Elliptic rearrangement for Poisson equations with mean-zero data

Abstract

The $L^p$ norm of a sign-changing Dirichlet solution of Poisson's equation is shown to be bounded by the maximum of the $L^p$ norms of the ball solutions obtained by signed rearrangement of the source and its negative, assuming the source term has mean value zero. This paired comparison on bounded open sets extends Talenti's elliptic rearrangement principle to solutions that change sign. In dimensions one and two, the paired comparison holds for all $1\le p\le\infty$, and it holds in every dimension for $p=\infty$. The inequalities are sharp and equality cases are characterized. In dimensions $n \geq 3$, paired comparison can fail for every finite $p$ although it remains valid for $p\ge(3n-1)/4$ in the special case of bang-bang sources with mean zero. The mean-zero source assumption is crucial when $p<\infty$. Without it, paired comparison can fail even for bang-bang sources. For sources of arbitrary mean in one dimension, however, the $L^\infty$ paired comparison still holds and so does a comparison for $L^p$ norms of the derivative when $1\le p\le3$.