\documentclass[twoside]{article}
\usepackage{amsfonts, amsmath} % used for R in Real numbers
\pagestyle{myheadings}
\markboth{Existence of solutions to an elliptic equation }
{ Marie-Fran\c{c}oise Bidaut-V\'{e}ron }
\begin{document}
\setcounter{page}{23}
\title{\vspace{-1in}\parbox{\linewidth}{\footnotesize\noindent
2001-Luminy conference on Quasilinear Elliptic and Parabolic Equations
and Systems,\newline
Electronic Journal of Differential Equations,
Conference 08, 2002, pp 23--34. \newline
http://ejde.math.swt.edu or http://ejde.math.unt.edu
\newline ftp ejde.math.swt.edu (login: ftp)}
\vspace{\bigskipamount} \\
%
Necessary conditions of existence for an elliptic equation with
source term and measure data involving $p$-Laplacian
%
\thanks{ {\em Mathematics Subject Classifications:} 35J70, 35B45, 35D05.
\hfil\break\indent
{\em Key words:} Degenerate quasilinear equations, measure data,
capacities, a priori estimates.
\hfil\break\indent
\copyright 2002 Southwest Texas State University. \hfil\break\indent
Published October 21, 2002.} }
\date{}
\author{Marie-Fran\c{c}oise Bidaut-V\'{e}ron}
\maketitle
\begin{abstract}
We study the nonnegative solutions to equation
$$
-\Delta_{p}u=u^{q}+\lambda\nu,
$$
in a bounded domain $\Omega$ of $\mathbb{R}^{N}$, where $1
p-1$, $\nu$ is a nonnegative Radon measure on $\Omega$,
and $\lambda>0$ is a parameter.
We give necessary conditions on $\nu$ for existence, with
$\lambda$ small enough, in terms of capacity. We also give a priori
estimates of the solutions.
\end{abstract}
\numberwithin{equation}{section}
\newtheorem{theorem}{Theorem}[section]
\section{Introduction}\label{intro}
Let $\Omega$ be a bounded regular domain in $\mathbb{R}^{N}$. We
denote by $\mathcal{M}(\Omega)$ the set of Radon measures on
$\Omega$, $\mathcal{M}^{+}(\Omega)$ the set of nonnegative ones,
and by $\mathcal{M}_{b}(\Omega),\mathcal{M}_{b}^{+}(\Omega)$ the
subsets of bounded ones. We consider the quasilinear elliptic
problem with a source term:
\begin{equation}
\begin{gathered}
-\Delta_{p}u=-\mathop{\rm div}(| \nabla u| ^{p-2}\nabla u)=| u|
^{q-1}u+\mu,\quad\text{in }\Omega,\\
u=0,\quad\text{on }\partial\Omega,
\end{gathered} \label{un}
\end{equation}
with $1
p-1$, and $\mu\in\mathcal{M}_{b}^{+}(\Omega)$. We
look for conditions on the measure $\mu$ ensuring that the
problem admits a nonnegative solution, and essentially in terms of
capacity. In order to take account of the size of the measure, we
will study the problem with
\[
\mu=\lambda\nu,\quad\text{ }\lambda\geq0,
\]
where $\nu\in\mathcal{M}_{b}^{+}(\Omega)$ is fixed and $\lambda$
is a parameter. Recall a result of \cite{BaPi} in case $p=2$ ,
$N\geq3$, which gives a necessary and sufficient condition for
existence:
\begin{theorem}[\cite{BaPi}]\label{bapi}The following problem:
\begin{equation}
\begin{gathered}
-\Delta u=u^{q}+\lambda\nu,\quad\text{in }\Omega,\\
u=0,\quad\text{on }\partial\Omega,
\end{gathered} \label{2la}%
\end{equation}
where $\nu\in\mathcal{M}_{b}^{+}(\Omega)$, $\nu\neq0$, has a
nonnegative solution (in the integral sense) if and only if
\begin{equation}
\lambda\int_{\Omega}\varphi d\nu\leq\frac{q-1}{q^{q'}}\int_{\Omega
}\varphi^{1-q'}(-\Delta\varphi)^{q'}dx,\label{star}
\end{equation}
for any $\varphi\in W_{0}^{1,\infty}(\Omega)\cap W^{2,\infty}(\Omega)$ such
that $-\Delta\varphi\geq0$, with compact support in $\Omega$.
