Theorem.
Cramer's rule
Consider $\boldsymbol{Ax}=\boldsymbol{b}$ where $\boldsymbol{A}$ is an invertible $n\times{n}$ matrix and $\boldsymbol{x}$ and $\boldsymbol{b}$ are vectors in $\mathbb{R}^n$. The components of $\boldsymbol{x}$ are given by:
$$x_i=\frac{\det(\boldsymbol{A}_i(\boldsymbol{b}))}
{\det(\boldsymbol{A})}$$
Where $\boldsymbol{A}_i(\boldsymbol{b})$ is a matrix that is identical to $\boldsymbol{A}$ except that the $i$-th column is replaced by $\boldsymbol{b}$.
Proof. Consider the following matrix:
$$\boldsymbol{A}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Similarly, consider the identity matrix $\boldsymbol{I}$ below:
$$\boldsymbol{I}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{e}_1&\boldsymbol{e}_2&\cdots&\boldsymbol{e}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Here, the columns of $\boldsymbol{I}$ are the standard unit vectors.
Now, by definition, $\boldsymbol{A}_i(\boldsymbol{b})$ is the same as matrix $\boldsymbol{A}$ except that the $i$-th column is replaced by $\boldsymbol{b}$, that is:
$$\boldsymbol{A}_i(\boldsymbol{b})=
\begin{pmatrix}
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\\
\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_{i-1}
&\boldsymbol{b}
&\boldsymbol{a}_{i+1}&\cdots&\boldsymbol{a}_{n}
\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Also, by definition, $\boldsymbol{I}_i(\boldsymbol{x})$ is:
$$\boldsymbol{I}_i(\boldsymbol{x})=
\begin{pmatrix}
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\\
\boldsymbol{e}_1&\boldsymbol{e}_2&\cdots&\boldsymbol{e}_{i-1}
&\boldsymbol{x}
&\boldsymbol{e}_{i+1}&\cdots&\boldsymbol{e}_{n}
\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Notice how the $i$-th row contains all zeros except for $x_i$. By Laplace expansion theoremlink, we can compute the determinant of $\boldsymbol{I}_i(\boldsymbol{x})$ by cofactor expansion along the $i$-th row:
$$\begin{equation}\label{eq:mc88MydaJ22kwUgV9Ox}
\begin{aligned}[b]
\det\big(\boldsymbol{I}_i(\boldsymbol{x})\big)
&=x_i\cdot\det(\boldsymbol{I})\\
&=x_i\cdot(1)\\
&=x_i
\end{aligned}
\end{equation}$$
Here, we used the fact that $\det(\boldsymbol{I})=1$ by theoremlink.
Now, the matrix product $\boldsymbol{A}\Big(\boldsymbol{I}_i(\boldsymbol{x})\Big)$ is:
$$\begin{equation}\label{eq:KlruqkGMOuMi0fTgm1X}
\begin{aligned}[b]
\boldsymbol{A}\Big(\boldsymbol{I}_i(\boldsymbol{x})\Big)&=
\boldsymbol{A}\begin{pmatrix}
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\\
\boldsymbol{e}_1&\boldsymbol{e}_2&\cdots&\boldsymbol{e}_{i-1}
&\boldsymbol{x}
&\boldsymbol{e}_{i+1}&\cdots&\boldsymbol{e}_{n}
\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}\\
&=\begin{pmatrix}
\vert&\vert&\cdots&\vert&\vert&\vert&\cdots&\vert\\
\boldsymbol{A}\boldsymbol{e}_1&\boldsymbol{A}\boldsymbol{e}_2&\cdots&\boldsymbol{A}\boldsymbol{e}_{i-1}
