\documentclass[twoside]{article}
\usepackage{amsfonts, amsmath }
\pagestyle{myheadings}
\markboth{\hfil High regularity for a nonlinear parabolic problem
\hfil EJDE--2002/48}
{EJDE--2002/48\hfil L. Barbu, Gh. Moro\c sanu, \& W. L. Wendland \hfil}
\begin{document}
\title{\vspace{-1in}\parbox{\linewidth}{\footnotesize\noindent
{\sc Electronic Journal of Differential Equations},
Vol. {\bf 2002}(2002), No. 48, pp. 1--12. \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} \\
%
High regularity of the solution of a nonlinear parabolic
boundary-value problem
%
\thanks{ {\em Mathematics Subject Classifications:}
35K60, 35K05, 35K20, 34G20, 47H05, 47H20.
\hfil\break\indent
{\em Key words:} Parabolic equation, nonlinear
boundary conditions, \hfil\break\indent
maximal monotone operator, subdifferential,
compatibility conditions.
\hfil\break\indent
\copyright 2002 Southwest Texas State University. \hfil\break\indent
Submitted April 11, 2002. Published May 29, 2002.
\hfil\break\indent
Gh. Moro\c sanu was supported
by the German Research Foundation DFG \hfil\break\indent
(Deutsche Forschungsgemeinschaft) under the project DFG We 659/35-2.
} }
\date{}
%
\author{Lumini\c{t}a Barbu, Gheorghe Moro\c sanu, \& Wolfgang L. Wendland}
\maketitle
\begin{abstract}
The aim of this paper is to report some results concerning
high regularity of the solution of a nonlinear parabolic
problem with a linear parabolic differential equation in
one spatial dimension and nonlinear boundary conditions.
We show that any regularity can be reached provided that
appropriate smoothness of the data and compatibility
assumptions are required.
\end{abstract}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{remark}{Remark}[section]
\numberwithin{equation}{section}
\section{Introduction}\label{s:1}
We consider the parabolic boundary value problem (BVP):
\begin{gather}
y_t-y_{xx}+ gy=f(x,t), \quad \mbox{in} \ D_T, \label{E}\\[2pt]
\begin{gathered}
y_x(0,t)\in {\beta}_1(y(0,t)),\\
-y_x(1,t)\in {\beta}_2(y(1,t)), \quad 00$ such that
$\int_a^{b-\delta}\Vert u(t+\delta)-u(t) \Vert^p_X\le C{\delta}^p $
for every $\delta\in(0, b-a]$.
\end{enumerate}
The implication $(i)\Rightarrow (ii)$ is
still valid for $p=1$. Moreover, $(ii)$ holds if $u$ is of bounded
variation on $[a,b]$ and $X$ is not necessarily reflexive.
\end{theorem}
Here and in what follows, $L^p$ and $W^{k,p}$ denote the usual
function and Sobolev spaces, respectively. Now, in order to
state the next result, let us consider two real Hilbert spaces $V$
and $H$, such that $V$ is densely and continuously embedded into
$H$. If $H$ is identified with its own dual, then we have
$V\subset H\subset V'$, algebraically and topologically. We have
denoted by $V'$ the dual of $V$. Denote also by $\langle \cdot ,
\cdot \rangle$ the pairing between $V$ and $V'$, i.e., $\langle
v,v^*\rangle := v^*(v), \quad \mbox{for} \quad v\in V, \quad v^*\in V'$.
For $v^*\in H' \equiv H , \quad \langle v, v^*\rangle $ reduces to
the scalar product in $H$ of $v$ and $v^*$. Following
\cite[Chapter 1]{LM}, for some $-\infty < a < b < \infty$ we set
$$
W(a,b):=\{u\in L^2(a,b;V);\, u'\in L^2(a,b;V')\}\,,
$$
where $u'$ is the distributional derivative of $u$, as a
$V$-valued distribution.
\begin{theorem}\label{th:2.2.}
Every $u\in W(a,b)$ has a representative $u_1\in C([a,b];H)$ and
so $u$ can be identified with such a continuous function.
