\wt C_{I*}$, in
290: which we have replaced
291: $\Phi$ by $\Phi+\delta$. From Lemma
292: \ref{derivation}, we get that for all vector fields $X_I$ tangent to
293: $\ol\Gamma_I\cup\ol\Gamma_I^*$, we
294: have $X_I^\zeta\wt C_{I*}^\delta \chi f=\wt
295: C_{I*}^\delta X_I^z \chi f_I+E_I \chi f$, where $E_I$ is a sum of operators of
296: type $(m,\epsilon,\delta)$, $\epsilon\geq 0$.
297: Let also $X_\cb$ be a complex tangential vector field to $M$.
298:
299: We deduce from Lemma \ref{estimbeta} that, for $\zeta\in \Gamma_I^*$,
300: $$|X_I^\zeta \wt C_{I*}^\delta
301: (\chi f)(\zeta) |\leq C~ \|X_I(\chi f_I)\|_{\beta}[{\rm
302: dist}(\zeta,M)+\delta]^{(\alpha-1)/2}$$
303: and
304: $$|X_I^\zeta X_\cb\wt C_{I*}^\delta
305: (\chi f)(\zeta) |\leq C'~ \|X_If_I\|_{\beta}[{\rm
306: dist}(\zeta,M)+\delta]^{\alpha/2-1},$$
307: where $C$ and $C'$ are constants
308: independent of $\delta$. Moreover $X_I^\zeta \wt C_{I*}^\delta f$
309: and $X_I^\zeta X_\cb\wt C_{I*}^\delta f$ converge respectively to
310: $X_I^\zeta \wt C_{I*} f$ and $X_I^\zeta X_\cb\wt C_{I*} f$ uniformly
311: on each compact subset of $\Gamma_I^*$, when $\delta$ tends to zero.
312: This implies that $X_I^\zeta \wt C_{I*}
313: f\in\bc^{(1-\alpha)/2}_*(\Gamma_I^*)$ and $X_I^\zeta X_\cb\wt C_{I*}
314: f\in\bc^{1-\alpha/2}_*(\Gamma_I^*)$ and by the classical
315: Hardy-Littlewood lemma that $\wt C_{I*} f\in\cc^{(\alpha+1)/2}_*(\ol
316: D\cap M)$ and $X_\cb\wt C_{I*} f\in\cc^{\alpha/2}_*(\ol D\cap M)$,
317: which means that $\wt C_{I*} f\in\ac^{1+\alpha}_*(\ol D\cap M)$.
318: This ends the proof of this case by definition of the operator
319: $R_{M}$.
320:
321: Assume now that $p=1$. Let $f\in\ac^{1+\alpha}_*(\ol D\cap
322: M)$, i.e. $f\in\cc^{(1+\alpha)/2}_*(\ol D\cap M)$ and $X_\cb
323: f\in\ac^{\alpha}_*(\ol D\cap M)$, where $X_\cb$ is complex tangent to
324: $M$.
325:
326: If we proceed as in the case $p=0$, but using Lemma
327: \ref{estimbeta-gradient} at the place of Lemma
328: \ref{estimbeta}, we can prove that, for each $I\in\ic'(k)$ and each
329: vector field $X$ tangent to $M$,
330: $\nabla_\zeta \wt C_{I*}(\chi X^z f)\in\bc^{1-\alpha/2}_*(D_I^*)$,
331: which implies by the classical
332: Hardy-Littlewood lemma that $\wt C_{I*}(\chi X^z
333: f)\in\cc^{\alpha/2}_*(\ol D\cap M)$.
334:
335: Using the case $p=0$, we get that if $X_\cb f\in\ac^{\alpha}_*(\ol
336: D\cap M)$, then $\wt R_{M} \chi X_\cb f\in\ac^{1+\alpha}_*(\ol D\cap
337: M)$. Moreover it follows from the proof of Theorem 5.6 in
338: \cite{BaLa} that if $X_\cb f$ is continuous then $X_\cb\wt R_{M}
339: (\chi f)$ and $\wt R_{M} \chi X_\cb f$ have the same regularity,
340: consequently $X_\cb\wt R_{M} (\chi f)\in\ac^{1+\alpha}_*(\ol D\cap
341: M)$.
342: By definition of the space $\ac^{2+\alpha}_*(\ol D\cap M)$, we have then
343: proved that $\wt R_{M} (\chi f)\in\ac^{2+\alpha}_*(\ol D\cap M)$.
344:
345: Since by the proof
346: of Theorem 5.6 in \cite{BaLa}, if
347: $X f$ is continuous for any vector field $X$ tangent to $M$, then $X\wt R_{M}
348: f$ and $\wt R_{M} \chi X f$ have the
349: same regularity, the theorem follows for $p\geq 2$, by a simple
350: induction, from the definition of the spaces $\ac^{p+\alpha}_*(\ol D\cap M)$.
351: \end{proof}
352:
353: \begin{rem}
354: Note that if we exchange the role played by the variables $z$ and
355: $\zeta$ in the operator $\wt R_{M}$, then Theorem \ref{estimlocale} is
356: valid for $0\leq p\leq 2l-6$
357: \end{rem}
358:
359: Theorem \ref{estimlocale} implies better regularity properties for the
360: operators involve in Proposition \ref{estmeeslocales} and Theorem
361: \ref{formuleglobale}.
