I am looking for (minimal) conditions, which guarantee that the problem

Lu = 0 in R^n,

where L is a second-order (uniformly) elliptic operator with analytic coefficients, has a unique global analytic solution. Does anybody have any references to relevant work?