Fix a finite set $X$ and two natural numbers $d$ and $n$.

For a partition $\lambda$ and a number $d$ denote by $s_\lambda^d(x_1,\dots,x_d)$ the Schur polynomial in $d$-many variables $x_1,\dots,x_d$. Denote by $\mathcal P_n(X)$ the set of partition-valued functions on $X$ of total size $n$, i.e. the set of functions $\lambda\colon X\to\{\text{partitions}\}$ such that $\|\lambda\|:=\sum_{x\in X}|\lambda(x)|=n$.

Let $a\colon X\to \mathbb{N}\backslash\{0\}$ be a function on $X$. EDIT: I am mostly interested in the case where $a$ is the constant function with value $1$.

I am trying to evaluate the expression $$F_X(d,a):=\sum_{\lambda\in \mathcal P_n(X)}\left(\prod_{x\in X}s^d_{\lambda(x)}(a(x),\dots,a(x))\right)^2. $$


  1. I have now good reason to believe that if we take for $a$ the constant function with value $1$ then we have $$F_X(d,a)=\left(\binom{d^2|X|}{n}\right)$$ Here $\left(\binom{m}{n}\right)$ denotes the multiset number

  2. More generally I am hoping that if $G$ is a finite group, $X$ is the set of irreducible characters (over $\mathbb C$) of $G$ and $a$ maps each character to its dimension then we have $$F_X(d,a)=\left(\binom{d^2|G^{\mathrm{ab}}|}{n}\right)$$ Here $G^\mathrm{ab}$ is the abelianisation of $G$. This reduces to case 1 if $G$ is taken to be an abelian group.

EDIT2: my conjecture 2 above is totally wrong (as can be seen by hand in small examples like $G=S_3$) and I have no good substitute hypothesis. I am now mostly interested in a proof of point 1 (where $a$ is the constant function $1$)

The special case where $X=\{\star\}$ is a one-element set can be done using the identity $$\prod_{i,j}(1-x_iy_j)^{-1}=\sum_n\sum_{|\lambda|=n} s_\lambda(x_1,\dots)s_\lambda(y_1,\dots)$$ by putting $x_i=y_j=t$ for $i,j=1,\dots,d$ (and zero otherwise) and picking out the coefficient of $t^{2n}$ obtaining $$c^d_nt^{2n}=\sum_{|\lambda|=n}s^d_\lambda(t,\dots,t)^2,$$ where $c^d_n$ is the coefficient of $t^n$ in $(1-t)^{-d^2}$ and can be calculated to be the multiset number $\left(\binom{d^2}{n}\right)$. All in all we obtain $F_{\{\star\}}(d,a)=\left(\binom{d^2}{n}\right)\cdot a(\star)^{2n}$

  • $\begingroup$ You rrally want the sum of the lengths to be $n $, not the sum of the sizes? $\endgroup$ – darij grinberg Feb 25 '16 at 15:55
  • $\begingroup$ you are right, of course. It should be the sizes $\endgroup$ – Tashi Walde Feb 25 '16 at 16:34

assume $a$ is the constant function with value $1$.

as in the case of $X=\{\star\}$ we have the equality $$(1-t^2)^{-d^2|X|}=\left(\sum_{\lambda\in \{\mathrm{partitions}\}}s^d_\lambda(t,\dots,t)^2\right)^{|X|}=\sum_{\lambda\colon X\to\{\mathrm{partitions}\}}\prod_{x\in X}s_{\lambda(x)}^d(t,\dots,t)^2$$ and picking out the monomial $t^2$ gives $$\left(\binom{d^2|X|}{n}\right)t^{2n}=\sum_{\lambda\in P_n(X)}\prod_{x\in X}s_{\lambda(x)}^d(t,\dots,t)^2$$ The result follows.


Note that $s_\lambda(1^d)$ can be computed explicitly via Weyls formula on wikipedia in terms of the parts of $\lambda$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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