The stability of the Rogers-Shephard inequality and of some related inequalities (Q703798): Difference between revisions

The author provides stability versions of some affine invariant inequalities where equality holds only for simplices. The main result is the following optimal stability version of the Rogers-Shephard inequality: If \(K\) is a convex body in \(n\)-dimensional euclidean space and \[ V(DK)=(1-\varepsilon)\binom{2n}n V(K) \] then \[ 1+ \frac1n\varepsilon \leq d_{BM}(K,T) \leq 1+ n^{50n^2}\varepsilon, \] where \(d_{BM}\) is the Banach-Mazur distance of \(K\) from any simplex \(T\). In the light of various other stability results it is remarkable that the error terms in the lower and the upper bound in the above inequality are of the same order. The proof of the main result depends on understanding how close \(K\cap (K+x)\) is to be homothetic to \(K\) for \(x\in \operatorname{int} DK\). Given a convex body \(C\), let \(R_K(C)\) denote the smallest number \(R\) such that a translate of \(RK\) contains \(C\). As a side result the author also proves the following inequality: If \(K\) is a convex body in \(n\)-dimensional euclidean space and \(d_{BM}(K,T)=1+\varepsilon\) then \[ R_K(K\cap (x+K)) \leq 1 - \| x \| _{DK} + \varepsilon \] for any \(x\in \operatorname{int}DK\), and there exists an \(x\in DK\) with \(\| x\| _{DK}=\frac12\) satisfying \[ R_K(K\cap(x+K))\geq \frac12 +n^{-38}\varepsilon. \] The method of proof of the main result can be applied to some related results where equality holds for simplices. The author exhibits this possibility both for the Zhang projection inequality and for the inclusions for the \(p\)th radial mean body, \((p\geq -1)\) verified by Gardner and Zhang.