Given $n\in \mathbb N$, how can I compute the determinant of $(a_{ij})_{1\leq i,j \leq n} \in \mathcal M_{n\times n}(\mathbb R)$ where, for each $1\leq i,j\leq n:$ $$a_{ij}= \begin{cases} \lambda &i=j \\ \mu & i\neq j \end{cases}$$ The only (or the only easy) way to do this I think involves using Laplace's formula, expanding along any row. But doing this everything gets too messy, I can't express it in a good way. I tried to get a recursive relation and prove it via induction but didn't get anywhere. Is there an easier way? Any hints appreciated.
-
The fact that $[1\ 1\ \cdots \ 1]^T$ is an eigenvector with eigenvalue $\lambda + (n-1)\mu$ might help. – Paul Sundheim Apr 19 '17 at 18:18
-
This seems to be the same question as here: https://math.stackexchange.com/questions/86644/determinant-of-a-specially-structured-matrix https://math.stackexchange.com/questions/165816/computing-determinant-of-a-specific-matrix – Martin Sleziak Apr 19 '17 at 18:56
2 Answers
Let $A$ be the matrix:
$$\begin{pmatrix} \lambda & \mu & \mu & \cdots & \mu \\ \mu & \lambda & \mu & \cdots & \mu \\ \mu & \mu & \lambda & \cdots & \mu \\ \vdots & \vdots & \vdots && \vdots \\ \mu & \mu & \mu & \cdots & \lambda \\ \end{pmatrix}$$
Consider:
$$\begin{pmatrix} \lambda & \mu & \mu & \cdots & \mu \\ \mu & \lambda & \mu & \cdots & \mu \\ \mu & \mu & \lambda & \cdots & \mu \\ \vdots & \vdots & \vdots && \vdots \\ \mu & \mu & \mu & \cdots & \lambda \\ \end{pmatrix} \begin{pmatrix}1\\1\\1\\\vdots\\1\end{pmatrix} = \begin{pmatrix}\lambda+(n-1)\mu\\\lambda+(n-1)\mu\\\lambda+(n-1)\mu\\\vdots\\\lambda+(n-1)\mu\end{pmatrix} = \left[\lambda+(n-1)\mu\right] \begin{pmatrix}1\\1\\1\\\vdots\\1\end{pmatrix}$$
Hence, $\left[\lambda+(n-1)\mu\right]$ is an eigenvalue of $A$.
Consider:
$$\begin{pmatrix} \lambda & \mu & \mu & \cdots & \mu \\ \mu & \lambda & \mu & \cdots & \mu \\ \mu & \mu & \lambda & \cdots & \mu \\ \vdots & \vdots & \vdots && \vdots \\ \mu & \mu & \mu & \cdots & \lambda \\ \end{pmatrix} \begin{pmatrix}1\\0\\0\\\vdots\\-1\end{pmatrix} = \begin{pmatrix}\lambda-\mu\\0\\0\\\vdots\\\mu-\lambda\end{pmatrix} = \left[\lambda-\mu\right] \begin{pmatrix}1\\0\\0\\\vdots\\-1\end{pmatrix}$$
Hence, $\left[\lambda-\mu\right]$ is an eigenvalue of $A$.
One can construct other eigenvalues using this method, so in total we will have $(n-1)$ eigenvalues which are $\left[\lambda-\mu\right]$.
$$\det A = \prod \lambda_k = \left[\lambda+(n-1)\mu\right]\left[\lambda-\mu\right]^{n-1}$$

- 3,014
- 12
- 20
-
@PaulSundheim I can hardly make you believe that I didn't copy your comment. – DHMO Apr 19 '17 at 18:22
-
Stupid me, didn't think of using eigenvales to calculate the determinant. Thanks to all answerers! – Staki42 Apr 19 '17 at 18:39
Let $v$ be a column vector of ones. Then $vv^T$ is a matrix of ones, and the matrix of interest can be expressed as $\mu (vv^T)+(\lambda-\mu)I_n.$ The determinant of this rank-one matrix can then be found in a number of ways e.g. using the matrix determinant lemma, considering the eigenvalues, etc.

- 15,842
-
@PaulSundheim I only just saw your comment, so I didn't intend to copy :) – Semiclassical Apr 19 '17 at 18:19
-