Statement

If KURnK\subset U\subseteq\mathbb R^n, with KK nonempty and compact and UU open, then some δ>0\delta>0 satisfies

{x:dist(x,K)<δ}U.\{x:\operatorname{dist}(x,K)<\delta\}\subset U.

If URnU\ne\mathbb R^n, equivalently

infxKdist(x,RnU)>0.\inf_{x\in K}\operatorname{dist}(x,\mathbb R^n\setminus U)>0.

This is a positive support margin.

Proof

The is continuous and strictly positive on KK. It attains a positive minimum by compactness. Any smaller positive number is a suitable δ\delta. If U=RnU=\mathbb R^n, every δ>0\delta>0 works. Empty KK causes no constraint.

Use in localization

Margins leave room for a cutoff to transition from one to zero before reaching a boundary. They also allow small translations or mollifications of compactly supported functions while keeping their supports in UU.