2
$\begingroup$

Let $n$ be a natural number. Let $\lambda \mapsto n$ , be a partition of n. So $\lambda=(\lambda_1, \lambda_2, \ldots ,\lambda_k)$, with $\lambda_1\leq \ldots \leq \lambda_k$, and $\sum_{i=1}^k \lambda_i=n$. Now, let $m_i(\lambda)$ denote the number of times $i$ occur in the partition $\lambda$. Consider the following

$$\sum_{\lambda \mapsto n} \frac{1}{\prod_{i=1}^{n} i^{m_i(\lambda)}m_i(\lambda) !}=1$$.

I want to know whether this identity is correct. If it is true, what can be a proof of this equality!

Thank you.

$\endgroup$
2
  • 2
    $\begingroup$ Hint: you can give a combinatorial proof for $$\sum_{\lambda \vdash n} \frac{n!}{\prod_{i=1}^{n} i^{m_i(\lambda)}m_i(\lambda) !}=n!$$ $\endgroup$ Commented Mar 26, 2019 at 15:43
  • $\begingroup$ For what is basically a full solution, see this answer. $\endgroup$ Commented Mar 26, 2019 at 18:52

1 Answer 1

3
$\begingroup$

Observe that the cycle index of the symmetric group is given by

$$Z(S_n) = \sum_{\lambda \vdash n} \frac{1}{\prod_{i=1}^n i^{m_i(\lambda)} m_i(\lambda)!} \prod_{i=1}^n a_i^{m_i(\lambda)}.$$

This is the unlabeled multiset combinatorial class

$$\def\textsc#1{\dosc#1\csod} \def\dosc#1#2\csod{{\rm #1{\small#2}}}\textsc{MSET}_{=n}.$$

Now we may use Burnside or PET. With the former when setting all $a_i$ to one i.e. computing

$$Z(S_n; 1,1,1,\ldots)$$

we obtain the number of colorings of $n$ interchangeable slots with one color. There is just one such coloring, proving the claim.

Alternatively using PET we count the number of multisets of $n$ elements drawn from a source of one element, which is $\textsc{MSET}_{=n}(\mathcal{Z})$ with generating function

$$\sum_{\lambda \vdash n} \frac{1}{\prod_{i=1}^n i^{m_i(\lambda)} m_i(\lambda)!} \prod_{i=1}^n z^{i\times m_i(\lambda)} = z^n \sum_{\lambda \vdash n} \frac{1}{\prod_{i=1}^n i^{m_i(\lambda)} m_i(\lambda)!}.$$

There is just one such multiset and we must have $f(z) = z^n$, proving that the scalar is one, and we once more have the claim.

The underlying principle here is that

$$\frac{n!}{\prod_{i=1}^n i^{m_i(\lambda)} m_i(\lambda)!} = \frac{n!}{\prod_{i=1}^n (i!)^{m_i(\lambda)}} \prod_{i=1}^n \left(\frac{i!}{i}\right)^{m_i(\lambda)} \prod_{i=1}^n \frac{1}{m_i(\lambda)!}$$

counts the number of permutations in the symmetric group on $n$ elements with the lengths of the cycles in the factorization of such a permutation into disjoint cycles being given by the elements of $\lambda$ i.e.

$\prod_{i=1}^n a_i^{m_i(\lambda)}.$

With this classification we obtain every permutation exactly once and the sum of the LHS over $\lambda\vdash n$ is indeed equal to $n!$ as remarked in the comments.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .