Non existence and other properties for solutions of quasilinear elliptic equations (Q1343195)

The authors present results about nonexistence and asymptotic behaviour for \(r \to \infty\) of positive solutions tending to zero as \(r \to \infty\) to variational equations of the form \[ {\left[ g(r) {dG \over dp} (u_ r) \right]}_ r + g(r)f(r,u) = 0, \quad r \geq 0. \tag{*} \] Beside smoothness they assume: \(G(0) = dG/dp(0) =0\), \(dG/dp(p)=O(| p |^{m-1})\) as \(p \nearrow 0\) for some \(m>1\), \(G\) is strictly convex, \(g(0) = 0\), \(g>0\), \(g' \geq 0\) on \(\mathbb{R}^ +\). In the special case \(g(r) = r^{n-1}\), \((*)\) is the polar form of some quasilinear elliptic equation for radially symmetric positive solutions in \(\mathbb{R}^ n\), where corresponding results have been obtained e.g. by Ni and Serrin. In particular the principal part of \((*)\) may be the degenerate \(m\)-Laplace operator or the mean curvature operator. Assuming \(r \mapsto g(r)\) to behave like \(r^ \beta\), \(\beta > m - 1\), \(f(r,u) \geq Cr^ \nu u^ q\), \(q>0\), \(\nu + m > 0\) for large \(r\) and small \(u>0\) and an integrability condition on \(f\) with respect to \(r\) they show: \(\bullet\) If \(q \leq (m-1) (\beta + \nu + 1)/(\beta + 1 - m)\) then there is no positive solution to \((*)\) tending to zero as \(r \to \infty\). \(\bullet\) If \(q > (m - 1)(\beta + \nu + 1)/(\beta + 1 - m)\) then every positive solution to \((*)\) tending to zero as \(r \to \infty\) behaves like \(u(r) = O(r^{-(m + \nu)/(q - m + 1)})\), \(r \to \infty\). In the second part of their paper, imposing more restrictive conditions on \(G\), \(g\) and \(f\), the authors are able to widen the range of \(q\), where nonexistence results hold. For example, if \(G(p) = | p |^ m\), \(g(r) = r^ \beta\), \(f(r,u) = u^ q\), where \(m > 1\), \(q > 0\), \(\beta > m - 1\), then for \(q < ((m-1) (\beta + 1) + m)/(\beta + 1 - m)\) there is no positive solution to \((*)\), which tends to 0 as \(r\to\infty\). The crucial device is a variational identity of \textit{P. Pucci} and \textit{J. Serrin} [Indiana Univ. Math. J. 35, 681-703 (1986; Zbl 0625.35027)], which develops ideas of Rellich and Pohožaev.