\end{theorem}
Thus if $q$ is subcritical, that means $q1$, $\mathop{\rm cap}{}_{m,r}$ is the capacity associated to the Sobolev space $W_{0}%
^{m,r}({\Omega})$, defined by
\[
\mathop{\rm cap}{}_{m,r}(K,{\Omega})=\inf\left\{ \left\| \psi\right\| _{W_{0}%
^{m,r}({\Omega})}^{r}:\psi\in\mathcal{D}({\Omega}),0\leq\psi
\leq1,\psi=1\text{ on }K\text{ }\right\} .
\]
In fact it was proved in \cite{AdPi} that (\ref{cap2}) is also sufficient:
\begin{theorem} [\cite{AdPi}]
Assume that $\nu$ has a compact support in $\Omega$. Then problem
(\ref{2la}) has a solution for any $\lambda\geq0$ small enough if and only if
there exists $C>0$ such that (\ref{cap2}) holds.
\end{theorem}
Condition (\ref{cap2}) implies that $\mu$ does not charge the sets
with $2,q'$- capacity zero. But it is stronger: if $q>N/(N-2)$ (resp.
$q=N/(N-2)$), there exists a function $\nu\in L^{s}(\Omega)$ with $1\leq
s0$. $\medskip$
Concerning problem (\ref{un}) with $p\neq2$, the question is much
harder, because the full duality argument used in \cite{BaPi}
cannot be used for the $p$-Laplacian. The first thing is to define
a notion of solution, as it is the case for the problem without
reaction term. In Section \ref{ens} we recall the usual notions of
entropy solutions, which suppose that the measure is bounded; this
leads to assume that $u^{q}\in L^{1}(\Omega)$. We denote by
\[
\overline{P}=\frac{N(p-1)}{N-p}%
\]
the critical exponent linked to the $p$-Laplacian, and we set
\[
q^{\ast}=q/(q-p+1),
\]
(hence $q^{\ast}=q'$ if $p=2$). In Section \ref{nec} we prove our main result:
\begin{theorem} \label{T1}
Let $\nu\in\mathcal{M}_{b}^{+}(\Omega)$ and $\lambda\geq0$. Assume
that problem%
\begin{equation}
\begin{gathered}
-\Delta_{p}u=u^{q}+\lambda\nu,\quad\text{in }\Omega,\\
u=0,\quad\text{on }\partial\Omega,
\end{gathered} \label{pla}
\end{equation}
has a nonnegative entropy solution (hence $u^{q}\in L^{1}(\Omega)$). Then for
any $R>pq^{\ast}$, there exists $C=C(N,p,q,R,\Omega)>0$ such that%
\begin{equation}
\lambda\int_{\Omega}\varphi d\nu+\int_{\Omega}u^{q}\varphi\,dx
\leq C\Big(\int_{\Omega}\varphi^{1-R}| \nabla\varphi|
^{R}dx\Big) ^{pq^{\ast}/R},\label{map}
\end{equation}
for any $\varphi\in W_{0}^{1,p}(\Omega)\cap W^{1,s}(\Omega)$ $(s>N)$ such that
$0\leq\varphi\leq1$ in $\Omega$. And for any $\alpha<0$, there exists
$C=C(\alpha,N,p,q,R,\Omega)>0$ such that
\begin{equation}
\int_{\Omega}(u+1)^{\alpha-1}|\nabla u|^{p}\varphi\, dx
\leq C\Big(1+\int_{\Omega}u^{q}\varphi\,dx\Big)
\Big( \int_{\Omega}\varphi^{1-R}| \nabla\varphi| ^{R}dx\Big) ^{p/R}.\label{mip}%
\end{equation}
\end{theorem}
This Theorem gives a priori estimate not only of the size of the
measure, but also of the integral $\int_{\Omega}u^{q}\varphi\,dx$,
independently on\textit{ }$u$. In the case $p=2$, this was first
remarked by \cite{DFLN} when $\mu=0$ ; it was the starting point
for proving $L^{\infty}$ universal estimates. It was also used in
\cite{BiVi} and \cite{BiY} for obtaining a priori estimates with a
general measure $\mu$. As a consequence we deduce the following:
\begin{theorem}
\label{T2} If problem (\ref{pla}) has a solution, then, for any $R>pq^{\ast}$, there exists $C=C(N,p,q,R,\Omega)>0$ such that%
\begin{equation}
\lambda\int_{K}d\nu\leq C\;(\mathop{\rm cap}{}_{1,R}
(K,{\Omega}))^{pq^{\ast}/R}, \quad\text{for every compact set
$K\subset\Omega$}.\label{lup}
\end{equation}
and if $\nu$ has a compact support in $\Omega$, there exists $C=C(N,p,q,R,\mu
)>0$ such that
\begin{equation}
\lambda\int_{K}d\nu\leq C\;(\mathop{\rm cap}{}_{1,R}(K,\mathbb{R}^{N}))^{pq^{\ast}/R}%
,\quad\text{for every compact set $K\subset\Omega$}.\label{lip}
\end{equation}
In particular, if $q>\overline{P}$, then $\nu$ does not charge the
point sets. Moreover for any $1\leq s0$,
problem (\ref{pla}) admits no solution.\medskip
\end{theorem}
In Section \ref{open}, we mention some partially or fully open
problems linked to this study. We refer to \cite{Bi2} for more
complete results for problem (\ref{un}) with possible signed
measure $\mu$, and for the problem
with an absorption term%
\begin{equation}
\begin{gathered}
-\Delta_{p}u+| u| ^{q-1}u=\mu,\quad\text{in }\Omega,\\
u=0,\quad\text{on }\partial\Omega.
