Invertibility and Determinants

Read an inverse as a transformation that undoes an invertible square matrix and a determinant as signed volume scaling. Connect invertibility to rank, null spaces, unique solutions, and numerical stability.

An inverse undoes a square linear transformation. A determinant records its signed volume scaling and reveals whether any dimension has collapsed. These ideas describe the same structural boundary from different viewpoints.

An Inverse Reverses a Transformation

For a square matrix $A$ , an inverse is a matrix $A^{-1}$ satisfying

A^{-1}A=AA^{-1}=I.

The two products matter because matrix multiplication is not generally commutative. If $A$ sends $\mathbf{x}$ to $\mathbf{b}$ , then $A^{-1}$ reverses that action:

A\mathbf{x}=\mathbf{b} \quad\Longrightarrow\quad \mathbf{x}=A^{-1}\mathbf{b}.

For example,

A=\begin{bmatrix}2&0\\0&4\end{bmatrix} \qquad\text{and}\qquad A^{-1}=\begin{bmatrix}1/2&0\\0&1/4\end{bmatrix}.

The first matrix scales the coordinate axes by $2$ and $4$ ; the second uses reciprocal scales to restore the original vector.

Exercise: Undo diagonal scaling

For $A=\operatorname{diag}(5,2)$ , what is the first diagonal entry of $A^{-1}$ ?

Check the inverse entry

Compute it first, then check your number.

HintMake the product equal one

The original and inverse diagonal entries must multiply to $1$ .

SolutionUse the reciprocal scale

The first coordinate is multiplied by $5$ , so the inverse must multiply it by

\frac{1}{5}=0.2.

Their composition scales that coordinate by $5(1/5)=1$ .

Invertibility Connects the Earlier Ideas

For a square $n\times n$ matrix, the following statements are equivalent:

$A^{-1}$ exists;
$A$ has rank $n$ ;
the columns of $A$ form a basis of $\mathbb{R}^n$ ;
$\operatorname{Col}(A)=\mathbb{R}^n$ ;
$\operatorname{Null}(A)=\{\mathbf{0}\}$ ;
$A\mathbf{x}=\mathbf{b}$ has exactly one solution for every $\mathbf{b}\in\mathbb{R}^n$ ;
$\det(A)\ne0$ .

These are not separate coincidences. Full column space gives existence for every target, while a trivial null space gives uniqueness. For a square matrix, both occur together exactly at full rank.

The matrix

B=\begin{bmatrix}1&2\\2&4\end{bmatrix}

is not invertible. Its second column repeats the first direction, its rank is $1$ , and $[-2,1]^{\mathsf T}$ lies in its null space. Once two distinct inputs produce the same output, no inverse can determine which input came first.

A Two-by-Two Inverse

For

A=\begin{bmatrix}a&b\\c&d\end{bmatrix},

the inverse, when it exists, is

A^{-1} =\frac{1}{ad-bc} \begin{bmatrix}d&-b\\-c&a\end{bmatrix}.

Multiplying the two matrices verifies the formula:

\begin{bmatrix}a&b\\c&d\end{bmatrix} \begin{bmatrix}d&-b\\-c&a\end{bmatrix} =(ad-bc)I.

The denominator $ad-bc$ is the determinant. If it is zero, division is impossible—and the matrix has already collapsed a direction.

Exercise: Find a two-by-two inverse entry

Let

A=\begin{bmatrix}2&1\\0&3\end{bmatrix}.

What is the upper-left entry of $A^{-1}$ ?

Check the entry

Compute it first, then check your number.

HintCompute the determinant first

Here $ad-bc=2(3)-1(0)=6$ .

SolutionApply the inverse formula

The inverse is

A^{-1}=\frac{1}{6} \begin{bmatrix}3&-1\\0&2\end{bmatrix}.

Its upper-left entry is $3/6=1/2=0.5$ .

The Determinant Is Signed Volume Scaling

For a two-dimensional matrix,

\det(A)=ad-bc.

