How to evaluate this determinant $$\det\begin{bmatrix} a& b&b &\cdots&b\\ c &d &0&\cdots&0\\c&0&d&\ddots&\vdots\\\vdots &\vdots&\ddots&\ddots& 0\\c&0&\cdots&0&d \end{bmatrix}?$$

I am looking for the different approaches.

  • 1
    Have you tried induction? – bgins Apr 1 '12 at 15:30
  • Is there a name for such matrix?! – user2468 Apr 1 '12 at 18:37
  • @J.D. Yes. See my answer. – J. M. is not a mathematician Apr 14 '12 at 10:57
up vote 5 down vote accepted

Your (upper) arrowhead matrix can be decomposed as follows:

$$\begin{pmatrix}a&b&b&\cdots&b\\c&d&0&\cdots&0\\c&0&d&\ddots&\vdots\\\vdots &\vdots&\ddots&\ddots&0\\c&0&\cdots&0&d\end{pmatrix}=\color{red}{\begin{pmatrix}a-b-c&&&&\\&d&&&\\&&d&&\\&&&\ddots&\\&&&&d\end{pmatrix}}+\color{blue}{\begin{pmatrix}1&c\\&c\\&c\\&\vdots\\&c\end{pmatrix}}\cdot\color{magenta}{\begin{pmatrix}b&b&b&\cdots&b\\1&&&&\end{pmatrix}}$$

Now, one can then use the Sherman-Morrison-Woodbury formula for determinants:

$$\det(\color{red}{\mathbf A}+\mathbf{\color{blue}{U}\color{magenta}{V^\top}}) = \det(\mathbf I + \color{magenta}{\mathbf V^\top}\color{red}{\mathbf A}^{-1}\color{blue}{\mathbf U})\det(\color{red}{\mathbf A})$$

to yield

$$\begin{align*} &\begin{vmatrix}a&b&b&\cdots&b\\c&d&0&\cdots&0\\c&0&d&\ddots&\vdots\\\vdots &\vdots&\ddots&\ddots&0\\c&0&\cdots&0&d\end{vmatrix}\\ &=\det\left(\mathbf I+\color{magenta}{\begin{pmatrix}b&b&\cdots&b\\1&&&\end{pmatrix}}\color{red}{\begin{pmatrix}\frac1{a-b-c}&&&\\&\frac1d&&\\&&\ddots&\\&&&\frac1d\end{pmatrix}}\color{blue}{\begin{pmatrix}1&c\\&c\\&\vdots\\&c\end{pmatrix}}\right)\color{red}{(a-b-c)d^{n-1}}\\ &=\det\left(\mathbf I+\color{magenta}{\begin{pmatrix}b&b&\cdots&b\\1&&&\end{pmatrix}}\color{orange}{\begin{pmatrix}\frac1{a-b-c}&\frac{c}{a-b-c}\\&\frac{c}{d}\\&\vdots\\&\frac{c}{d}\end{pmatrix}}\right)\color{red}{(a-b-c)d^{n-1}}\\ &=\det\left(\mathbf I+\color{green}{\begin{pmatrix}\frac{b}{a-b-c}&bc\left(\frac1{a-b-c}+\frac{n-1}{d}\right)\\\frac1{a-b-c}&\frac{c}{a-b-c}\end{pmatrix}}\right)\color{red}{(a-b-c)d^{n-1}}\\ &=\frac{bc(1-n)+ad}{d(a-b-c)}(a-b-c)d^{n-1}\\ &=(bc(1-n)+ad)d^{n-2} \end{align*}$$

  • What a fascinating solution. – Sunni Apr 19 '12 at 19:13

If the dimension of the matrix is $2$, it's only $ad-bc$. If it's $\geq 3$, and $d=0$ the determinant is $0$. If $d\neq 0$, do $C_1\leftarrow C_1-\frac cdC_j$, $2\leq j\leq n$. The first column becomes $\pmatrix{a-(n-1)\frac{bc}d\\\ 0\\\ \vdots\\\ 0}$, and the determinant is $d^{n-1}\left(a-(n-1)\frac{bc}d\right)=d^{n-2}(ad-(n-1)bc)$.

  • Thanks, bgins's solution works also for general lambda matrices....I am looking for other approaches. – Sunni Apr 1 '12 at 16:46
  • You can see the matrix as a block matrix, and use the formula, if $a\neq 0$, $\det\pmatrix{a&b\mathbf 1^T\\\ c\mathbf 1&dI_{n-1}}=a\det(dI_{n-1}-\frac{bc}O)$, where $O$ is the matrix which has only ones. – Davide Giraudo Apr 1 '12 at 19:29

Let $A_n=(a_{ij})$ be the $n\times n$ matrix with $$ a_{ij}=\left\{\matrix{a&i=j=1\\b&i=1\ne j\\c&i\ne1=j\\d&i=j\ne1\\0&\text{otherwise}}\right. $$ and $\Delta_n=\det A_n$. Then ($\Delta_1=a$ according to my definition above, which may differ from your implicit definition), $\Delta_2=ad-bc$ and for $n\ge2$, $$ \Delta_{n+1}=d\,\Delta_n-bc\,d^{n-1}=\left(\Delta_n-bcd^{n-2}\right)d $$ expanding on the bottom row (or equivalently the rightmost column, but with the matrix below transposed and with $c$ in stead of $b$). This is because the $n\times n$ below has determinant $$\det\begin{bmatrix} b&b &\cdots&b&b\\ d &0&\cdots&0&0\\ 0&d&\ddots&\vdots&0\\ \vdots&\ddots&\ddots&0&0\\ 0&\cdots&0&d&0 \end{bmatrix} = (-1)^{n-1}bd^{n-1}\,, $$ and is multiplied by $(-1)^nc$ from $a_{n+1,1}=c$, giving the second term above. Thus $$ \eqalign{ \frac{\Delta_{n+1}}{d^{n-1}}&= \frac{\Delta_{n} }{d^{n-2}}-bc \quad \text{and} \quad \frac{\Delta_1}{d^{-1}}=ad \quad \implies \\\\ \frac{\Delta_{n}}{d^{n-2}}&= ad-(n-1)bc \qquad \implies \\\\ \Delta_{n}&=\big[ad-(n-1)bc\big]d^{n-2} } $$ for $n\ge1$ by induction.

  • Thanks, but there is a small error in your calculation. ... – Sunni Apr 1 '12 at 16:45
  • Thanks @Sunni, I think it's fixed now. – bgins Apr 1 '12 at 18:24

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.