\end{gathered} \label{abs}
\end{equation}
\section{Entropy solutions}\label{ens}
First recall some well-known results concerning the problem
\begin{equation}
\begin{gathered}
-\Delta_{p}u=\mu,\quad\text{in }\Omega,\\
u=0,\quad\text{on }\partial\Omega,
\end{gathered} \label{basic}
\end{equation}
with $\mu\in\mathcal{M}_{b}(\Omega)$. We set
\[
P_{0}=\frac{2N}{N+1},\quad P_{1}=2-\frac{1}{N},
\]
so that $1P_{0}$
$\Longleftrightarrow\overline{P}>1$. When $p>P_{1}$, problem
(\ref{basic}) admits at least a solution $u$ in the sense of
distributions, such that $u\in W_{0}^{1,r}(\Omega)$ for any $1\leq
r<\overline{P}$ . In the general case, one can define a notion of
entropy or renormalized solutions in four equivalent ways, see
\cite{DMOP}, which allow to give a sense to the gradient in any
case: they are solutions such that $\nabla T_{k}(u)\in
L_{loc}^{1}(\Omega)$ for any $k>0$, where
\begin{equation}
T_{k}(s)=\begin{cases}
s, &\text{if } |s| \leq k,\\
k \mathop{\rm sign}(s),& \text{if }|s| >k,
\end{cases} \label{tk}
\end{equation}
and the gradient of $u$, denoted by $y=\nabla u$ is defined by
\begin{equation}
\nabla(T_{k}(u))=y\times1_{\left\{ | u| \leq k\right\} }%
\quad\text{a.e. in }\Omega.\label{yg}%
\end{equation}
For any $p>1$ there exists at least an entropy solution of (\ref{basic}), and
it is unique if $\mu\in L^{1}(\Omega)$. Moreover any entropy solution
satisfies the equation in the sense of distributions. The role of $P_{0}$ and
$P_{1}$ is shown by the estimates
\begin{gather*}
u^{p-1} \in L^{s}(\Omega),\quad\text{for any }1\leq sP_{1}$
and $u$ itself is in $L^{1}(\Omega)$ if and only if $p>P_{0}$. \medskip
Recall that any measure $\mu\in\mathcal{M}_{b}(\Omega)$ can be decomposed as
\[
\mu=\mu_{0}+\mu_{s}^{+}-\mu_{s},%
\]
where $\mu_{0}\in\mathcal{M}_{0,b}(\Omega)$, set of bounded measures such
that
\begin{equation}
\mu_{0}(B)=0\quad\text{for any Borel set }B\subset\Omega\text{ such that
}\mathop{\rm cap}{}_{1,p}(B,\Omega)=0;\label{mu0}%
\end{equation}
and $\mu_{s}^{+},\mu_{s}^{-}$ are nonnegative and concentrated on a set $E$
with $\mathop{\rm cap}{}_{1,p}(E,\Omega)=0$. If $\mu\in\mathcal{M}_{b}^{+}(\Omega)$, then
$\mu_{0}$ is nonnegative, and $\mu=\mu_{0}+\mu_{s}^{+}$. \medskip
We will use one of the four equivalent definitions of solution: $u$ is an
entropy solution if $u$ is measurable and finite $a.e$. in $\Omega$, and
\begin{equation}
T_{k}(u)\in W_{0}^{1,p}(\Omega)\quad\text{for every }k>0,\label{gtk}%
\end{equation}
and the gradient defined by (\ref{yg}) satisfies
\begin{equation}
| \nabla u| ^{p-1}\in L^{r}(\Omega),\quad\text{for any }1\leq
rN$, such
that $h(u)\varphi\in W_{0}^{1,p}(\Omega).\medskip$
In the same way, for given $\mu=\mu_{0}+\mu_{s}^{+}\in\mathcal{M}_{b}%
^{+}(\Omega)$, a nonnegative entropy solution $u$ of problem
(\ref{un}) will be a measurable function $u$ such that $u^{q}\in
L^{1}(\Omega)$ and $u$ is an entropy solution of problem
\[
\begin{gathered}
-\Delta_{p}u=\mu-u^{q}\quad\text{in }\Omega,\\
u=0\quad\text{on }\partial\Omega.
