\documentclass[twoside]{article} \usepackage{amsfonts} % used for R in Real numbers \pagestyle{myheadings} \markboth{ Existence of non-negative solutions } { Cecilia S. Yarur } \begin{document} \setcounter{page}{359}{ \title{\vspace{-1in}\parbox{\linewidth}{\footnotesize\noindent USA-Chile Workshop on Nonlinear Analysis, \newline Electron. J. Diff. Eqns., Conf. 06, 2001, pp. 359--367.\newline http://ejde.math.swt.edu or http://ejde.math.unt.edu \newline ftp ejde.math.swt.edu or ejde.math.unt.edu (login: ftp)} \vspace{\bigskipamount} \\ % Existence of non-negative solutions \\ for a Dirichlet problem % \thanks{ {\em Mathematics Subject Classifications:} 35C20, 35D10. \hfil\break\indent {\em Key words:} Dirichlet Problem, Non-negative solutions. \hfil\break\indent \copyright 2001 Southwest Texas State University. \hfil\break\indent Published January 8, 2001. \hfil\break\indent Partially supported by Fondecyt grant 1990877 and DICYT } } \date{} \author{ Cecilia S. Yarur } \maketitle \begin{abstract} The aim of this paper is the study of existence of non-negative solutions of fundamental type for some systems without sign restrictions on the non linearity. \end{abstract} \newtheorem{theorem}{Theorem}[section] \newtheorem{corollary}[theorem]{Corollary} \renewcommand{\theequation}{\thesection.\arabic{equation}} \catcode`@=11 \@addtoreset{equation}{section} \catcode`@=12 \newcommand{\ds}{\displaystyle} \newcommand{\no}{\noindent} \newcommand{\ep}{\varepsilon} \newcommand{\la}{\lambda} \newcommand{\al}{\alpha} \newcommand{\be}{\beta} \newcommand{\de}{\delta} \newcommand{\R}{\mathbb{R}} \section{Introduction} We study the existence of non-negative non-trivial solutions to the boundary-value problem \begin{eqnarray}\label{example0} &\Delta u = a_2 v^{p_2}-a_1 v^{p_1} \quad\mbox{in } B'&\nonumber\\ &\Delta v = b_2 u^{q_2}-b_1 u^{q_1} \quad\mbox{in }B' &\\ & u= v= 0 \quad \mbox{on} \partial B\,, &\nonumber \end{eqnarray} where $a_i,b_i$ are non-negative constants, $p_i>0$, $q_i>0$ for $i=1,2$, $B$ is the unit ball centered at zero in $\mathbb{R}^N$, $N\ge 3$, and $B'= B\setminus \{0\}$. The above problem involves many problems of a quite different nature depending on the values of $a_i$, $b_i$. For instance, if $a_2=b_2=0$ the solutions $u,v$ are sub-harmonic functions, while if $a_1=b_1=0$ the solutions are super-harmonic. For a better understanding of this system, we recall that P.L. Lions \cite{li}, Ni and Sacks \cite{nsa}, and Ni and Serrin \cite{ns}, studied conditions for existence or non existence of non-negative solutions $u$ to \begin{equation}\label{lions} -\Delta u = u^q \quad \mbox{in $B'$,} \quad u = 0 \quad\mbox{on $\partial B$}. \end{equation} The range of existence of solutions to (\ref{lions}) is $ q < (N+2)/(N-2)$. On the other hand, the problem $$ \Delta u = u^q \quad \mbox{in $B'$,} \quad