11
$\begingroup$

I would like to know the proof for:

The determinant of the block matrix

$$\begin{pmatrix} A & B\\ C& D\end{pmatrix}$$

equals

$$(D-1) \det(A) + \det(A-BC) = (D+1) \det(A) - \det(A+BC)$$

when $A$ is a square matrix, $D$ is a scalar, $C$ is a row vector and $B$ is a column vector.

I appreciate your help.

$\endgroup$
2
  • 4
    $\begingroup$ The discussion of block matrices on Wikipedia's Determinant page is relevant here. $\endgroup$ Sep 11 '14 at 4:17
  • $\begingroup$ Yes. But I could not find the proof there and could not work it out how it is derived. $\endgroup$
    – Sani
    Sep 11 '14 at 4:26
15
$\begingroup$

We add an extra column and row to this matrix:

$$\left(\begin{array}{cc} A & B & 0 \\ C & D & 0 \\ 0 & 0& 1\end{array}\right).$$

This new matrix has the same determinant as the original.

Now we perform some row and column operations that don't change the determinant. First add $D-1$ times the last row to the second-to-last, and then subtract the last column from the second-to-last column. The result is

$$ \left(\begin{array}{ccc} A & B & 0 \\ C & 1 & D-1\\ 0 & -1 & 1\end{array}\right).$$

By linearity of the determinant in each row and column, we can write the determinant of the above as

$$ \begin{align} &\det\left(\begin{array}{ccc} A & B & 0 \\ C & 1 & 0\\ 0 & -1 & 1\end{array}\right) + \det\left(\begin{array}{ccc} A & B & 0 \\ C & 1 & D-1\\ 0 & -1 & 0\end{array}\right) \\ &=\det \left(\begin{array}{cc} A & B \\ C & 1\end{array}\right) + (-1)(D-1)\det(\left(\begin{array}{cc} A & B \\ 0 & -1\end{array}\right) \\ &= \det(A-BC) + (D-1) \det(A).\end{align}$$

The other formula can be derived by replacing $D-1$ with $D+1$ in the second step and proceeding in a similar way.

$\endgroup$
0
2
$\begingroup$

Let

$$\mathrm M := \begin{bmatrix} \mathrm A & \mathrm c\\ \mathrm r^\top & \alpha\end{bmatrix}$$

where $\mathrm A \in \mathbb R^{n \times n}$ and $\mathrm c, \mathrm r \in \mathbb R^n$. Assuming that $\mathrm A$ is invertible,

$$\underbrace{\begin{bmatrix} \mathrm I_n & 0_n\\ -\mathrm r^\top \mathrm A^{-1} & 1\end{bmatrix}}_{\det (\cdot) = 1} \begin{bmatrix} \mathrm A & \mathrm c\\ \mathrm r^\top & \alpha\end{bmatrix} = \begin{bmatrix} \mathrm A & \mathrm c\\ 0_n^\top & \alpha - \mathrm r^\top \mathrm A^{-1} \mathrm c\end{bmatrix}$$

Hence,

$$\begin{array}{rl} \det (\mathrm M) &= \det \begin{bmatrix} \mathrm A & \mathrm c\\ 0_n^\top & \alpha - \mathrm r^\top \mathrm A^{-1} \mathrm c\end{bmatrix}\\\\ &= \det (\mathrm A) \cdot (\alpha - \mathrm r^\top \mathrm A^{-1} \mathrm c)\\\\ &= (\alpha - 1) \cdot \det (\mathrm A) + \det (\mathrm A) \cdot (1 - \mathrm r^\top \mathrm A^{-1} \mathrm c)\\\\ &= (\alpha - 1) \cdot \det (\mathrm A) + \det (\mathrm A) \cdot \det(\mathrm I_n - \mathrm A^{-1} \mathrm c \mathrm r^\top)\\\\ &= (\alpha - 1) \cdot \det (\mathrm A) + \det(\mathrm A - \mathrm c \mathrm r^\top)\end{array}$$

where we used the Weinstein-Aronszajn determinant identity to conclude that

$$1 - \mathrm r^\top \mathrm A^{-1} \mathrm c = \det(\mathrm I_n - \mathrm A^{-1} \mathrm c \mathrm r^\top)$$

$\endgroup$

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.