3
$\begingroup$

Let $\lambda=(\lambda_1,\lambda_2,\dotsc,\lambda_{\ell(\lambda)})$ be an integer partition of positive numbers where $\ell(\lambda)$ is the length of the partition. One may associate a Ferrer diagram or Young diagram $Y$ to $\lambda$. I read this concept Ferrer's matrix on OEIS but nowhere else.

First, compute $m=\max\{\lambda_1,\ell(\lambda)\}$ and then construct an $m\times m$ matrix by inserting a $1$ in each box of $Y$ while inserting $0$ elsewhere. For example, if $\lambda=(4,3,1)\vdash 8$ then $m=4$ and the corresponding matrix becomes $\begin{pmatrix} 1&1&1&1 \\ 1&1&1&0 \\ 1&0&0&0 \\ 0&0&0&0 \end{pmatrix}$. Now, add elements of all anti-diagonals to get $(1,2,3,2,0,0,0)$. Re-sorting (and ignoring $0$'s) gives $\mu=(3,2,2,1)\vdash 8$.

If one does apply the procedure to all (ordered) partitions of $n=4$, i.e. $\{4, 31, 22, 211, 1111\}$, the resulting partitions form the multi-set $\{1111, 211, 211, 211, 1111\}$. The new statistic (frequencies) reads $2, 3$. Here is a short table from OEIS: $$\begin{array}l 1 \\ 2 \\ 2, 1 \\ 2, 3 \\ 2, 2, 3 \\ 2, 2, 6, 1 \\ 2, 2, 4, 3, 4. \end{array}$$ On a different route, given any partition $\lambda$ of $n$, compute its hook-lengths and add them to generate some statistic. As an example, take $n=4$. Associate the set hook-lengths $\{(4, 3, 2, 1), (4, 1, 2, 1), (3, 2, 2, 1), (4, 2, 1, 1), (4, 3, 2, 1)\}$. The resulting multi-set of sums is $\{10, 8, 8, 8, 10\}$ with frequencies $2, 3$.

QUESTION. Is there a bijection to explain the total agreement between the two statistics (see above table)? I would also be nice if there is a generating function here.

REMARK. The sum of hooks of $\lambda$ equals sum of squares of parts of $\lambda$.

UPDATE. Peter Taylor (comments below 1 2) identified an error in the above correspondence. So, the answer to my question is negative, unfortunately.

$\endgroup$
10
  • $\begingroup$ Are you saying you know that for all values of $n$ these two statistics have the same frequencies on partitions of $n$? $\endgroup$ Apr 9, 2022 at 21:04
  • $\begingroup$ @SamHopkins: I was able to compute the latter data and it agrees with the one shown. Is that what you ask? $\endgroup$ Apr 9, 2022 at 21:16
  • $\begingroup$ it might make sense to ask findstat, but I can probably only do so on monday. $\endgroup$ Apr 9, 2022 at 22:05
  • 1
    $\begingroup$ The dynamics of the map $\lambda \mapsto \mu$ might still be interesting to explore further. $\endgroup$ Apr 10, 2022 at 18:04
  • 2
    $\begingroup$ @SamHopkins, it turns out to be idempotent. $\endgroup$ Apr 12, 2022 at 8:33

1 Answer 1

3
$\begingroup$

First, compute $m=\max\{\lambda_1,\ell(\lambda)\}$ and then construct an $m\times m$ matrix by inserting a $1$ in each box of $Y$ while inserting $0$ elsewhere. For example, if $\lambda=(4,3,1)\vdash 8$ then $m=4$ and the corresponding matrix becomes $\begin{pmatrix} 1&1&1&1 \\ 1&1&1&0 \\ 1&0&0&0 \\ 0&0&0&0 \end{pmatrix}$. Now, add elements of all anti-diagonals to get $(1,2,3,2,0,0,0)$. Re-sorting (and ignoring $0$'s) gives $\mu=(3,2,2,1)\vdash 8$.

The sorting is a bit of a red herring. Denote the anti-diagonal frequency vector, $(1,2,3,2,0,0,0)$ in the example, as $(ad_1, ad_2, ad_3, \ldots)$. It's easy to see that $ad_k > ad_{k-1}$ requires $\forall 1 \le i \le k: ad_i = i$. So if $A = \max(\mu)$ the unique "unsorting" to a legal anti-diagonal frequency vector which produces it is $(1, 2, 3, \ldots, A, ad_{A+1}, ad_{A+2}, \ldots)$ with $A \ge ad_{A+1} \ge ad_{A+2} \ge \cdots$

Let $H$ denote the hook length sum. Suppose we build the Ferrers diagram starting from the unique partition of $1$ and adding a cell at a time in such a way that we always have a valid Ferrers diagram. When we add a cell at $(r, c)$ we increase $ad_{r+c-1}$ by $1$ and $H$ by $(r-1)+(c-1)+1 = r+c-1$, so $H - \sum_k k \cdot ad_k$ is invariant. The initial partition has $H=1$, $ad_1 = 1$, so $$H = \sum_k k \cdot ad_k$$

Thus for every $\mu$ there is a unique hook length sum $H_{\mu}$ which all partitions having that anti-diagonal frequency partition share, but there's nothing to prevent different $\mu$ sharing their hook length sum, and e.g. anti-diagonal frequency vectors $(1,2,2,2,2) \to 1 + 4 + 6 + 8 + 10 = 29$ and $(1, 2, 3, 1, 1, 1) \to 1 + 4 + 9 + 4 + 5 + 6 = 29$ collide.

$\endgroup$
1
  • $\begingroup$ The number of distinct $\mu$ among partitions of $n$ is A000009: see Jon Perry's comment of Sep 21 2005. $\endgroup$ Apr 12, 2022 at 11:48

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.