12
$\begingroup$

For $\{h_{n}\}_{n=0}^{\infty}$ a real sequence, detone by $H_{n}$ the $n\times n$ Hankel matrix of the form $$ H_{n}:=\begin{pmatrix} h_{0} & h_{1} & \dots & h_{n-1}\\ h_{1} & h_{2} & \dots & h_{n}\\ \vdots & \vdots & \ddots & \vdots\\ h_{n-1} & h_{n} & \dots & h_{2n-2} \end{pmatrix}. $$ Just by a mathematical curiosity, I am interested in formulas for $\det H_{n}$ where $$ h_{n}=\binom{pn}{n} $$ where $p\geq2$ is an integer.

If $p=2$, the Hankel determinant (also called the Hankel transform) of central binomial coefficients is well-known and is not very difficult to evaluate. It reads $$ \det H_{n}=2^{n-1}. $$

For $p=3$, the formula (and its derivation) is more complicated, though it is also known that $$ \det H_{n}=3^{n-1}\left(\prod_{i=0}^{n-1}\frac{(3i+1)(6i)!(2i)!}{(4i)!(4i+1)!}\right), $$ see, for example, this paper for the proof.

Nevertheless, if $p\geq4$, I have not found any result about the corresponding Hankel determinant, nor I was able to even guess a form of the possible identity for $p=4$ while experimenting with Mathematica.

My questions are as follows:

  1. Is there something known about $\det H_{n}$ for $p\geq4$?
  2. Is anybody able to at least guess a formula for the case $p=4$?
  3. Is there any reason which might indicate that an explicit formula for $\det H_{n}$, with $p\geq4$, (like the one for $p=3$) need not exists?

Remark: I would like to remark that I now Krattenthaller's survey on advanced determinant calculus but I didn't find an answer therein.

$\endgroup$
  • 3
    $\begingroup$ One thing you can say about it: since ${4n \choose n}$ is divisible by $4$ for $n \ge 1$, the $2$-adic order $\nu_2(\det H_n) \ge 2n-2$. That inequality is an equality for $n = 1, 2, 4, 5, 6, 16, 17, 18, 20, 21, 22, 64, 65, 66, 68, 69, 70, 80, 81, 82, 84, 85, 86, \ldots$. Perhaps for all powers of $4$? $\endgroup$ – Robert Israel Dec 1 '16 at 19:37
  • 1
    $\begingroup$ Is $H_n > 0$ for all $n$ regardless of the value of $p$? This is, I think, equivalent to asking whether ${pn \choose n}$ is always a moment sequence, and in that case there is a standard technique to try related to orthogonal polynomials (see e.g. qchu.wordpress.com/2012/09/18/…). When $p = 2$ we get the moment sequence of an arcsin distribution; I already have no idea what distribution we get when $p = 3$, although the Hankel determinants do appear to be positive in this case. $\endgroup$ – Qiaochu Yuan Dec 2 '16 at 4:37
  • 1
    $\begingroup$ Perhaps it would be easier to consider instead of $\binom{4n}{n}$ more generally $\sum\binom{2n}{j}\binom{2n}{n-j}t^j.$ $\endgroup$ – Johann Cigler Dec 2 '16 at 11:53
  • 3
    $\begingroup$ Note that ${np \choose n}$ is a moment sequence. It follows from the "sin-omials" solution that ${np \choose n}=\mathbb{E}(Y_{1/p})^{np}=\mathbb{E}(Y_{1/p}^p)^n$ where $Y_\alpha=\frac{\sin(X)}{(\sin(\alpha X))^\alpha (\sin((1-\alpha)X))^{1-\alpha}}$ and $X$ is uniform on $[0,\pi]$. Thus $\det(H_n)>0$ (the distribution of $Y_{1/p}$ is not finitely supported). $\endgroup$ – esg Dec 7 '16 at 19:57
  • 1
    $\begingroup$ One can also ask for the Smith normal form (SNF) when $p=2$ and $p=3$. For $p=2$ this is easy. For $p=3$ it is an open problem. One intriguing observation when $p=3$ is that the number $d_n$ of 3's in the SNF satifies $|d_n-\lfloor \frac 23n\rfloor|\leq 1$ for $n<224$, but this fails for $n=224$. See page 5 of math.mit.edu/~rstan/papers/snf_survey.pdf. $\endgroup$ – Richard Stanley Feb 14 '17 at 23:02
3
$\begingroup$

This is not an answer.

Although there might be no "closed formula" for the case $p=4$, I wish to add the following to the case $p=2$ for which the determinants evaluate to $2^{n-1}$.

Suppose we decide to compute the determinant of the $n\times n$ Hankel matrix with left-most corner is at any position; in detail, consider the matrix $H_n(a)$ whose $(i,j)$ entry is given by $$\binom{2(i+j+a)}{i+j+a} \qquad \text{for} \qquad 0\leq i,j\leq n-1$$ then we have $$\det H_n(a)=\prod_{j=0}^{n-1}\binom{2a+2j}{a+j}\binom{a+n+j-1}{2j}^{-1}.$$ In particular, the original matrix $H_n=H_n(0)$ and we get a peculiar identity $$\prod_{j=0}^{n-1}\binom{2j}j\binom{n+j-1}{2j}^{-1}=2^{n-1}.$$

$\endgroup$

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.