u = 0 \quad\mbox{on $ \partial B$}, $$ has a non-negative non-trivial solution if and only if $q < N/(N-2)$, see \cite{bv} for the non-existence and \cite{v} for existence and related problems. We state next some known results concerning particular cases of problem (\ref{example0}). Assume first that $ a_1= b_1 =0$. Thus, we are concerned with \begin{eqnarray}\label{example01} &-\Delta u = a_2 v^{p_2} \quad \mbox{in $B'$}&\nonumber\\ &-\Delta v = b_2 u^{q_2} \quad \mbox{in $B'$}&\\ &u=v=0 \quad \mbox{on $\partial B$.}&\nonumber \end{eqnarray} The following result is well known. \begin{theorem} Assume that $a_1=b_1=0$, $p_2q_2>1$ and $a_2>0$, $b_2>0$. Then, there exists a classical solution to (\ref{example01}) if and only if $$ \frac{N}{p_2+1}+ \frac{N}{q_2+ 1 } > N-2.$$ \end{theorem} Troy \cite{tr} proved radial symmetry of positive classical solutions to problem (\ref{example01}). The existence of positive classical solutions of (\ref{example01}) was studied by Hulshof and van der Vorst \cite{hv} and de Figueiredo and Felmer \cite{ff}. The behavior of solutions was studied by Bidaut-V\'eron in \cite{biv1}. The existence of some singular solutions, that is solutions with either \\ $\limsup_{x\to 0}u(x)=+\infty$ or $\limsup_{x\to 0}v(x)=+\infty$, is given by Garc\'{\i}a-Huidobro, Man\'asevich, Mitidieri and Yarur, see \cite{gmmy}. Using Pohozaev-Pucci-Serrin type identity, Mitidieri \cite{m1,m2} and van der Vorst \cite{vv} proved non existence of classical solutions of (\ref{example01}). Non-existence of radially symmetric singular positive solutions was given by Garcia-Huidobro, Man\'asevich, Mitidieri and Yarur in \cite{gmmy}. We note that since $u$ and $v$ are super-harmonic functions, and due to a result of Brezis and P.L.Lions \cite{bl}, $u^{q_2}\in L^1(B)$ , $v^{p_2}\in L^1(B)$ and there exist $c\ge 0$ and $d\ge 0$ such that \begin{eqnarray*} &-\Delta u = a_2 v^{p_2}+ c\delta_0 \quad \mbox{in $\mathcal{D'}(B)$}& \\ &-\Delta v = b_2u^{q_2} + d\delta_0 \quad \mbox{in $\mathcal{D'}(B)$}&\\ & u= v= 0 \quad \mbox{on $\partial B$.}& \end{eqnarray*} If $(c, d) \not= (0, 0)$ we call this singularity of {\em fundamental} type. Let us consider now $ a_1= b_2 = 0$, in (\ref{example0}). Hence, we are looking for the solutions of: \begin{eqnarray}\label{example02} &-\Delta u = a_2 v^{p_2} \quad \mbox{in $B'$} &\nonumber\\ &\Delta v = b_1 u^{q_1} \quad \mbox{in $B'$}&\\ & u= v=0 \quad \mbox{on $\partial B$.}&\nonumber \end{eqnarray} Since $v$ is sub-harmonic, there exists no non-negative classical solutions to (\ref{example02}). The following result is given in \cite{cidy1} for $p_2q_1>1$ and in \cite{cidy2} for $p_2q_1 <1$. \begin{theorem} Assume $a_1=b_2=0$, $p_2>0$, $q_1>0$, and $p_2q_1 \not=1$. Then there exists a non-trivial non-negative solution to (\ref{example02}) if and only if $$ \frac{N}{p_2+1} +\frac{N-2}{q_1 + 1} > N-2, \quad \mbox{and}\quad p_2 < N/(N-2).