Linear Transformations and Matrices

SECTION 1

Historical Context & Motivation

Long before the modern language of matrices existed, mathematicians grappled with the problem of describing how geometric objects change shape, position, and orientation under systematic rules. The ancient Greeks studied reflections and rotations in the context of Euclidean geometry, but they lacked an algebraic framework powerful enough to unify these seemingly disparate operations. The critical insight — that every linear transformation of the plane can be captured by a compact array of numbers — emerged gradually through centuries of mathematical development, culminating in the matrix algebra we use today.

1750s

Cramer's Rule & Systems of Equations

Gabriel Cramer published systematic methods for solving systems of linear equations using determinants, laying foundational algebraic groundwork for matrix theory.

1858

Cayley Formalizes Matrices

Arthur Cayley introduced the concept of a matrix as a mathematical object in its own right, defining addition, multiplication, and the inverse for rectangular arrays of numbers.

1888

Peano Axiomatizes Vector Spaces

Giuseppe Peano formalized the axioms of vector spaces, providing the precise setting in which linear transformations operate and connecting algebraic structure to geometric intuition.

1900s–1950s

Linear Algebra in Physics & Computing

Quantum mechanics and early computer science demanded large-scale matrix computations, driving both theoretical advances and the development of efficient numerical algorithms.

Today

Ubiquitous Applications

From 3-D graphics rendering and machine learning to robotics and data science, linear transformations and their matrix representations permeate modern technology.

The central question this lesson addresses is deceptively simple: How can we represent geometric operations — rotations, reflections, scalings, and shears — as algebraic objects that we can compute with? The answer lies in the deep correspondence between linear transformations of vectors and multiplication by matrices, a connection that transforms abstract geometry into concrete arithmetic.

SECTION 2

Core Principles & Definitions

A linear transformation is a function T from ℝ² to ℝ² (or more generally from ℝⁿ to ℝᵐ) that satisfies two defining properties: it preserves vector addition and it preserves scalar multiplication. In formal notation, T(u + v) = T(u) + T(v) for all vectors u and v, and T(cv) = cT(v) for every scalar c and vector v. These two conditions together ensure that the transformation respects the linear structure of the space — straight lines remain straight, the origin stays fixed, and parallel lines stay parallel. Any function meeting these criteria can be completely described by a matrix, a rectangular array of numbers that encodes how basis vectors are mapped.

Additivity

T(u + v) = T(u) + T(v). The image of a sum equals the sum of the images. This guarantees that parallelogram structures are preserved under the transformation.

Homogeneity

T(cv) = cT(v). Scaling a vector before or after applying T yields the same result. This ensures proportional relationships between vectors are maintained.

Origin Preservation

T(0) = 0 follows directly from homogeneity (set c = 0). The origin is always a fixed point of any linear transformation, distinguishing them from affine transformations.

Matrix Representation

Every linear transformation T: ℝ² → ℝ² corresponds to a unique 2 × 2 matrix A such that T(v) = Av. The columns of A are the images of the standard basis vectors e₁ and e₂.

Composition as Multiplication

Applying transformation T₁ then T₂ corresponds to multiplying their matrices: the composite transformation has matrix A₂A₁. Order matters — matrix multiplication is not commutative.

✦ KEY TAKEAWAY

Think of a linear transformation as a machine that reshapes the entire coordinate grid uniformly — like stretching a rubber sheet that is pinned at the origin. No matter where you are on the sheet, the rule for how your neighborhood distorts is the same. The matrix is the instruction manual for that machine: its columns tell you where the two reference directions (the standard basis vectors) end up, and everything else follows by linearity.

SECTION 3

Visualizing Linear Transformations

The most powerful way to understand a linear transformation is to watch what it does to the standard coordinate grid. Before the transformation, the grid is composed of evenly spaced horizontal and vertical lines meeting at right angles. After the transformation, these lines may stretch, compress, rotate, reflect, or shear — but they always remain straight and evenly spaced. The diagram below shows the original basis vectors e₁ = (1, 0) and e₂ = (0, 1), and their images under a sample transformation, illustrating how the columns of the matrix directly encode the destinations of these basis vectors.