\end{gathered}
\]
In particular
\begin{equation*}
\int_{\Omega}| \nabla u| ^{p-2}\nabla u.\nabla(h(u)\varphi
)dx+\int_{\Omega}u^{q}h(u)\varphi\,dx=\int_{\Omega}h(u)\varphi d\mu
_{0}+h(+\infty)\int_{\Omega}\varphi d\mu_{s}^{+}, \label{ploc}%
\end{equation*}
for any $h$ and $\varphi$ as above.
\section{Proofs and comments}\label{nec}
\paragraph{Proof of Theorem \ref{T1}}
Let $\mu=\lambda\nu=\mu_{0}+\mu_{s}^{+}$, where
$\mu_{0}\in\mathcal{M}_{0,b}(\Omega)$ and $\mu_{s}^{+}$ is
singular, and let $\alpha\in\left( 1-p,0\right) $ be a parameter.
For any $k>0$, we set $u_{k}=T_{k}(u)$, and, for any
$\varepsilon\in\left( 0,k\right) $,
\[
h_{\alpha,k,\varepsilon}(r)=(T_{k}(r^{+})+\varepsilon)^{\alpha}
=\begin{cases}
\varepsilon^{\alpha}, &\text{if }r\leq0,\\
(r+\varepsilon)^{\alpha}, &\text{if }0\leq r\leq k,\\
(k+\varepsilon)^{\alpha}, &\text{if }r\geq k.
\end{cases}
\]
We choose in (\ref{ploc}) the test functions
$h=h_{\alpha,k,\varepsilon}$, and $\varphi\in
W_{0}^{1,p}(\Omega)\cap W^{1,s}(\Omega)$, with $s>N$ and
$\varphi\geq0$ in $\Omega$, and obtain%
\begin{align*}
\int_{\Omega}(u_{k}&+\varepsilon)^{\alpha}\varphi d\mu_{0}+(k+\varepsilon
)^{\alpha}\int_{\Omega}\varphi d\mu_{s}^{+}+\int_{\Omega}(u_{k}+\varepsilon
)^{\alpha}u^{q}\varphi\,dx\\
+| \alpha|& \int_{\Omega}\int_{\Omega}(u_{k}+\varepsilon
)^{\alpha-1}|\nabla u_{k}|^{p}\varphi\,dx\\
=&\int_{\Omega}(u_{k}+\varepsilon
)^{\alpha}| \nabla u| ^{p-2}\nabla u.\nabla\varphi\,dx\\
\leq&\int_{\Omega}(u_{k}+\varepsilon)^{\alpha}| \nabla u_{k}|
^{p-1}| \nabla\varphi| dx+\int_{\left\{ u\geq k\right\} }%
(u_{k}+\varepsilon)^{\alpha}| \nabla u| ^{p-1}|
\nabla\varphi| dx\\
\leq&\frac{| \alpha| }{2}\int_{\Omega}(u_{k}+\varepsilon
)^{\alpha-1}|\nabla u_{k}|^{p}\text{ }\varphi\,dx+C\int_{\Omega}(u_{k}%
+\varepsilon)^{\alpha+p-1}\varphi^{1-p}| \nabla\varphi|
^{p}\text{ }dx\\
& +(k+\varepsilon)^{\alpha}\int_{\left\{ u\geq k\right\} }| \nabla
u| ^{p-1}| \nabla\varphi| dx,
\end{align*}
where $C=C(\alpha)>0$.
Now from H\"{o}lder inequality, setting
$\theta=q/(p-1+\alpha)>1$,
\begin{multline*}
\int_{\Omega}(u_{k}+\varepsilon)^{\alpha+p-1}\varphi^{1-p}|
\nabla\varphi| ^{p}\text{ }dx \\
\leq \Big( \int_{\Omega}(u_{k}+\varepsilon)^{q}
\varphi\,dx\Big)^{1/\theta}\Big( \int_{\Omega}
\varphi^{1-p\theta'}| \nabla\varphi| ^{p\theta'
}dx\Big) ^{1/\theta'}.
