The \(L_p\) John ellipsoids for negative indices (Q6165164)

The authors provide a definition of \(L_p\) John ellipsoid of a convex body \(K\) (i.e. a compact and convex subset of the \(n\) dimensional Euclidean space \(\mathbb{R}^n\)), for negative values of \(p\), in the range \((-n,0)\), under some assumptions on the boundary of \(K\). They achieve this result by solving an appropriate variational problem. The notion of \(L_p\) John ellipsoid, which was previously known for \(p\geq 0\), is therefore extended to the range \(p\in(-n,0)\). A convex body \(K\) in \(\mathbb{R}^n\) is said to be of class \(C^2_+\) if its boundary is of class \(C^2\) and the Gauss curvature is strictly positive at each point of the boundary. Given \(p<0\), a convex body \(K\) containing the origin in its interior, and an ellipsoid \(E\) centered at the origin, the authors consider the quantity \[ \overline{V}_p(K,E)=\left[\frac1{V(K)}\int_{\mathbb{S}^{n-1}}\left(\frac{h_E}{h_K}\right)^p dV_K \right]. \] Here \(V\) denotes the volume (Lebesgue measure), \(h_E\) and \(h_K\) are the support functions of \(E\) and \(K\), respectively, and \(V_K\) is the cone volume measure of \(K\). In order to define the \(L_p\) John ellipsoid, the following problem is considered. Problem A. Given \(p<0\) and a convex body \(K\) containing the origin in its interior, among all centered ellipsoids, find an ellipsoid \(E\) which solves \[ \max_{E}V(E)\text{ under the constraint }\overline{V}_p(K,E)\leq 1. \] Coherently with the case \(p\geq 0\), an ellipsoid which solves this problem is called the \(L_p\) John ellipsoid of \(K\). The following result is proved. Theorem 1. Let \(p\in(-n,0)\). Let \(K\) be a convex body of class \(C^2_+\), containing the origin in its interior. Then there exists an ellipsoid \(E\), solving Problem A. The authors also establish a negative result when \(p<-n\). Theorem 2. Let \(p<-n\). Let \(K\) be a convex body of class \(C^2_+\), containing the origin in its interior. Then Problem A admits no maximizer.