Furthermore, if $u,\, \tilde u\in W(a,b)$, then the function
$t\mapsto \langle u(t),\tilde u(t)\rangle$
is absolutely continuous on $[a,b]$ and
$$
{\frac{d}{dt}}\langle u(t), \tilde u(t)\rangle = \langle u(t),
\tilde u'(t)\rangle + \langle \tilde u(t), u'(t)\rangle \quad \mbox{for a.a.} \quad t\in (a,b),
$$
and, hence, in particular,
$$
{\frac{d}{dt}}\Vert u(t)\Vert ^2_H = 2\langle u(t), u'(t)\rangle \quad \mbox{for a.a.} \quad t\in (a,b)\,.
$$
\end{theorem}
Finally, we recall the following theorem due to H. Attouch and A.
Damlamian \cite{AD} which will be needed for the derivation of
higher regularity of the solution of the BVP:
\begin{theorem}\label{th:2.3.} Let
$A(t)=\partial\phi(t,\cdot)$ for $0\let\le T$,
where $\phi(t,\cdot)\,:\,H\to (-\infty,+\infty]$ are proper,
convex, and lower semi-continuous functions, with a domain of
definition $D(\phi(t,\cdot))=D$ which is independent of $t$. Here
$H$ is a real Hilbert space. Assume further that there exists a
nondecreasing function $\gamma$$:$$[0,T]$$\to $$\mathbb{R}$ and
some real constants $C_1$, $C_2$ such that
%
\begin{equation}\label{eq:2.1}
\phi(t,v)\le\phi(s,v)+[\gamma (t) - \gamma (s)]\cdot [\phi (s,v) +
C_1\Vert v\Vert ^2_H +C_2]
\end{equation}
%
for all $v\in D$ and $0\le s\le t\le T$. Then, for every $u_0\in
D$ and $f\in L^2(0,T; H)$, there exists a unique solution $u\in
W^{1,2}(0,T; H)$ of the equation $u'(t)+A(t)u(t)=f(t)$ for a.a.
$t\in (0,T)$ with the initial condition $u(0)=u_0$. Moreover, there
exists a function $h\in L^1(0,T)$ such that
%
\begin{equation}\label{eq:2.2}
\phi (t,u(t))\le \phi (s,u(s))+\int _s^t h(r)dr \quad \mbox{for all} \
\ 0\le s\le t\le T\,.
\end{equation}
\end{theorem}
Here, we denote by $\partial \phi (t, \cdot )$ the subdifferential
of the function $\phi (t, \cdot )$. In what follows, we shall use
the theory of evolution equations associated with monotone
operators in Hilbert spaces. For details we refer to
\cite{Ba,Br2,M}.
\section{High Regularity of Solutions}\label{s:3}
\setcounter{equation}{0} If $\beta _1, \ \beta _2$ are maximal
monotone mappings, then existence and uniqueness of the solution
to the BVP is well known. The most important results, even for
$n$ dimensions, were established by H. Brezis \cite{Br1,Br2}. Our
problem can be expressed as a Cauchy problem in $L^2(0,1)$,
associated with a subdifferential and, hence, existence is
available (see, e.g., Brezis's theorem in \cite[p. 56]{M}, where
the regularizing effect of the subdifferential on the initial data
is pointed out). For a more general problem than ours see
\cite[Chapter 3, \S 3]{M}, where the existence to a higher order,
one-dimensional, parabolic equation is discussed.
\begin{theorem}\label{th:3.1.}
Assume that
%
\begin{gather}\label{eq:3.1}
f\in W^{1,1}(0,T;L^2(0,1))\,;\\
\label{eq:3.2}
{\beta}_1, \ {\beta}_2 \quad \mbox{are both maximal monotone}\,; \\
\label{eq:3.3} y_0\in H^2(0,1)\,;
\end{gather}
and that the following first-order compatibility conditions are
fulfilled:
%
\begin{equation}\label{eq:3.4}
y'_0(0)\in {\beta}_1(y_0(0)), \quad -y'_0(1)\in
{\beta}_2(y_0(1))\,.
\end{equation}
%
Then the BVP (\ref{E}), (\ref{BC}), (\ref{IC}) has a unique strong solution
%
\begin{equation}\label{eq:3.5}
y\in W^{1,\infty}(0,T;L^2(0,1))\cap L^\infty(0,T;H^2(0,1))\cap
W^{1,2}(0,T;H^1(0,1))\,.
