Why Does ad − bc Measure Area?

A graphics program stretches a one-unit square until its edges become the vectors

$$\mathbf u=(2,1) \qquad\text{and}\qquad \mathbf v=(-1,3).$$

The square has become a slanted parallelogram. How much area does the program now cover?

Multiplying the vector lengths will not work because the vectors are not perpendicular. Drawing the height will work, but there is a shorter calculation:

$$(2)(3)-(-1)(1)=7.$$

The expression on the left is the determinant of the matrix whose columns are $\mathbf u$ and $\mathbf v$. The number $7$ is the parallelogram’s area, but the determinant carries one extra piece of information that ordinary area discards: which vector comes first as you turn around the origin.

Bend the parallelogram

The diagram begins with $\mathbf u=(2,1)$ and $\mathbf v=(-1,3)$. Move any coordinate with the sliders. The solid and dashed edges distinguish the two vectors without requiring color.

Signed-area explorer

Move two vectors; watch one number track the parallelogram

Vector u = (a, c)

a

2.00

c

1.00

Vector v = (b, d)

b

−1.00

d

3.00

Signed area7.00

Ordinary area7.00 square units

Orientationcounterclockwise

The hatched region vanishes when the two vectors point along the same line. Continue moving one vector across that line and the region opens again, but the determinant has the opposite sign. The magnitude measures area; the sign records orientation.

The claim, with every symbol named

Let

$$A= \begin{pmatrix} a&b\\ c&d \end{pmatrix}$$

be a $2\times2$ matrix. The first column is the vector $\mathbf u=(a,c)$, and the second column is the vector $\mathbf v=(b,d)$. Their endpoints, together with the origin and $\mathbf u+\mathbf v$, form a parallelogram.

The determinant of $A$ is

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

This number is the parallelogram’s signed area:

$$\text{ordinary area}=|\det(A)|.$$

The sign uses the standard orientation of the coordinate plane:

• $\det(A)>0$ when the shorter turn from $\mathbf u$ to $\mathbf v$ is counterclockwise;
• $\det(A)<0$ when that turn is clockwise;
• $\det(A)=0$ when the vectors are collinear, so the parallelogram has collapsed to a segment or a point.

The order matters. Swapping the two columns changes $ad-bc$ into $bc-ad$, which is its negative. The geometric region has the same area after a swap, but the trip from first vector to second vector runs in the opposite direction.

Where the subtraction comes from

The formula $ad-bc$ is easier to remember once it is rebuilt from base times height.

Start with the first vector

$$\mathbf u=(a,c).$$

Rotating $\mathbf u$ counterclockwise by $90^\circ$ gives

$$\mathbf u^{\perp}=(-c,a).$$

The superscript $\perp$ means “perpendicular.” The rotated vector has the same length as $\mathbf u$. Therefore

$$\frac{\mathbf u^{\perp}}{\lVert\mathbf u\rVert}$$

is a unit vector perpendicular to the base, provided $\mathbf u$ is not the zero vector. Here $\lVert\mathbf u\rVert$ denotes the length of $\mathbf u$.

Projecting $\mathbf v$ onto that perpendicular unit vector gives a signed height:

$$h=\mathbf v\mathbin{\boldsymbol\cdot} \frac{\mathbf u^{\perp}}{\lVert\mathbf u\rVert}.$$

The symbol $\boldsymbol\cdot$ is the dot product. Multiply the signed height by the base length $\lVert\mathbf u\rVert$:

$$\begin{aligned} \text{signed area} &=\lVert\mathbf u\rVert h\\ &=\lVert\mathbf u\rVert \left( \mathbf v\mathbin{\boldsymbol\cdot} \frac{\mathbf u^{\perp}}{\lVert\mathbf u\rVert} \right)\\ &=\mathbf v\mathbin{\boldsymbol\cdot}\mathbf u^{\perp}. \end{aligned}$$

Now substitute $\mathbf v=(b,d)$ and $\mathbf u^{\perp}=(-c,a)$:

$$\begin{aligned} \mathbf v\mathbin{\boldsymbol\cdot}\mathbf u^{\perp} &=(b,d)\mathbin{\boldsymbol\cdot}(-c,a)\\ &=b(-c)+d(a)\\ &=ad-bc. \end{aligned}$$

If $\mathbf u$ is the zero vector, the parallelogram is already collapsed and both $ad-bc$ and its area equal zero. The same formula therefore covers every case.

Worked example: seven square units, checked twice

Return to

$$A= \begin{pmatrix} 2&-1\\ 1&3 \end{pmatrix}.$$

Read the entries carefully: $a=2$, $b=-1$, $c=1$, and $d=3$. Then

$$\begin{aligned} \det(A) &=ad-bc\\ &=(2)(3)-(-1)(1)\\ &=6-(-1)\\ &=7. \end{aligned}$$

The sign is positive, so the turn from $\mathbf u=(2,1)$ to $\mathbf v=(-1,3)$ is counterclockwise. The ordinary area is $|7|=7$ square units.

