Determinant of 3×3 matrix — cofactor expansion: Visual walkthrough

We only assume you can read a table of numbers. Everything else — even what "determinant" means — we rebuild here.

Step 1 — What is a matrix, and what does its determinant measure?

WHAT. A matrix is just a grid of numbers arranged in rows (going across) and columns (going down). A 3×3 matrix has 3 rows and 3 columns — nine numbers total. We name each number $a_{ij}$ : the letter $i$ is the row it sits in, and $j$ is the column.

$A = a_{11} a_{21} a_{31} a_{12} a_{22} a_{32} a_{13} a_{23} a_{33}$

Here $a_{11}$ (read "a-one-one") lives in row 1, column 1 — top-left. $a_{23}$ lives in row 2, column 3.

WHY. We read the three rows as three arrows pointing into 3D space. The first row is one arrow, the second row another, the third row a third. The determinant is a single number that measures the volume of the slanted box (a parallelepiped — a squished shoe-box) that these three arrows span. Big box → big number. Flat box → zero.

PICTURE. Three arrows leaving the origin; the box they cage in is the thing whose volume we want.

Figure — Determinant of 3×3 matrix — cofactor expansion

Step 2 — The tool we already own: the 2×2 area

WHAT. Before 3D volume, recall 2D area. Two arrows in a flat plane cage a parallelogram (a leaning rectangle). For a 2×2 matrix its area is one clean formula, from the 2×2 determinant:

$a c b d = a d - b c$

Term by term: $a d$ is the main-diagonal product (top-left times bottom-right). $b c$ is the anti-diagonal product (top-right times bottom-left). We subtract because $b c$ measures the "wrong-way" overlap that must be cancelled.

WHY this tool. The whole strategy of the 3×3 is to not invent anything new. If we can chop the 3D box into pieces whose sizes are just 2×2 areas, we're done — we already know areas.

PICTURE. The parallelogram, its diagonal-product rectangles shaded, one in red.

Recall

Value of $5 - 1 62$ ? ::: $5 \cdot 2 - 6 \cdot (- 1) = 10 + 6 = 16$

Step 3 — Slice the box along the top row

WHAT. Look at the first row, $a_{11}, a_{12}, a_{13}$ . We claim the whole volume is built by taking each of these three numbers, one at a time, and multiplying it by the area of a smaller 2×2 slice.

WHY. Volume of a box = (how far it reaches in one direction) × (area of the cross-section it drags along). The top-row entries are the "how far" pieces; the 2×2 areas are the "cross-sections." So we walk across the top row and collect three volume-slabs, then add them.

The 2×2 slice belonging to $a_{1 j}$ is called its minor $M_{1 j}$ : cover up row 1 (because we've already used it) and column $j$ (because $a_{1 j}$ lives there), and take the determinant of the four numbers left over.

PICTURE. Row 1 highlighted; for $a_{11}$ we grey out its row and column and the surviving 2×2 block glows red.

Step 4 — The three slices, drawn out

WHAT. Do the covering for each top-row entry in turn:

$M_{11} = a_{22} a_{32} a_{23} a_{33}, M_{12} = a_{21} a_{31} a_{23} a_{33}, M_{13} = a_{21} a_{31} a_{22} a_{32}$

$M_{11}$ : cover row 1 & column 1 → columns 2,3 survive.
$M_{12}$ : cover row 1 & column 2 → columns 1,3 survive (notice the gap — column 2 is gone).
$M_{13}$ : cover row 1 & column 3 → columns 1,2 survive.

WHY. Each minor is the cross-sectional area left after we "spend" one top-row number in one direction. Three numbers, three cross-sections.

PICTURE. The 3×3 shown three times side by side, each with a different column struck through in red.

Step 5 — Where the minus sign is born

WHAT. If we naively added $a_{11} M_{11} + a_{12} M_{12} + a_{13} M_{13}$ we would get the wrong answer. The true rule alternates sign:

$det (A) = + a_{11} M_{11} - a_{12} M_{12} + a_{13} M_{13}$

The middle term is subtracted.

