Picard principle for linear elliptic differential operators (Q1061923)

Consider the linear operator \(L=\Delta +b(x)\cdot \nabla +c(x)\) whose coefficients are assumed to belong respectively to the spaces \(C^ 2(\bar B\setminus \{0\})\) and \(C^ 1(\bar B\setminus \{0\})\); where \(B=B(0,1)\). Denote then by ex P the extreme points of the convex set P consisting of the non-negative solutions u of \(Lu=0\), vanishing on B and verifying the normalization condition \(\int_{\partial B}D_ nu dS=1.\) The author shows that P has a unique extreme point when the coefficients b and c satisfy the following estimates \(| b(x)| \leq C_ 1| x|^{-1} ;\quad | \nabla \cdot b(x)| \leq C_ 2| x|^{-2} ;\quad | c(x)| \leq C_ 3| x|^{-2}\) on a suitable sequence of disjoint annuli converging to \(x=0\). The present paper generalizes a previous result by \textit{M. Kawamura} [J. Math. Soc. Japan 31, 354-371 (1979; Zbl 0399.30032)].