An infinite double bubble theorem

Abstract

The classical double bubble theorem characterizes the minimizing partitions of ${\mathbb{R}}^n$ into three chambers, two of which have prescribed finite volume. In this paper we prove a variant of the double bubble theorem in which two of the chambers have infinite volume. Such a configuration is an example of a (1,2)-cluster, or a partition of ${\mathbb{R}}^n$ into three chambers, two of which have infinite volume and only one of which has finite volume. A $(1,2)$-cluster is locally minimizing with respect to a family of weights $\{c_{jk}\}$ if for any $B_r(0)$, it minimizes the interfacial energy ${\sum }_{j<k}{c}_{jk}{\mathcal{H}}^{n-1}(\partial \mathcal{X}(j)\cap \partial \mathcal{X}(k)\cap B_r(0))$ among all variations with compact support in $B_r(0)$ which preserve the volume of $\mathcal{X}(1)$. For $(1,2)$ clusters, the analogue of the weighted double bubble is the weighted lens cluster, and we show that it is locally minimizing. Furthermore, under a symmetry assumption on $\{c_{jk}\}$ that includes the case of equal weights, the weighted lens cluster is the unique local minimizer in ${\mathbb{R}}^n$ for $n\le 7$, with the same uniqueness holding in ${\mathbb{R}}^n$ for $n\ge 8$ under a natural growth assumption. We also obtain a closure theorem for locally minimizing $(N,2)$-clusters.