6
$\begingroup$

Let $a_{n,k}$ be the coefficient of $$X_1^{\frac{k(n-1)}{2}}X_2^{\frac{k(n-1)}{2}}\cdots X_n^{\frac{k(n-1)}{2}}$$ in the expansion of the real polynomial $$\left(\prod\limits_{1\leq i<j\leq n}(X_j-X_i)\right)^k,$$where $n,k$ are positive integers such that $n>1$ and $k(n-1)\equiv0 \pmod 2$.

Since $$\prod\limits_{1\leq i<j\leq n}(X_j-X_i)=\sum\limits_{\sigma\in S_n}sgn(\sigma)\prod\limits_{i=1}^nX_{\sigma_i}^{i-1},$$ here the sum is computed over all permutations $\sigma$ of the set $\{1,2,\cdots,n\}$, we can get the following simple results:

$(1)a_{n,1}=0$ for every odd positive integer $n>1$;

$(2)a_{n,2}\neq 0$ for every positive integer $n>1$.

I want to ask whether $a_{n,k}$ is equal to 0 or not when $k>2$.

$\endgroup$
  • 1
    $\begingroup$ Your polynomial is the kth power of the Vandermonde polynomial. A naive google search yielded this article arxiv.org/abs/1201.4572 by C. Ballantine which discusses the coefficients when decomposing into Schur functions. I could not immediately parse the notation but it may be useful to you. $\endgroup$ – j.c. Mar 7 '16 at 17:57
7
$\begingroup$

Let's call your polynomial $\Delta$. Then $\Delta$ is anti-symmetric in the $x_i$ (if you switch $x_i$ to $x_j$ then $\Delta$ turns into $-\Delta$). If $k$ is odd, then $\Delta^k$ also has the property, so $a_{n,k}=0$ whenever $k$ is odd.

I will now show that $(-1)^{\binom{n}{2} m} a_{n,2m}>0$. Write $k=2m$. Let $T$ be the torus $\theta_1 + \theta_2 + \cdots + \theta_n=0$ inside $(\mathbb{R}/2 \pi \mathbb{Z})^n$, equipped with the Haar measure $d V$ with volume one.

Since $\Delta^{k}$ is homogenous of degree $k \binom{n}{2}$, it cannot contain monomials of the form $(x_1 x_2 \cdots x_n)^e$ for any $e$ other than $k \tfrac{n-1}{2}$. Therefore, we can extract $a_{n,k}$ by a contour integral over $T$:

$$a_{n,k} = \int_T \Delta(e^{i \theta_1}, \ldots, e^{i \theta_n})^{2m} dV$$ $$=\int_T \prod_{a<b} \left( e^{i \theta_a} - e^{i \theta_b} \right)^{2m} dV$$ $$=\int_T \prod_{a<b} \left( e^{i2 \theta_a} - 2 e^{i (\theta_a+\theta_b)} + e^{i 2 \theta_b} \right)^{m} dV$$ $$=\int_T e^{i (n-1) \sum \theta_a} \prod_{a<b} \left( e^{i (\theta_a- \theta_b)} - 2 + e^{i (\theta_b-\theta_a)} \right)^{m} dV$$ $$=(-2)^{\binom{n}{2} m} \int_T \prod_{a<b} \left( 1 - \cos (\theta_a - \theta_b) \right)^{m} dV.$$ Going from the second-to-last line to the last line, I used that $\sum \theta_a=0$ on $T$.

The integrand is nonnegative (and positive except on a set of measure zero) so the final integral is positive.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ On reflection, it probably would have been easier to integrate $(x_1 \ldots x_n)^{-k(n-1)/2} \Delta^k$ over $(\mathbb{R}/2 \pi \mathbb{Z})^n$, but the final result is the same either way. $\endgroup$ – David E Speyer Mar 7 '16 at 19:36
  • $\begingroup$ David Speyer, thank you very much for your answer! I am not familiar with the Haar measure and I will learn about it. I want to ask you first whether the following is right: $$\int_T(e^{i\theta_1})^{a_1}\cdots (e^{i\theta_n})^{a_n}dV=0$$for all nonnegative integers $a_1,\cdots,a_n$ satisfy $a_1+\cdots+a_n=\frac{kn(n-1)}{2}$ except for $a_1=\cdots=a_n=\frac{k(n-1)}{2}$. $\endgroup$ – user173856 Mar 12 '16 at 6:40
  • $\begingroup$ Yes, that's right. $\endgroup$ – David E Speyer Mar 12 '16 at 16:08
10
$\begingroup$

This is not an answer but a conjecture based on some trials. $$a_{n,k} = \begin{cases} 1, & n=1 \\ 0, & n\gt 1, k\text{ odd} \\ (-1)^{t\binom n2}\displaystyle\frac{(tn)!}{(t!)^n}, & k=2t. \end{cases} $$ Note that $\frac{(tn)!}{(t!)^n}$ is the number of ways of partitioning a set of size $tn$ into $n$ sets of size $t$.

Oh, now I see that this is a special case of Macdonald's constant-term conjecture, see this article, which has references.

| cite | improve this answer | |
$\endgroup$

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.