$$ \end{theorem} The above result is based on the results given in \cite{bg}. We say that $(u, v)$ has a {\em strong} singularity at $0$ if either $$\limsup_{x\to 0} |x|^{N-2}u(x)= +\infty \quad \mbox{ or} \quad \limsup_{x\to 0} |x|^{N-2}v(x)= +\infty\,.$$ It can be proved that there exists a region in the plane $p_2-q_1$ where there exist both {\em strong} and {\em fundamental} non-negative singular solutions, see \cite{cidy2}. This region is given by $$ \frac{N-2}{p_2+ 1} + \frac{N}{q_1 + 1} > N-2, \quad \mbox{and}\quad p_2 < N/(N-2) < q_1 \,.$$ Assume now that $ a_2= b_2 =0$, and thus the problem (\ref{example0}) is \begin{eqnarray}\label{example03} &\Delta u = a_1 v^{p_1} \quad \mbox{in $B'$}&\nonumber\\ &\Delta v = b_1 u^{q_1} \quad \mbox{in $B'$}&\\ &u=v=0 \quad\mbox{on $\partial B$,}&\nonumber \end{eqnarray} Since $u$ and $v$ are sub-harmonic we have non existence of non-negative solutions with either $u$ or $v$ bounded. In \cite{bg2} and \cite{y} it was proved non existence of positive solutions if either $$ \frac{N}{p_1+1} + \frac{N-2}{q_1 +1} \le N-2, \quad \mbox{or}\quad \frac{N-2}{p_1+1} + \frac{N}{q_1 + 1}\le N-2.$$ If $ a_1= 0$ ( similarly for $b_1=0$) we have \begin{eqnarray}\label{example04} &-\Delta u = a_2 v^{p_2} \quad \mbox{in $B'$}&\nonumber\\ & -\Delta v =b_2u^{q_2}- b_1 u^{q_1}\quad \mbox{in $B'$}&\\ &u=v=0 \quad\mbox{on $\partial B$} \end{eqnarray} The following result was proved in \cite{cidy3}. \begin{theorem} \label{corollaryintro} Let $p_2>0$, $q_1>0$ and $ q_2>0$. Let $a_1=0$, $a_2\ge 0$, $b_1\ge 0$ and $b_2 \ge 0$. Assume that for $i=1,2$ we have \begin{equation}\label{hipotesis0} p_2 < \frac{N}{N-2}, \quad \frac{N}{p_2+1}+\frac{N-2}{q_i+1} > N-2. \end{equation} Assume that one of the following holds: \begin{enumerate} \item[(i)] $p_2q_i > 1$ for all $i=1, 2$. \item[(ii)] $p_2q_i < 1$, for all $i=1,2$. \item[(iii)] If $p_2q_i=1$ for some $i=1,2$ then $a_2^{p_2}b_i$ is sufficiently small. \item [(iv)] $p_2q_i < 1 < p_2q_j$, for some $i,j= 1, 2$, $i\not=j$, and $a_2^{p_2}b_i$ is sufficiently small. \end{enumerate} Then, there exist $d_*\ge 0$, $d^*> 0$ with $d_* < d^*$ such that for any $d\in (d_*, d^*)$, there exists $(u , v)$ a non-negative solution to (\ref{example04}) satisfying $$\lim_{x \to 0}|x|^{N-2}(u(x), v(x))= (0, d).$$ Moreover, if $p_2q_i \ge 1,$ $i=1, 2$ then $d_*= 0$, and if $p_2q_i \le 1$, $i=1, 2$ then $d^*= \infty$. \end{theorem} For the general case we have the following previous result, see \cite{cidy3}. \begin{theorem} \label{corollaryintro1} Let $p_1>0$, $p_2>0$, $q_1>0$ and $ q_2 >0$. Let $a_i$, $b_i$ $i=1, 2$ be non-negative constants. Assume that \begin{equation}\label{hipotesis} p_i < \frac{N}{N-2}, \quad q_i < \frac{N}{N-2}, \quad i= 1, 2\end{equation} Assume that one of the following holds: \begin{enumerate} \item[(i)] $p_iq_j > 1$, for all $i,j=1, 2$. \item[(ii)] $p_iq_j < 1$, for all $i,j =1,2$. \item[(iii)] If $p_iq_j=1$ for some $i=1,2$ and some $j=1,2$ then $a_i^{p_i}b_j$ is sufficiently small. \item [(iv)] $p_iq_j <1 0$, $d>0$ and $(u , v)$ a non-negative solution to (\ref{example0}) such that $$\lim_{x \to 0}|x|^{N-2}(u(x), v(x))= (c, d).