3
$\begingroup$

I'm having troubles solving the following determinant:

$$\left| \begin{array}{cccc} a_1 & x & \ldots & x \\ x & a_2 & \ldots & \vdots \\ \vdots & \ldots & \ddots & x \\ x & \ldots & x & a_n \\ \end{array} \right|$$

I have tried coming up with recurrent formula, but there's always one lingering row and column. I also tried finding values of x for which determinant is 0, hoping I could get a nice polynomial solution, with no luck.

By subtracting first row from every other I get somewhat nicer result, but there's still that one row/column which I can't solve. If all other columns are added to the first one and then I subtract the first row, then the column doesn't depend on x anymore, but I haven't found 1..n much more useful.

$\endgroup$
  • $\begingroup$ Possible duplicate of How to calculate the following determinants (all ones, minus $I$) $\endgroup$ – amd Jan 23 '17 at 20:06
  • $\begingroup$ @amd the diagonal entries are different here. $\endgroup$ – StubbornAtom Jan 23 '17 at 20:12
  • $\begingroup$ Even though the question is different, the answer is similar; however, I cannot pull out any factor after adding everything to the first row, since I have different numbers on the diagonal (I have $$a_i+(n-1)x$$ instead of $$a+(n-1)x$$) $\endgroup$ – Luke Jan 23 '17 at 20:13
  • 4
    $\begingroup$ Your matrix is $D + xk^T k$, where $D$ is the diagonal matrix $\operatorname{diag}\left(a_1-x,a_2-x,\ldots,a_n-x\right)$, and $k$ is the row vector $\left(1,1,\ldots,1\right)$. Now, use the fact so inventively and unambiguously christened the "matrix determinant lemma" ( en.wikipedia.org/wiki/Matrix_determinant_lemma#Statement , see the version without the invertibility requirement). $\endgroup$ – darij grinberg Jan 23 '17 at 20:18
  • 1
    $\begingroup$ @darij grinberg You should give your comment as an answer. $\endgroup$ – Jean Marie Jan 23 '17 at 21:20
7
$\begingroup$

As requested, here is the answer I hinted to in my comment:

Theorem 1. Let $n\in\mathbb{N}$. Let $\mathbb{K}$ be a commutative ring. Let $a_{1},a_{2},\ldots,a_{n}$ be $n$ elements of $\mathbb{K}$. Let $x\in\mathbb{K}$. Let $A\in\mathbb{K}^{n\times n}$ be the $n\times n$-matrix $\left( \begin{array} [c]{ccccc} a_{1} & x & x & \cdots & x\\ x & a_{2} & x & \cdots & x\\ x & x & a_{3} & \cdots & x\\ \vdots & \vdots & \vdots & \ddots & \vdots\\ x & x & x & \cdots & a_{n} \end{array} \right) $ (this is the $n\times n$-matrix whose diagonal entries are $a_{1},a_{2},\ldots,a_{n}$, while all its other entries are $x$). For each $i\in\left\{ 1,2,\ldots,n\right\} $, set $b_{i}=\prod\limits_{\substack{k\in \left\{ 1,2,\ldots,n\right\} ;\\k\neq i}}\left( a_{k}-x\right) $. Then,

$\det A=\prod\limits_{i=1}^{n}\left( a_{i}-x\right) +x\sum\limits_{i=1}^{n}b_{i}$.

To prove this, we recall the following fact (known as the matrix determinant lemma, although it would not surprise me if it has several contenders for its rather generic name):

Lemma 2. Let $n\in\mathbb{N}$. Let $\mathbb{K}$ be a commutative ring. Let $\mathbf{A}\in\mathbb{K}^{n\times n}$ be an $n\times n$-matrix. Let $\mathbf{u}\in\mathbb{K}^{n\times1}$ and $\mathbf{v}\in\mathbb{K}^{n\times1}$ be two column vectors. Then,

$\det\left( \mathbf{A}+\mathbf{uv}^{T}\right) =\det\mathbf{A}+\mathbf{v} ^{T}\left( \operatorname*{adj}\mathbf{A}\right) \mathbf{u}$.

(Here and in the following, $\operatorname*{adj}\mathbf{A}$ denotes the adjugate of $\mathbf{A}$.)

See Matrix determinant lemma with adjugate matrix for a proof of Lemma 2.

We also shall use the following simple fact:

Lemma 3. Let $n\in\mathbb{N}$. Let $\mathbb{K}$ be a commutative ring. Let $a_{1},a_{2},\ldots,a_{n}$ be $n$ elements of $\mathbb{K}$. For each $i\in\left\{ 1,2,\ldots,n\right\} $, set $b_{i}=\prod\limits_{\substack{k\in \left\{ 1,2,\ldots,n\right\} ;\\k\neq i}}a_{k}$. Then,

$\operatorname*{adj}\left( \left( a_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}\right) =\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}$.

(Here, as usual, $\delta_{i,j}$ is the Kronecker delta of $i$ and $j$.)

