Let $f(x)=(p_1-x)\cdots (p_n-x)$ $p_1,\ldots, p_n\in \mathbb R$ and let $a,b\in \mathbb R$ such that $a\neq b$

Prove that $\det A={bf(a)-af(b)\over b-a}$ where $A$ is the matrix:

$$\begin{pmatrix}p_1 & a & a & \cdots & a \\ b & p_2 & a & \cdots & a \\ \vdots & \vdots & \vdots & \ddots & \vdots \\ b & b & b & \cdots & p_n \end{pmatrix}$$

that is the entries $k_{ij}=a$ if $i<j$, $k_{ij}=p_i$ if $i=j$ and $k_{ij}=b$ if $i>j$

I tried to do it by induction over $n$. The base case for $n=2$ is easy $\det(A)=p_1p_2-ba={bf(a)-af(b)\over b-a}$

The induction step is where I don´t know what to do. I tried to solve the dterminant by brute force(applying my induction hypothesis for n and prove it for n+1) but I don´t know how to reduce it. It gets horrible.

I would really appreciate if you can help me with this problem. Any comments, suggestions or hints would be highly appreciated


Here is a possible proof without induction. The idea is to consider $\det A$ as a function of $p_n$.

We define the function $F: \Bbb R \to \Bbb R$ as $$ F(p) = \begin{vmatrix} p_1 &a &\ldots &a &a \\ b &p_2 &\ldots &a &a \\ \vdots &\vdots &\ddots &\vdots &\vdots \\ b &b &\ldots &p_{n-1} &a\\ b &b &\ldots &b &p \end{vmatrix} $$

$F$ is a linear function of $p$ and therefore completely determined by its values at two different arguments.

$F(a)$ and $F(b)$ can be computed easily: By subtracting the last row from all previous rows we get $$ F(a) = \begin{vmatrix} p_1 &a &\ldots &a &a \\ b &p_2 &\ldots &a &a \\ \vdots &\vdots &\ddots &\vdots &\vdots \\ b &b &\ldots &p_{n-1} &a\\ b &b &\ldots &b &a \end{vmatrix} = \begin{vmatrix} p_1-b &a-b &\ldots &a-b &0\\ 0 &p_2-b &\ldots &a-b &0 \\ \vdots &\vdots &\ddots &\vdots &\vdots\\ 0 &0 &\ldots &p_{n-1}-b &0 \\ b &b &\ldots &b &a \end{vmatrix} \\ $$ i.e. $$ F(a) = a(p_1-b)\cdots (p_{n-1}-b) \, . $$ In the same way (or by using the symmetry in $a$ and $b$) we get $$ F(b) = b (p_1-a)\cdots (p_{n-1}-a) \, . $$

Now we can compute $\det A = F(p_n)$ with linear interpolation: $$ \det A = \frac{b-p_n}{b-a} F(a) + \frac{p_n-a}{b-a} F(b) \\ = \frac{- a(p_1-b)\cdots (p_{n-1}-b)(p_n-b) + b(p_1-a)\cdots (p_{n-1}-a)(p_n-a) }{b-a} \\ = \frac{-af(b) + bf(a)}{b-a} \, . $$

  • $\begingroup$ Of course the same proof works for complex numbers as well. $\endgroup$ – Martin R Oct 20 '17 at 20:06
  • $\begingroup$ Works over the polynomial ring $\mathbb{Z}[a,b,p_1,\ldots,p_n]$ too, I believe, and thus for any values over any ring. (with the appropriate care taken in the cases where $b-a$ isn't cancellable; e.g. to divide before specializing the variables) $\endgroup$ – user14972 Oct 20 '17 at 22:07

Hint. To emphasise the dependence of the $A$ and $f$ on $n$ as well as the $p_i$s, rewrite them as $A(p_1,\ldots,p_n)$ and $f(x; p_1,\ldots,p_n)$.

Now, in the induction step, subtract the last column by the last but one column. Then apply Laplace expansion. The $(n,n-1)$-th minor is $\det A(p_1,\ldots,p_{n-2},b)$, while the $(n,n)$-th minor is $\det A(p_1,\ldots,p_{n-1})$. Hence \begin{align} \det A &= (p_{n-1}-a)\det A(p_1,\ldots,p_{n-2},b)\\ &\phantom{=}+ (p_n-b)\det A(p_1,\ldots,p_{n-1})\\ &= (p_{n-1}-a)\frac{bf(a; p_1,\ldots,p_{n-2},b)-af(b; p_1,\ldots,p_{n-2},b)}{b-a}\\ &\phantom{=}+ (p_n-b)\frac{bf(a; p_1,\ldots,p_{n-1})-af(b; p_1,\ldots,p_{n-1})}{b-a}. \end{align} The rest is straightforward if you write each $f$ as a product of linear factors according to the definition.


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.