For $\chi$ being an irreducible character of the symmetric group $S_n$ and being $M$ a complex $n\times n$-matrix, I would like to show

$$ \sum_{\sigma, \rho \in S_n} \overline{\chi(\sigma)} \chi(\rho) \prod_{j=1}^n M_{\sigma_j, \rho_j} = \frac{ n! }{\chi(e)} \sum_{\sigma \in S_n} \chi(\sigma) \prod_{j=1}^n M_{j, \sigma_j} , $$ where the right-hand-side is proportional to the immanant of $M$, $$ \text{imm}(M) = \sum_{\sigma \in S_n} \chi(\sigma) \prod_{j=1}^n M_{j, \sigma_j} $$ and $e$ is the identity element in $S_n$. This expression appears as the scalar product of many-body quantum states with exotic exchange symmetries.

For one-dimensional representations (i.e. the trivial constant or the alternating representation), we get the identities $$ \sum_{\sigma, \rho \in S_n} \prod_{j=1}^n M_{\sigma_j, \rho_j} = n! ~ \text{perm}(M) \\ \sum_{\sigma, \rho \in S_n} \text{sgn}(\sigma) \text{sgn}(\rho) \prod_{j=1}^n M_{\sigma_j, \rho_j} = n! ~ \text{det}(M) $$ for the permanent and the determinant (related to bosons and fermions), respectively, which are shown using $\chi(\sigma \rho)=\chi(\sigma) \chi(\rho)$ and $\chi(e)=1$ for one-dimensional $\chi$.


I hope this is not homework.

\begin{eqnarray*} \text{lhs}&=&\sum_{\sigma,\rho} \overline{\chi(\sigma)}\chi(\rho)\left(\prod_{i=1}^nM_{\sigma(i),\rho(i)}\right)\\ &=&\sum_{\sigma,\rho} \overline{\chi(\sigma)}\chi(\rho)\left(\prod_{i=1}^nM_{i,\rho\sigma^{-1}(i)}\right)\\ &=&\sum_{\tau}\left(\sum_{\sigma}\overline{\chi(\sigma)}\chi(\tau\sigma) \right)\prod_{i=1}^n M_{i,\tau(i)}\\ &=&\sum_{\tau}\frac{n!}{\chi(e)} \chi(\tau)\prod_{i=1}^nM_{i,\tau(i)}. \end{eqnarray*}

  • $\begingroup$ thanks, is there a simple way to see/understand the very last equality, i.e. why that sum over $/sigma$ resolves this way? $\endgroup$ – Malte May 15 '15 at 14:11
  • 1
    $\begingroup$ That result is an adaptation of en.wikipedia.org/wiki/Schur_orthogonality_relations -- so that in particular, if $G$ is a finite group and $\chi, \xi$ are irreducible chars, then $$\sum_{\sigma \in G} \overline{\chi(\sigma)}\xi(\sigma\tau) = \begin{cases} |G|\chi(\tau)/\chi(e), & \text{if}\ \chi=\xi;\\ 0 & \text{otherwise}.\end{cases}$$ $\endgroup$ – Suvrit May 15 '15 at 14:39

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.