&\boldsymbol{A}\boldsymbol{x}&\boldsymbol{A}\boldsymbol{e}_{i+1}&\cdots&\boldsymbol{A}\boldsymbol{e}_{n}\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}\\
&=\begin{pmatrix}
\vert&\vert&\cdots&\vert&\vert&\vert&\cdots&\vert\\
\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_{i-1}
&\boldsymbol{b}&\boldsymbol{a}_{i+1}&\cdots&\boldsymbol{a}_{n}\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}\\
&=\boldsymbol{A}_i(\boldsymbol{b})
\end{aligned}
\end{equation}$$
Note that the second equality holds because of theoremlink. Now, we take the determinant of both sides of \eqref{eq:KlruqkGMOuMi0fTgm1X} to get:
$$\det\Big(\boldsymbol{A}\big(\boldsymbol{I}_i(\boldsymbol{x})\big)\Big)=
\det\big(\boldsymbol{A}_i(\boldsymbol{b})\big)$$
From the multiplicative property of determinantslink, we have that:
$$\det(\boldsymbol{A})\cdot\det\big(\boldsymbol{I}_i(\boldsymbol{x})\big)=
\det\big(\boldsymbol{A}_i(\boldsymbol{b})\big)$$
Substituting \eqref{eq:mc88MydaJ22kwUgV9Ox} gives:
$$\begin{align*}
\det(\boldsymbol{A})\cdot{x_i}&=
\det\big(\boldsymbol{A}_i(\boldsymbol{b})\big)\\
x_i&=\frac{\det\big(\boldsymbol{A}_i(\boldsymbol{b})\big)}{\det(\boldsymbol{A})}
\end{align*}$$
This completes the proof.
■
Example.
Solving systems of linear equations using Cramer's rule
Solve the following system of linear equations using Cramer's rule:
$$\begin{cases}
x_1+2x_2&=7\\
x_1+x_2&=3
\end{cases}$$
Solution. The system of linear equations can be expressed as:
$$\begin{pmatrix}
1&2\\1&1
\end{pmatrix}
\begin{pmatrix}
x_1\\x_2
\end{pmatrix}=
\begin{pmatrix}
7\\3
\end{pmatrix}$$
Let $\boldsymbol{A}$ represent the matrix on the left and $\boldsymbol{b}$ represent the vector on the right. To use Cramer's rule, we first need to compute the determinant of $\boldsymbol{A}$ like so:
$$\begin{align*}
\det(\boldsymbol{A})&=
(1)(1)-(2)(1)\\
&=1-2\\
&=-1
\end{align*}$$
Next, $\boldsymbol{A}_1(\boldsymbol{b})$ and $\boldsymbol{A}_2(\boldsymbol{b})$ are:
$$\boldsymbol{A}_1(\boldsymbol{b})=
\begin{pmatrix}
7&2\\3&1
\end{pmatrix},\;\;\;\;\;\;
\boldsymbol{A}_2(\boldsymbol{b})=
\begin{pmatrix}
1&7\\1&3
\end{pmatrix}$$
We now compute the determinant of each:
$$\begin{align*}
\det\big(\boldsymbol{A}_1(\boldsymbol{b})\big)
&=(7)(1)-(2)(3)\\
&=1\\\\
\det\big(\boldsymbol{A}_2(\boldsymbol{b})\big)
&=(1)(3)-(7)(1)\\
&=-4
\end{align*}$$
Now, we use Cramer's rule to solve the system:
$$\begin{align*}
x_1
&=\frac{\boldsymbol{A}_1(\boldsymbol{b})}
{\det(\boldsymbol{A})}\\
&=\frac{1}{-1}\\
&=-1\\\\
x_2
&=\frac{\boldsymbol{A}_2(\boldsymbol{b})}
{\det(\boldsymbol{A})}\\
&=\frac{-4}{-1}\\
&=4
\end{align*}$$
Therefore, the solution is:
$$\begin{pmatrix}
x_1\\x_2
\end{pmatrix}=
\begin{pmatrix}
-1\\4\end{pmatrix}$$
Finally, just to confirm that this is indeed the solution to our system of linear equations, let's substitute $x_1$ and $x_2$ into the system:
$$\begin{cases}
-1+2(4)&=7\\
(-1)+4&=3
\end{cases}$$
These are indeed the solutions to the system 🎉!
■
Definition.