\end{multline*}
In particular for any $k>1$,
\begin{multline}
\frac{| \alpha| }{2}\int_{\Omega}\int_{\Omega}(u_{k}
+\varepsilon)^{\alpha-1}|\nabla u_{k}|^{p}\varphi\,dx\\
\leq C\Big( \int_{\Omega}(u_{k}+\varepsilon)^{q}\varphi\,dx\Big)
^{1/\theta}\Big( \int_{\Omega}\varphi^{1-p\theta'}|
\nabla\varphi| ^{p\theta'}dx\Big) ^{1/\theta'}
+\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx. \label{mir}
\end{multline}
Letting $\varepsilon$ tend to $0$, we get
\begin{align}
\frac{| \alpha| }{2}\int_{\Omega}u_{k}^{\alpha-1}|\nabla
u_{k}|^{p}\varphi\,dx
\leq &C\Big( \int_{\Omega}u_{k}^{q}\varphi\,dx\Big)
^{1/\theta}\Big( \int_{\Omega}\varphi^{1-p\theta'}|
\nabla\varphi| ^{p\theta'}dx\Big) ^{1/\theta'
}\nonumber\\
& +\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx. \label{mil}
\end{align}
Choosing now $h(u)=1$ in (\ref{ploc}), with the same $\varphi$, we
find
\begin{align}
\int_{\Omega}\varphi d\mu_{0}&+\int_{\Omega}\varphi d\mu_{s}^{+}+\int_{\Omega
}u^{q}\varphi\,dx=\int_{\Omega}| \nabla u| ^{p-2}\nabla
u.\nabla\varphi\,dx \nonumber\\
\leq&\int_{\Omega}u_{k}^{(\alpha-1)/p'}| \nabla u|
^{p-1}u_{k}^{(1-\alpha)/p'}| \nabla\varphi|
dx+\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx\nonumber\\
\leq&\Big( \int_{\Omega}u_{k}^{\alpha-1}| \nabla u_{k}|
^{p}\text{ }\varphi\,dx\Big) ^{1/p'}\Big( \int_{\Omega}
u_{k}^{(1-\alpha)(p-1)}\varphi^{1-p}| \nabla\varphi|
^{p}dx\Big) ^{1/p}\nonumber\\
& +\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx.\label{zer}%
\end{align}
Since $q>p-1$, we can fix $\alpha\in(1-p,0)$ such that
$\tau=q/(1-\alpha)(p-1)>1$.
From (\ref{mil}) and (\ref{zer}), we derive%
\begin{align*}
\int_{\Omega}&\varphi d\mu+\int_{\Omega}u^{q}\varphi\,dx\\
\leq&\Big( \int_{\Omega}u_{k}^{\alpha-1}| \nabla u_{k}|
^{p}\text{ }\varphi\,dx\Big) ^{1/p'}\Big( \int_{\Omega}u_{k}%
^{q}\varphi\,dx\Big) ^{1/\tau p}\Big( \int_{\Omega}\varphi^{1-\tau
'p}| \nabla\varphi| ^{\tau'p}dx\Big)
^{1/\tau'p}\\
& +\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx\\
\leq&\Big( C\Big( \int_{\Omega}u_{k}^{q}\varphi\,dx\Big) ^{1/\theta
}\Big( \int_{\Omega}\varphi^{1-p\theta'}| \nabla
\varphi| ^{p\theta'}dx\Big) ^{1/\theta'}
+\int_{\left\{ u\geq k\right\} }| \nabla u| ^{p-1}|
\nabla\varphi| dx\Big) ^{1/p'}\\
& \times\Big( \int_{\Omega}u_{k}^{q}\varphi\,dx\Big) ^{1/\tau p}\left(
\int_{\Omega}\varphi^{1-\tau'p}| \nabla\varphi|
^{\tau'p}dx\right) ^{1/\tau'p}+\int_{\left\{ u\geq
k\right\} }| \nabla u| ^{p-1}| \nabla\varphi| dx.