362:
363: \begin{cor}\label{estimanis}
364: Under the hypotheses of Proposition \ref{estmeeslocales} and Theorem
365: \ref{formuleglobale}, the operators $T_r$, $1\leq r\leq q$ and
366: $n-k-q+1\le r\le n-k$, in Proposition \ref{estmeeslocales} are
367: bounded linear operators from $\ac^{p+\alpha}_{n,r}(M)$ into
368: $\ac^{p+1+\alpha}_{n,r-1}(\ol U)$ and the operators $A_r$, $1\leq
369: r\leq q$ and $n-k-q+1\le r\le n-k$, in Theorem \ref{formuleglobale}
370: are bounded linear operators from $\ac^{p+\alpha}_{n,r}(M,E)$ into
371: $\ac^{p+1+\alpha}_{n,r-1}(M,E)$.
372: \end{cor}
373: \begin{proof}
374: Since the operators $T_r$ from Proposition \ref{estmeeslocales} are
375: defined as the restrictions to $\cc^0_{n,r}(M)$ of the operators
376: $\wt R_{M}$, if $1\le r\le q-1$, and as the restrictions to
377: $\cc^0_{n,r}(M)$ of the operators $\wt R_{M}$ after exchanging the
378: variables $z$ and $\zeta$ in the kernels, if $n-k-q+1\le r\le n-k$,
379: they are in fact bounded from $\ac^{p+\alpha}_{n,r}(M)$ into
380: $\ac^{p+1+\alpha}_{n,r-1}(\ol U)$.
381:
382: Assume $M$ is a compact $q$-concave generic CR submanifold of a
383: complex manifold $X$ and $E$ is an holomorphic vector bundle on X.
384: Let us go back to the construction of the global operators $A_r$
385: (cf. \cite{LaLeglobal} and \cite{BroLaLe}). First by globalizing the
386: homotopy formula \ref{homotopielocale} by mean of a partition of
387: unity, one get new linear operators $\wt T_r$ bounded from
388: $\ac^{p+\alpha}_{n,r}(M,E)$ into $\ac^{p+1+\alpha}_{n,r-1}(M,E)$ and
389: $K_r$ bounded from $\ac^{p+\alpha}_{n,r}(M,E)$ into
390: $\ac^{p+1+\alpha}_{n,r}(M,E)$ such that
391: \begin{equation}\label{homotopiecompact}
392: f+K_rf=\begin{cases}\wt T_{1}\opa_b f&\text{if }r=0\,,\\\opa_b\wt
393: T_rf+\wt T_{r+1}\opa_b f\qquad&\text{if }1\leq r\leq r-1 \text{ or }
394: n-k-q+1\le r\le n-k\,.
395: \end{cases}
396: \end{equation}
397: From functional analysis arguments we get linear operators $T'_r$,
398: $1\leq r\leq q$ and $n-k-q+1\le r\le n-k$, bounded from
399: $\cc^0_{n,r}(M,E)$ into $\cc^\infty_{n,r-1}(M,E)$ and $K'_r$, $0\leq
400: r\leq q-1$ and $n-k-q+1\le r\le n-k$, bounded from
401: $\cc^0_{n,r}(M,E)$ into $\cc^\infty_{n,r}(M,E)$ such that if
402: \begin{equation}
403: N_r=I+K_r+K'_r
404: \end{equation}
405: then
406: $$\cc^0_{n,r}(M,E)={\rm Im}N_r\oplus{\rm Ker}N_r,~{\rm Ker}N_r\subset
407: \cc^\infty_{n,r}(M,E).$$ We set $\widehat N_r=N_r+P_r=I+R_r$, where
408: $P_r$ denotes the linear projection with ${\rm Im}P_r={\rm Ker}N_r$
409: and ${\rm Ker}P_r={\rm
410: Im}N_r$.
411: The operator $R_r$ is continuous from $\ac^{p+\alpha}_{n,r}(M,E)$
412: into $\ac^{p+1+\alpha}_{n,r}(M,E)$ and hence compact as an
413: endomorphism of $\ac^{p+\alpha}_{n,r}(M,E)$. Consequently $\widehat
414: {N_r}_{|_{\ac^{p+\alpha}_{n,r}(M,E)}}$ is a Fredholm endomorphism
415: with index $0$ and as ${\rm Ker}\widehat N_r=\{0\}$, it is an
416: isomorphism. Moreover $\widehat N_r^{-1}= I-R_r\widehat N_r^{-1}$,
417: this implies that $\widehat
418: {N_r^{-1}}_{|_{\ac^{p+1+\alpha}_{n,r}(M,E)}}$ is a continuous
419: endomorphism of $\ac^{p+1+\alpha}_{n,r}(M,E)$. The operators $A_r$
420: are then defined by
421: \begin{equation}
422: A_r=\begin{cases} &\widehat N_{r-1}^{-1}(T_r+T_r')\,,\qquad 1\le
423: r\le q \\
424: &(T_r+T_r')\widehat N_{r}^{-1}\,,\qquad n-k-q+1\le r\le n-k\,,
425: \end{cases}
426: \end{equation}
427: and hence continuous from $\ac^{p+\alpha}_{n,r}(M,E)$ into
428: $\ac^{p+1+\alpha}_{n,r}(M,E)$.
429: \end{proof}