Adjugate of a matrix
The adjugate (or classical adjoint) of a matrix $\boldsymbol{A}$ is defined as the transpose of a matrix containing the cofactors of $\boldsymbol{A}$, that is:
$$\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
C_{11}&C_{21}&\cdots&C_{n1}\\
C_{12}&C_{22}&\cdots&C_{n2}\\
\vdots&\vdots&\ddots&\vdots\\
C_{1n}&C_{2n}&\cdots&C_{nn}\\
\end{pmatrix}$$
The adjugate of a matrix is also known as the adjunct matrix or adjoint.
Example.
Finding the adjugate of a 2x2 matrix
Find the adjugate of the following matrix:
$$\boldsymbol{A}=
\begin{pmatrix}
3&1\\2&4
\end{pmatrix}$$
Solution. The adjugate of $\boldsymbol{A}$ is:
$$\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
C_{11}&C_{21}\\C_{12}&C_{22}
\end{pmatrix}$$
We now need to find the cofactors:
$$\begin{align*}
C_{11}&=4\\
C_{21}&=-1\\
C_{12}&=-2\\
C_{22}&=3\\
\end{align*}$$
Therefore, the adjugate of $\boldsymbol{A}$ is:
$$\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
4&-1\\-2&3
\end{pmatrix}$$
■
Theorem.
Finding the inverse matrix using the adjugate of a matrix
If $\boldsymbol{A}$ is an invertible matrix, then its inverse can be computed by:
$$\boldsymbol{A}^{-1}
=
\frac{1}{\det(\boldsymbol{A})}
\;\mathrm{adj}(\boldsymbol{A})$$
Where $\text{adj}(\boldsymbol{A})$ is the adjugate of $\boldsymbol{A}$.
Proof. By definitionlink, a matrix $\boldsymbol{A}$ is invertible if and only if there exists another matrix $\boldsymbol{B}$ such that:
$$\boldsymbol{AB}=\boldsymbol{I}$$
Once again, we treat matrices as a collection of columns:
$$\boldsymbol{A}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix},
\;\;\;\;\;\;\boldsymbol{B}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{b}_1&\boldsymbol{b}_2&\cdots&\boldsymbol{b}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix},
\;\;\;\;\;\;
\boldsymbol{I}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{e}_1&\boldsymbol{e}_2&\cdots&\boldsymbol{e}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Where the column vectors in $\boldsymbol{I}$ represent the standard unit vectors. From theoremlink, the matrix product $\boldsymbol{AB}$ can be expressed as:
$$\boldsymbol{AB}=
\begin{pmatrix}
\vert&\vert&\cdots&\vert\\
\boldsymbol{A}\boldsymbol{b}_1&\boldsymbol{A}\boldsymbol{b}_2&\cdots&\boldsymbol{A}\boldsymbol{b}_n\\
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
Aligning the columns of $\boldsymbol{AB}$ and $\boldsymbol{I}$ gives:
$$\begin{align*}
\boldsymbol{A}\boldsymbol{b}_1&=\boldsymbol{e}_1\\
\boldsymbol{A}\boldsymbol{b}_2&=\boldsymbol{e}_2\\
&\vdots\\
\boldsymbol{A}\boldsymbol{b}_n&=\boldsymbol{e}_n
\end{align*}$$
Let's consider the $j$-th equality:
$$\begin{equation}\label{eq:NzWlAObw1vbirulI81f}
\boldsymbol{A}\boldsymbol{b}_j=
\boldsymbol{e}_j
\end{equation}$$
This can be thought of as a system of linear equations. The components of $\boldsymbol{b}_j$ can be expressed as:
$$\begin{equation}\label{eq:QHoHxlY9MaHnXYENUgz}
