A matrix is a machine that stretches, rotates, and shears the whole plane at once — and the determinant is how much it scales area.
You already know vectors — arrows you can add and scale — and you know how to give one coordinates. Linear algebra takes one more step and treats a whole plane of arrows as a single object you can transform all at once. The tool that does the transforming is a matrix: feed in a vector, a stretched-and-rotated vector comes back. The astonishing part is how compact this is — two columns of numbers encode an entire reshaping of space, their determinant tells you exactly how much area got scaled, and that very same picture quietly solves systems of equations. This is the doorway to a field that runs through computer graphics, data science, quantum mechanics, and Google's search — and it all begins with arrows.
Start with the simplest possible promise. A vector space is a collection of arrows for which two operations always stay inside the collection: you can add any two arrows tip-to-tail, and you can scale any arrow by a number (longer, shorter, or reversed). Add two arrows in the plane and you land on another arrow in the plane; scale an arrow and you stay in the plane. That "you never leave" property is called being closed under addition and scaling, and it is the whole definition in spirit.
Nothing here is new to you. In vectors & direction you added arrows and stretched them; the plane of all such arrows is the vector space ℝ². What linear algebra adds is a change of attitude: stop thinking about one arrow and start thinking about all of them at once — the entire grid — as the thing you operate on.
A vector space is a set of vectors closed under addition (u + v stays in) and scaling (c·v stays in). The plane ℝ² is the example we will live inside.
Here is the surprise that makes everything else possible: you do not need infinitely many arrows to describe the plane — you need exactly two. Choose the two unit arrows along the axes,
e₁ = (1, 0) and e₂ = (0, 1).
Scale e₁ by some number c₁, scale e₂ by some number c₂, add the results, and you can land on any point in the plane:
v = c₁e₁ + c₂e₂ = (c₁, c₂).
An expression like c₁e₁ + c₂e₂ is a linear combination. The two building arrows are a basis, and the two numbers (c₁, c₂) are the vector's coordinates in that basis. The point worth pausing on: the arrow itself is a real, physical thing, but its coordinates depend on which basis you chose — change the basis and the same arrow gets new numbers.
The arrow to (2, 3) is the linear combination v = 2e₁ + 3e₂: walk 2 steps east, then 3 steps north. Its coordinates in the standard basis are exactly (2, 3) — the weights you used.
Dial the weights c₁ and c₂. The blue and amber arrows chain tip-to-tail to reach the green target — those two weights are the coordinates.
Now for the central idea of the whole subject. A 2×2 matrix is a rule that takes every vector x in the plane and sends it to a new vector Mx. It does this linearly: it respects addition and scaling, so it never bends straight lines and never moves the origin. Because of that, the matrix is completely determined by what it does to just the two basis arrows — and that is the secret to reading any matrix on sight:
The columns of M are the images of e₁ and e₂. If M sends e₁ to (a, c) and e₂ to (b, d), then M = ⟨a b; c d⟩, and every other vector follows because M(c₁e₁+c₂e₂) = c₁(Me₁) + c₂(Me₂).
So a matrix doesn't move one vector — it moves the entire grid, dragging every point along to keep the lines straight and evenly spaced. Watch what each famous matrix does to the square grid below.
Pick a matrix. The slate grid is the original plane; the green parallelogram is the image of the unit square; the blue arrow is Me₁ and the amber arrow is Me₂ — the two columns.
If a matrix is an action, then doing one action after another should itself be an action — a single matrix. It is, and it is exactly what matrix multiplication computes. The product AB means "do B first, then A" — read right to left, like nested functions A(B(x)). Each column of AB is just A applied to the corresponding column of B.
And here is the famous trap: order matters. Rotating then shearing is not the same as shearing then rotating, so in general
AB ≠ BA.
Let A = rotate 90° = ⟨0 −1; 1 0⟩ and B = shear = ⟨1 1; 0 1⟩. Then AB = ⟨0 −1; 1 1⟩ but BA = ⟨1 −1; 1 0⟩ — different matrices. Check the first column: AB sends e₁ to (0, 1) while BA sends it to (1, 1). Same two ingredients, different order, different result.
Choose A and B from the menus; the figure shows the unit square after AB (do B then A) and after BA (do A then B). The readout computes both products — compare them.
Every reshaping of the plane changes areas, and the matrix carries that number in plain sight. The determinant of M = ⟨a b; c d⟩ is
det M = ad − bc,
and it has a single, vivid meaning: it is the factor by which area is multiplied. The unit square has area 1; after M it becomes a parallelogram of area |det M|. Three cases tell the whole story:
The shear ⟨1 1; 0 1⟩ has det = (1)(1) − (1)(0) = 1. It tilts the square into a slanted parallelogram, but base × height is unchanged — a shear preserves area. The reflection ⟨1 0; 0 −1⟩ has det = (1)(−1) − 0 = −1: same area, opposite orientation.
