Matrix Transforms — How Space Bends
How 2×2 matrices rotate, scale, and reshape space itself
This concept builds on Vectors — Direction and Magnitude, Sine on the Unit Circle and in a Right Triangle, Cosine — The Other Side. Consider exploring them first for the best experience.
The Identity Matrix
Start with [[1, 0], [0, 1]]. The blue î and green ĵ arrows sit on the axes and the purple square matches its dashed ghost. Drag anywhere on the plane to move the orange input vector v: the pink image M·v lands exactly on top of it, because the identity is the "do nothing" transform every other matrix is measured against.
Scaling Matrix
Move the 'Scale factor s' slider to build [[s, 0], [0, s]]. Every point moves s times farther from the origin: the square grows, and the pink M·v arrow stretches along the same line as v. The "Area scale" card reads s squared, because both sides grow. Values below 1 shrink space; values above 1 stretch it.
Rotation Matrix
Move the 'Rotation θ (degrees)' slider. The matrix [[cos θ, −sin θ], [sin θ, cos θ]] turns every point by θ around the origin, and the cos θ and sin θ cards show its entries. Watch the blue î arrow: its tip is the first column, (cos θ, sin θ). Columns are destinations — they tell you where each axis lands.
Composing Transforms
Choose a rotation and a scale together and the canvas shows both at once, while the formula box multiplies them into one matrix. Matrix multiplication chains transforms: the rightmost matrix is applied first, the leftmost last. Rotating then scaling equals scaling then rotating here only because the scale is uniform — with a stretch or a shear, order matters.
Shear and General Transforms
Press 'Build Your Own Matrix' and move the 'b (row 1, col 2)' slider away from 0 while a = d = 1: the square shears into a parallelogram instead of turning. Then press 'Reflect' — the "Orientation" card reads Flipped, because a negative determinant means the shape comes out mirror-imaged, like a reflection.
Worked Example: Stretching (1, 1) by [[2, 0], [0, 1]]
Press 'Build Your Own Matrix', then press 'Stretch x' to load [[2, 0], [0, 1]], and drag v to (1, 1). Across: 2·1 + 0·1 = 2. Up: 0·1 + 1·1 = 1. So the "Image M·v" card reads (2, 1) — one copy of the first column (2, 0) plus one of the second (0, 1). The determinant is 2·1 − 0·0 = 2, so the "Area scale" card reads 2.00×: every area doubles.
How It Works — Step by Step
-
1
The Identity Matrix
Start with [[1, 0], [0, 1]]. The blue î and green ĵ arrows sit on the axes and the purple square matches its dashed ghost. Drag anywhere on the plane to move the orange input vector v: the pink image M·v lands exactly on top of it, because the identity is the "do nothing" transform every other matrix is measured against.
-
2
Scaling Matrix
Move the 'Scale factor s' slider to build [[s, 0], [0, s]]. Every point moves s times farther from the origin: the square grows, and the pink M·v arrow stretches along the same line as v. The "Area scale" card reads s squared, because both sides grow. Values below 1 shrink space; values above 1 stretch it.
-
3
Rotation Matrix
Move the 'Rotation θ (degrees)' slider. The matrix [[cos θ, −sin θ], [sin θ, cos θ]] turns every point by θ around the origin, and the cos θ and sin θ cards show its entries. Watch the blue î arrow: its tip is the first column, (cos θ, sin θ). Columns are destinations — they tell you where each axis lands.
-
4
Composing Transforms
Choose a rotation and a scale together and the canvas shows both at once, while the formula box multiplies them into one matrix. Matrix multiplication chains transforms: the rightmost matrix is applied first, the leftmost last. Rotating then scaling equals scaling then rotating here only because the scale is uniform — with a stretch or a shear, order matters.
-
5
Shear and General Transforms
Press 'Build Your Own Matrix' and move the 'b (row 1, col 2)' slider away from 0 while a = d = 1: the square shears into a parallelogram instead of turning. Then press 'Reflect' — the "Orientation" card reads Flipped, because a negative determinant means the shape comes out mirror-imaged, like a reflection.
-
6
Worked Example: Stretching (1, 1) by [[2, 0], [0, 1]]
Press 'Build Your Own Matrix', then press 'Stretch x' to load [[2, 0], [0, 1]], and drag v to (1, 1). Across: 2·1 + 0·1 = 2. Up: 0·1 + 1·1 = 1. So the "Image M·v" card reads (2, 1) — one copy of the first column (2, 0) plus one of the second (0, 1). The determinant is 2·1 − 0·0 = 2, so the "Area scale" card reads 2.00×: every area doubles.
Key Takeaways
Columns Are Destinations
The first column of a 2×2 matrix is where [1, 0] (the x-axis) lands. The second column is where [0, 1] (the y-axis) lands. Read the columns to instantly understand any linear transform.
Rotation Needs Two Functions
cos θ and sin θ appear in every rotation matrix because rotating a unit vector by θ produces (cos θ, sin θ) — the very definition of sine and cosine.
Order Matters
Matrix multiplication is not commutative: rotate then scale ≠ scale then rotate (in general). When composing transforms, the order in which you multiply matrices determines the result.
Follow One Vector
The image of (x, y) is x copies of the first column plus y copies of the second. [[2, 0], [0, 1]] sends (1, 1) to (2, 0) + (0, 1) = (2, 1), and its determinant 2 says every area doubles.
Connected Concepts
Dot Product
Matrix multiplication generalises the dot product to transform entire spaces.
Vectors
Matrices are machines that take vectors in and output transformed vectors.
Sine & Cosine
The rotation matrix is built entirely from sin and cos — trig in disguise.
Fourier Series
Fourier analysis decomposes signals using orthogonal basis vectors — a matrix idea at heart.
Determinant
The "Area scale" and "Orientation" readouts here are exactly the determinant — its magnitude and its sign.
Eigenvalues & Eigenvectors
Every matrix you build here has special directions it only stretches, never rotates off their line — that’s what eigenvalues find.
Gaussian Elimination
Solving A·x = b means finding which vector this transform sends to b — row operations on the augmented matrix [A | b] find it step by step.
Cramer's Rule
Cramer's rule solves A·x = b by dividing areas — and det(A), the area-scale factor of this transform, is always the denominator.
Updated
- — Drag an input vector to see its image under the matrix, plus a worked example and a new quiz question.