5
$\begingroup$

What is the determinant of the sum of two matrices

$$\det (G + S)$$

where $S$ is all zeros except for a single column of $1$'s?

$$S = \begin{bmatrix} 0 & ... & 0 & 1 & 0 & ... & 0 \\ 0 & ... & 0 & 1 & 0 & ... & 0 \\ \vdots & & \vdots & \vdots & \vdots & & \vdots \\ 0 & ... & 0 & 1 & 0 & ... & 0 \\ \end{bmatrix}$$

I understand this can be solved by breaking up the determinant into columns, but I am unsure of how to do this. Also, $S$ is clearly singular - is there a general rule for the determinant of the sum of a singular and non-singular matrix (i.e. $G$ orthogonal $S$ singular)? Any help greatly appreciated

Im essentially asking what happens to the determinant of a matrix when you add $1$ to each entry in a column. Specifically, I am interested in the case where $G$ is orthogonal with $\det(G) = -1$. I would also be interested in the case where we add $1$ to just a single entry.

$\endgroup$
3
$\begingroup$

You can use the following two properties of determinants to answer this: \begin{align} &1.&\mathrm{det}\left[a_1,a_2,...,a_k+v_k,...,a_n\right] &= \mathrm{det}\left[a_1,a_2,...,a_k,...,a_n\right] + \mathrm{det}\left[a_1,a_2,...,v_k,...,a_n\right] \\ &2.&\mathrm{det}\left[a_1,a_2,...,\lambda a_k,...,a_n\right] &= \lambda \mathrm{det}\left[a_1,a_2,...,a_k,...,a_n\right] \end{align} where $a_i$, $v_k$ are columns of appropriate sizes and $\lambda \in \mathbb{R}$. Now you can generalize your problem to the following question:

Assume $G$ is orthogonal with $\mathrm{det}\left(G\right) = -1$, and let $g_i$ be the columns of $G$ then how do we compute $\mathrm{det}\left[g_1,g_2,...,g_k+v_k,...,g_n\right]$, where $v_k$ is some constant vector?

First, since $G$ is orthogonal, there exists $\alpha_i$ such that $v_k = \sum^{n}_{i=1} g_i \alpha_i$. Now you can use the first property and get: \begin{align} \mathrm{det}\left[g_1,g_2,...,g_k+v_k,...,g_n\right] &= \mathrm{det}\left[g_1,g_2,...,g_k+\sum^{n}_{i=1} g_i \alpha_i,...,g_n\right] \\ &=\mathrm{det}\left[g_1,g_2,...,\sum^{n}_{i=1,i\neq k} g_i \alpha_i,...,g_n\right]+\mathrm{det}\left[g_1,g_2,...,(1+\alpha_k)g_k,...,g_n\right]\\ &= (1+\alpha_k)\mathrm{det}\left[g_1,g_2,...,g_k,...,g_n\right]\\ &=-(1+\alpha_k) \end{align} You can also express $\alpha_k$ using orthogonal projections, $\alpha_k = \left(g_k^\mathrm{T}g_k\right)^{-1}g_k^\mathrm{T}v_k$ and thus get \begin{align} \mathrm{det}\left[g_1,g_2,...,g_k+v_k,...,g_n\right] &= -1-\left(g_k^\mathrm{T}g_k\right)^{-1}g_k^\mathrm{T}v_k \end{align}

In your case $v_k$ consists only of ones as elements. Furthermore, this discussion can also be extended to general matrices $A$ of course, but since orthogonal matrices were your concern, this is shorter.

$\endgroup$
1
$\begingroup$

Suppose that we have an invertible matrix $\mathrm A \in \mathbb R^{n \times n}$. Using Sylvester's determinant identity,

$$\det ( \,\mathrm A + 1_n \mathrm e_i^\top ) = \det(\mathrm A) \cdot \det ( \mathrm I_n + \mathrm A^{-1} 1_n \mathrm e_i^\top ) = \det(\mathrm A) \cdot ( 1 + \mathrm e_i^\top \mathrm A^{-1} 1_n )$$

If $\rm A$ is orthogonal, then $\det (\mathrm A) = \pm 1$ and $\mathrm A^{-1} = \mathrm A^\top$. Hence,

$$\det ( \,\mathrm A + 1_n \mathrm e_i^\top ) = \pm ( 1 + \mathrm e_i^\top \mathrm A^\top 1_n ) = \pm ( 1 + 1_n^\top \mathrm A \, \mathrm e_i ) = \color{blue}{\pm ( 1 + 1_n^\top \mathrm a_i )}$$

where $\mathrm a_i$ is the $i$-th column of $\rm A$. Thus,

$$\det ( \,\mathrm A + \mathrm e_i \mathrm e_j^\top ) = \pm ( 1 + \mathrm e_j^\top \mathrm A^\top \mathrm e_i ) = \pm ( 1 + \mathrm e_i^\top \mathrm A \, \mathrm e_j ) = \color{blue}{\pm ( 1 + a_{ij} )}$$

$\endgroup$

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.