\documentclass[twoside]{article} \usepackage{amssymb, amsmath} \pagestyle{myheadings} \markboth{\hfil On the ellipticity and solvability \hfil EJDE--2001/57} {EJDE--2001/57\hfil Abdelillah El Haial \& Rabah Labbas \hfil} \begin{document} \title{\vspace{-1in}\parbox{\linewidth}{\footnotesize\noindent {\sc Electronic Journal of Differential Equations}, Vol. {\bf 2001}(2001), No. 57, pp. 1--18. \newline ISSN: 1072-6691. URL: http://ejde.math.swt.edu or http://ejde.math.unt.edu \newline ftp ejde.math.swt.edu (login: ftp)} \vspace{\bigskipamount} \\ % On the ellipticity and solvability of an abstract second-order differential equation % \thanks{ {\em Mathematics Subject Classifications:} 34G10, 35J25, 47D03. \hfil\break\indent {\em Key words:} Abstract elliptic equations. \hfil\break\indent \copyright 2001 Southwest Texas State University. \hfil\break\indent Submitted November 15, 1998. Published August 27, 2001.} } \date{} % \author{ Abdelillah El Haial \& Rabah Labbas } \maketitle \begin{abstract} In this work we give some new results on a complete abstract second-order differential equation of elliptic type in the non-homogeneous case. Existence, uniqueness, and maximal regularity of the strict solution are proved under some natural assumptions which imply the ellipticity of the differential equation. \end{abstract} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \allowdisplaybreaks \section{Introduction} In this paper, we study the second-order abstract differential equation \begin{equation} u''(t)+2Bu'(t)+Au(t)=f(t),\quad t\in (0,1) \end{equation} under the non-homogeneous boundary conditions \begin{equation} u(0)=\varphi , \quad u(1)=\psi , \end{equation} where $\varphi ,\psi $ and $f(t)$ belong to a complex Banach space $E$, and $A$, $B$ are two closed linear operators with domains $D(A)$, $D(B)$. We are interested in the existence, uniqueness, and maximal regularity of the strict solution $u$ when $f$ is regular (for instance $f$ \ is H\"{o}lder continuous function). We recall that $u\in C([0,1] ;E)$ is a strict solution of (1)-(2) if and only if \[ u\in C^{2}([0,1] ;E)\cap C^1([0,1] ;D(B))\cap C([0,1] ;D(A)) \] and $u$ satisfies (1) and (2). Throughout this paper we assume the following hypotheses: \begin{enumerate} \item[H1)] Operator $B$ generates a strongly continuous group on $E$ \item[H2)] There exists $K>0$ such that for all $\lambda \geqslant 0$, $\| (A-B^{2}-\lambda I)^{-1}\|_{L(E)}\leqslant K/(1+\lambda)$ \item[H3)] For all $\mu \in \mathbb{R}$, and all $\lambda \geqslant 0$, $(A-B^{2}-\lambda I)^{-1}(B-\mu I)^{-1}-(B-\mu I)^{-1}(A-B^{2} -\lambda I)^{-1}=0$ \item[H4)] There exists $K>0$ such that for all $\lambda \geqslant 0$,\\ i)\quad $\| B(A-B^{2}-\lambda I)^{-1}\|_{L(E)}\leqslant K/(1+\sqrt{\lambda })$, \\ ii)\quad $\| B^{2}(A-B^{2}-\lambda I)^{-1}\|_{L(E)}\leqslant K$,\\ iii)\quad $\| A(A-B^{2}-\lambda I)^{-1}\|_{L(E)}\leqslant K$. \end{enumerate} Note that H1 implies $\overline{D(B)}=E$, but the domain $D(A-B^{2})$ may be not dense. On the other hand, it is not difficult to see that assumptions H1, H2, H3 permit us to apply the Da Prato and Grisvard \cite{d1} sum theory and deduce that, necessarily, \[ (A-\lambda I) ^{-1}\in L(E) \quad\mbox{and}\quad \|(A-\lambda I)^{-1}\|_{L(E)}\leqslant K/(1+\lambda ),\,\mbox{for all }\lambda \geqslant 0. \] (See the proof of Theorem \ref{thm1}, the end of statement i)). Several authors have studied equation (1) when it is regarded as an abstract Cauchy problem, that is under the following initial data $u(0)=\varphi $, $u'(0)=\psi $. See, for instance, Favini \cite{f1}, Neubrander \cite{n1}, Liang and Xiao \cite{l3}. The techniques used in these papers are based on the parabolicity of (1), that is the parabolicity of the operator pencil defined by \[ P(\lambda ):D(A)\cap D(B)\to E;\quad x\mapsto P(\lambda )x=(\lambda ^{2}I+2\lambda B+A)x. \] In this paper we try to provide an unified treatment of some class of second order partial differential equations when they are regarded as equations of elliptic type. The principal part of these equations can be written in the form \[ \mathcal{P}u=\partial_{t}^{2}u+2b(.)\partial_{xt}^{2}u+a(.)\partial _{x}^{2}u, \] or more generally \[ \mathcal{P}u=\partial_{t}^{2}u+\mathcal{B}(x,\partial_{x})\partial_{t}u+ \mathcal{A}(x,\partial_{x})\partial_{x}^{2}u, \] where $\mathcal{B}(x,\partial_{x})$ and $\mathcal{A}(x,\partial_{x})$ are linear differential operators in $\mathbb{R}^{n}$ with some natural assumptions on their coefficients of order $m$ and $l$. So, our objective is to express the ellipticity of these concrete operators in general abstract form, find a representation of the solution of the problem (1)-(2) and study its maximal regularity. Some particular equations of type (1) can be used to describe the singularities of the solutions in elliptic problems on singular domains. Let us give the following example of Laplacian operator \begin{gather*} -\Delta u=g,\quad g\in L^{p}(Q) \\ u=0\quad \mbox{on } \partial Q, \end{gather*} in a conical domain $Q=\left\{ \rho \sigma ~:~\rho >0,\sigma \in G\right\} \subset \mathbb{R}^{n}$, where $G$ is an open subset of the sphere $S^{n-1}$. In the three dimensional domain, the boundary $\partial Q$ has a vertex at $O$ and also edges depending on whether the boundary $\partial G$ has corners or not. Let us assume that the variational solution $u$ of this Dirichlet problem exists. So, we would have $u=u_{r}+u_{s}$, with a regular part $u_{r}$ $\in W^{2,p}(Q) $ and a singular part $u_{s}$ in $\exp $licit form. Then, one performs the change of variable $\rho =e^{t}$ which maps $Q$ onto the infinite cylinder $\mathbb{R}\times G$. The Laplacian equation in polar coordinates is \[ D_{\rho }^{2}u+\frac{n-1}{\rho }D_{\rho }u+\frac{1}{\rho ^{2}}\Delta 'u=g \] where $\Delta '$ denotes the Laplace-Beltrami operator on $S^{n-1}$. The above mentioned change of variable leads us to \[ D_{t}^{2}u+D_{t}u+\Delta 'u=e^{2t}g=g_{1}, \] then by puttting $v(t)(\sigma )=e^{\varpi t}u(e^{t}\sigma )$ and $g_{2}(t)(\sigma )=e^{\varpi t}g_{1}$ \ we have \begin{gather*} D_{t}^{2}v+(n-2-2\varpi ) D_{t}v+\varpi (\varpi -n+2) v+\Delta 'v=g_{2} \\ v=0\quad \mbox{on }\partial (\mathbb{R}\times G) \end{gather*} where $\varpi =-2+n/p$, precisely the opposite of the Sobolev exponent of $W^{2,p}$ . This final equation can be regarded as a particular form of our abstract general problem (1). See, for details, Labbas-Moussaoui-Najmi \cite{l2}. Equations (1)-(2) can be illustrated, for instance, by the following simple model differential problem \begin{equation} \begin{gathered} \frac{\partial ^{2}u}{\partial t^{2}}(t,x)+2b\frac{\partial ^{2}u}{ \partial x\partial t}(t,x)+a\frac{\partial ^{2}u}{\partial x^{2}} (t,x)=f(t,x),\quad (t,x)\in \Sigma , \\ u(0,x)=\varphi (x),\quad \,u(1,x)=\psi (x),\quad x\in \mathbb{R}, \end{gathered} \end{equation} where $\Sigma =\left] 0,1\right[ \times \mathbb{R}$ and \begin{equation} a-b^{2}>0. \end{equation} Then we can choose in $E=L^{p}(\mathbb{R})$, $1
0}
r^{\theta }\|L(L-rI)^{-1}\varphi \|_{E}<\infty \right\} ,
\]
(see Grisvard \cite{g2}).