\end{equation}
\end{theorem}
\paragraph{Proof:} Let $H=L^2(0,1)$ with
$$ \langle p,q\rangle = \int_0^1 p(x)q(x)\, dx, \quad \Vert p
{\Vert}_H^2 = \int_0^1 p(x)^2dx\, .
$$
We define the operator $A\,:\,D(A)\subseteq H\to H$ as
$$
Ap = -p'' + gp
$$
on the domain of definition
%
\begin{eqnarray*}
D(A)&=&
\big\{p\in H^2(0,1) ; p(0)\in D(\beta _1), \\
&& p(1)\in D(\beta _2), p'(0)\in {\beta}_1(p(0)),
-p'(1)\in {\beta}_2(p(1))\big\}\,.
\end{eqnarray*}
%
Then the BVP may be written as the Cauchy problem in $H$,
%
\begin{equation}\label{eq:3.6}
\begin{gathered}
y'(t)+Ay(t)=f(t), \quad \ 00$, we
obtain from $(3.22)$ and $(3.23)$ the two estimates
%
\begin{eqnarray}\label{eq:3.24}
\lefteqn{
\frac{1}{2}{\frac{d}{dt}}\Vert z(\cdot , t+h)-z(\cdot ,t)\Vert _H^2 + \Vert z_x(\cdot ,t+h) - z_x(\cdot ,t)\Vert _H^2}\notag\\
&\le&
\Vert f_t(\cdot ,t+h) -f_t(\cdot ,t)\Vert _H\cdot \Vert z(\cdot ,t+h)-z(\cdot ,t)\Vert _H \\
&&+
{\frac{1}{2}}(\Vert z_x(\cdot , t+h)-z_x(\cdot ,t)\Vert _H^2 +
\Vert
z(\cdot ,t+h) - z(\cdot ,t)\Vert _H^2) + {\frac{1}{2}}C_4h^2\,,\notag
\end{eqnarray}
and
\begin{eqnarray} \label{eq:3.25}
\lefteqn{
{\frac{1}{2}}{\frac{d}{dh}}\Vert z(\cdot ,h) - z_0\Vert _H^2 +
\Vert z_x(\cdot ,h)-z_0'\Vert _H^2}\notag\\
& \le& \Vert f_h(\cdot ,h)
-B(0)z_0\Vert _H\cdot \Vert z(\cdot ,h)-z_0\Vert _H\\
&&+{\frac{1}{2}}(\Vert z_x(\cdot ,h) - z_0'\Vert _H^2 + \Vert z(\cdot ,h) - z_0\Vert _H^2) +
{\tfrac{1}{2}}C_5h^2\,, \notag
\end{eqnarray}
%
where $C_4$ and $C_5$ are some positive constants. Therefore,
%
\begin{eqnarray}\label{eq:3.26}
\lefteqn{ {\tfrac{d}{dt}}(e^{-t}\Vert z(\cdot ,t+h)-z(\cdot ,t)\Vert _H^2) +e^{-t}\Vert z_x(\cdot ,t+h)-z_x(\cdot ,t)\Vert _H^2 }\notag\\
&\le&
2e^{-t}\Vert f_t(\cdot ,t+h)-f_t(\cdot ,t)\Vert _H\cdot \Vert z(\cdot
,t+h)-z(\cdot ,t)\Vert _H + C_4h^2\,,
\end{eqnarray}
%
and
%
\begin{equation}\label{3.27}
{\tfrac{d}{dh}}(e^{-h}\Vert z(\cdot ,h)-z_0\Vert _H^2)\le
C_5h^2+2e^{-h}\Vert f_h(\cdot ,h)-B(0)z_0\Vert _H \cdot \Vert z(\cdot
,h)-z_0\Vert _H\,.
\end{equation}
%
If we drop the second term in the left--hand side of $(3.26)$ and
integrate the resulting inequality over $[0,t]$, then, by using a
Gronwall type lemma (see, e.g., \cite[p. 47]{M}) we arrive at the
estimate
%
\begin{eqnarray}\label{eq:3.28}
\Vert z(\cdot ,t+h) - z(\cdot ,t)\Vert _H
&\le&
C_6\{ \Vert z(\cdot ,h)-z_0\Vert _H + h \notag\\
&&\mbox{}
+\int_0^t\Vert f_s(\cdot ,s+h)-f_s(\cdot ,s)\Vert _Hds\}\,.
\end{eqnarray}
%
Now, from $(3.25)$ we obtain in a similar way
%
\begin{equation}\label{eq:3.29}
\Vert z(\cdot ,h) -z_0\Vert _H \le C_7(h^{3/2} + \int_0^h\Vert
f_s(\cdot ,s) -B(0)z_0\Vert _Hds)\,.