WHY. Watch the columns of the surviving block. For $M_{11}$ we kept columns in order $(2, 3)$ — natural order. For $M_{12}$ we kept $(1, 3)$ : to line them up with the "natural" left-to-right box we had to swap column 1 past column 2. One swap flips the box to its mirror image, which flips the sign of the volume. That single swap is the minus sign. For $M_{13}$ , columns $(1, 2)$ need an even number of swaps back, so it's plus again.

We package this bookkeeping into $(- 1)^{i + j}$ . For the top row $i = 1$ :

$(- 1)^{1 + 1} = + 1, (- 1)^{1 + 2} = - 1, (- 1)^{1 + 3} = + 1$

The number $a_{ij}$ times its sign times its minor is the cofactor $C_{ij} = (- 1)^{i + j} M_{ij}$ .

PICTURE. The checkerboard sign grid, with the top-row cells $+ - +$ picked out in red, and a little "one swap = flip" cartoon.

Step 6 — Assemble the full formula and test it

WHAT. Putting Steps 3–5 together gives the parent's headline result:

$det (A) = a_{11} a_{22} a_{32} a_{23} a_{33} - a_{12} a_{21} a_{31} a_{23} a_{33} + a_{13} a_{21} a_{31} a_{22} a_{32}$

WHY test. A picture is worth nothing if the arithmetic breaks. Take $A = 25 - 1 - 3 04 162 .$

$M_{11} = 0462 = 0 \cdot 2 - 6 \cdot 4 = - 24$ , sign $+$ .
$M_{12} = 5 - 1 62 = 10 + 6 = 16$ , sign $-$ .
$M_{13} = 5 - 1 04 = 20 - 0 = 20$ , sign $+$ .

$det (A) = 2 (- 24) - (- 3) (16) + 1 (20) = - 48 + 48 + 20 = 20$

PICTURE. Bar chart of the three signed slabs $- 48, + 48, + 20$ stacking up to the red total $20$ .

Step 7 — Any row works (and why you should choose zeros)

WHAT. Nothing forced us to slice the top row. The same argument slices any row or column. The general rule:

$det (A) = \sum_{j = 1}^{3} a_{ij} C_{ij} (pick any fixed row i)$

Here the $\sum$ ("sigma") means "add these up for $j = 1, 2, 3$ ." All six choices (three rows, three columns) give the identical number — see the properties.

WHY. Because they all measure the same box's volume. So pick the row/column with the most zeros — a zero kills its whole slab before you compute the minor. For $B = 314050207,$ column 2 has two zeros, so only the middle survives: $det (B) = 5 \cdot (- 1)^{2 + 2} 3427 = 5 (21 - 8) = 65$

PICTURE. Column 2 highlighted; its two zero-slabs faded to nothing, only the red middle slab remains.

Step 8 — The degenerate case: a flat box gives zero

WHAT. What if the three arrows don't reach into all of 3D — if they lie in a single flat sheet? Then the box has no thickness, volume $= 0$ , so $det = 0$ . Example: $C = 147258369, row 3 = 2 \cdot row 2 - row 1 .$

Because row 3 is built from the other two, its arrow lies in the plane of the first two. Cofactor expansion confirms it: $det (C) = 3 (- 3) + 6 (6) + 9 (- 3) = - 9 + 36 - 27 = 0$

WHY it matters. A zero determinant is the signal that the transformation collapses 3D onto a flat plane — the matrix has no inverse, and systems like Cramer's rule break exactly here.

PICTURE. Three arrows lying flat in one tilted plane; the "box" is a red pancake with zero height.

The one-picture summary

Everything above in a single frame: top row across the top, each entry firing a signed 2×2 slab downward, the three slabs summing into the box's volume.

Recall Feynman retelling — say it plain

Three rows of numbers are three arrows; the determinant is the volume of the slanty box they cage. To measure it, I walk across the top row. Each top number tells me how far the box reaches in one direction; the little 2×2 area left after I hide that number's row and column tells me the cross-section it drags. Distance times area = a slab of volume. I collect three slabs — but the middle one comes out mirror-flipped (one column had to hop past another), so I subtract it: plus, minus, plus. Add the slabs and I have the volume. I can start from any row or column and get the same answer, so I start where the zeros are to save work. And if the three arrows all lie flat in one sheet, the box is a pancake with no height — volume zero, determinant zero.