5
$\begingroup$

$V_n(a_1,a_2\dots, a_n)$ is a $n\times n$ Vandermonde matrix = $$\left|\begin{array}[cccc] 11&a_1&\cdots&a^{n-1}_1\\ 1&a_2&\cdots&a^{n-1}_2\\ 1&a_3&\cdots&a^{n-1}_3\\ \vdots&\vdots&\ddots&\vdots\\ 1&a_n&\cdots&a_n^{n-1} \end{array}\right|$$

Replace $a_1$ by $x$ so that the first row is $1,x, \dots ,x^{n-1}$

$$V_n(x,a_2\dots, a_n) = \left|\begin{array}[cccc] 11&x&\cdots&x^{n-1}\\ 1&a_2&\cdots&a^{n-1}_2\\ 1&a_3&\cdots&a^{n-1}_3\\ \vdots&\vdots&\ddots&\vdots\\ 1&a_n&\cdots&a_n^{n-1} \end{array}\right|$$

Let $P(x) = V_n(x,a_2\dots, a_n)$.

(a) Show that $P(x)$ is a polynomial in $x$ of degree $\leq n-1$.

$z = 1\det() -x\det() + x^2\det() -\cdots+ (-1)^{n-1}x^{n-1} \cdot \det()$, where each $\det()$ stands for the determinant of a smaller matrix after removing the appropriate columns, rows.

Is this right? But isn't $z$ a polynomial of degree $n$?

(b) Show that $P(x)$ has $n-1$ distinct roots $a_2, \dots a_n$ and therefore has factorization $P(x) = k \prod_{i=2}^n(x-a_i)$ where the constant factor $k$ is the coefficient of $x^{n-1}.$

I'll be honest, I have no clue how to do this. Not even sure where to start. Nor do I know where to start on the rest.

(c) Show that $k = (-1)^{n-1}V_{n-1}(a_2,\dots a_n)$.

(d) Use parts (b) and (c) to deduce the recursion formula $$V_n(a_1, \dots a_n) = \left(\prod_{i=2}^n(a_i-a_1)\right)V_{n-1}(a_2,\dots a_n)$$

(e) Use part (d) to deduce that $V_n(a_1,a_2,\dots a_n) = \prod^n_{1\leq i< j\leq n} (a_j - a_i)$

$\endgroup$
  • $\begingroup$ You seem to have mixed up $\;a_i$'s and $\;z_i$'s ... $\endgroup$ – DonAntonio Apr 2 '14 at 3:46
2
$\begingroup$

(a) Show that $P(x)$ is a polynomial in $x$ of degree $\le n−1$.

Use the Leibniz formula for determinants, and show that each sum contains only one factor of $x^k$ from the top row.

(b) Show that $P(x)$ has $n−1$ distinct roots $a_2,~ \dots ,~ a_n$ and therefore has factorization $P(x)=k\prod^n_{i=2}(x−a_i)$ where the constant factor $k$ is the coefficient of $x^{n−1}.

Plug $x = a_i$ into your matrix, and compare the first row to the $i^{th}$ row. Show that the determinant must be zero in that case.

(c) Show that $k = (-1)^{n-1}V_{n-1}(a_2,\dots a_n)$

This follows immediately from the Laplace expansion that you did in part (a).

(d) Use parts (b) and (c) to deduce the recursion formula...

Just plug $x=a_1$ and part (c) into the formula in (b).

(e) Use part (d) to deduce that...

Use the fact that $$\bigg(\prod_{x \in X} f_x\bigg)\bigg(\prod_{y \in Y} g_y\bigg) = \prod_{(x,y) \in X \times Y} f_x \cdot g_y$$

$\endgroup$
2
$\begingroup$

For part (a), this is just development (Laplace expansion) of the determinant by the first row. Actually the $\det()$ factors should have alternating signs. Since the only occurrences of $x$ are in that first row, all the $\det()$ expressions are constants, and one gets a polynomial of degree at most $n-1$ (from the final term) in $x$.

For part (b), that $P(a_i)=0$ for $i=2,3,\ldots,n$ is just the fact that $P(a_i)$ equals the determinant of the matrix obtained by substituting $a_i$ for $x$, so from the original matrix $a_1$ has been replaced by $a_i$, and as this matrix has its rows $1$ and $i$ identical, its determinant vanishes. all this uses is that the Laplace expansion used commutes with such substitution. Furthermore a polynomial of degree at most $n-1$ with $n-1$ specified roots $a_2,\ldots,a_n$ can only be a scalar multiple of $(x-a_2)\ldots(x-a_n)$.

For part (c), this is just remarking that the $\det()$ in question is $(-1)^{n-1}$ times the determinant of the lower-left $(n-1)\times(n-1)$ submatrix, which determinant precisely matches the definition of $V_{n-1}(a_2,\ldots,a_n)$.

For part (d) write $(-1)^{n-1}\prod_{i=2}^n(x-a_i)=\prod_{i=2}^n(a_i-x)$ to get $$ V(x,a_2,\ldots,a_n)= (-1)^{n-1}V_{n-1}(a_2,\ldots,a_n)\prod_{i=2}^n(x-a_i) =V_{n-1}(a_2,\ldots,a_n)\prod_{i=2}^n(a_i-x), $$ and then set $x=a_1$ to get $$ V(a_1,a_2,\ldots,a_n) =V_{n-1}(a_2,\ldots,a_n)\prod_{i=2}^n(a_i-a_1), $$

Part (e) applies induction on $n$ to $V_{n-1}(a_2,\ldots,a_n)$ (the starting case is $V_0()=1=\prod_{1\leq i<j\leq 0}1$, an empty product, or if you fear $n=0$ it is $V_1(a)=1=\prod_{1\leq i<j\leq 1}1$, still an empty product), to get $$ V(a_1,a_2,\ldots,a_n) =\left(\prod_{2\leq i<j\leq n}(a_j-a_i)\right)\prod_{j=2}^n(a_j-a_1) =\prod_{1\leq i<j\leq n}(a_j-a_i). $$

$\endgroup$

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.