Finding a Basis - Linear Algebra

SECTION 1

Historical Context & Motivation

Imagine you have a huge box of LEGO bricks. Some bricks are copies or can be built from smaller ones. Wouldn't it be great to find the smallest set of unique bricks that lets you build everything in the box? That's exactly what mathematicians wanted when they developed the idea of a basis for a vector space. Over centuries, thinkers realized that complicated collections of arrows, forces, and equations could be described using just a few essential pieces.

1637 • Descartes & Coordinate Geometry

René Descartes introduced the coordinate plane, allowing points and directions to be described with numbers. This was the first step toward thinking of directions as 'building blocks.'

1844 • Grassmann's Extension Theory

Hermann Grassmann wrote about spaces of any dimension and the idea that a few independent directions could generate an entire space. His work was ahead of its time.

1888 • Peano's Axioms for Vector Spaces

Giuseppe Peano gave the first rigorous definition of a vector space, setting the stage for a formal definition of basis and dimension.

1900s • Modern Linear Algebra Takes Shape

Mathematicians like Steinitz proved that every vector space has a basis, and that all bases for the same space have the same number of vectors. These results became cornerstones of modern math and science.

The central question that drove all of this history is simple: What is the minimum set of vectors you need to describe every vector in a space? Answering that question is what "finding a basis" is all about.

SECTION 2

Core Principles & Definitions

Before we can find a basis, we need to understand three key ideas: vectors, linear combinations, and linear independence. Think of a vector as an arrow that has both a direction and a length. A linear combination is what you get when you scale some vectors by numbers and add them together — for example, 3·v₁ + 2·v₂ is a linear combination of v₁ and v₂. Vectors are linearly independent when none of them can be written as a combination of the others — each one brings something genuinely new to the table. These two ideas — spanning (can we reach everywhere?) and independence (is every vector truly necessary?) — are the foundation for what comes next.

Span

The span of a set of vectors is every vector you can create by scaling and adding those vectors together. If a set spans a space, you can reach any point in that space.

Linear Independence

Vectors are linearly independent if no vector in the set is redundant — you can't build any one of them from the others. Removing any one vector would shrink the span.

Basis

A basis is a set of vectors that is both linearly independent AND spans the entire space. It is the most efficient 'toolkit' for building every vector in the space.

Dimension

The dimension of a vector space is the number of vectors in any basis. Every basis for the same space has the same size, so dimension is a fixed property of the space itself.

✦

KEY TAKEAWAY

Think of a basis like a set of primary colors. By mixing primary colors in different amounts, you can create a vast range of other colors — and no primary color can be mixed from the remaining ones, so each is truly independent. A basis works the same way: it is the smallest independent set of vectors that, when scaled and added, can produce every vector in the space. Just as the number of primary colors needed depends on the color system you use, the number of basis vectors equals the dimension of the specific space — 2 vectors for a 2D plane, 3 for 3D space, and so on.

SECTION 3

Visual Explanation

Let's look at what a basis means in two dimensions — the flat plane you draw on in math class. In 2D, a basis consists of exactly two linearly independent vectors. The diagram below shows how two basis vectors, e₁ and e₂, can be scaled and added to reach any point on the plane.

The cyan arrow e₁ and the violet arrow e₂ form the standard basis for 2D. The pink dashed arrow shows how any vector v can be reached by scaling and adding the two basis vectors. Here, v = 3·e₁ + 2·e₂.

Notice that the two basis vectors point in completely different directions — they are linearly independent. If both arrows pointed along the same line, you could only reach points on that line, not the whole plane. That's why independence matters: each basis vector must add a genuinely new direction.

SECTION 4

Mathematical Framework

Now let's put the idea into precise mathematical language. Recall that a linear combination of vectors v₁, v₂, …, vₙ means choosing numbers c₁, c₂, …, cₙ and forming c₁v₁ + c₂v₂ + … + cₙvₙ. A set of vectors {v₁, v₂, …, vₙ} is a basis for a vector space V if two conditions are met: (1) the vectors span V, and (2) the vectors are linearly independent.

