How neighborly can a centrally symmetric polytope be? (Q2505226): Difference between revisions

A polytope~\(P\) is called \textit{\(k\)-neighborly} if every subset of at most~\(k\) of its vertices is the vertex set of a face of~\(P\). It is well-known that a \(d\)-dimensional polytope different from the simplex can be at most \(\lfloor d/2\rfloor\)-neighborly. A polytope~\(Q\) is \textit{centrally symmetric} if \(Q=-Q\). Centrally symmetric polytopes (with more than two vertices) can only be \(1\)-neighborly in the general sense. Therefore, one calls a centrally symmetric polytope~\(Q\) \textit{\(k\)-neighborly} if every subset~\(S\) of at most~\(k\) vertices with \(S\cap (-S)=\varnothing\) is the vertex set of a face of~\(Q\). Denoting by \(k(d,n)\) the maximal~\(k\) for which a \(k\)-neighborly \(d\)-dimensional polytope with~\(n\) vertices exists, the authors prove that there are constants \(C_1,C_2>0\) with \[ \frac{C_1 d}{1+\log((n+d)/d)}\;\leq\;k(d,n)\;\leq\;\frac{C_2 d}{1+\log((n+d)/d)} \] for all~\(d\) and~\(n\). In particular, the authors prove that a \(d\)-dimensional \(2\)-neighborly centrally symmetric polytope can have at most \(2^d\) vertices.