\boldsymbol{b}_j=
\begin{pmatrix}
b_{1j}\\b_{2j}\\\vdots\\b_{nj}
\end{pmatrix}
\end{equation}$$
We can find each component using Cramer's rulelink like so:
$$\begin{equation}\label{eq:VhAat4egrXF78WCsobx}
b_{ij}=
\frac{\det\big(\boldsymbol{A}_i(\boldsymbol{e}_j)\big)}
{\det(\boldsymbol{A})}
\end{equation}$$
The numerator of \eqref{eq:VhAat4egrXF78WCsobx} is:
$$\boldsymbol{A}_i(\boldsymbol{e}_j)=
\begin{pmatrix}
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\\
\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_{i-1}
&\boldsymbol{e}_j
&\boldsymbol{a}_{i+1}&\cdots&\boldsymbol{a}_{n}
\\
\vert&\vert&\cdots&\vert&
\vert&\vert&\cdots&\vert
\end{pmatrix}$$
By definition, column $\boldsymbol{e}_j$ has a $1$ in the $j$-th entry and zero in all other entries. We now perform cofactor expansion along the $i$-th column to obtain the determinant of $\boldsymbol{A}_i(\boldsymbol{e}_j)$ like so:
$$\begin{equation}\label{eq:QwllPQsCa6VKNAJxQPS}
\det\big(\boldsymbol{A}_i(\boldsymbol{e}_j)\big)=
(-1)^{i+j}\cdot\det(\boldsymbol{A}_{ji})
\end{equation}$$
Where $\boldsymbol{A}_{ji}$ represents the sub-matrix in which the $j$-th row and the $i$-th column of $\boldsymbol{A}$ are removed. For notational convenience, we write the right-hand side as the cofactor $C_{ji}$ of entry $a_{ji}$ to get:
$$\begin{equation}\label{eq:DZT5wYfD8XE7rMz7g0f}
\det\big(\boldsymbol{A}_i(\boldsymbol{e}_j)\big)=
C_{ji}
\end{equation}$$
Substituting \eqref{eq:DZT5wYfD8XE7rMz7g0f} into \eqref{eq:VhAat4egrXF78WCsobx} gives:
$$\begin{equation}\label{eq:YSZHqoEIbE2fcJGaKue}
b_{ij}=
\frac{C_{ji}}
{\det(\boldsymbol{A})}
\end{equation}$$
Using \eqref{eq:YSZHqoEIbE2fcJGaKue}, we now express \eqref{eq:QHoHxlY9MaHnXYENUgz} as:
$$\boldsymbol{b}_j=
\begin{pmatrix}
C_{j1}/\det(\boldsymbol{A})\\
C_{j2}/\det(\boldsymbol{A})\\
\vdots\\
C_{jn}/\det(\boldsymbol{A})\\
\end{pmatrix}=\frac{1}{\det(\boldsymbol{A})}
\begin{pmatrix}
C_{j1}\\
C_{j2}\\
\vdots\\
C_{jn}\\
\end{pmatrix}$$
Remember, $\boldsymbol{b}_j$ represents the $j$-th column of $\boldsymbol{B}$. We can now express $\boldsymbol{B}$ in its entirety:
$$\begin{equation}\label{eq:vrmyga8REQfbqxH3cqg}
\boldsymbol{B}=
\frac{1}{\det(\boldsymbol{A})}
\begin{pmatrix}
C_{11}&C_{21}&\cdots&C_{n1}\\
C_{12}&C_{22}&\cdots&C_{n2}\\
\vdots&\vdots&\ddots&\vdots\\
C_{1n}&C_{2n}&\cdots&C_{nn}\\
\end{pmatrix}
\end{equation}$$
The matrix in \eqref{eq:vrmyga8REQfbqxH3cqg} is called the adjugatelink of $\boldsymbol{A}$, that is:
$$\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
C_{11}&C_{21}&\cdots&C_{n1}\\
C_{12}&C_{22}&\cdots&C_{n2}\\
\vdots&\vdots&\ddots&\vdots\\
C_{1n}&C_{2n}&\cdots&C_{nn}\\
\end{pmatrix}$$
Finally, $\boldsymbol{B}$ is the inverse of $\boldsymbol{A}$, so let's write it as $\boldsymbol{A}^{-1}$ instead. Therefore, we end up with the following result:
$$\boldsymbol{A}^{-1}
=
\frac{1}{\det(\boldsymbol{A})}
\;\mathrm{adj}(\boldsymbol{A})$$
This completes the proof.