\end{equation}
%
Finally, $(3.28)$ and $(3.29)$ imply that $z_t\in
L^{\infty}(0,T;H)$. Using this conclusion together with $(3.24)$,
we
get, by using of Theorem 2.1, that $z_x\in W^{1,2}(0,T; H)$.
This concludes the proof.
\begin{remark}\label{r:3.2} \rm
If $f$ in Theorem 3.2 is assumed to be more regular with respect
to $x$, then $y$ is also more regular with respect to $x$, because
$$
y_{tt}=y_{xxt}-gy_t+f_t = y_{xxxx}-gy_{xx}+f_{xx}-gy_t+f_t.
$$
On the other hand, by reasoning similarly as in the proof of
Theorem 3.2, one can show that $y\in W^{3,2}(0,T;H^1(0,1))\cap
W^{3,\infty}(0,T;L^2(0,1))$ under appropriate assumptions on the
smoothness of $\beta _1, \ \beta _2, \ y_0, \ f$ and
compatibility. The proof needs a slight modification since the
boundary conditions corresponding to $(3.20)$ are now
inhomogeneous. Fortunately, the inhomogeneous terms there are
$H^1$-functions and so Theorem 2.3 is again applicable, with a
slight change of $\phi $. The corresponding $t$-dependent operator
is nonlinear because its domain is an affine subset of $L^2(0,1)$.
Here we will not further present these details. Of course, higher
regularity with respect to $x$ can also be obtained at the expense
of additional regularity and compatibility assumptions. Actually,
our procedure can be repeated as many times as we want to and so
any regularity of the solution with respect to $t$ and $x$ can be
reached under sufficient smoothness of the data and compatibility
conditions. More precisely, the $H^k(D_T)$-regularity can be shown
for every $k$ and, thus, the $C^k(\overline{D_T})$-regularity can
be obtained as well for every $k$, due to Sobolev's embedding
theorem.
\end{remark}
%
\begin{remark}\label{3.3} \rm
Theorem 3.2 is formulated in such a way that the next level of
regularity can be obtained. Note, however, that this is a
parabolic problem so that, in order to get, for instance,
$$
y\in W^{2,2}(0,T;L^2(0,1))\cap W^{1,\infty}(0,T; H^1(0,1)),
$$
we can relax our assumptions. More precisely, it suffices to
assume that
$$
f\in W^{1,2}(0,T;L^2(0,1)), \quad f(\cdot ,0) \in H^1(0,1);
$$
$\beta _1$ and $\beta _2$ satisfy $(3.11)$; $y_0\in H^3(0,1)$, and
$y_0$ satisfies $(3.4)$ with equalities instead of inclusions.
\end{remark}
\begin{remark}\label{r:3.4} \rm
Here, we do not discuss nonlinear cases of the equation (\ref{E}),
since this would require a special treatment. However, let us
point out some immediate extensions of the above results.
\end{remark}
%
Assume that the linear term $gy$ of Equation (\ref{E}) is replaced by
a nonlinear one, say $g(y)$, where $g:\mathbb{R}\longrightarrow \mathbb{R}$
is a single-valued, continuous and nondecreasing function. Then
Theorem 3.1 is still valid. If, in addition, $g$ is a
$W_{loc}^{2,\infty }(\mathbb{R})$-function, then Theorem 3.2 is
also valid for this nonlinear case with
$$
z_0(x)=f(x,0)+y_0''(x)-g(y_0(x)).
$$
Indeed, $(3.17)$ remains true, because $g$ is Lipschitzian on
bounded sets and $y$ takes values in a bounded set. So, again,
$z=y_t\in W^{1,2}(0,T;V')$. The remainder of the proof of Theorem
3.2 continues with slight modifications. In particular, in Eq.
$(3.20)_1$ the term $gz$ must be replaced by $g'(y)z$ and $B(t)$
becomes
$$
B(t)p=-p'' + g'(y(\cdot ,t))p,
$$
with the same domain of definition as before. This means that in
the definition of the energy function $\phi (t,p)$, the term
$(1/2)g\int _0^1p(x)^2dx$ should be replaced by
$(1/2)\int _0^1g'(y(x,t))p(x)^2dx$. Finally, here $\gamma $ can be
chosen in the specific form
$$
\gamma (t) = K\int _0^t(\vert g_1'(\tau )\vert + \vert
g_2'(\tau )\vert + \Vert g'' (y(\cdot
,\tau))y_{\tau}(\cdot ,\tau)\Vert _{L^{\infty}(0,1)})d\tau .
$$
Then all the conclusions can be derived by similar arguments as
above.
