Liouville theorems and Harnack inequality for a class of semilinear elliptic equations (Q7007463)

The author gives logarithmic gradient estimate and Harnack inequality for a class of semilinear elliptic equations on Riemannian manifolds with (Bakry-Emery) Ricci curvature bounded below. More specifically he considers the general semilinear elliptic equation \N\[\N\Delta u+f(u)=0\text{ on }M^n,\N\]\Nwhere \(M^n\) is an \(n\)-dimensional Riemannian manifold and \(f\in C^1(0,\infty)\), obtaining the following result:\N\NTheorem. Let \((M^n,g)\) be an \(n\)-dimensional complete Riemannian manifold with \(\mbox{Ric}\geq -K\, g\), where \(K\geq 0\). If \(u\) is a classical positive solution of the previous equation on \(B(x_0,2R)\) and there exists \(\alpha\in (1, (n+3)/(n-1))\) such that \(u^{-\alpha} f(u)\) is nonincreasing on \((0,\infty)\), then\N\[\N\sup_{B(x_0,R)}\frac{\vert\nabla u\vert^2}{u^2}\leq C\left(K+\frac{1}{R^2}\right),\N\]\Nwhere \(C=C(n,\alpha)\) and \(B(x_0,R)\) denotes the open geodesic ball with center \(x_0\) and radius \(R\).\N\NThe direct corollaries of this theorem are Harnack inequality and Liouville theorem:\N\NCorollary. Let \((M^n,g)\) be an \(n\)-dimensional complete Riemannian manifold with \(\mbox{Ric}\geq -K\, g\), where \(K\geq 0\). If \(u\) is a classical positive solution of the previous equation on \(B(x_0,2R)\) and there exists \(\alpha\in (1, (n+3)/(n-1))\) such that \(u^{-\alpha} f(u)\) is nonincreasing on \((0,\infty)\), then\N\[\N\sup_{B(x_0,R)}u\leq e^{C\left(\sqrt{K}R+1\right)}\cdot \inf_{B(x_0,R)}u,\N\]\Nwhere \(C=C(n,\alpha)\).\N\NTheorem. Let \((M^n,g)\) be an \(n\)-dimensional complete Riemannian manifold with nonnegative Ricci curvature. If \(u\) is a classical positive solution of the previous equation on \(M^n\) and there exists \(\alpha\in (1, (n+3)/(n-1))\) such that \(u^{-\alpha} f(u)\) is nonincreasing on \((0,\infty)\), then \(u\) is contant.\N\NIn the last section, the author considers a slight generalization of the previous equation as follows.\N\[\N\Delta_V u+f(u)=0,\N\]\Nwhere \(\Delta_V u=\Delta u +\langle\nabla u, V\rangle\), \(V\) is a \(C^1\) vector field. Of course, when \(V\) vanishes this equation becomes the previous one. The author obtains similar results for such equation.