■
Theorem.
Finding the inverse of a 2x2 matrix using its determinant
Consider the following $2\times2$ matrix:
$$\boldsymbol{A}=
\begin{pmatrix}
a&b\\
c&d
\end{pmatrix}$$
If $\mathrm{det}(\boldsymbol{A})\ne0$, that is, if $\boldsymbol{A}$ is invertible, then the inverse of $\boldsymbol{A}$ is computed as:
$$\boldsymbol{A}^{-1}=
\frac{1}{ad-bc}
\begin{pmatrix}
d&-b\\
-c&a
\end{pmatrix}$$
Note that if $\mathrm{det}{(\boldsymbol{A})}=0$, then the inverse of $\boldsymbol{A}$ does not exist.
Proof. By theoremlink, we have that:
$$\begin{equation}\label{eq:gZWXPHh86uoVaBPL25D}
\boldsymbol{A}^{-1}
=
\frac{1}{\det(\boldsymbol{A})}
\;\mathrm{adj}(\boldsymbol{A})
\end{equation}$$
By theoremlink, we know that the determinant of a $2\times2$ matrix is:
$$\begin{equation}\label{eq:S6Nx8o9E4W7RCn1bWUG}
\det(\boldsymbol{A})=ad-bc
\end{equation}$$
The adjugate of $\boldsymbol{A}$ is:
$$\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
C_{11}&C_{21}\\C_{12}&C_{22}
\end{pmatrix}$$
The cofactors are:
$$\begin{align*}
C_{11}&=d\\
C_{21}&=-b\\
C_{12}&=-c\\
C_{22}&=a\\
\end{align*}$$
The adjugate of $\boldsymbol{A}$ is therefore:
$$\begin{equation}\label{eq:IprqJxAw30U05ioX3DD}
\mathrm{adj}(\boldsymbol{A})=
\begin{pmatrix}
d&-b\\-c&a
\end{pmatrix}
\end{equation}$$
Substituting \eqref{eq:S6Nx8o9E4W7RCn1bWUG} and \eqref{eq:IprqJxAw30U05ioX3DD} into \eqref{eq:gZWXPHh86uoVaBPL25D} gives:
$$\boldsymbol{A}^{-1}
=
\frac{1}{ad-bc}
\begin{pmatrix}
d&-b\\-c&a
\end{pmatrix}
$$
This completes the proof.
■
Example.
Computing the inverse of a 2x2 matrix
Compute the inverse of the following matrix:
$$\boldsymbol{A}=
\begin{pmatrix}
2&1\\
4&3\\
\end{pmatrix}$$
Solution. The inverse of matrix $\boldsymbol{A}$ is:
$$\begin{align*}
\boldsymbol{A}^{-1}
&=\frac{1}{(2)(3)-(1)(4)}
\begin{pmatrix}
3&-1\\
-4&2\\
\end{pmatrix}\\
&=\frac{1}{2}
\begin{pmatrix}
3&-1\\
-4&2\\
\end{pmatrix}\\
&=
\begin{pmatrix}
1.5&-0.5\\
-2&1\\
\end{pmatrix}
\end{align*}$$
Let's confirm that this is actually the inverse of $\boldsymbol{A}$ by computing $\boldsymbol{AA}^{-1}$ like so:
$$\begin{align*}
\boldsymbol{A}\boldsymbol{A}^{-1}
&=\begin{pmatrix}2&1\\4&3\\\end{pmatrix}
\begin{pmatrix}1.5&-0.5\\-2&1\\\end{pmatrix}\\
&=\begin{pmatrix}(2)(1.5)+(1)(-2)&2(-0.5)+(1)(1)\\
(4)(1.5)+(3)(-2)&(4)(-0.5)+(3)(1)\\\end{pmatrix}\\
&=\begin{pmatrix}1&0\\0&1\\\end{pmatrix}
\end{align*}$$
Indeed, the $\boldsymbol{A}^{-1}$ that we found is the inverse of $\boldsymbol{A}$.
■