Rotation covariant local tensor valuations on convex bodies

TL;DR

This paper classifies local tensor valuations on convex bodies that are covariant under special orthogonal transformations, revealing new tensor measures in low dimensions and extending prior classification results.

Contribution

It provides a complete classification of local tensor valuations with SO(n) covariance, highlighting new tensor measures in dimensions two and three.

Findings

New tensor-valued support measures in dimensions two and three.

Complete classification of SO(n)-covariant local tensor valuations.

Extension of previous O(n)-covariance classification results.

Abstract

For valuations on convex bodies in Euclidean spaces, there is by now a long series of characterization and classification theorems. The classical template is Hadwiger's theorem, saying that every rigid motion invariant, continuous, real-valued valuation on convex bodies in $\mathbb{R}^n$ is a linear combination of the intrinsic volumes. For tensor-valued valuations, under the assumptions of isometry covariance and continuity, there is a similar classification theorem, due to Alesker. Also for the local extensions of the intrinsic volumes, the support, curvature and area measures, there are analogous characterization results, with continuity replaced by weak continuity, and involving an additional assumption of local determination. The present authors have recently obtained a corresponding characterization result for local tensor valuations, or tensor-valued support measures (generalized curvature measures), of convex bodies in $\mathbb{R}^n$. The covariance assumed there was with respect to the group ${\rm O}(n)$ of orthogonal transformations. This was suggested by Alesker's observation, according to which in dimensions $n> 2$, the weaker assumption of ${\rm SO}(n)$ covariance does not yield more tensor valuations. However, for tensor-valued support measures, the distinction between proper and improper rotations does make a difference. The present paper considers, therefore, the local tensor valuations sharing the previously assumed properties, but with ${\rm O}(n)$ covariance replaced by ${\rm SO}(n)$ covariance, and provides a complete classification. New tensor valued support measures appear only in dimensions two and three.