$$ \end{theorem} Here we prove the following general existence result of non negative non-trivial solutions to (\ref{example0}). Set \begin{equation}\label{defgamma} \Gamma(p,q):=\frac{N-2}{p+1}+\frac{N}{q+1}-(N-2). \end{equation} \begin{theorem} \label{teorema} Let $p_i, \ q_i$, $i=1,2$, positive numbers. Then, there exists a nonnegative nontrivial solution $(u, v)$ of (\ref{example0}) if one of the following holds: \begin{itemize} \item[(i)] $a_1>0$, $b_1>0$, and $p_2 < N/(N-2), \ q_2< N/(N-2)$ $$\min\{\Gamma(p_1, q_1), \Gamma(q_1, p_1), \Gamma(p_2, q_1),\Gamma(q_2, p_1)\}>0,$$ with small coefficient $a_j$ ( respectively $b_j$ ) for some $j=1,2$ if $p_j \le 1$ ( respectively $q_j\le 1$) and $1 \le \max\limits_{i=1,2}\{p_i, q_i\}$. \item[(ii)] $a_1=0$, $b_1>0$, $p_2 < N/(N-2)$ and $$\min\{\Gamma(q_1, p_2), \Gamma( q_2, p_2)\} >0,$$ with small coefficient $a_2$ ( respectively $b_j$) if $p_2\le 1$ (respectively $q_j\le1$) for some $j= 1, 2$, and $ 1 \le \max\limits_{i=1,2}\{p_2, q_i\}$. \item[(iii)] $a_1=0 = b_1$, and $$\max\{\Gamma(p_2, q_2), \Gamma(q_2, p_2)\}>0,$$ with small coefficient $a_2$ ( respectively $b_2$) if $p_2\le 1$ (respectively $q_2\le1$) and $1 \le\max\{p_2, q_2\}$. \end{itemize} \end{theorem} \section{Proof of Theorem \ref{teorema}} We note that for $p$ and $q$ non-negative numbers the condition $\Gamma(p,q)>0$ is equivalent to $$ p(2-(N-2)q) + N >0.$$ Moreover, if $pq>1$, $$\displaylines{ \Gamma(p,q) = (\zeta -(N-2))(pq-1), \quad \zeta=\frac{2(p+1)}{pq-1}\cr \Gamma(q,p) = (\xi -(N-2))(pq-1),\quad \xi= \frac{2(q+1)}{pq-1}. }$$ Recall that $u(x)=C_1|x|^{-\zeta}$, $v(x)= C_2 |x|^{ -\xi}$ for some positive constants $C_1$ and $C_2$ is a non-negative solution of $$ -\Delta u = v^p, \quad -\Delta v = u^q $$ if $\Gamma(p,q) < 0$ and $\Gamma(q,p)<0$. This particular solution also plays a fundamental role for example for the system $$ -\Delta u = v^p, \quad \Delta v= u^q,$$ where this solution exists if $\Gamma(p, q) < 0$ and $\Gamma(q, p)>0$. \paragraph{Proof of Theorem \ref{teorema}.} Set $$ f_i(t)=a_i t^{p_i}, \ g_i(t)= b_i t^{q_i}, \ i=1,2.