The magnitude $|\det(A)|$ is the factor by which $A$ scales area. In three dimensions, the magnitude gives the volume-scaling factor; the same interpretation extends to $n$ -dimensional volume.

The sign records orientation:

$\det(A)>0$ preserves orientation;
$\det(A)<0$ reverses orientation;
$\det(A)=0$ collapses volume into a lower-dimensional set.

Thus a negative determinant does not mean negative geometric area. Its magnitude gives the area scale, while its sign records a reversal such as a reflection.

For diagonal and triangular matrices, the determinant is the product of the diagonal entries. For example,

\det\begin{bmatrix}2&1\\0&3\end{bmatrix}=2\cdot3=6.

Exercise: Detect a collapsed transformation

Compute the determinant of $\begin{bmatrix}1&2\\2&4\end{bmatrix}$ .

Check the determinant

Compute it first, then check your number.

HintSubtract the diagonal products

Compute $1(4)-2(2)$ .

SolutionThe signed area scale is zero

The determinant is

1\cdot4-2\cdot2=4-4=0.

A two-dimensional region is flattened into a line, so the matrix cannot be inverted.

Determinants Respect Composition

Applying $B$ and then $A$ gives the composed transformation $AB$ . Their volume scales multiply:

\det(AB)=\det(A)\det(B).

This rule explains two useful facts. Since $I$ preserves volume, $\det(I)=1$ . If $A$ is invertible, then

1=\det(I)=\det(AA^{-1}) =\det(A)\det(A^{-1}),

and therefore

\det(A^{-1})=\frac{1}{\det(A)}.

Exercise: Find an inverse determinant

If $\det(A)=-4$ , what is $\det(A^{-1})$ ?

Check the determinant

Compute it first, then check your number.

HintUse the identity composition

Since $AA^{-1}=I$ , their determinants must multiply to $1$ .

SolutionTake the reciprocal

The inverse determinant is

\det(A^{-1})=\frac{1}{-4}=-\frac14=-0.25.

The negative sign remains because both transformations reverse orientation.

Solve Systems Without Forming the Inverse

The identity $\mathbf{x}=A^{-1}\mathbf{b}$ is valuable for proofs and reasoning. In numerical code, however, explicitly constructing $A^{-1}$ merely to solve $A\mathbf{x}=\mathbf{b}$ usually performs unnecessary work and can amplify rounding error. A linear-system solver instead uses an appropriate factorization.

Solve directly and verify

Edit the matrix or target. The residual checks the computed solution without explicitly constructing the inverse.

Ready to run.

Exercise: Choose a numerical operation

In numerical code, which operation is normally preferred for finding $\mathbf{x}$ in $A\mathbf{x}=\mathbf{b}$ ?

Choose the preferred method

Solve the linear system directly.

Always compute the full inverse first.

Divide b by the determinant.

Select one choice, then check.

HintCompute only what is needed

The goal is one solution vector, not the full inverse transformation.

SolutionUse a linear-system solver

A solver uses a suitable factorization to compute $\mathbf{x}$ without forming all of $A^{-1}$ . This approach is generally more efficient and numerically reliable.

Near Singularity Requires More Than a Determinant Check

A nonzero determinant proves invertibility in exact mathematics, but it does not by itself guarantee a stable numerical solve. A matrix can be invertible yet send one direction to a very small scale. Rounding or measurement error in that direction may then be strongly amplified by the inverse.

For numerical work, do not test det(A) == 0 as a general stability check. Matrix scale, singular values, and the condition number provide more useful information. Later lessons on singular value decomposition and numerical computation develop these tools.

Before Calling a Matrix Invertible

Confirm that the matrix is square, then use the structural test suited to the problem: full rank, trivial null space, unique solvability, or nonzero determinant. Use an inverse to reason about undoing a transformation, but use a linear-system solver when the computational task is to find $\mathbf{x}$ from $A\mathbf{x}=\mathbf{b}$ .

The next lesson considers systems for which exact inversion is unavailable or inappropriate. Least squares finds the closest reachable output, and the pseudoinverse extends inverse-like reasoning to rectangular or rank-deficient matrices.