In section 2, we build the natural representation of the solution of
(1)-(2) by using the operational calculus and the Dunford integral.
In section 3, we prove an essential lemma, which allow us to justify and
to study the optimal smoothness of the previous representation; we then give
necessary and sufficient conditions on $\varphi ,\psi $, $f(0)$ and $f(1)$
for having a strict solution when $f$ is H\"{o}lder continuous function.
Finally, in section 4, we give an example, to which our abstract results
can be applied.
\section{Construction of the solution}
If $A$ and $B$ are two scalars such that $B^{2}-A=-\lambda $ is
strictly positive, then the solution of (1) is given by
\[
u(t)=\frac{\sinh \sqrt{-\lambda }(1-t)}{\sinh \sqrt{-\lambda }}
e^{-tB}\varphi +\frac{\sinh \sqrt{-\lambda }t}{\sinh \sqrt{-\lambda }}
e^{(1-t)B}\psi \medskip -\int_0^1G_{\sqrt{-\lambda }}(t,s)f(s)ds,
\]
where
\[
G_{\sqrt{-\lambda }}(t,s)=\left\{
\begin{array}{c}
G_{\sqrt{-\lambda }}^{+}(t,s)=\frac{\sinh \sqrt{-\lambda }(1-t)\sinh \sqrt{
\lambda }s}{\sqrt{-\lambda }\sinh \sqrt{-\lambda }},~~0\leqslant
s\leqslant t\medskip \\
G_{\sqrt{-\lambda }}^{-}(t,s)=\frac{\sinh \sqrt{-\lambda }(1-s)\sinh \sqrt{
\lambda }t}{\sqrt{-\lambda }\sinh \sqrt{-\lambda }},\,~~t\leqslant
s\leqslant 1.
\end{array}
\right.
\]
Now, it is well known that H2 implies the existence of $\delta
_0\in ]0,\pi /2[$ and $\varepsilon_0>0$ such that the resolvent set of
$A-B^{2}$ contains the following sector of the complex plane
\[
S(\delta_0,\varepsilon_0)=\left\{ z\in \mathbb{C}
:\left| \arg (z)\right| \leqslant \delta_0\right\} \cup B(
0,\varepsilon_0) ,
\]
where $B(0,\varepsilon_0) $ is the open ball of radius
$\varepsilon_0$. If $\gamma $ denotes the sectorial boundary curve of
$S(\delta_0,\varepsilon_0)$ oriented positively, then the natural
representation of the solution of (1)-(2), in the abstract case, is
given by the Dunford integral
\begin{eqnarray}
u(t) &=&\frac{e^{-tB}}{2\pi i}\int\nolimits_{\gamma }g_{\sqrt{-\lambda }
}(t)(L-\lambda I)^{-1}\varphi d\lambda \nonumber \\
&&+\frac{e^{(1-t)B}}{2\pi i}\int\nolimits_{\gamma }g_{\sqrt{-\lambda }
}(1-t)(L-\lambda I)^{-1}\psi d\lambda \\
&&-\frac{1}{2\pi i}\int\nolimits_{\gamma }\int_0^1G_{\sqrt{-\lambda }
}(t,s)e^{-(t-s)B}(L-\lambda I)^{-1}f(s)ds\,d\lambda , \nonumber
\end{eqnarray}
where $L=A-B^{2}$, $D(L)=D(A)\cap D(B^{2})$,
\[
g_{\sqrt{-\lambda }}(t)=\frac{\sinh \sqrt{-\lambda }(1-t)}{\sinh \sqrt{
\lambda }}\quad t\in (0,1),\lambda \in \gamma .
