1. Recap: Matrices as Transformations
- A linear transformation is fixed by where and land.
- A vector lands on .
- Record the new , as the columns of a matrix; matrix–vector multiplication applies the transformation.
2. Composition of Transformations
Definition 2.1 – Composition
Applying one linear transformation and then another yields a new linear transformation, called the composition. Like any linear transformation, it has its own matrix — found by tracking and through both steps.
Worked example — rotate 90° CCW, then shear.
| After rotation | After shear | |
|---|---|---|
Definition 2.2 – Matrix product
Applying then to any :
So the product is defined to be the matrix of the composition.
Multiplying two matrices = applying one transformation, then the other.
3. Reading Order: Right to Left
Rule 3.1
In , apply first (right), then (left) — same as function notation .
⚠️ Pitfall: It’s easy to read left-to-right out of habit. The rightmost matrix acts first.
4. Computing a Product Numerically
Theorem 4.1 – Columns of the product
- Column 1 of — where ends up.
- Column 2 of — where ends up.
General formula:
where e.g. the first column is .
Worked example.
- : lands on after ; then .
- : lands on after ; then .
5. Properties Explained Geometrically
Property 5.1 – Order matters (not commutative)
In general .
Example.
| Order | Matrix | ||
|---|---|---|---|
| Rotate, then shear () | |||
| Shear, then rotate () |
In the basis vectors end up close together; in they end up farther apart — clearly different transformations. You can see this by imagining the motions, no arithmetic needed.
Property 5.2 – Associativity
Proof by thinking: both sides mean “apply , then , then ” — the same three steps in the same order. The parentheses don’t change anything. This is a genuine proof and an explanation of why it’s true (versus an unenlightening symbol-pushing proof).
⚠️ Pitfall: Associativity does not mean you can reorder the matrices — only regroup them.
Exercises: Imagine two transformations, predict the composite visually, then verify numerically — e.g. compute and for = 90° rotation and ; more practice on the lesson page.