$$ We will construct radially symmetric non-negative solutions to (\ref{example0}), by monotone iteration as follows. Let $d>0$, $(u_1,v_1)= (0, d m), $ where $m(r):= |x|^{2-N}-1$ and let $(u_n, v_n)$ be given by $(u_{n+1}, v_{n+1}) = T(u_n, v_n)$ where $T=(T_1, T_2)$ is the operator given by \begin{eqnarray}\label{operador2} T_1(u, v)(r)&=&\int_r^1s^{1-N}\ds\int_s^1 t^{N-1}f_1(v(t))dt ds \nonumber\\ && + \int_r^1s^{1-N}\int_0^s t^{N-1}f_2(v(t))dt ds ,\\ T_2(u,v)(r)&=& d m(r)+ \int_r^1s^{1-N}\ds\int_s^1 t^{N-1}g_1(u(t))dt ds \nonumber\\ &&+ \int_r^1s^{1-N}\int_0^s t^{N-1}g_2(u(t))dt ds,\nonumber \end{eqnarray} We are looking for $\alpha$, $\delta$ and $C$ such that $$T_1(C r^{\alpha},Cr^{\delta}) \le C r^{\alpha}, \ T_2(C r^{\alpha},Cr^{\delta})\le Cr^{\delta} $$ and $v_1= d m(r) \le C r^{\delta}$. Hence, the sequence $(u_n, v_n)$ satisfies $$u_n \le C r^{\alpha}, \ v_n\le Cr^{\delta} \ \mbox{ for all $n\in \mathbb{N}$}, $$ and the convergence of $(u_n, v_n)$ to a solution of (\ref{example0}) follows. To find $C$, $d$, $\alpha$ and $\delta$ we use the following: Let $\kappa $ be any number such that $\kappa + N\not=0$, and define $$\phi(\kappa):= \min\{ 2-N, \kappa+ 2 \}. $$ Then \begin{equation}\label{bound11} m_{\kappa}(r):=\int_r^1s^{1-N}\ds\int_s^1 t^{N-1+\kappa}dt ds\le K r^{\phi(\kappa)}, \end{equation} where $K=K(N, \kappa)$. Moreover, for any $\kappa$ satisfying $\kappa+N>0$, and $\kappa +2 \not= 0, $ set $$\psi(\kappa):= \min\{0, \kappa + 2 \}. $$We have \begin{equation}\label{bound10} h_{\kappa}:=\int_r^1s^{1-N}\ds\int_0^s t^{N-1+\kappa}dt ds\le K r^{{\psi(\kappa)}}, \end{equation} where $K=K(N, \kappa)$. Hence, $$ T_1(Cr^{\alpha},Cr^{\delta})= a_1C^{p_1} m_{p_1\delta} + a_2C^{p_2} h_{p_2\delta}.$$ From (\ref{bound11}) and (\ref{bound10}) and if we choose $p_1\delta +N\not=0$ and $p_2\delta +N >0$ we obtain \begin{equation}\label{looking} T_1(Cr^{\alpha},Cr^{\delta})\le K\left( a_1C^{p_1} r^{\phi(p_1\delta)} + a_2C^{p_2} r^{\psi(p_2\delta)}\right). \end{equation} We note that $\phi(p_1\delta) \le 2-N < \psi(p_2\delta), $ and thus \begin{equation} \label{found1} T_1(Cr^{\alpha},Cr^{\delta})\le K\left( a_1C^{p_1} + a_2C^{p_2}\right) r^{\sigma}, \end{equation} where $$\sigma:= \left\{ \begin{array}{lll} \phi(p_1\delta) && \mbox{if $a_1\not=0$,}\\ \psi(p_2\delta) &&\mbox{if $a_1= 0$}.\end{array}\right. $$ Therefore, if $\alpha\le \sigma$ and $K( a_1C^{p_1} + a_2C^{p_2})\le C$, we obtain $T_1(Cr^{\alpha},Cr^{\delta})\le C r^{\alpha}$. Arguing as above, we have \begin{equation}\label{looking2} T_2(Cr^{\alpha},Cr^{\delta})\le dr^{2-N}+ K\left( b_1C^{q_1} r^{\phi(q_1\alpha)} + b_2C^{q_2} r^{\psi(q_2\alpha)}\right), \end{equation} with $q_1\alpha + N \not=0$, $q_2\alpha + 2\not=0$, and $q_2\alpha+N >0$. Therefore, \begin{equation} \label{found2} T_2(Cr^{\alpha},Cr^{\delta})\le \left( d+ K\left( b_1C^{q_1} + b_2C^{q_2}\right)\right) r^{\rho}, \end{equation} where $$\rho:= \left\{ \begin{array}{lll} \phi(q_1\alpha) && \mbox{if $b_1\not=0,$}\\ 2-N &&\mbox{if $b_1= 0$.}\end{array}\right.