A new cubical \(h\)-vector (Q1924354): Difference between revisions

The \(h\)-vectors of simplicial convex \(d\)-polytopes have proved central in the description of possible numbers of faces of such polytopes. Cubical polytopes (that is, those whose faces are combinatorial cubes) possess analogous properties to simplicial polytopes; there is thus the expectation that corresponding \(h\)-vectors would play a similar rôle for them. The author defines two \(h\)-vectors here, and describes some of their properties. More generally, let \(K\) be a cubical \((d-1)\)-complex (its faces are combinatorial cubes). If \(f_i\) is the number of its \(i\)-faces, then \((f_0, \dots, f_{d-1})\) is the \(f\)-vector of \(K\). The short cubical \(h \)-vector \((h^{\text{(s)}}_0, \dots, h_{d-1}^{\text{(sc)}})\) of \(K\) is defined by \[ \sum^{d-1}_{i= 0} h_i^{\text{(sc)}} q^i: =\sum^{d-1}_{j=0} f_j(2q)^j (1-q)^{d-j-1}, \] and the (long) cubical \(h\)-vector \((h_0^{\text{(c)}}, \dots, h_d^{\text{(c)}})\) is given recursively by \[ h_i^{\text{(sc)}}= : h_i^{\text{(c)}} + h_{i+1}^{\text{(c)}} \] for \(i=0, \dots, d-1\), with the initial value \(h_0^{\text{(c)}}: =2^{d-1}\). Both \(h\)-vectors are clearly integral. More importantly, the author shows that, if \(K\) (together with an adjoined maximal element) is Eulerian, then the (long) \(h\)-vector is symmetric (the ``Dehn-Sommerville equations''), and if \(K\) is shellable, then it is nonnegative. It is natural to ask whether the analogy with simplicial \(h\)-vectors extends further. For example, the author poses the question: is it true that \(h_{i-1}^{\text{(c)}} \leq h_i^{\text{(c)}}\) when \(1\leq i \leq d/2\), for the boundary complex of a cubical (convex) \(d\)-polytope? A result of \textit{G. Blind} and \textit{R. Blind} [Isr. J. Math. 71, No. 2, 129-134 (1990; Zbl 0719.52009)] gives the case \(i=1\) (it just says that such a polytope has at least \(2^d\) vertices).