SPANNING CONDITION

For every vector w in V, there exist scalars c₁, c₂, …, cₙ such that w = c₁v₁ + c₂v₂ + … + cₙvₙ

This says every vector w in the space can be written as a combination of the basis vectors. The numbers c₁, c₂, … are called coordinates of w with respect to that basis.

INDEPENDENCE CONDITION

c₁v₁ + c₂v₂ + … + cₙvₙ = 0 implies c₁ = c₂ = … = cₙ = 0

The only way to combine the basis vectors to get the zero vector is to use all-zero scalars. If you could use non-zero scalars, it would mean at least one vector is redundant.

TESTING INDEPENDENCE WITH A MATRIX

Row-reduce the matrix A = [ v₁ | v₂ | … | vₙ ]. If every column has a pivot, the vectors are linearly independent.

Place each vector as a column in a matrix. Then use row reduction (also called Gaussian elimination) to simplify. A pivot is the leading non-zero entry in each row of the reduced matrix. If every column has a pivot, the vectors are independent.

💡

Quick Checklist for Finding a Basis

Step 1: Gather the vectors that describe your space (or subspace). Step 2: Form a matrix with these vectors as columns. Step 3: Row-reduce the matrix. Step 4: The columns corresponding to pivot positions in the original matrix form a basis. Non-pivot columns are redundant and can be removed.

SECTION 5

Types of Bases & How They Compare

A vector space can have many different bases, just as there are many ways to choose primary colors. Some bases are especially convenient. The diagram below compares the standard basis (the usual unit vectors along the axes) with a non-standard basis that still spans the same plane but points in different directions.

Left: the standard basis with perpendicular unit vectors. Right: a non-standard basis where the vectors are not perpendicular but still point in different enough directions to span the plane. Both are valid bases for ℝ².

Property	Standard Basis	Non-Standard Basis
Vectors	e₁ = (1, 0), e₂ = (0, 1)	b₁ = (3, 1), b₂ = (1, −2)
Perpendicular?	Yes	No
Linearly independent?	Yes	Yes
Spans ℝ²?	Yes	Yes
Easy coordinates?	Yes — coordinates match the entries	Harder — requires solving a system

Comparison of two valid bases for ℝ²

The key point is that a basis does not have to be the standard one. Any set that is linearly independent and spans the space works. However, the number of vectors in any basis for the same space is always the same. For the 2D plane ℝ², every basis has exactly 2 vectors. For 3D space ℝ³, every basis has exactly 3.

SECTION 6

Worked Example

Let's walk through a complete example. Suppose we have three vectors in ℝ³ and we want to determine whether they form a basis, or if one is redundant.

Find a basis from the set { v₁ = (1, 2, 3), v₂ = (4, 5, 6), v₃ = (2, 1, 0) }

1
Step 1 — Build the matrix

Place each vector as a column in a matrix A. We get: A = [ 1 4 2 ] [ 2 5 1 ] [ 3 6 0 ]
2
Step 2 — Row reduce (R₂ ← R₂ − 2R₁, R₃ ← R₃ − 3R₁)

Subtract 2 times Row 1 from Row 2, and 3 times Row 1 from Row 3: [ 1 4 2 ] [ 0 −3 −3 ] [ 0 −6 −6 ]
3
Step 3 — Continue reducing (R₃ ← R₃ − 2R₂)

Subtract 2 times Row 2 from Row 3: [ 1 4 2 ] [ 0 −3 −3 ] [ 0 0 0 ]

Row 3 is all zeros. This means we have only 2 pivots, not 3.
4
Step 4 — Identify pivot columns

The pivots are in columns 1 and 2. This means v₁ and v₂ are linearly independent and form a basis for the subspace spanned by the three vectors. The third vector v₃ is a combination of v₁ and v₂, so it is redundant.