$$ Hence, $$T_2(Cr^{\alpha},Cr^{\delta})\le C r^{\delta},$$ if $\delta \le \rho$ and $d+ K(b_1C^{q_1} + b_2C^{q_2})\le C$. Next we prove the existence of $\alpha$, $\delta$, $C$ and $d$ under the hypothesis of the theorem. The existence of $C$ and $d$, is classical. We can choose $d= C/2$ and for $i=1, 2$ $$ Ka_iC^{p_i}\le C/2, \quad Kb_iC^{q_i}\le C/4. $$ Therefore, if either for all $i$ , $p_i< 1$ and $q_i < 1$, or $p_i>1$ and $q_i>1$, the existence of $C$ follows. By the contrary if $\max\{p_i, q_i, i=1,2 \} \ge 1$ and $\min\{p_i, q_i, i=1,2 \}\le 1$, we obtain existence with a restriction on the coefficients. We summarize the conditions that $\alpha$ and $\delta$ must satisfy as follows: $$ \alpha \le \left\{ \begin{array}{ll} \min\{2-N, p_1\delta +2 \} & \mbox{if $a_1\not= 0$ }\\ \min\{0, p_2\delta+2\} &\mbox{if $a_1=0$ } \end{array}\right. $$ $$ \delta \le \left\{ \begin{array}{ll} \min\{2-N, q_1\alpha +2 \} & \mbox{if $b_1\not= 0$ }\\ 2-N & \mbox{if $b_1=0$}. \end{array}\right. $$ Moreover, we need that \begin{equation}\label{hiperbolas2} \ p_2\delta + N >0, \quad q_2\alpha + N >0, \end{equation} We also used that $ p_1\delta + N \not=0, \ q_1\alpha + N \not=0, \ p_2\delta + 2\not=0, \ q_2\alpha +2 \not=0$. These last conditions are not relevant since we can take $\alpha$ and $\delta$ smaller and hence these new $\alpha$ and $ \delta$ satisfy the conditions. \paragraph{Case (i).} Assume first that $a_1>0$ and $b_1>0$. If $p_1 < N/(N-2)$, and since $ \Gamma(p_1, q_1) >0$, and $\Gamma(p_2, q_1)>0$, we can take $$\alpha = 2-N, \ \delta= \min\{2-N, 2-(q_1-\varepsilon)(N-2) \},$$ where $\varepsilon>0$ is such that $$\Gamma(p_1, q_1-\varepsilon)>0, \ \mbox{ and $ \Gamma(p_2, q_1-\varepsilon) >0.$}$$ Now, since $q_20$. From $p_2< N/(N-2)$ and $\Gamma(p_2, q_1-\varepsilon)>0$, we also have $p_2\delta+N>0$. It remains to prove that $\alpha=2-N \le p_1\delta +2$, which follows easily from $\Gamma(p_1, q_1-\varepsilon)>0$. If $p_1 \ge N/(N-2)$, from $\Gamma(p_1, q_1)>0$ we deduce that $q_1< N/(N-2)$. Thus, we may proceed as before but now with $$ \delta= 2-N, \ \alpha= p_1(2-N)+2.$$ \paragraph{Case (ii).} Assume that $a_1= 0$ and $b_1 >0$. Let us choose $$\delta = 2-N,\ \alpha= \min\{0, 2-(p_2-\varepsilon)(N-2)\},$$ where $\varepsilon>0$ is such that $\Gamma(q_1, p_2-\varepsilon)>0$ and $\Gamma(q_2, p_2-\varepsilon) >0$. Then the conclusion follows as in the above case. \paragraph{Case (iii).} Assume that $b_1=0$ and $a_1=0$. Assume that $p_2\le q_2$. Since $\Gamma(q_2, p_2)>0$, then $p_20$. \begin{thebibliography}{00} \bibitem{bl} H. Brezis, P.L. Lions, A note on isolated singularities for linear elliptic equations, {\em Jl. Math. Anal. and Appl.}, {\bf 9A} (1981), 263--266. \bibitem{biv1} M. F. Bidaut-V\'eron, Local behaviour of solutions of a class of nonlinear elliptic systems, {\em Adv. Differential Equations}, to appear. \bibitem{bg} M.F.Bidaut-V\'eron and P. Grillot, Asymptotic behaviour of elliptic system with mixed absorption and source terms, {\em Asymptot. Anal.,} (1999), {\bf 19}, 117-147. \bibitem{bg2} M.F.Bidaut-V\'eron and P. Grillot, Singularities in elliptic systems, preprint. \bibitem{bv} H. Brezis and L. V\'eron, Removable singularities of some nonlinear elliptic equations {\em Archive Rat. Mech. Anal.