Left: the standard basis vectors e₁ (cyan) and e₂ (violet) on the original grid. Right: after applying the transformation T, e₁ maps to T(e₁) = (1, 0.5) shown in pink and e₂ maps to T(e₂) = (−0.5, 1) shown in amber. These two image vectors become the columns of the transformation matrix.

Observe that the key information in the diagram reduces to four numbers: the x- and y-components of T(e₁) and T(e₂). These four numbers fill the entries of a 2 × 2 matrix. Every other vector v = (v₁, v₂) is a linear combination v₁e₁ + v₂e₂, so by linearity its image is T(v) = v₁T(e₁) + v₂T(e₂). This is precisely the computation performed by matrix–vector multiplication, which is why a single matrix encodes the entire transformation.

SECTION 4

Mathematical Framework

The algebraic core of this topic rests on the relationship between a linear transformation T: ℝ² → ℝ² and its standard matrix A. Given the standard basis vectors e₁ = (1, 0) and e₂ = (0, 1), the matrix A is constructed by placing T(e₁) as the first column and T(e₂) as the second column. Once A is known, applying T to any vector v amounts to computing the matrix–vector product Av.

MATRIX–VECTOR PRODUCT

T(v) = Av = [ a b ] [ v₁ ] = [ a·v₁ + b·v₂ ] [ c d ] [ v₂ ] [ c·v₁ + d·v₂ ]

Here a, b, c, d are the entries of the 2 × 2 matrix A, and v = (v₁, v₂) is the input vector. The first column (a, c) is the image of e₁, and the second column (b, d) is the image of e₂.

COMPOSITION OF TRANSFORMATIONS

(T₂ ∘ T₁)(v) = A₂(A₁v) = (A₂A₁)v

When two linear transformations are composed, the resulting matrix is the product A₂A₁. Note the order: the transformation applied first (T₁) appears on the right. Matrix multiplication is associative but generally not commutative, meaning A₂A₁ ≠ A₁A₂ in most cases.

DETERMINANT AND AREA SCALING

det(A) = ad − bc

The determinant of A measures how the transformation scales areas. If |det(A)| = 2, every region's area doubles. If det(A) < 0, the transformation reverses orientation (like a reflection). If det(A) = 0, the transformation collapses the plane onto a line or point and is non-invertible.

INVERSE TRANSFORMATION

A⁻¹ = (1 / det(A)) × [ d −b ] [ −c a ]

When det(A) ≠ 0, the transformation is invertible and A⁻¹ reverses its effect. The inverse exists precisely when the transformation does not collapse any dimension — geometrically, when the two column vectors are not parallel.

SECTION 5

Classification of Planar Transformations

Different choices of matrix entries produce fundamentally different geometric effects. The diagram below catalogs the most important families of 2 × 2 linear transformations, showing both their matrices and their action on the unit square. Understanding these families allows you to recognize a transformation's geometric meaning at a glance and to construct matrices for desired effects.

Six families of 2 × 2 transformations. Dashed squares represent the original unit square; colored outlines show the transformed shape. Note how the determinant distinguishes area-preserving (|det| = 1), area-scaling (|det| ≠ 1), and dimension-collapsing (det = 0) transformations.

Each family in the diagram above highlights a different geometric property. Rotations preserve both distances and angles, making them rigid motions. Scalings stretch or compress along the coordinate axes independently. Reflections flip orientation, changing the sign of the determinant. Shears tilt the grid without changing area — their determinant equals 1. Projections collapse a dimension entirely, yielding a determinant of zero and a non-invertible transformation. Any general 2 × 2 matrix can be understood as a combination of these elementary types.