\end{align*}
Now we can let $k$ tend to $\infty$, since $u^{q}+| \nabla u|
^{p-1}\in L^{1}(\Omega)$. It follows that
\begin{align} \label{cst}
\int_{\Omega}\varphi d\mu+\int_{\Omega}u^{q}\varphi\,dx
&\leq C\Big( \int_{\Omega}u^{q}\varphi\,dx\Big)
^{1/p'\theta+1/\tau p} \\
&\times \Big( \int_{\Omega}\varphi^{1-p\theta'}|
\nabla\varphi| ^{p\theta'}dx\Big) ^{1/p'\theta'} \Big(
\int_{\Omega}\varphi^{1-\tau'p}| \nabla\varphi| ^{\tau'p}dx\Big)
^{1/\tau'p},\nonumber
\end{align}
with a new $C=C(\alpha,N,p,q)$. Since
$1/\theta'p'+1/\tau'p=1/q^{\ast}=1-(1/\theta p'+1/\tau p)$, we
find in particular
\begin{align*}
\Big( \int_{\Omega}&u^{q}\varphi\,dx\Big) ^{1-(p-1)/q} \\
=&\Big(\int_{\Omega}u^{q}\varphi\,dx\Big) ^{(1/p'\theta'+1/\tau'p)}\\
\leq& C\Big( \int_{\Omega}\varphi^{1-p\theta'}|
\nabla\varphi| ^{p\theta'}dx\Big) ^{1/p'%
\theta'}\Big( \int_{\Omega}\varphi^{1-\tau'p}| \nabla\varphi|
^{\tau'p}dx\Big) ^{1/\tau'p}.
\end{align*}
Consequently
\begin{align*}
\int_{\Omega}&u^{q}\varphi \,dx\\
\leq &C\Big( \int_{\Omega}\varphi^{1-p\theta
'}| \nabla\varphi| ^{p\theta'}dx\Big)
^{\tau'p/(\tau'p+p'\theta')}\Big(
\int_{\Omega}\varphi^{1-\tau'p}| \nabla\varphi|
^{\tau'p}dx\Big) ^{p'\theta'/(\tau'p+p'\theta')}.
\end{align*}
Notice that $\tau0$. Moreover, from (\ref{cst}) and (\ref{aut}),
\[
\int_{\Omega}\varphi d\mu\leq C\Big( \int_{\Omega}\varphi^{1-\tau'%
p}| \nabla\varphi| ^{\tau'p}dx\Big) ^{(q^{\ast
}-1+1/p'+1/p)/\tau'},%
\]
then
\[
\int_{\Omega}\varphi d\mu+\int_{\Omega}u^{q}\varphi \,dx\leq C\Big(
\int_{\Omega}\varphi^{1-\tau'p}| \nabla\varphi|
^{\tau'p}dx\Big) ^{q^{\ast}/\tau'}.
\]
We can choose $| \alpha| $ sufficiently small, such that
\[
pq^{\ast}2q'$ has also
a $2,q'$- capacity zero, see \cite{AdHe}. The capacity involved in
Theorem \ref{T1} is of order 1 instead of 2, because we cannot use
the full duality argument of the linear case. However, observe
that a point set has a $1,2q'$- capacity zero if and only if
$q>N/(N-2)$, that means if and only if it has a $2,q'$- capacity
zero.
\paragraph{Proof of Theorem \ref{T2}}
Let $\psi_{n}\in$ $\mathcal{D}({\Omega})$ such that
$0\leq\psi_{n}\leq1$ and $\psi_{n}\geq\chi_{K}$ and $\left\|
\psi_{n}\right\| _{W^{1,R}({\Omega})}^{R}$ tends to $\mathop{\rm cap}{}_{1,R}%
(K,\Omega)$ as $n$ tends to $\infty$. Choosing
$\varphi=\psi_{n}^{R}$ in (\ref{map}), we deduce that
\[
\lambda\int_{K}d\nu\leq C\Big( \int_{\Omega}| \nabla\psi_{n}|
^{R}dx\Big) ^{pq^{\ast}/R}\leq C\left\| \psi_{n}\right\| _{W^{1,R}%
({\Omega})}^{R},
\]
with new constants $C=C(N,p,q,R,\Omega)$, and (\ref{lup}) follows.