} {\bf 75} (1980), . \bibitem{cidy1} C. Cid and C. Yarur, A sharp existence result for a Dirichlet problem - The superlinear case, to appear in Nonlinear Anal. \bibitem{cidy2} C. Cid and C. Yarur, Existence of solutions for a sublinear system of elliptic equations, {\em Electron. J. Diff. Eqns.}, {\bf 2000} (2000), No. 33, 1--11. (http://ejde.math.swt.edu) \bibitem{cidy3} C. Cid and C. Yarur, in preparation. \bibitem{ff} P. Felmer and D.G. de Figueiredo, On superquadratic elliptic systems, {\em Trans. Amer. Math. Soc.} {\bf 343} (1994), 99--116. \bibitem{gmmy} M. Garc\'{\i}a-Huidobro, R. Man\'asevich, E. Mitidieri, and C. Yarur Existence and noexistence of positive singular solutions for a class of semilinear systems, {\em Archive Rat. Mech. Anal.} {\bf 140} (1997), 253-284. \bibitem{hv} J. Hulsholf and R.C.A.M. van der Vorst, Differential systems with strongly indefinite variational structure, {\em J. Func. Anal.} {\bf 114} (1993), 32-58. \bibitem{li} P.L. Lions, Isolated singularities in semilinear problems, {\em J. Differential Equations} {\bf 38} (1980),441-450. \bibitem{m1} E. Mitidieri, Nonexistence of positive solutions of semilinear elliptic systems in $\mathbb{R}^N$, {\em Quaderno Matematici} {\bf 285}, University of Trieste 1992. \bibitem{m2} E. Mitidieri, A Rellich type identity and applications, {\em Comm. in P.D.E.} {\bf 18 } (1993), 125-151. \bibitem{nsa} W.M. Ni and P. Sacks, Singular behavior in nonlinear parabolic equations, {\em Trans. Amer. Mat. Soc. } {\bf 287} (1985),657-671. \bibitem{ns} W.M. Ni and J. Serrin, Nonexistence theorems for singular solutions of quasilinear partial differential equations, {\em Comm. Pure Appl. Math.} {\bf 38} (1986), 379-399. \bibitem{tr} W.C. Troy, Symmetry properties of semilinear elliptic systems, {\em J. Differential Equations} {\bf 42} (1981), 400--413. \bibitem{v} L. V\'eron, Singular solutions of some nonlinear elliptic equations, {\em Nonlinear Analysis, Methods \& Applications } {\bf 5} (1981), 225-242. \bibitem{vv} Van der Vorst, R.C.A.M., Variational identities and applications to differential systems, {\em Arch. Rational Mech. Anal.} {\bf 116 } (1991), 375--398. \bibitem{y} C. Yarur, Nonexistence of positive singular solutions for a class of semilinear elliptic systems, {\em Electron. J. Diff. Eqs.}, {\bf 1996} (1996), No. 8, 1--22. (http://ejde.math.swt.edu) \end{thebibliography} \noindent{\sc Cecilia S. Yarur}\\ Departamento de Matematicas \\ Universidad de Santiago de Chile \\ Casilla 307, Correo 2, Santiago, Chile \\ email: cyarur@fermat.usach.cl \end{document}