SECTION 6

Worked Example: Composing Rotation and Scaling

Suppose we wish to first rotate every point in the plane by 90° counterclockwise and then scale the result by a factor of 2 in both directions. We will find the single matrix that accomplishes this composite transformation and apply it to the vector v = (3, 1).

Composite Transformation: Rotation then Scaling

Step 1 — Write the Rotation MatrixFor a counterclockwise rotation by θ = 90°, we use the rotation matrix formula. Since cos 90° = 0 and sin 90° = 1, the rotation matrix is R = [ 0 −1 ; 1 0 ]. This matrix sends e₁ = (1, 0) to (0, 1) and e₂ = (0, 1) to (−1, 0), consistent with a quarter-turn.

R = [ 0 −1 ; 1 0 ]

Step 2 — Write the Scaling MatrixScaling both coordinates by a factor of 2 corresponds to a diagonal matrix with both diagonal entries equal to 2. This matrix doubles the length of every vector while preserving its direction.

S = [ 2 0 ; 0 2 ]

Step 3 — Compute the Composite MatrixSince we apply R first and then S, the composite matrix is M = S × R. Performing the multiplication: M = [ 2 0 ; 0 2 ] × [ 0 −1 ; 1 0 ]. The (1,1) entry is 2×0 + 0×1 = 0. The (1,2) entry is 2×(−1) + 0×0 = −2. The (2,1) entry is 0×0 + 2×1 = 2. The (2,2) entry is 0×(−1) + 2×0 = 0.

M = [ 0 −2 ; 2 0 ]

Step 4 — Apply M to the Vector v = (3, 1)We compute Mv = [ 0 −2 ; 2 0 ] × [ 3 ; 1 ]. The first component is 0×3 + (−2)×1 = −2. The second component is 2×3 + 0×1 = 6.

T(v) = (−2, 6)

Step 5 — Verify with the Determinantdet(M) = 0×0 − (−2)×2 = 4. Since det(R) = 1 and det(S) = 4, we expect det(M) = det(S) × det(R) = 4 × 1 = 4. ✓ The area scaling factor is |det(M)| = 4, consistent with doubling both dimensions (2 × 2 = 4). The positive determinant confirms that orientation is preserved.

det(M) = 4 — area scales by factor 4, orientation preserved ✓

SECTION 7

Strengths & Limitations of the Matrix Framework

Representing linear transformations as matrices is an extraordinarily powerful technique, but it is important to understand both its capabilities and its boundaries. The table below summarizes the key strengths alongside notable limitations that arise when the assumptions of linearity break down.

Strengths and limitations of the 2 × 2 matrix representation of linear transformations.

Aspect	Strength	Limitation
Expressiveness	Captures rotations, reflections, scalings, shears, and projections in a unified format.	Cannot directly represent translations (shifts), which do not fix the origin. Affine transformations or augmented matrices are needed.
Composition	Composing transformations reduces to matrix multiplication — fast, systematic, and scalable to long chains.	Matrix multiplication is not commutative: the order of operations matters. Rotating then scaling differs from scaling then rotating.
Invertibility	If det(A) ≠ 0, the inverse matrix A⁻¹ provides an exact reversal of the transformation.	When det(A) = 0, information is irreversibly lost (e.g., projection onto a line). No inverse exists.
Scope	Extends naturally to higher dimensions (3 × 3 for 3-D, n × n for ℝⁿ) using the same rules.	Restricted to linear (and via augmented matrices, affine) transformations. Non-linear maps like bending or warping require other tools.

✦ KEY TAKEAWAY

Matrices are to linear transformations what blueprints are to buildings: they encode all the essential instructions in a compact, computable format. Just as a blueprint cannot represent the color of paint or the feel of carpet, a 2 × 2 matrix cannot capture non-linear effects like translations or curvature. Knowing these boundaries helps you choose the right mathematical tool for the task at hand.

