A poset view of the major index. (Q472919)

This paper extends the combinatorial major index statistic to posets. The main object of study is the \textit{major MacMahon map} \(\Theta\colon\mathbb Z\langle\mathbf{a,b}\rangle\to\mathbb Z[q]\), which is defined on a word in the noncommuting variables \(\mathbf a\) and \(\mathbf b\) by \(\Theta(w):=\prod_{i:w_i=\mathbf b}q^i\) and is extended to the ring of non-commuting polynomials in \(\mathbf a\) and \(\mathbf b\) by linearity. The connection to posets comes about in the following way. Let \(P\) be a graded poset of rank \(n+1\) with a minimal element \(\widehat 0\) and a maximal element \(\widehat 1\). The \(\mathbf{ab}\)-index of \(P\), \(\Psi(P)\in\mathbb Z\langle\mathbf{a,b}\rangle\), encodes useful combinatorial information about \(P\). Hence it is interesting to apply the major MacMahon map to the \(\mathbf{ab}\)-index of such a poset. The authors show that if \(P\) is a graded simplicial poset, then \(\Theta(\Psi(P))=[n]!\cdot h(P)\), where \(h(P)\) is the \(h\)-polynomial of \(P\) (written as a polynomial in the variable \(q\)) and \([n]!\) is the \(q\)-analogue of \(n!\). Next, the authors show that if \(P\) is a graded poset of rank \(m\) and \(Q\) is a graded poset of rank \(n\), then \[ \Theta(\Psi(P\times Q))={m+n\choose n}\cdot\Theta(\Psi(P))\cdot\Theta(\Psi(Q)). \] They apply this to products of chain posets to give a poset proof of MacMahon's classical result that if \(\alpha=(\alpha_1,\ldots,\alpha_k)\) is a composition of \(n\), then \[ \sum_\pi q^{\text{maj}(\pi)}=\frac{[n]!}{[\alpha_1]!\cdot [\alpha_2]!\cdots [\alpha_k]!}, \] where the sum is taken over all permutations of the multiset containing \(\alpha_i\) instances of the number \(i\) for all \(1\leq i\leq k\).