Proof of Lemma 3. Lemma 3 simply says that the adjugate of the diagonal matrix with diagonal entries $a_{1},a_{2},\ldots,a_{n}$ is the diagonal matrix with diagonal entries $b_{1},b_{2},\ldots,b_{n}$. This is easy to check (since each $\left( n-1\right) \times\left( n-1\right) $-submatrix of a diagonal matrix either is a diagonal matrix itself, or has a zero row).

Now, we can prove Theorem 1:

Proof of Theorem 1. Let $\mathbf{A}$ be the $n\times n$-matrix $\left( \left( a_{i}-x\right) \delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n} $. (This $\mathbf{A}$ is the diagonal matrix with diagonal entries $a_{1}-x,a_{2}-x,\ldots,a_{n}-x$). Thus, $\det\mathbf{A}=\prod\limits_{i=1} ^{n}\left( a_{i}-x\right) $ (since the determinant of a diagonal matrix is the product of its diagonal entries). But Lemma 3 (applied to $a_{1} -x,a_{2}-x,\ldots,a_{n}-x$ instead of $a_{1},a_{2},\ldots,a_{n}$) yields that

$\operatorname*{adj}\left( \left( \left( a_{i}-x\right) \delta _{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}\right) =\left( b_{i} \delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}$.

Since $\left( \left( a_{i}-x\right) \delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}=\mathbf{A}$, this rewrites as

$\operatorname*{adj}\mathbf{A}=\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}$.

Let $\mathbf{v}\in\mathbb{K}^{n\times1}$ be the column vector $\left( 1,1,\ldots,1\right) ^{T}$. Then, $\mathbf{vv}^{T}=\left( 1\right) _{1\leq i\leq n,\ 1\leq j\leq n}$, so that $x\mathbf{vv}^{T}=x\left( 1\right) _{1\leq i\leq n,\ 1\leq j\leq n}=\left( x\right) _{1\leq i\leq n,\ 1\leq j\leq n}$. Hence, it is easy to see that $A=\mathbf{A}+x\mathbf{vv}^{T}$.

Recall that $\mathbf{v}=\left( 1,1,\ldots,1\right) ^{T}$. Thus, for every $n\times n$-matrix $B$, we have

$\mathbf{v}^{T}B\mathbf{v}=\left( \text{the sum of all entries of }B\right) $.

Applying this to $B=\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}$, we obtain

$\mathbf{v}^{T}\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}\mathbf{v}$

$=\left( \text{the sum of all entries of }\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}\right) $

$=\sum\limits_{i=1}^{n}\sum\limits_{j=1}^{n}b_{i}\delta_{i,j}=\sum\limits_{i=1}^{n}b_{i} \underbrace{\sum\limits_{j=1}^{n}\delta_{i,j}}_{=1}=\sum\limits_{i=1}^{n}b_{i}$.

But from $A=\mathbf{A}+x\mathbf{vv}^{T}$, we obtain

$\det A=\det\left( \mathbf{A}+x\mathbf{vv}^{T}\right) $

$=\underbrace{\det\mathbf{A}}_{=\prod\limits_{i=1}^{n}\left( a_{i}-x\right) }+\mathbf{v}^{T}\underbrace{\left( \operatorname*{adj}\mathbf{A}\right) }_{=\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n} }x\mathbf{v}$

(by Lemma 2, applied to $\mathbf{u}=x\mathbf{v}$)

$=\prod\limits_{i=1}^{n}\left( a_{i}-x\right) +\mathbf{v}^{T}\left( b_{i} \delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}x\mathbf{v}$

$=\prod\limits_{i=1}^{n}\left( a_{i}-x\right) +x\underbrace{\mathbf{v}^{T}\left( b_{i}\delta_{i,j}\right) _{1\leq i\leq n,\ 1\leq j\leq n}\mathbf{v}} _{=\sum\limits_{i=1}^{n}b_{i}}$

$=\prod\limits_{i=1}^{n}\left( a_{i}-x\right) +x\sum\limits_{i=1}^{n}b_{i}$.

This proves Theorem 1.

$\endgroup$
  • $\begingroup$ @becko: Your answer uses division, which in general requires a Zariski-density-style argument; I made the conscious choice to avoid that in my post. Other than this, your answer is indeed simpler. $\endgroup$ – darij grinberg Sep 26 '18 at 16:41
  • $\begingroup$ This answer looks overly complicated. The OP question can be answered more directly, please see math.stackexchange.com/a/2931799/10063. Also (as in your comment), the result can be computed simply applying the Sherman Morrison formula. $\endgroup$ – becko Sep 26 '18 at 16:42
  • $\begingroup$ @becko: The Sherman-Morrison uses division, too :) $\endgroup$ – darij grinberg Sep 26 '18 at 16:43
  • $\begingroup$ I meant for real matrices. I think your proof is much more ambitious. $\endgroup$ – becko Sep 26 '18 at 16:47

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.