SECTION 8

Connection to Advanced Theory

The ideas you have studied in this lesson form the gateway to one of the most far-reaching branches of mathematics: linear algebra. In a full linear algebra course, these 2 × 2 concepts generalize to arbitrary dimensions, and powerful decomposition theorems reveal the internal structure of any matrix. The table below maps each concept from this lesson to its advanced counterpart.

How core concepts from this lesson extend into advanced mathematics and applications.

This Lesson (ℝ²)	Advanced Linear Algebra	Where It Appears
2 × 2 transformation matrix	m × n matrices representing T: ℝⁿ → ℝᵐ	College linear algebra, data science
Determinant as area scale factor	n-dimensional determinant as volume scale factor; connection to wedge products	Multivariable calculus (Jacobian), differential geometry
Rotation and reflection matrices	Orthogonal matrices (O(n)), unitary matrices (U(n))	Quantum mechanics, signal processing, computer graphics
Eigenvalues via det(A − λI) = 0	Spectral theorem, eigendecomposition, singular value decomposition (SVD)	Machine learning (PCA), vibrations analysis, Google PageRank
Composition as matrix multiplication	Matrix groups, Lie groups, representation theory	Particle physics, robotics (kinematic chains)

Perhaps the most exciting forward connection involves eigenvalues and eigenvectors. An eigenvector of a matrix A is a non-zero vector v such that Av = λv for some scalar λ — meaning the transformation simply scales v without changing its direction. Finding these special directions unlocks deep structural information about the transformation, from the principal axes of an ellipse to the dominant modes of a vibrating system. When you encounter eigenvalues in a future course, remember that the intuition starts right here: every matrix reshapes the plane, and eigenvectors are the directions along which that reshaping is simplest.

SECTION 9

Practice Problems

A linear transformation T: ℝ² → ℝ² maps every vector to the zero vector: T(v) = 0 for all v. Which of the following is the standard matrix of T?

Let A = [ 2 −1 ; 3 4 ]. Compute the image of the vector v = (5, 2) under the transformation T(v) = Av. Which of the following is T(v)?

A linear transformation T rotates every vector in the plane by 60° counterclockwise. Which of the following is the standard matrix for T? (Use exact values: cos 60° = 1/2, sin 60° = √3/2.)

PROBLEM 4 — APPLIED

A computer graphics program applies two transformations in sequence to every vertex of a polygon: first a horizontal shear with matrix S = [ 1 2 ; 0 1 ], then a reflection over the x-axis with matrix R = [ 1 0 ; 0 −1 ]. (a) Find the single composite matrix M = RS that accomplishes both operations. (b) Apply M to the vertex P = (1, 3) and state the final coordinates. (c) Compute det(M) and interpret its geometric meaning.

PROBLEM 5 — CRITICAL THINKING

Let T be a linear transformation with matrix A = [ a b ; c d ] where det(A) = 0. (a) Explain geometrically what det(A) = 0 implies about the image of the unit square under T. (b) Show that there exists a non-zero vector v such that Av = 0, and describe what this means about T. (c) Can T be undone by another linear transformation? Justify your answer using both the determinant and the geometric interpretation.

SUMMARY

Lesson Summary

A linear transformation is a function that preserves vector addition and scalar multiplication, keeping the origin fixed and straight lines straight. Every such transformation from ℝ² to ℝ² has a unique 2 × 2 standard matrix whose columns are the images of the standard basis vectors e₁ and e₂. Key transformation families include rotations, reflections, scalings, shears, and projections.

The determinant det(A) = ad − bc measures the area scaling factor and indicates whether orientation is preserved (det > 0) or reversed (det < 0). A zero determinant signals a non-invertible transformation that collapses dimension. Composition of transformations corresponds to matrix multiplication (applied right to left), and the inverse matrix reverses a transformation when it exists. These concepts form the foundation for higher-dimensional linear algebra, eigenvalue theory, and countless applications in science and engineering.