An independent cut-and-count check reaches the same number. The parallelogram’s vertices are

$$(0,0),\quad(2,1),\quad(1,4),\quad(-1,3).$$

Those four points fit inside the rectangle from $x=-1$ to $x=2$ and from $y=0$ to $y=4$. The rectangle has area

$$3\times4=12.$$

The four gaps around the parallelogram are right triangles with areas $1$, $3/2$, $1$, and $3/2$. Their total is $5$, leaving

$$12-5=7.$$

The algebra and the dissection agree.

Why a matrix inherits the area

The matrix $A$ represents the linear transformation that sends the horizontal unit vector $(1,0)$ to its first column $\mathbf u$ and the vertical unit vector $(0,1)$ to its second column $\mathbf v$. The unit square is built from those two original directions, so its image is exactly the parallelogram in the diagram.

Every small square in a grid is transformed in the same way. Each square’s area is multiplied by $|\det(A)|$. Approximating a region with finer grid cells extends that scale factor to general planar regions. Thus:

• $|\det(A)|=2$ doubles every area;
• $|\det(A)|=1$ preserves every area;
• $0<|\det(A)|<1$ shrinks every area;
• $\det(A)=0$ flattens the entire plane onto a line or a point.

That last case also explains invertibility. A flattened plane has lost a direction, so no inverse transformation can recover every original point. A $2\times2$ matrix is invertible exactly when its determinant is nonzero.

Try the formula

Problem 1

Let $\mathbf u=(1,2)$ and $\mathbf v=(3,1)$. Find the signed area, the ordinary area, and the orientation.

Reveal the solution

The vectors are the columns of

$$A= \begin{pmatrix} 1&3\\ 2&1 \end{pmatrix}.$$

Therefore

$$\det(A)=(1)(1)-(3)(2)=1-6=-5.$$

The signed area is $-5$. The ordinary area is $|-5|=5$ square units. The negative sign says that the turn from $\mathbf u$ to $\mathbf v$ is clockwise.

Problem 2

The vectors $\mathbf u=(2,1)$ and $\mathbf v=(x,3)$ must enclose a counterclockwise signed area of $5$. Find $x$.

Reveal the solution

The matrix has entries $a=2$, $b=x$, $c=1$, and $d=3$. Set its determinant equal to the required signed area:

$$\begin{aligned} ad-bc&=5\\ (2)(3)-x(1)&=5\\ 6-x&=5\\ x&=1. \end{aligned}$$

Checking gives $\det\begin{pmatrix}2&1\\1&3\end{pmatrix}=6-1=5$, which is positive as required.

A formula older than matrices

Determinants did not begin as decorations attached to matrices. Problems about simultaneous equations came first. The Chinese text The Nine Chapters on the Mathematical Art, compiled during the Han dynasty, arranged coefficients on counting boards and used elimination methods centuries before modern notation. The MacTutor history of matrices and determinants traces that earlier lineage.

In 1683, the Japanese mathematician Seki Takakazu developed determinant calculations in tables while studying polynomial equations. Ten years later, Gottfried Wilhelm Leibniz used determinant conditions for simultaneous equations in Europe. The two developments were independent; MacTutor’s biography of Seki notes that Seki’s treatment was more general.

The geometric meaning arrived later. Joseph-Louis Lagrange used a determinant expression for the volume of a tetrahedron in 1773. Augustin-Louis Cauchy used the word determinant in its modern sense in 1812 and organized much of the early theory. James Joseph Sylvester introduced the word matrix only in 1850. The historical order reverses the order in many textbooks: the determinant was useful before the matrix had a name.

Where the signed scale factor is used

The determinant answers three practical questions about a linear transformation at once: how much it scales area, whether it reverses orientation, and whether it destroys information by flattening the plane. Wolfram MathWorld’s determinant reference summarizes the same algebraic test for unique solutions of linear systems.

In computational geometry, the sign of a $2\times2$ determinant tells whether three points make a left turn, a right turn, or lie on one line. That orientation test supports convex-hull algorithms, polygon winding checks, and segment-intersection routines.

In multivariable calculus, a nonlinear map looks approximately linear when examined close enough to one point. Its matrix of local derivatives is the Jacobian matrix. The absolute value of the Jacobian determinant supplies the local area or volume correction in a change of variables; MathWorld’s Jacobian reference states that ratio-of-volumes interpretation directly.

The same idea reaches computer graphics, robotics, continuum mechanics, and differential equations wherever a local map can stretch, flip, or collapse space.

One idea to remember

For column vectors $(a,c)$ and $(b,d)$, the quantity $ad-bc$ is not an arbitrary crossing of products. It is base times signed height: the area scale factor of the transformation, with orientation still attached.

To continue, compare this lesson with counting area on a lattice, then see how multiplication by $i$ rotates the plane without changing area. A next step in linear algebra is to prove the multiplication rule $\det(AB)=\det(A)\det(B)$: applying one transformation after another multiplies their area scale factors.

Sources

• MacTutor History of Mathematics, “Matrices and determinants”
• MacTutor History of Mathematics, “Takakazu Seki”
• Wolfram MathWorld, “Determinant”
• Wolfram MathWorld, “Jacobian”