Let $\lambda$ be the partition of integer $d$. The Frobenius coordinate of $\lambda$ is given $$ (a_1 ,\ldots,a_{d(\lambda)}|b_1,\ldots,b_{d(\lambda)}),$$ where $d(\lambda)$ denote the diagonal of $\lambda$. Let $a_i'= a_i +\frac12$ and $b_i'= b_i +\frac12$ are modified Frobenius coordinates. By a classical theorem of Frobenius, it states that

$f(\lambda)=\frac{|C_{2,1,\ldots,1}|}{dim \lambda}\chi_{2,1,\ldots,1}^{\lambda}=\frac12\sum_{i=1}^{d(\lambda)}(a_i')^2-(b_i')^2$

Notice that the above example we treat the L.H.S as a function of $\lambda$ as we fix the partition of the form $2,1^{d-2}$

If there is any generalisation of the above theorem? That is if we replace $(2,1,\ldots,1)$ with any other partition of $d$ then say $\mu$ then what will be

$$\frac{|C_{2^{d/2}}|}{dim \lambda}\chi_{2^{d/2}}^{\lambda}$$ For example say $\mu$ is of the form $2^{d/2}$

in Frobenius coordinate? If it is already studied can anyone please site the reference.


2 Answers 2


There is a generalisation of the character formula, although it is usually stated in terms of contents rather than Frobenius coordinates. Also, it applies to conjugacy classes labelled by partitions $(1^{m_1} 2^{m_2} \cdots)$, where $m_2, m_3, \ldots$ are fixed and $m_1$ varies with $n$ (as in the example you gave, where there is one part of size $2$, and $n-2$ parts of size 1). So in particular, it does not apply directly to the partitions $(2^{n/2})$.

For the benefit of the uninitiated, let me state right away that $\frac{|C_\mu| \chi_{\mu}^\lambda}{dim(\lambda)}$ is a central character. The sum of all elements of $S_n$ of cycle type $\mu$ is a central element of the group algebra of $S_n$, and the central character is the scalar by which the central element acts on the irreducible representation labelled by $\lambda$. Central characters are important in modular representation theory, which might provide a little motivation for why the quantity in question is a natural thing to consider.

Suppose that we have a partition $\lambda = (\lambda_1, \ldots, \lambda_l)$. Suppose that $d$ is the largest integer such that $\lambda_d \geq d$ (the Durfee size of $\lambda$). Then the Frobenius coordinates of $\lambda$ are $(\alpha_1, \ldots, \alpha_d | \beta_1, \ldots, \beta_d)$, where $\alpha_i = \lambda_i - i$ and $\beta_i = \lambda_i^\prime - i$ (here $\lambda^\prime$ is the transpose partition of $\lambda$). So the $\alpha_i$ and $\beta_i$ count the number of boxes in past the diagonal in the $i$-th row and $i$-th column respectively (and the number of each is $d$). On the other hand, the contents of a box in row $j$ and column $i$ is $i-j$. It is not difficult to see that the multiset of contents of $\lambda$ consists of $0$ with multiplicity $d$, and $\{1,2,\ldots, \alpha_i\}$ for each $i$ together with $\{-1, -2, \ldots, -\beta_i\}$ for each $i$. So in particular, the sum of $r$-th powers of the contents of $\lambda$ can be expressed as polynomials in the $\alpha_i$ and $\beta_i$ by Faulhaber's Formula. For example, the sum of the contents ($r=1$) is

$$\sum_i (1 + 2 + \cdots + \alpha_i) - (1 + 2 + \cdots + \beta_i) = \sum_i {\alpha_i + 1 \choose 2} - {\beta_i + 1 \choose 2},$$

which is precisely the quantity appearing in the example you gave. We write $cont(\lambda)$ for the multiset of contents of $\lambda$.

Now, for a partition $\mu = (\mu_1, \mu_2, \cdots, \mu_k)$, let $\bar{\mu} = (\mu_1 - 1, \mu_2 - 1, \ldots, \mu_k - 1)$ be the partition obtained from $\mu$ by subtracting 1 from each nonzero part of $\mu$ (and ignoring trailing zeros). For example, if $\mu = (2,1, \ldots, 1)$ as in your example, then $\bar{\mu} = (1)$.

Theorem There exists a family $f_{\rho}$ of elements of $\mathbb{Q}[n] \otimes \Lambda$ (where $n$ is a polynomial variable) and $\Lambda$ is the ring of symmetric functions with the property that

$$\frac{|C_\mu| \chi_{\mu}^\lambda}{dim(\lambda)} = f_{\bar{\mu}}(|\lambda|, cont(\lambda))$$

By this notation I mean that we evaluate $f_{\bar{\mu}}$ at $n=|\lambda|$ and its symmetric function variables at $cont(\lambda)$.

It turns out that $f_{(1)} = p_1$ (the first power-sum symmetric function) which doesn't actually depend on $n$. Then putting this together, we see

$$\frac{|C_{(2,1,\ldots,1)}| \chi_{(2,1,\ldots,1)}^\lambda}{dim(\lambda)} = p_1(cont(\lambda)) = \sum_i {\alpha_i + 1 \choose 2} - {\beta_i + 1 \choose 2},$$

which recovers the result you stated. In general, you can take $f_{\bar{\mu}}$, express it as a polynomial in $n$ and the power-sums $p_1, p_2, \ldots \in \Lambda$, and upon evaluating at the contents of $\lambda$ you will obtain a polynomial in $n$ and the Frobenius coordinates of $\lambda$. For example,

$$f_{(2)} = p_2 - {n \choose 2},$$

which tells us that the central character corresponding to the conjugacy class $(3,1,\ldots, 1)$ may be computed using the formula

$$\left( \sum_{i} \frac{\alpha_i(\alpha_i + 1)(2\alpha_i + 1)}{6} + \frac{\beta_i (\beta_i + 1)(2\beta_i + 1)}{6} \right) - {|\lambda| \choose 2}.$$

There is a lot more to be said about the $f_\rho$. For example, they form a $\mathcal{R}$-basis of $\mathcal{R} \otimes \Lambda$, wher $\mathcal{R}$ is the subring of $\mathbb{Q}[n]$ consisting of integer-valued polynomials. But in this post I worked over $\mathbb{Q}$ because I wanted to use power-sum symmetric functions to turn contents-evaluaion into polynomials in Frobenius coordinates.

References you might find helpful (each of which contains an explanation of the above theorem):

Corteel, Sylvie; Goupil, Alain; Schaeffer, Gilles, Content evaluation and class symmetric functions, Adv. Math. 188, No. 2, 315-336 (2004). ZBL1059.05104.

Section 5.4 of

Ceccherini-Silberstein, Tullio; Scarabotti, Fabio; Tolli, Filippo, Representation theory of the symmetric groups. The Okounkov-Vershik approach, character formulas, and partition algebras., Cambridge Studies in Advanced Mathematics 121. Cambridge: Cambridge University Press (ISBN 978-0-521-11817-0/hbk). xv, 412 p. (2010). ZBL1230.20002.

(although this is based on the material of the above paper)

Section 3 of my own paper (which doesn't appear in the citation engine, so I'm providing an arXiv link instead) Stable Centres I: Wreath Products


Frobenius coordinates is very similar to multirectangular coordinates. There has been some studies on how (normalized) characters behave in these coordinates, see e.g. these slides by V. Feray.


Your Answer

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