On directional convexity (Q5944936)

Let \(D\subset{\mathbb R}^d\). A function \(f:{\mathbb R}^d\to{\mathbb R}\) is \(D\)-convex if its restriction to any line parallel to a nonzero vector in \(D\) is a convex function. The \(D\)-convex hull of a compact set \(A\subset{\mathbb R}^d\) (denoted by co\(^D(A)\)) is the intersection of the zero sets of all nonnegative \(D\)-convex functions that are zero on \(A\). It is proved that co\(^D(A)\) equals the zero set of the \(D\)-convex envelope of the distance function of \(A\). The definition of the \(D\)-convex hull can be very nonrobust. Namely, for each \(n\geq 2\) there exists an \(n\)-point set \(A\subset{\mathbb R}^2\) where the \(D\)-convex envelope of the distance function is exponentially close to zero at points lying relatively far from co\(^D(A)\). Some general properties of \(D\)-convexity are also established. If \(D\) is the orthonormal basis of \({\mathbb R}^d\), then \(D\)-convexity is called separate convexity. In this case arbitrarily large finite sets \(A\subset{\mathbb R}^3\) such that co\(^D(A)\neq A\) and all proper subsets \(B\subset A\) are equal to co\(^D(B)\) are constructed. This implies the existence of analogous sets for rank-one convexity and for quasiconvexity on \(3\times 3\) (or larger) matrices. Express Carathéodory's theorem in the form: The Carathéodory number for convexity (in the usual sense) in \({\mathbb R}^d\) is \(d+1\). The existence of sets \(A\) described above means that the Carathéodory number for separate convexity in dimensions 3 and higher is infinite (it is known that the Carathéodory number for separate convexity in \({\mathbb R}^2\) is 5).