If $\nu$ has a compact support $X$ in $\Omega$, then (\ref{lip})
holds after localization on a neighborhood of $X$. Assume moreover
that $q>\overline{P}$, then we can choose $R$ such that
$pq^{\ast}0$ small enough, we derive
\begin{equation}
\lambda\int_{B(x_{0},r)}d\nu\leq Cr^{N-R},\label{tru}
\end{equation}
with $C=C(N,p,q,R,x_{0},\Omega)$. For any $1\leq s0$,
$\lambda\nu$ does not satisfy (\ref{tru}) for $pq^{\ast}P_{0}$,
the existence of solutions of problem (\ref{un}), with possibly
signed measure $\mu$, is shown in \cite{Gr}. In the supercritical
case, the problem is entirely open, even for $L^{s}$ functions. In
particular it would be interesting to extend to the case $p\neq2$
a consequence of Theorem \ref{bapi}:
\begin{theorem}
[\cite{BaPi}]Assume that $N\geq3$, and $\nu\in L^{s}(\Omega)$, for some
$s\geq1$. If $q>N/(N-2)$ and $s\geq N/2q'$, or $q=N/(N-2)$ and
$s>N/2q'$, then problem (\ref{2la}) has a solution for $\lambda$ small enough.\medskip
\end{theorem}
\noindent\textbf{Problem 2:} Can we solve problems (\ref{basic})
and
(\ref{pla}) if $\mu$ is not bounded?\\
Let us begin by the case without reaction term. For any
$x\in\Omega$, denote by $\rho(x)$ the distance from $x$ to
$\partial\Omega$. When $p=2$, problem (\ref{basic}) is well posed
in fact for any measure $\mu$, possibly unbounded, such that
$\int_{\Omega}\rho d| \mu| <\infty:$ it admits a unique integral
solution
\begin{equation}
u(x)=G(\mu)=\int_{\Omega}\mathcal{G}(x,y)d\mu(y),\label{int}
\end{equation}
where $\mathcal{G}$ is the Green kernel. And $u$ is also the weak
solution of the problem in the sense that $u\in L^{1}(\Omega)$ and
\begin{equation}
\int_{\Omega}u(-\Delta\xi)dx=\int_{\Omega}\xi d\mu,\label{weak}%
\end{equation}
for any $\xi\in C^{1}(\overline{\Omega})$ vanishing on
$\partial\Omega$ with $\nabla\xi$ is Lipschitz continuous, see
\cite{BiVi}. The case where $\mu$ is a function $f$, such that
$\int_{\Omega}\rho fdx<\infty$, was first considered by
Br\'{e}zis, see \cite{Ve}. The problem is open when $p\neq2:$ up
to now we have no existence results concerning equation
(\ref{basic}) when $\mu$ may be unbounded, even in the case
$p>P_{1}$, where the definition of the gradient does not cause any
problem.
Now let us consider the problem with source term. When $p=2$, it
was studied in \cite{KaVb} and specified in \cite{BrCa}:
\begin{theorem}[\cite{KaVb}]
Let $\nu\in\mathcal{M}^{+}(\Omega)$, $\nu\neq0$ such that
$\int_{\Omega}\rho d\nu<\infty$. Then problem (\ref{2la}) has a
solution such that $G(u^{q})<\infty$, $a.e$. in $\Omega$, for any
$\lambda\geq0$ small enough, if and only if there exists $C>0$
such that
\begin{equation}
G(G^{q}(\nu))\leq CG(\nu),\quad a.e.\text{ in }\Omega.\label{gns}%
\end{equation}
\end{theorem}
Notice that condition $G(u^{q})<\infty$ $a.e$. in $\Omega$, is
satisfied as soon as $\int_{\Omega}\rho fu^{q}dx<\infty$, and the
solutions are taken in the integral sense. More recently new
existence results and a priori estimates were given in \cite{BiY},
covering the case of measures $\mu$ such that
$\int_{\Omega}\rho^{\gamma}d\mu<\infty$ for some
$0\leq\gamma\leq1$. Condition (\ref{gns}) allows to obtain a
supersolution, and then a solution by using an iterative scheme.
It is available for much more general linear operators, see
\cite{KaVb} and \cite{Vb}. It seems to be difficult to extend to
nonlinear ones, since it is based on a representation formula.
However Kalton and Verbitski \cite{KaVb} also gave necessary and
sufficient in terms of capacity with weights, extending the result
of \cite{AdPi} to general measures:
\begin{theorem}[\cite{KaVb}]
Let $\nu\neq0$ be a nonnegative Radon measure on $\Omega$. Then
problem (\ref{2la}) has a solution for any $\lambda\geq0$ small enough if and
only if there exists $C>0$ such that
\[
\int_{K}d\nu\leq C\mathop{\rm cap}{}_{2,q',\rho}(K), \quad\text{
for every compact set $K\subset\Omega$},
\]
where
\[
\mathop{\rm cap}{}_{2,q',\rho}(K)=\inf\Big\{ \int_{\Omega}w^{q'}%
\rho^{1-q'}dx:w\geq0,\quad Gw\geq\rho\chi_{K}\quad a.e.\text{ in
}\Omega\Big\} .