Basis = { (1, 2, 3), (4, 5, 6) }
5
Step 5 — Verify the dimension

Our basis has 2 vectors, so the subspace spanned by v₁, v₂, and v₃ is a 2-dimensional plane inside 3D space. This makes sense — three vectors that are not all independent can at most span a plane, not all of ℝ³.

Dimension of the subspace = 2

ℹ

Why did v₃ become redundant?

We can verify that v₃ is redundant by finding coefficients c₁ and c₂ such that v₃ = c₁·v₁ + c₂·v₂. Reading from the reduced matrix: the third column of the row-reduced form is (2, −3, 0)ᵀ (before the zero row), which means column 3 satisfies the same linear relationships as the pivot columns. Specifically, from the reduced system: Row 2 gives −3c₂ = −3, so c₂ = 1; Row 1 gives c₁ + 4c₂ = 2, so c₁ = 2 − 4 = −2. Therefore v₃ = −2·v₁ + 1·v₂. Let's verify: −2·(1, 2, 3) + (4, 5, 6) = (−2 + 4, −4 + 5, −6 + 6) = (2, 1, 0) ✓. This confirms v₃ is exactly a linear combination of v₁ and v₂, so removing it leaves a valid basis.

SECTION 7

Methods for Finding a Basis

There are several methods you might use to find a basis, depending on what you're given. Each has its strengths and limitations. The table below compares the most common approaches you'll encounter.

Method	When to Use	Strengths	Limitations
Row Reduction	Given a set of vectors; want to find which are independent	Systematic, works every time, gives the dimension automatically	Can be tedious with large matrices by hand
Inspection	Simple sets with obvious relationships (e.g., one vector is a scalar multiple of another)	Fast, builds intuition	Only works for small, simple sets; easy to make mistakes
Determinant Test	Checking if n vectors form a basis for ℝⁿ (square matrix)	Quick yes/no answer: if det ≠ 0, they're a basis	Only works when the number of vectors equals the dimension; doesn't extract a subset
Gram-Schmidt Process	When you want an orthogonal (perpendicular) basis	Produces perpendicular basis vectors, which simplify many calculations	More complex; typically studied after mastering basic basis-finding

Comparison of common methods for finding a basis

✦

KEY TAKEAWAY

Row reduction is like using a sieve to filter out redundant vectors. You pour all your vectors in, shake (row reduce), and what's left — the pivot columns — is a clean, independent basis. The other vectors fall through because they were just mixtures of the ones that remained.

SECTION 8

Connection to Advanced Topics

Finding a basis is not just a classroom exercise — it's a gateway to many powerful ideas in mathematics, science, and engineering. Once you know how to find a basis, you can describe transformations, solve systems of equations more efficiently, and even compress images and data.

Concept You Know	Where It Leads
Basis & dimension	Rank of a matrix — the dimension of the column space tells you how many independent equations you really have
Coordinates relative to a basis	Change of basis — you can switch between different coordinate systems, which is essential in computer graphics and physics
Orthogonal basis	Fourier series and signal processing — sounds and images are broken down into perpendicular 'frequency' basis vectors
Eigenvalues & eigenvectors	Special bases where a transformation just stretches each basis vector. Used in Google's PageRank and vibration analysis

How finding a basis connects to advanced topics

As you move forward in linear algebra, you'll see that nearly every important concept circles back to the idea of a basis. Mastering this skill now gives you a strong foundation for everything that comes next.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

In your own words, explain why the set { (1, 0), (2, 0) } is NOT a basis for ℝ². Which requirement does it fail?

Show answer

The set { (1, 0), (2, 0) } fails the spanning requirement. Both vectors lie along the x-axis, so their span is just the x-axis — a single line, not the entire 2D plane. It also fails the linear independence requirement because (2, 0) = 2 × (1, 0), meaning the second vector is just a scaled copy of the first. A basis for ℝ² needs two vectors that point in genuinely different directions.

Question 2: BASIC CALCULATION

Determine whether the set { (1, 3), (2, 7) } is a basis for ℝ². Use the determinant test.