On the other hand, if the (\ref{BC}) are inhomogeneous and
(\ref{E}) is nonlinear, then a transformation like (\ref{eq:1.1}) would lead
us to an evolution equation where the spatial operator becomes
time-dependent. Instead, we still can keep the inhomogeneous form
of (\ref{BC}) and associate with our problem an energy functional
$\phi (t, \cdot )$ which also includes the inhomogeneous
terms.
\begin{thebibliography}{99} \frenchspacing
\bibitem{AD}
H. Attouch, A. Damlamian, Strong solutions for parabolic
variational inequalities, {\it Nonlinear Anal. TMA} \textbf{2}
(1978), No. 3, 329--353.
%
\bibitem{BWCM}
L. Barbu, E. Cosma, Gh. Moro\c sanu, W.L. Wendland, Asymptotic
analysis of the telegraph system with nonlinear boundary
conditions, in Vol: {\it Partial Differential Equations, Integral
Equations, and Optimal Control Problems} (S. Aizicovici and N.H.
Pavel, eds), Dekker, (2001) 63--76.
%
\bibitem{BM1}
L. Barbu, Gh. Moro\c sanu, Asymptotic analysis of the telegraph
equations with non-local boundary value conditions,
{\it PanAmerican Math. J.} \textbf{8} (1998), No. 4, 13--22.
%
\bibitem{BM2}
L. Barbu, Gh. Moro\c sanu, A first order asymptotic expansion of
the solution of a singularly perturbed problem for the telegraph
equations , {\it Applicable Analysis} \textbf{72} (1999), No. 1-2,
111--125.
%
\bibitem{Ba}
V. Barbu, {\it Nonlinear Semigroups and Differential Equations in
Banach Spaces}, Noordhoff, Leyden, 1976.
%
\bibitem{Br1}
H. Brezis, Problemes unilateraux, {\it J. Math. Pures Appl.}
\textbf{5} (1972), 1--168.
%
\bibitem{Br2}
H. Brezis, {\it Operateurs Maximaux Monotones et Semigroupes de
Contractions dans les Espaces de Hilbert}, North Holland,
Amsterdam, 1973.
%
\bibitem{GK}
M.S. Ghausi, J.J. Kelly, {\it Introduction to
Distributed-Parameter Networks With Applications to Integrated
Circuits}, Robert E. Krieger Publ. Co., Inc., Huntington, New
York, 1977.
%
\bibitem{LSU}
O.A. Ladyzenskaja, V.A. Solonnikov, and N.N. Ural'ceva, {\it
Linear and Quasilinear Equations of Parabolic Type}, Translations
of Mathematical Monographs, Vol. \textbf{23}, AMS, Providence, Rhode
Island, 1968.
%
\bibitem{LM}
J.L. Lions, E. Magenes, {\it Problemes aux Limites Non Homogenes
et Applications}, Vol. 1, Dunod, Paris, 1968.
%
\bibitem{MN}
C.A. Marinov and P. Neittaanmaki, {\it Mathematical Models in
Electrical Circuits: Theory and Applications}, Kluwer, Dordrecht,
1991.
%
\bibitem{M}
Gh. Moro\c sanu, {\it Nonlinear Evolution Equations and
Applications}, Reidel, Dordrecht, 1988.
\end{thebibliography}
\bigskip
\noindent\textsc{Lumini\c{t}a Barbu}\\
Department of Mathematics and Informatics\\
Ovidius University\\
Blvd. Mamaia 124\\
8700 Constan\c ta, Romania\\
e-mail: lbarbu@univ-ovidius.ro
\medskip
\noindent\textsc{Gheorghe Moro\c sanu}\\
Department of Mathematics\\
``Al.~I.~Cuza'' University\\
Blvd. Carol I, 11\\
6600 Ia\c{s}i,
Romania\\
e-mail: gmoro@uaic.ro
\medskip
\noindent\textsc{ Wolfgang L. Wendland} \\
Mathematisches Institut A\\
University of Stuttgart\\
Pfaffenwaldring 57\\
70569 Stuttgart, Germany\\
e-mail: wendland@mathematik.uni-stuttgart.de
\end{document}