\]
\end{theorem}
One can ask if results of this type can be obtained for the
$p$-Laplacian, using capacities of order 1 with suitable weights.
\begin{thebibliography}{99}
\bibitem{AdHe}D. R. Adams and L. I. Hedberg, Functions spaces and potential
theory, Grundlehren der Mathematischen wissenchaften, Springer Verlag, Berlin,
314 (1996).
\bibitem{AdPi}D. Adams et M. Pierre, \textit{Capacitary strong type estimates
in semilinear problems, }Ann. Inst. Fourier, 41 (1991), 117-135.
\bibitem{BaPi}P. Baras and M. Pierre, \textit{Crit\`{e}re d'existence de
solutions positives pour des \'{e}quations semi-lin\'{e}aires non monotones,}
Ann. Inst. H. Poincar\'{e}, Anal. non Lin. 2 (1985), 185-212.
\bibitem{Bi0}M-F. Bidaut-V\'{e}ron, \textit{Local and global behavior of
solutions of quasilinear equations of Emden-Fowler type,} Arc. Rat. Mech.
Anal. 107 (1989), 293-324.
\bibitem{Bi2}M-F. Bidaut-V\'{e}ron, \textit{Removable singularities and
existence for a quasilinear equation with absorption or source
term and measure data}, Adv. in Nonlinear Studies (to appear).
\bibitem{BiP}M-F. Bidaut-V\'{e}ron and S. Pohozaev, \textit{Nonexistence
results and estimates for some nonlinear elliptic problems,} J. Anal. Math.,
84 (2001), 1-49.
\bibitem{BiVi}M-F. Bidaut-V\'{e}ron and L. Vivier, \textit{An elliptic
semilinear equation with source term involving boundary measures: the
subcritical case, }Revista Matematica Iberoamericana, 16 (2000), 477-513.
\bibitem{BiY}M-F. Bidaut-V\'{e}ron and C. Yarur, \textit{Semilinear elliptic
equations and systems with measure data: existence and a priori estimates,
}Advances in Diff. Equ., 7 (2002), 257-296.
\bibitem{BrCa}H. Br\'{e}zis and X.\ Cabr\'{e}, \textit{Some simple nonlinear
PDE's without solutions}, Boll. Unione Mat. Ital. 8 (1998), 223-262.
\bibitem{BrL}H. Br\'{e}zis and P.L.\ Lions, \textit{A note on isolated
singularities for linear elliptic equation}, Math. Anal. Appl. Adv. Math.
Suppl. Stud. 7A (1981), 263-266.
\bibitem{DMOP}Dal Maso, F. Murat, L. Orsina and A. Prignet,
\textit{Renormalized solutions of elliptic equations with general measure
data}, Ann. Sc. Norm. Sup. Pisa, 28 (1999), 741-808.
\bibitem{DFLN}De Figuereido, P.L. Lions and R.D. Nussbaum, \textit{A priori
estimates and existence of positive solutions of semilinear elliptic
equations,} J. Math. Pures et Appl., 61 (1982), 41-63.
\bibitem{Gr}N. Grenon, \textit{Existence results for some semilinear elliptic
equations with measure data, }Ann. Inst. H. Poincar\'{e}, 19
(2002), 1-11.
\bibitem{KaVb}N. Kalton and I.\ Verbitsky, Nonlinear equations and weighted
norm inequalities, Trans. Amer. Math. Soc., 351, 9 (1999), 3441-3497.
\bibitem{Li}P. L. Lions, \textit{Isolated singularities in semilinear
problems,} J. Diff. Eq., 38 (1980), 441-450.
\bibitem{Vb}I. Verbitski, \textit{Superlinear equations, potential theory and
weighted norm inequalities,} Nonlinear Analysis, Funct. Spaces
Appl., Vol. 6. (1999), 223-269.
\bibitem{Ve}L. V\'{e}ron, \textit{Singularities of solutions of second order
quasilinear equations,} Pitman Research Notes in Math., Longman
Sci. \& Tech. 353 (1996).
\end{thebibliography}
\noindent\textsc{Marie-Fran\c{c}oise Bidaut-V\'{e}ron}\newline
Laboratoire de Math\'{e}matiques et Physique Th\'{e}orique,\\
CNRS UMR 6083, Facult\'{e} des Sciences,\\
Parc de Grandmont, 37200 Tours, France\\
e-mail: veronmf@univ-tours.fr
\end{document}