Show answer

Form the 2×2 matrix with these vectors as columns: A = [[1, 2], [3, 7]]. Compute the determinant: det(A) = (1)(7) − (2)(3) = 7 − 6 = 1. Since the determinant is not zero, the two vectors are linearly independent. Because we have 2 independent vectors in ℝ² (which has dimension 2), they also span ℝ². Therefore, { (1, 3), (2, 7) } is indeed a basis for ℝ².

Question 3: INTERMEDIATE

Given the vectors v₁ = (1, 0, 1), v₂ = (0, 1, 1), and v₃ = (1, 1, 2), determine if they form a basis for ℝ³. If not, find a basis from this set for the subspace they span.

Show answer

Form the matrix A = [[1, 0, 1], [0, 1, 1], [1, 1, 2]] and row reduce. R₃ ← R₃ − R₁ gives [[1, 0, 1], [0, 1, 1], [0, 1, 1]]. Then R₃ ← R₃ − R₂ gives [[1, 0, 1], [0, 1, 1], [0, 0, 0]]. There are only 2 pivots (in columns 1 and 2), so the three vectors are NOT linearly independent and do NOT form a basis for ℝ³. A basis for the subspace they span is { v₁, v₂ } = { (1, 0, 1), (0, 1, 1) }, which spans a 2-dimensional plane inside ℝ³. Note that v₃ = v₁ + v₂, confirming it is redundant.

Question 4: APPLIED

A drone can move in three directions: forward (1, 0, 0), right (0, 1, 0), and diagonally up-right (0, 1, 1). Can the drone reach any point in 3D space using combinations of these three movements? Explain using the concept of basis.

Show answer

We need to check if { (1, 0, 0), (0, 1, 0), (0, 1, 1) } forms a basis for ℝ³. Form the matrix and compute the determinant: det = 1·(0·1 − 0·1) − 0·(0·1 − 0·1) + 0·(0·1 − 1·1). Expanding along the first row: det = 1·(0 − 0) − 0 + 0 = 0... Wait, let's recalculate carefully: det = 1·(0·1 − 1·0) − 0·(0·1 − 1·0) + 0·(0·0 − 1·0) = 1·0 = 0. Actually, using cofactor expansion along column 1: det = 1·det[[1,0],[1,1]] = 1·(1−0) = 1 ≠ 0. Since the determinant is 1, the three vectors are linearly independent and form a basis for ℝ³. The drone can indeed reach any point in 3D space by combining forward, right, and diagonally-up-right movements with the right scaling.

Question 5: CRITICAL THINKING

Suppose you have 4 vectors in ℝ³. Can they ever form a basis for ℝ³? Explain why or why not using the relationship between the number of vectors in a basis and the dimension of the space.

Show answer

No, 4 vectors can never form a basis for ℝ³. By definition, a basis must be both linearly independent and spanning. The dimension of ℝ³ is 3, which means every basis for ℝ³ has exactly 3 vectors. If you have 4 vectors in ℝ³, at least one must be a linear combination of the others — they cannot all be independent. You could find a basis from the set by row reducing and keeping the pivot columns, but that basis would have at most 3 vectors. This illustrates a deep principle: the dimension is an invariant of the space, and no basis can have more or fewer vectors than the dimension.

SUMMARY

Summary

A basis is the most efficient set of building blocks for a vector space. It must satisfy two conditions: the vectors must be linearly independent (no vector is redundant) and they must span the entire space (every vector can be written as a combination of the basis vectors). The number of vectors in any basis is the dimension of the space, and this number never changes no matter which basis you choose.

To find a basis from a given set of vectors, use row reduction: place the vectors as columns in a matrix, reduce to echelon form, and keep the original vectors corresponding to pivot columns. For a quick check when the number of vectors equals the dimension, compute the determinant — if it's not zero, you have a basis. This concept is the foundation for nearly everything else in linear algebra, from solving systems to understanding transformations.