The matrix is ⟨1 1; 0 d⟩; slide the bottom-right entry d. The green parallelogram is the image of the unit square; its area equals |det|, computed live. Watch it collapse to a segment and flip sign.
Now collect the reward. A system of two linear equations,
ax + by = e, cx + dy = f,
is exactly the single matrix equation Mx = k with M = ⟨a b; c d⟩ and right-hand side k = (e, f). The determinant decides everything:
So the determinant you met as "area scale" is the very same number that tells you whether a system can be solved. One idea, two faces.
Solve 2x + y = 5 and x + 3y = 10. Here M = ⟨2 1; 1 3⟩, so det = (2)(3) − (1)(1) = 5 ≠ 0 — a unique answer exists. Solving gives x = (1, 3). Check: 2(1) + 3 = 5 ✓ and 1 + 3(3) = 10 ✓. The two lines meet at the single point (1, 3).
Pick a system. Each equation is a slate line; their crossing (green dot) is the solution, found by solve2. The last choice has det = 0 — see the lines go parallel and the solution vanish.
AB ≠ BA — transformations rarely commute, so you cannot swap the order of matrices. det = 0 is the danger sign: the machine flattens the plane, there is no inverse, and the system loses its unique solution. And remember a vector is the arrow itself — its coordinates are only labels that depend on the basis you picked.
| Idea | What it says | The picture |
|---|---|---|
| Vector space | closed under + and scaling | arrows you never leave |
| Basis & coordinates | v = c₁e₁ + c₂e₂ | two weights locate any point |
| Matrix | columns = images of e₁, e₂ | the grid reshapes |
| Multiplication | AB = do B then A; AB ≠ BA | actions composed |
| Determinant | ad − bc = area scale | + keep, 0 flatten, − flip |
| Solving Mx = b | det ≠ 0 ⇒ unique x | lines cross once |
Write the vector (4, −1) as a linear combination of e₁ = (1, 0) and e₂ = (0, 1). What are its coordinates?
(4, −1) = 4e₁ + (−1)e₂: four steps east, one step south. Its coordinates in the standard basis are (4, −1) — the weights you used.
A matrix M sends e₁ to (3, 0) and e₂ to (0, 2). Write M, and find M applied to (1, 1).
The columns are the images, so M = ⟨3 0; 0 2⟩. Then M(1, 1) = 1·(3, 0) + 1·(0, 2) = (3, 2) — it stretches x by 3 and y by 2.
Compute the determinant of ⟨2 1; 4 3⟩. By what factor does it scale area, and does it flip orientation?
det = (2)(3) − (1)(4) = 6 − 4 = 2. Area is multiplied by |2| = 2; since the determinant is positive, orientation is preserved (no flip).
With A = ⟨1 1; 0 1⟩ (shear) and B = ⟨2 0; 0 1⟩ (scale x), compute AB and BA. Are they equal?
AB = ⟨2 1; 0 1⟩ and BA = ⟨2 2; 0 1⟩. They are not equal — the off-diagonal entry differs (1 vs 2). Order matters: shearing a stretched square is not the same as stretching a sheared one.
Does the system 3x + 6y = 9, x + 2y = 4 have a unique solution? Decide with the determinant.
M = ⟨3 6; 1 2⟩, so det = (3)(2) − (6)(1) = 6 − 6 = 0. No unique solution. The lines are parallel: 3x + 6y = 9 means x + 2y = 3, while the other says x + 2y = 4 — the constants disagree, so there is no solution.
Solve x − y = 1 and x + y = 5 using the determinant idea, then check your answer.
M = ⟨1 −1; 1 1⟩, det = (1)(1) − (−1)(1) = 2 ≠ 0, so a unique solution exists. Adding the equations: 2x = 6 ⇒ x = 3; then y = 5 − 3 = 2. Solution (3, 2). Check: 3 − 2 = 1 ✓, 3 + 2 = 5 ✓.
Six questions to lock it in. Tap the answer you think is right.
This lesson is a first, honest glimpse of college linear algebra, organized around the one idea that makes the subject click: a matrix is a transformation of the plane, its columns are the images of the basis vectors, and its determinant is the factor by which area scales. It aligns with HSN-VM.C (matrices as transformations of the plane, the determinant as area, and solving linear systems with matrices) and builds directly on the vector work in Stage 27. From here the road forks toward eigenvectors, higher dimensions, and the linear algebra behind graphics, data, and machine learning — every number in the widgets is computed by the same matrix engine (det2, matMul, solve2), so the pictures never lie. A door, opening.