5
$\begingroup$

This is a continuation of this question. Is there a simple formula for $$I(\sigma_1,\cdots,\sigma_p)=(-1)^\sigma\left(\left(\det\left(\frac{\partial}{\partial U_{ij}}\right)_{i,j=1}^n\right)^p \prod_{i=1}^n\prod_{j=1}^p U_{i\sigma_j(i)}\right)|_{U=0},$$ where all $\sigma_i\in S_n$? Equivalently $$I(\sigma_1,\cdots,\sigma_p)=\sum_{\pi\in C}\sum_{\tau_1,\tau_2\in R}(-1)^{\sigma\pi} [\tau_1\pi\tau_2=\sigma],$$ where $C$ and $R$ are the Young subgroups of $S_{pn}$ $$C=\mathrm{Sym}(1,\dots,n)\times\cdots\times\mathrm{Sym}((p-1)n+1,\dots,pn)\cong S_n^p$$ and $$R=\mathrm{Sym}(1,\cdots,(p-1)n+1)\times\cdots\times\mathrm{Sym}(n,\cdots,pn)\cong S_p^n$$ and $\sigma=\sigma_1\oplus\cdots\oplus\sigma_p\in C$.


In the linked question, Carlo Beenakker guessed and Abdelmalek Abdesselam and David Speyer both proved that $I(\sigma_1,\sigma_2)=2^{\#\mathrm{cyc}(\sigma_1\sigma_2^{-1})}$ when $p=2$. This can also be expressed in the group algebra $\mathbb{C}[S_n]$ as $$\sum_{\sigma\in S_n} I(1,\sigma)\sigma=(2+J_1)\cdots(2+J_n),$$ where the $J_k$ are the Jucys-Murphy elements.

Note that $I$ is symmetric and $I(\sigma_1,\cdots,\sigma_p)=I(\rho\sigma_1,\cdots,\rho\sigma_p)$ for all $\rho$. When $(p,n)\ge(4,3)$ or $(3,5)$, $I$ can be negative. It appears that the prime factors of $I$ are at most $p$.

Here are the values of $I(\sigma_1,\sigma_2,1)/12$ when $p=3$ and $n=2$.

Here are the values of $I(\sigma_1,\sigma_2,1)/24$ when $p=3$ and $n=3$.

Here are the values of $I(\sigma_1,\sigma_2,1)/24$ when $p=3$ and $n=4$.

When $p=3$ and $n=5$, $I/24$ takes the values $-1,0,1,2,4,6,12,36,108,324$.

$\endgroup$

1 Answer 1

6
$\begingroup$

Perhaps for small values of $p$ one can find reasonable formulas but not for general $p$, I think. Take the cycle $\sigma_1=(12\cdots n)$ and then let $\sigma_j=(\sigma_1)^j$. Then, $I$ is, up to scale, $$ \int_{SU(n)} \prod_{i=1}^{n}\prod_{j=1}^{n} U_{ij}\ dU $$ which counts the difference between even and odd latin squares. It is an open problem, called the Alon-Tarsi conjecture to show that it is nonzero for $n$ even. See this article by Kumar and Landsberg.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.