Bipartition orders and statistics on words (Q1918878)

Let \(X\) be any alphabet. For any subset \(U\) of \(X\times X\), \(x\succ y\) if \((x, y)\in U\), and \(x\not\succ y\) if \((x, y)\not\in U\); \(U\) is an ``ordre bipartitionnaire'' if, for all \(x, y, z\in X\), \(z\succ y\) and \(y\succ x\Rightarrow z\succ x\), and \(z\succ y\) and \(x\not\succ y\Rightarrow z\succ x\). For any word \(w= x_1 x_2\dots x_m\in X^*\), the free monoid generated by \(X\), \(\text{inv}_U w= \text{Card}\{(i, j)\mid 1\leq i< j\leq m\) and \(x_i\succ x_j\}\), \(\text{maj}_U w= \sum \{i\mid 1\leq i\leq n- 1\) and \(x_i\succ x_{i+ 1}\}\). \textit{D. Foata} and \textit{D. Zeilberger} [Graphical major indices, submitted for publication] proved Theorem 1: Statistics \(\text{maj}_U\) and \(\text{inv}_U\) are equidistributed iff \(U\) is an ordre bipartitionnaire. For \(x, y\in X\), the (generalized) cyclic interval \(] ]x, y] ]_U\) is defined to equal \(\{z\in X\mid z\succ x\) and \(z\not\succ y\}\) if \(x\not\succ y\), and \(\{z\in X\mid z\succ x\) or \(z\not\succ y\}\) if \(x\succ y\). Where the symbol \(\infty\) is adjoined to an alphabet, one assumes that, for all \(x\), \(x\not\succ \infty\) an \(\infty\succ x\). For a word \(w= x_1 x_2\dots x_m\) one defines, with the convention \(x_{n+ 1}= \infty\), \(\text{maj} 2_U w= \sum^m_{j= 1} \langle x_1 x_2\dots x_{j- 1}, ] ] x_j, x_{j+ 1}] ]_U\rangle\). ``In this note we prove the following theorem: Theorem 2. The statistics \(\text{maj}_U\) and \(\text{maj} 2_U\) are equivalent iff the subset \(U\) is an ordre bipartitionnaire''.