\]
Here $\sqrt{-\lambda }$ is the analytic determination defined by
$Re\sqrt{-\lambda }>0$. From writing
\[
\left| g_{\sqrt{-\lambda }}(t)\right| =
\big| e^{-\sqrt{-\lambda }t}\Big(
\frac{1-e^{-2\sqrt{-\lambda }(1-t)}}{1-e^{-2\sqrt{-\lambda }}}\Big)
\big| ,
\]
we deduce that there exists two constants $c_0$ and $K_0$ such that
\begin{equation}
\begin{gathered}
\left| g_{\sqrt{-\lambda }}(t)\right| \leqslant K_0e^{-c_0\left| \lambda
\right| ^{1/2}t}\quad \forall \lambda \in \gamma ,\; \forall t\in
]0,1], \\
\left| g_{\sqrt{-\lambda }}(1-t)\right| \leqslant K_0e^{-c_0\left|
\lambda \right| ^{1/2}(1-t)}\quad \forall \lambda \in \gamma ,\; \forall
t\in [0,1[,
\end{gathered}
\end{equation}
where $c_0=\cos (\pi /2-\delta_0/2)$ and
$K_0=2/\big(1 -\exp(-2\mathbf{c}_0\sqrt{\varepsilon_0})\big)$.
On the other hand, for any $f\in C([0,1] ;E) $, we
can see that
\begin{equation}
\|\int_0^1G_{\sqrt{-\lambda }}(t,s)f(s)ds\|_{E}\leqslant
\frac{1}{\left| \lambda \right| \cos (\delta_0/2) }
\|f\|_{C(E)},\quad \forall \lambda \in \gamma .
\end{equation}
According to H1, H2 and estimates (6)-(7), all the
integrals in (5) converge absolutely for every $t\in ]0,1[$.
\section{Smoothness of the solution}
Let us set, for $\varphi $ and $\psi $ in $E$,
\begin{gather*}
U(t,-B,L)\varphi =\frac{e^{-tB}}{2\pi i}\int_\gamma g_{
\sqrt{-\lambda }}(t)(L-\lambda I)^{-1}\varphi d\lambda \quad t\in ]0,1], \\
U(1-t,B,L)\psi =\frac{e^{(1-t)B}}{2\pi i}\int_\gamma g_{
\sqrt{-\lambda }}(1-t)(L-\lambda I)^{-1}\psi d\lambda \quad t\in [0,1[,
\end{gather*}
then we have the essential lemma
\begin{lemma} \label{lm1}
Under assumptions H1 and H2
\begin{enumerate}
\item[i)] there exists a positive constant $K=K(\varepsilon_0,\delta_0)$
such that
\begin{gather*}
\|U(t,-B,L)\varphi \|_{E}\leqslant K\|\varphi \|
_{E},\quad \forall \varphi \in E, \forall t\in ]0,1], \\
\|U(1-t,B,L)\psi \|_{E}\leqslant K\|\psi \|
_{E},\quad \forall \psi \in E, \forall t\in \lbrack 0,1[,
\end{gather*}
\item[ii)] $U(.,-B,L)\varphi $ $\in C([0,1] ;E)$ if and only if
$\varphi \in \overline{D(L)}$,
\item[iii)] $U(1-.,B,L)\psi \in C([0,1] ;E)$ if and only if
$\psi \in \overline{D(L)}$.
