4
$\begingroup$

Usually, I like working with determinants related to the Vandermonde matrix, i.e. $$\det(x_j^{i-1})=\prod_{i<j}(x_j-x_i).$$ However, I run into some unusual matrix and its determinant. Define the $(2n)\times (2n)$ matrix $\mathbb{M}_{2n}$ with entries $$\mathbb{M}_{2n}(i,j)= \begin{cases} (x_j-x_i)(x_i-x_{2n}) \qquad \text{if $i<j$}, \\ (x_j-x_i)(x_j-x_{2n}) \qquad \text{if $i\geq j$}. \end{cases}$$

QUESTION. Is this true? $$\det\mathbb{M}_{2n}=(x_1-x_2)^2(x_2-x_3)^2\cdots(x_{2n-1}-x_{2n})^2(x_{2n}-x_1)^2.$$

$\endgroup$
8
  • $\begingroup$ Apparently, you can remove the "$-x_n$" part and the determinant becomes $x_1^2 \left(x_1-x_2\right)^2 \left(x_2-x_3\right)^2 \cdots \left(x_{2n-1}-x_{2n}\right)^2$. This should generalize the corrected version of your question. $\endgroup$ Commented Feb 12, 2022 at 23:31
  • 1
    $\begingroup$ Note also that the matrix is antisymmetric, so it is not surprising that its determinant is a square. The question is why its Pfaffian is this nice. $\endgroup$ Commented Feb 12, 2022 at 23:31
  • $\begingroup$ This is reminiscent of (4.2) in Donald E. Knuth, Overlapping Pfaffians, Electronic Journal of Combinatorics 32 (1996), issue 2. $\endgroup$ Commented Feb 12, 2022 at 23:35
  • $\begingroup$ Also, the true "little brother" of this claim is not the Vandermonde determinant, but the known fact that $\det\left(\left(x_{\min\left\{i,j\right\}}\right)_{1\leq i,j\leq m}\right) = x_1 \left(x_1 - x_2\right) \left(x_2 - x_3\right) \cdots \left(x_{m-1}-x_m\right)$ for any $m \geq 1$. $\endgroup$ Commented Feb 12, 2022 at 23:45
  • 1
    $\begingroup$ @SamHopkins: I think all determinants/Pfaffians in the paper (you mentioned) involves the Vandermonde determinant, not the "cyclic type" in the question. $\endgroup$ Commented Feb 13, 2022 at 16:23

1 Answer 1

6
$\begingroup$

The answer is yes for any $n \geq 2$. (The $n=1$ case should be treated separately.)

The determinant, as a polynomial, contains $(x_m−x_{m+1})$ $(m=1,2,...,2n-1)$ and $(x_{2n}-x_1)$ as factors, because setting $x_m=x_{m+1}$ will make the $m$th and $(m+1)$th column of $\mathbb{M}_{2n}$ equal, and setting $x_{2n}=x_1$ will make the $(2n)$th and first column of $\mathbb{M}_{2n}$ equal, making the determinant zero.

Since the determinant of a skew-symmetric matrix can always be written as the square of a polynomial in the matrix entries that only depend on the size of the matrix, the determinant contains the factor $(x_1−x_2)^2(x_2−x_3)^2⋯(x_{2n−1}−x_{2n})^2(x_{2n}−x_1)^2$. As the determinant cannot have degree higher than that of this factor, the only thing left is to plug in values of $x_k$ such that LHS=RHS≠0 in the last equation stated in the question.

If we let $x_k=1$ if $k$ is odd and $0$ otherwise, then

$\mathbb{M}_{2n}(i,j)= \begin{cases} \phantom{-}1 \qquad \text{if $i<j$},i \text{ odd}, j \text{ even} \\ -1 \qquad \text{if $i > j$}, i \text{ even}, j \text{ odd}\\ \phantom{-}0 \qquad \text{otherwise} \end{cases}$.

The $(2n-1)$th row has only one nonzero element, namely $\mathbb M_{2n-1,2n}=1$, and the $(2n-1)$th column has only one nonzero element, namely $\mathbb M_{2n,2n-1}=-1$. By applying the Leibniz formula for determinants, we see that $\det \mathbb M_{2n}=\det \mathbb M_{2n-2}$. By mathematical induction, we have $\det \mathbb{M}_{2n}=1$ for any $n$. As the RHS also evaluates to $1$, we claim that LHS=RHS for any $x_k$s.

$\endgroup$
1
  • 3
    $\begingroup$ Note: This breaks for $n = 1$ since two factors are equal. It is best to remove the "$-x_{2n}$"s in the OP and prove the resulting more general formula (from which you can derive the OP by subtracting $x_{2n}$ from each $x_i$). $\endgroup$ Commented Feb 13, 2022 at 17:50

You must log in to answer this question.

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