3
$\begingroup$

Recall that a permutohedron is a graph on the set of permutations $S_n$ with an edge between $\sigma$ and $\tau$ if they differ by one adjacent transposition: $\tau = (i,i+1) \circ \sigma$ for some $i \in [n-1]$. $\tau < \sigma$ if $\sigma(i) < \sigma(i+1)$, that is, if $\tau$ has one more disarranged pair than $\sigma$. This relation can be extended to a maximal partial order on $S_n$ in the obvious manner. So for instance, $1234 \geq \sigma \geq 4321$ for any $\sigma \in S_4$. This is essentially the metric $b_4$ in this paper, modified to allow only corrections of type $b_3$; but this is not important.

I would like to compute the number of pairs $(\sigma_1,\sigma_2) \in S_n^2$ for which $\sigma_1 > \sigma_2$. This allows me to know the number of pairs that are incomparable. Intuitively, $\sigma_1 > \sigma_2$ are comparable if one can traverse the permutohedron from $\sigma_1$ to $\sigma_2$ in a strictly descending path as show in wikipedia here.

$\endgroup$
4
  • 1
    $\begingroup$ I don't know that this question has a particularly satisfying answer, but I do know that the order you're using is opposite "weak Bruhat order", and recommend you reverse it before confusing people. Have you computed the first few numbers and checked the OEIS? $\endgroup$ Jul 14, 2014 at 8:05
  • 2
    $\begingroup$ People have looked at this kind of question, but I believe your particular question is open. See for instance Exercise 3.185 of *Enumerative Combinatorics", vol. 1. Your question is closely related to mathoverflow.net/questions/173042, since the number of pairs $\tau\leq\sigma$ is equal to $\sum_\sigma \#[\mathrm{id},\sigma]$. $\endgroup$ Jul 14, 2014 at 13:52
  • 1
    $\begingroup$ The sequence starts with $17,151,1899, 31711$, which is oeis.org/A007767. I don't understand what the avoidance condition there should be (avoiding $12$ means for me the same as being the permutation $n,\ldots,2,1$, which is clearly not the case here.) $\endgroup$ Jul 14, 2014 at 15:11
  • 2
    $\begingroup$ Lower and upper bounds are given in a paper by Hammett and Pittel. They conjecture that the real answer is qualitatively close to the upper bound $(n!)^2 (0.362)^n$ obtained in the paper. $\endgroup$ Jul 18, 2014 at 22:17

1 Answer 1

1
$\begingroup$

Recall that a permutation $\sigma$ of order $n$ is uniquely determined by its inversion table: $T(\sigma)=(t_1,\dots,t_n)$, where $t_k$ is the number of integers larger than $k$ appearing in $\sigma$ ahead of $k$. The range for $t_k$ in the inversion table is $[0,n-k]$. In particular, the inversion table for the identity permutation is zero vector, while $T((n,n-1,\dots,1)) = (n-1,n-2,\dots,0,0)$.

Claim. If $\sigma_1 > \sigma_2$, then $T(\sigma_1)$ is dominated (component-wise) by $T(\sigma_2)$.

Proof. First consider two permutations that are one transposition away from each other. If $\sigma_2$ is obtained from $\sigma_1$ by exchanging $\sigma_1(i)$ and $\sigma_1(i+1)$, then:

  • if $\sigma_1(i) > \sigma_1(i+1)$, then $T(\sigma_2)$ is the same as $T(\sigma_1)$, except for the $\sigma_1(i+1)$-th element which is one less than in $T(\sigma_1)$;

  • if $\sigma_1(i) < \sigma_1(i+1)$, then $T(\sigma_2)$ is the same as $T(\sigma_1)$, except for the $\sigma_1(i)$-th element which is one more than in $T(\sigma_1)$.

That is, $\sigma_1 > \sigma_2$ if and only if in $T(\sigma_1)$ one element is decreased by one as compared to $T(\sigma_2)$.

If $\sigma_1 > \sigma_2$ and they are $k$ transpositions away from each other, then $T(\sigma_1)$ can be obtained from $T(\sigma_2)$ by $k$ decrements of its elements, which implies that $T(\sigma_1)$ is dominated by $T(\sigma_2)$. QED

Now, every pair of permutations $\sigma_1 \geq \sigma_2$ from $S_n$ corresponds to a vector of 2-combinations with repetitions (composed from the corresponding elements of their inversion tables), where the $k$-th 2-combination is chosen from the range $[0,n-k]$. The number of such vectors is $$\binom{n+1}{2}\cdot \binom{n}{2} \cdots \binom{2}{2} = \frac{(n+1)\cdot n!^2}{2^n}$$ and this gives an upper bound for the number of such pairs of permutations.

$\endgroup$

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.