\end{enumerate} \end{lemma}
\paragraph{Proof.} i) For $t>0$ we can write
\begin{eqnarray*}
\lefteqn{U(t,-B,L)\varphi }\\
&=&\frac{e^{-tB}}{2\pi i}\int_{\gamma_{+}^{t}}
g_{\sqrt{-\lambda }}(t)(L-\lambda I)^{-1}\varphi d\lambda
+\frac{e^{-tB}}{2\pi i}\int_{\gamma_{-}^{t}} g_{\sqrt{
-\lambda }}(t)(L-\lambda I)^{-1}\varphi d\lambda \\
&=&I_{+}+I_{-},
\end{eqnarray*}
where
\begin{equation}
\gamma_{+}^{t}=\left\{ z\in \gamma ;\,\left| z\right| \geqslant
1/t^{2}\right\} , \quad \gamma_{-}^{t}=\left\{ z\in \gamma
;\,\left| z\right| \leqslant 1/t^{2}\right\} .
\end{equation}
Then we have
$$
\|I_{+}\|_{E} \leqslant K_0(\int_{1/t^{2}}^{+\infty }
\frac{e^{-c_0\left| \lambda \right| ^{1/2}t}}{\left| \lambda \right| }
d\left| \lambda \right| ) \|\varphi \|_{E}
\leqslant 2K_0\int_{1}^{+\infty }\frac{e^{-c_0\sigma }}{\sigma }
d\sigma \|\varphi \|_{E}\leqslant K\|\varphi \|
_{E},
$$
and
\begin{eqnarray*}
I_{-} &=&\frac{e^{-tB}}{2\pi i}\int_{\gamma_{-}^{t}} (g_{
\sqrt{-\lambda }}(t)-g_{\sqrt{-\lambda }}(0)) (L-\lambda
I)^{-1}\varphi d\lambda \\
&&+\frac{e^{-tB}}{2\pi i}\int_{\gamma_{-}^{t}} (L-\lambda
I)^{-1}\varphi d\lambda \\
&=&-\frac{e^{-tB}}{2\pi i}\int_{\gamma_{-}^{t}} (
\int_0^{t}\frac{\sqrt{-\lambda }\cosh \sqrt{-\lambda }(1-s)}{\sinh \sqrt{
\lambda }}ds) (L-\lambda I)^{-1}\varphi d\lambda \\
&&-\frac{e^{-tB}}{2\pi i}\int_{C_{1/t^2}} (L-\lambda I)^{-1}\varphi d\lambda \\
&=&I_{-}'+I_{-}'',
\end{eqnarray*}
where
\[
\,C_{1/t^{2}}=\left\{ z : \left| \arg z\right| \leqslant \delta_0\mbox{
and }\left| z\right| =1/t^{2}\right\} .
\]
Then
\begin{gather*}
\|I_{-}'\|_{E}\leqslant \frac{K_0}{2\pi }
\int_{\varepsilon_0}^{1/t^{2}}\frac{\left| \lambda \right| ^{1/2}t}{
\left| \lambda \right| }d\left| \lambda \right| .\|\varphi \|
_{E}\leqslant K\|\varphi \|_{E}, \\
\|I_{-}''\|_{E}\leqslant \frac{1}{2\pi }
\int_{-\delta_0}^{+\delta_0}\|(L-\frac{e^{i\theta }}{
t^{2}}I) ^{-1}\varphi \|_{E}\frac{d\theta }{t^{2}}\leqslant
K\|\varphi \|_{E}.
\end{gather*}
Similarly we obtain
\[
\|U(1-t,B,L)\psi \|_{E}\leqslant K\|\psi \|
_{E},\quad \forall \psi \in E\mbox{ , }\forall t\in [0,1[.
\]
ii) Fix $\varepsilon >0$ and let $\varphi \in \overline{D(L)}$, then there
exists $y\in D(L)$ such that
\begin{equation}
\|\varphi -y\| \leqslant \varepsilon .
\end{equation}
Using the identity
\[
(L-\lambda I)^{-1}y=\frac{(L-\lambda I)^{-1}Ly}{\lambda }-\frac{y}{\lambda },
\]
we have
\[
U(t,-B,L)y=\frac{e^{-tB}}{2\pi i}\int_\gamma g_{\sqrt{
\lambda }}(t)\frac{(L-\lambda I)^{-1}}{\lambda }Lyd\lambda ,
\]
which gives
\begin{equation}
\|U(t,-B,L)y\|_{E}\leqslant \frac{1}{2\pi }\int_{\varepsilon
_0}^{\infty }\frac{K}{\left| \lambda \right| ^{2}}d\left| \lambda \right|
\|Ly\|_{E}\leqslant K\|Ly\|_{E}~\forall
t\in [0,1].
\end{equation}
Now, from the equality
\[
U(t,-B,L)\varphi -\varphi =U(t,-B,L)\varphi
-U(t,-B,L)y+U(t,-B,L)y-y+y-\varphi ,
\]
and the estimates (9), (10) we deduce that
\[
U(t,-B,L)\varphi -\varphi \to 0 \quad\mbox{as}\quad t\to 0^+.
\]
The continuity in $t>0$ is easily verified.
Conversely, assume $U(.,-B,L)\varphi\in C([0,1] ;E)$,
then $\lim_{t\to 0^+} U(t,-B,L)\varphi $ exists and is
necessarily equal to $\varphi$; however
\[
g_{\sqrt{-\lambda }}(t)e^{-tB}(L-\lambda I)^{-1}\varphi \in D(L),
\]
which implies that $U(t,-B,L)\varphi \in \overline{D(L)}$.
For statement iii) it is enough to substitute $1-t$ for $t$.
\hfill$\Box$\medskip
Let us consider, for $\theta \in \left] 0,1\right[ $, the well known real
interpolation space between $D(L)$ and $E$ characterized by
\[
D_{L}(\theta ;+\infty )=\big\{ \varphi \in E : \sup_{r>0}
r^{\theta }\|L(L-rI)^{-1}\varphi \|_{E}<\infty \big\} .
\]
(See Grisvard$ \cite{g1})$.
\begin{lemma} \label{lm2}
Under assumptions H1, H2, H3 and for $\theta \in
]0,1/2[ $ we have
\[
U(.,-B,L)\varphi \in C^{2\theta }([0,1] ;E)
\quad\mbox{if and only if}\quad \varphi \in D_{L}(\theta ;+\infty ).
\]
\end{lemma}
\paragraph{Proof.}
Let $\varphi \in D_{L}(\theta ;+\infty )$ and $\tau ,t\in ]0,1[$ such that
$\tau 0$. Then we know that
\begin{eqnarray*}
D_{L}(\theta ;+\infty ) &=&(D(L);E)_{1-\theta ,\infty } \\
&=&(W^{2m,p}(\mathbb{R});L^{p}(\mathbb{R}
))_{1-\theta ,\infty } \\
&=&\mathcal{B}_{p,\infty }^{2m\theta }(\mathbb{R}).
\end{eqnarray*}
Where this last space is a Besov space defined, for instance, in Grisvard
\cite{g2}. It is not difficult to see that H1, H2, H3, and H4
are satisfied. Applying Theorem \ref{thm1} we have the following statement.
\begin{theorem} \label{thm2}
Let $f\in C^{2\theta }([0,1];L^{p}(\mathbb{R}))$ such that for $j=0,1$
the mappings $x\mapsto f(j,x)$ belong to
$\mathcal{B}_{p,\infty }^{2m\theta }(\mathbb{R})$ and assume that
$\varphi $, $\psi $ belong to $W^{2m,p}(\mathbb{R})$ and
$\varphi ^{(2m)}$, $\psi ^{(2m)}$ belong to
$\mathcal{B}_{p,\infty }^{2m\theta }(\mathbb{R})$. Then the problem
\begin{gather*}
\partial_{t}^{2}u(t,x)+\mathcal{B}(x,\partial_{x})\partial_{t}u(t,x)+
\mathcal{A}(x,\partial_{x})\partial_{x}^{2}u(t,x)=f(t,x) \\
u(0,x)=\varphi (x), \quad u(1,x)=\psi (x),
\end{gather*}
has a unique solution $u$ satisfying \begin{enumerate}
\item[i)] $u\in C^{2}([0,1];L^{p}(\mathbb{R}))\cap
C^1([0,1],W^{m,p}(\mathbb{R}))\cap C([0,1],W^{2m,p}(
\mathbb{R}))$,
\item[ii)] $\partial_{x}^{2m}u$, $\partial_{x}^{m}\partial_{t}u$ and $\partial
_{t}^{2}u$ belong to $C^{2\theta }([0,1];L^{p}(\mathbb{R}))$.
