Explicit formulae for Kerov polynomials (Q617328)

If a Young diagram (or partition) \(\lambda\) is represented as a function in the plane \(\mathbb R^2\), let \(x_0,\dots ,x_m\) and \(y_1,\dots ,y_m\) denote the sequences of \(x\)-coordinates of the local maxima and minima, respectively, of \(\lambda\), and put \({\mathcal H_{\lambda}(z)=\frac{\prod_{i=0}^m(z-x_i)}{\prod_{i=1}^m(z-y_i)}}\). If \(\mathcal K_{\lambda}(z)=\mathcal F_{\lambda}^{\langle -1\rangle}(z)\), i.e., \(\mathcal K_{\lambda}(\mathcal H_{\lambda}(z))=\mathcal H_{\lambda}(\mathcal K_{\lambda}(z))=z\), then the \(n\)th free cumulant \(R_n(\lambda)\) is the coefficient of \(z^{n-1}\) in the formal Laurent series expansion of \(\mathcal K_{\lambda}(z)\). The \(k\)th Kerov polynomial \(\Sigma_k(R_2,\dots ,R_{k+1})\) satisfies \[ {\Sigma_k(R_2(\lambda),\dots ,R_{k+1}(\lambda))=(n)_k\frac{\chi_{\lambda}(k,1^{n-k})}{\chi_{\lambda}(1^n)}}, \] where \((n)_k=n(n-1)\cdots(n-k+1)\) and \(\chi_{\lambda}(k,1^{n-k})\) denotes the value of the irreducible character of the symmetric group \(\mathfrak G_n\) indexed by the partition \(\lambda\) on \(k\)-cycles. \(\Sigma_k\) is a ``universal polynomial'' in that it depends neither on \(\lambda\) nor on \(n\). Its coefficients are nonnegative integers, and an explicit combinatorial description of its coefficients is known. The authors prove two new formulae for \(\Sigma_k\). Let \(NC_{k+1}\) denote the set of all noncrossing partitions of the set \([k+1]=\{1,2,\dots ,k+1\}\), and \(NC_{k+1}^{\text{irr}}\) the subset of those partitions which are irreducible (i.e. having \(1\) and \(k+1\) in the same block) with special partial ordering \(\leq^{\text{irr}}\). Theorem 4.1 gives \({\Sigma_k=\sum_{\tau\in NC^{\text{irr}}_{k+1}}[\sum_{\tau\leq^{\text{irr}}\pi}(-1)^{\ell_{\pi}-1}W_{\tau}(\pi)]R_{\dot\tau}}\), where \(\ell_{\pi}\) is the number of blocks of \(\pi\), \(W_{\tau}(\pi)\) is a suitable weight depending on \(\tau\) and \(\pi\), and \(R_{\dot\tau}=\prod_B R_{|B|}\) where \(B\) ranges over the blocks of \(\tau\) having at least two elements; Theorem 4.2 gives \({\Sigma_k=\sum_{\tau\in NC_{k+1}}[(-1)^{d_{\tau}-1}V_{\tau}]R_{\dot\tau}}\), where \(d_{\tau}\) is the number of irreducible components of \(\tau\) and \(V_{\tau}\) is a suitable weight depending on \(\tau\). The authors also give a combinatorial description of a family of symmetric functions \(\mathbf g_{\mu}(x_1,\dots ,x_k)\) specializing (when \(x_i=i\)) in the coefficients of \(\Sigma_k\) (Remark 4.1).