Link Search Menu Expand Document

Determinants

Table of contents
  1. Determinant
    1. Singular Matrix
    2. Properties of the Determinant
  2. Laplace Expansion
    1. Singleton Matrix Case
    2. General Case
      1. Expansion Along a Row
      2. Expansion Along a Column
    3. 2x2 Matrix Application
  3. Rule of Sarrus
  4. Determinant of a Triangular Matrix
  5. Leibniz Formula
  6. Geometrical Interpretation

Determinant

Determinant of a $n \times n$ square matrix $\boldsymbol{A}$ is denoted:

$$ \det(\boldsymbol{A}) \quad \text{or} \quad |\boldsymbol{A}| $$

Only defined for square matrices.

Singular Matrix

The following is an important property of the determinant.

For a square matrix $\boldsymbol{A}$,

$$ \det(\boldsymbol{A}) = 0 \quad \iff \quad \boldsymbol{A} \text{ is singular} \quad \iff \nexists \boldsymbol{A}^{-1} $$

Thus a singular matrix is not invertible, and the system of linear equations $\boldsymbol{A}\boldsymbol{x} = \boldsymbol{b}$ has no unique solution.

Properties of the Determinant

Let $\boldsymbol{A}$, $\boldsymbol{B}$, $\boldsymbol{C}$ be square matrices of same size and $k$ a scalar:

  • $\det(k\boldsymbol{A}) = k^n \det(\boldsymbol{A})$
    • Multiplying a single row or column by $k$ multiplies the determinant by $k$.
    • Multiplying all rows or columns by $k$ multiplies the determinant by $k^n$.
  • If $\boldsymbol{A}$, $\boldsymbol{B}$, $\boldsymbol{C}$ only differ by a single (same) row or column $i$, and the different row or column of $\boldsymbol{C}_i = \boldsymbol{A}_i + \boldsymbol{B}_i$, then $\det(\boldsymbol{C}) = \det(\boldsymbol{A}) + \det(\boldsymbol{B})$.
  • If $\boldsymbol{A}$ is obtained by swapping two rows or columns of $B$, then $\det(\boldsymbol{A}) = -\det(\boldsymbol{B})$.
  • If $\boldsymbol{A}$ has two identical rows or columns, then $\det(\boldsymbol{A}) = 0$.
  • $\det(\boldsymbol{A}) \neq 0$ if and only if $\boldsymbol{A}$ has full rank.
  • Adding a multiple of one row or column to another row or column does not change the determinant.
  • $\det(\boldsymbol{A}) = \det(\boldsymbol{A}^T)$
  • $\det(I) = 1$
  • If any row or column of $\boldsymbol{A}$ is all zeros, then $\det(\boldsymbol{A}) = 0$.
  • Multiplicativity of determinants: $\det(\boldsymbol{A}\boldsymbol{B}) = \det(\boldsymbol{A}) \det(\boldsymbol{B})$
  • $\det(\boldsymbol{A}^{-1}) = \frac{1}{\det(\boldsymbol{A})}$
  • $\det(\operatorname{adj}(\boldsymbol{A})) = \det(\boldsymbol{A})^{n-1}$
  • If $\boldsymbol{A}$ is an orthogonal matrix, $|\det(\boldsymbol{A})| = 1$

Laplace Expansion

You need to understand minor (determinants) and cofactor before continuing.

Laplace expansion is a recursive method to calculate the determinant of a matrix.

Singleton Matrix Case

For a $1 \times 1$ matrix $\boldsymbol{A} = [a]$, the determinant is defined as simply:

$$ a $$

General Case

With the base case in mind, we can recursively calculate the determinant of a matrix by expanding along a row or a column.

Expansion Along a Row

For a $n \times n$ matrix $\boldsymbol{A}$, we can fix a row $i$ and expand along that row.

Usually, we expand along the first row $i = 1$.

The determinant of $\boldsymbol{A}$ is the sum of the products of the elements on the row $i$ and their cofactors.

$$ \det(\boldsymbol{A}) = \sum_{j=1}^n a_{ij} \cdot (-1)^{i+j} \boldsymbol{M}_{ij} $$

Expansion along the row is more common than expansion along the column.

Expansion Along a Column

Same idea, we can fix a column $j$ and expand along that column.

The determinant of $\boldsymbol{A}$ is the sum of the products of the elements on the column $j$ and their cofactors.

\[\det(\boldsymbol{A}) = \sum_{i=1}^n a_{ij} \cdot (-1)^{i+j} M_{ij}\]

2x2 Matrix Application

Let’s apply the Laplace expansion to a $2 \times 2$ matrix $\boldsymbol{A}$.

\[\boldsymbol{A} = \begin{bmatrix} a & b \\ c & d \end{bmatrix}\]

Expansion along the first row $i = 1$.

\[\begin{align*} \det(\boldsymbol{A}) &= \sum_{j=1}^2 a_{1j} \cdot (-1)^{1+j} M_{1j} \\ &= a \cdot (-1)^{1+1} M_{11} + b \cdot (-1)^{1+2} M_{12} \\ &= a \cdot M_{11} - b \cdot M_{12} \\ &= a \cdot d - b \cdot c \end{align*}\]

The determinant of a $2 \times 2$ matrix $\boldsymbol{A}$ is thus defined as:

$$ \det(\boldsymbol{A}) = ad - bc $$


Rule of Sarrus

This is a quick method for the determinant of a $3 \times 3$ matrix.

Let $\boldsymbol{A}$ be a $3 \times 3$ matrix:

$$ \det(\boldsymbol{A}) = \begin{vmatrix} a & b & c \\ d & e & f \\ g & h & i \end{vmatrix} = a(ei - fh) - b(di - fg) + c(dh - eg) $$


Determinant of a Triangular Matrix

See triangular matrices.

For a triangular matrix $\boldsymbol{T}$,

$$ \det(\boldsymbol{T}) = \prod_{i=1}^n t_{ii} $$

i.e. the determinant is the product of the diagonal elements.


Leibniz Formula

Leibniz formula is another way to calculate the determinant of a square matrix.

It comes from the observation that the full determinant expansion can be written as a special sum of products of elements.

Explanation assumes expansion along row $i$ for simplicity.

The key takeaway is that each recursive product in the sum ends up looking like the following:

\[\prod_{i=1}^n a_{i\sigma(i)}\]

That is, while the row $i$ increases from $1$ to $n$ (or is an identity permutation), the column is actually permuted by some $\sigma$.

Like $a_{12} \cdot a_{23} \cdot a_{31}$, where the column $j$ is permuted to $2, 3, 1$.

In addition, the sign of the recursive product ends up equal to the parity of the permutation $\sigma$.

Therefore the determinant of $\boldsymbol{A}$ can be written as:

$$ \det(\boldsymbol{A}) = \sum_{\sigma \in S_n} \text{sgn}(\sigma) \prod_{i=1}^n a_{i\sigma(i)} $$

Or equivalently,

\[\det(\boldsymbol{A}) = \sum_{\sigma \in S_n} \text{sgn}(\sigma) \prod_{i=1}^n a_{\sigma(i)i}\]

Geometrical Interpretation

In 2D, the determinant of a $2 \times 2$ matrix can be interpreted as the signed area of the parallelogram formed by the two column vectors.

In 3D, the determinant of a $3 \times 3$ matrix can be interpreted as the signed volume of the parallelepiped formed by the three column vectors.

A parallelepiped is a 3D generalization of a parallelogram (i.e. a linearly transformed cube).

If the columns are linearly dependent, the area or volume becomes zero.