Generalized intersection bodies are not equivalent (Q2480308)

In [Geom. Funct. Anal. 10, 1507--1526 (2000; Zbl 0974.52002)], \textit{A.\ Koldobsky} asked whether two types of generalization of the notion of an intersection body are in fact equivalent. The questions arises from the generalized Busemann-Petty problem: If \(K\) and \(L\) are origin-symmetric convex bodies in \({\mathbb R}^n\), and the volume of \(K\cap H\) is smaller than the volume of \(L\cap H\) for every \(k\)-codimensional subspace \(H, 1< k <n\), does it follow that the volume of \(K\) is smaller than the volume of \(L\)? The first class of generalized intersection bodies, \({\mathcal BP}_k^n\), was introduced by \textit{G. Zhang} [Am. J. Math. 118, 319--340 (1996; Zbl 0854.52004)] using dual Radon transforms. Zhang proved that the \(k\)-codimensional Busemann-Petty problem has a positive answer if and only if all centrally symmetric convex bodies belong to \({\mathcal BP}_k^n\). The second class, \({\mathcal I}_k^n\), was introduced by Koldobsky [op. cit.], who proved that \({\mathcal BP}_k^n\subset {\mathcal I}_k^n\) and that for \(k\geq n-3\) all origin-symmetric convex bodies belong to \({\mathcal I}_k^n\). The author constructs a counter-example, which provides a negative answer to Koldobsky's question in a strong sense. The generalized Busemann-Petty problem remains unsolved for \(k=n-2, n-3\) for \(n\geq5\). Milman's counter-example also implies the existence of non-trivial non-negative functions in the range of the spherical Radon transform, and the existence of non-trivial spaces which embed in \(L_p\) for certain negative values of \(p\).