\end{enumerate}
\end{theorem}
\paragraph{Acknowledgments.} The authors would like to express our thanks to
Professor Jerry L. Bona for his valuable remarks and suggestions concerning
this work.
\begin{thebibliography}{99} \frenchspacing
\bibitem{d1} Da Prato, G., Grisvard, P.: \textit{Sommes d'op\'{e}rateurs
lin\'{e}aires et \'{e}quations diff\'{e}rentielles op\'{e}rationnelles}, J.
Math. Pures et Appl., 54 (1975), 305-387.
\bibitem{f1} Favini, A.: \textit{Parabolicity of second order
differential equations in Hilbert space}, Semigroup Forum, 42\thinspace
(1991), 303-312.
\bibitem{g1} Grisvard, P.: \textit{Commutativit\'{e} de deux foncteurs d}$^{
\mbox{'}}$\textit{interpolation et applications}, J. Math. Pures et Appli.,
45\thinspace (\thinspace 1966),\thinspace 143-290.
\bibitem{g2} Grisvard, P.: \textit{Spazi di tracce e applicazioni, Rendiconti
di Matematica}, 4, vol. 5, s\'{e}rie VI (1972), 657-729.
\bibitem{k1} Krein, S.G.: Linear differential equations in Banach spaces,
Moscou, 1967.
\bibitem{l1} Labbas, R.: \textit{Equation elliptique abstraite du second
ordre et \'{e}quation parabolique pour le probl\`{e}me de Cauchy abstrait},
C.R. Acd. Sci. Paris, t. 305 S\'{e}rie I, (1987), 785-788, .
\bibitem{l2} Labbas, R., Moussaoui, M. and Najmi, M.:\textit{\ Singular
behavior of the Dirichlet problem in H\"{o}lder spaces of the solutions to
the \ Dirichlet problem in a cone, }Rend. Istit. Mat. Univ. Trieste,\ XXX,
(1998), 155-179.
\bibitem{l3} Liang, J., Xiao, T.: \textit{Wellposedness results for certain
classes of higher order abstract Cauchy problems connected with integrated
semigroups}, Semigroup Forum 56\thinspace (1998), 84-103.
\bibitem{l4} Lions, J.L. et Peetre, J.: \textit{Sur une classe d'espaces
d'interpolation}, Inst. Hautes Etudes Sci. Publ. Math., 19\thinspace (1964),
5-86.
\bibitem{n1} Neurbrander, F.: \textit{Integrated semigroups and their
applications to\thinspace complete second order Cauchy problems}, Semigroup
Forum 39 (1989), 233-251.
\bibitem{s1} Stone, M. H.: \textit{On one-parameter unitary groups in Hilbert
space}, Ann. Math. 33 (1932) 643-648.
\end{thebibliography}
\noindent\textsc{Abdelillah El Haial} \\
Universit\'{e} du Havre, I.U.T, \\
B.P 4006, 76610 Le Havre, France
\smallskip
\noindent\textsc{Rabah Labbas} \\
Laboratoire de Math\'{e}matiques, Facult\'{e} des Sciences
et Techniques, \\
Universit\'{e} du Havre,
B.P 540, 76058 Le Havre Cedex, France\\
e-mail: Labbas@univ-lehavre.fr
\end{document}