Highly neighborly centrally symmetric spheres (Q2189507): Difference between revisions

A simplicial \(d\)-sphere is called \(j\)-neighborly if any \(j\)-tuple of vertices spans a \((j-1)\)-face. Except for the simplex this is possible only for \(j\le (d+1)/2\). Similarly, a centrally-symmetric (cs) simplicial \(d\)-sphere is called cs-\(j\)-neighborly if any \(j\)-tuple of vertices spans a \((j-1)\)-face provided that it does not contain a pair of ``antipodal'' vertices, denoted by \(x\) and \(-x\). An infinite family of cs-2-neighborly 3-spheres was constructed by \textit{W. Jockusch} in [J. Comb. Theory, Ser. A 72, No. 2, 318--321 (1995; Zbl 0844.52006)]. In the article under review the authors construct cs \(d\)-spheres with \(2n\) vertices for any \(d\ge 3\) and \(n\ge d+1\) that are (maximally) cs-\(d/2\)-neighborly or cs-\((d+1)/2\)-neighborly, respectively. As a consequence, the upper bound problem for cs simplicial spheres is solved. By generalizing Jockusch's construction, the authors use the following induction step for raising the number of vertices while keeping the cs-\(i\)-neighborliness: Assume the \(d\)-sphere admits a decomposition into two \(d\)-balls \(B\) and \(-B\) that are cs-\((d-1)\)-neighborly and \((i-1)\)-stacked. Then \(B\) and \(-B\) are replaced by \(\partial B*v_{n+1}\) and \(\partial(-B)*(-v_{n+1})\), respectively. However, the details of the induction are complicated. Since this construction is not compatible with automorphism groups, the following Question 4.4 from the 1999 thesis of \textit{F. H. Lutz} remains open: Are there maximally cs-neighborly \(d\)-spheres for any \(n\) and any odd \(d\) that admit a vertex transitive action of the cyclic automorphism group?