Change of Basis | Varsity Tutors

SECTION 1

Historical Context & Motivation

Imagine you are giving someone directions to a friend's house. You could say "go three blocks east and two blocks north," or you could say "walk straight toward the big oak tree for five minutes." Both descriptions lead to the same place, but they use different reference points. In mathematics, a change of basis works the same way — it lets you describe the same vector (the same point or arrow in space) using a different set of reference directions.

Throughout history, mathematicians realized that choosing the right coordinate system could turn a messy problem into a simple one. The idea of switching between coordinate systems has roots stretching back centuries, and it became a formal tool as linear algebra grew into a fully developed branch of mathematics.

1637

Descartes Introduces Coordinates

René Descartes published his idea of using two perpendicular number lines (the x-axis and y-axis) to locate points in a plane. This Cartesian coordinate system became the "default" basis most students learn first.

1750s

Euler & Rotated Axes

Leonhard Euler studied how rotating a coordinate system could simplify equations in physics and astronomy. His work showed that different bases reveal different structures in the same problem.

1844

Grassmann's Theory of Extensions

Hermann Grassmann introduced abstract ideas about vector spaces and linear combinations, laying the groundwork for thinking about bases as interchangeable building blocks.

1888

Peano Formalizes Vector Spaces

Giuseppe Peano gave the first rigorous definition of a vector space, making the notion of a basis — and therefore changing between bases — precise and general.

1900s–Today

Modern Applications

Change of basis is now used everywhere: in computer graphics to rotate 3-D models, in data science (PCA) to simplify big datasets, and in quantum mechanics to switch between measurement setups.

The central question this lesson answers is: If you know a vector's coordinates in one basis, how do you find its coordinates in a different basis? Understanding this will give you a powerful tool for simplifying problems across many areas of math and science.

SECTION 2

Core Principles & Definitions

Before you can change a basis, you need to understand what a basis actually is. Think of a basis as a set of building-block arrows. Every vector in the space can be built by stretching and combining those arrows. Here are the key ideas you need.

Basis

A basis is a set of vectors that (1) are linearly independent (none is a combination of the others) and (2) span the whole space (you can build any vector from them).

Coordinates

The coordinates of a vector are the numbers that tell you how much of each basis vector to use. Change the basis, and the coordinates change — even though the actual vector stays the same.

Standard Basis

In 2-D, the standard basis is {e₁, e₂} = {(1, 0), (0, 1)}. These are the familiar x- and y-direction unit vectors.

Change-of-Basis Matrix

The change-of-basis matrix (often called P) is a special matrix that converts coordinates from one basis to another. Multiply a coordinate vector by P (or P⁻¹) and you switch bases.

The Vector Itself Doesn't Change

This is the most important idea: a change of basis does not move the vector. It only changes the labels (coordinates) we use to describe it.

✦ KEY TAKEAWAY

Think of it like language translation. The word "water" in English and "agua" in Spanish refer to the exact same liquid. Similarly, a vector might be (3, 2) in one basis and (1, 4) in another — but it is the same arrow in space. A change of basis is just a translation between two mathematical "languages."

SECTION 3

Visual Explanation

The diagram below shows the same vector v (the pink arrow) described in two different bases. On the left, you see the standard basis (blue arrows along the x- and y-axes). On the right, you see a new basis (violet arrows) that point in different directions. Notice how the pink vector hasn't moved — only the grid lines and the coordinate numbers change.

_{in basis B}c₁×b₁c₂×b₂

Left: the vector v has coordinates (3, 3) in the standard basis {e₁, e₂}. Right: the same vector v is expressed as a linear combination c₁·b₁ + c₂·b₂ in basis B = {b₁, b₂} where b₁ = (2, 1) and b₂ = (1, 3). The pink arrow is identical in both panels — only the coordinate labels change. Section 4 shows how to compute the exact values of c₁ and c₂ using the change-of-basis matrix.

In the left panel, you can see how the vector v is built by going 3 units along e₁ and 3 units along e₂, placing its tip at (3, 3) in standard coordinates. In the right panel, the same tip of the arrow is reached by combining b₁ and b₂ with appropriate scalar weights c₁ and c₂. To find those weights precisely, we use the change-of-basis matrix P whose columns are b₁ and b₂ — the full calculation is carried out in Section 4.

👁️ Visual Intuition

When you change the basis, imagine stretching and rotating the entire grid. The vector stays pinned in place while the grid lines shift around it. The new grid lines give the vector new coordinates.

SECTION 4

Mathematical Framework

Now let's see the actual formulas. Suppose you have a vector v whose coordinates in the standard basis are known, and you want to find its coordinates in a new basis B = {b₁, b₂}. The key tool is the change-of-basis matrix.

Building the Change-of-Basis Matrix

Write each new basis vector as a column in a matrix. If your new basis in 2-D is B = {b₁, b₂}, then the change-of-basis matrix P is:

CHANGE-OF-BASIS MATRIX

P = [ b₁ | b₂ ]

Each column of P is one of the new basis vectors, written in standard coordinates.

Converting Standard → New Basis

If you know the vector's coordinates in the standard basis (call them [v]_standard) and you want the coordinates in basis B (call them [v]_B), you solve:

STANDARD TO NEW BASIS

[v]_B = P⁻¹ × [v]_standard

P⁻¹ is the inverse of the change-of-basis matrix. It "undoes" the new basis to reveal the new coordinates.

Converting New Basis → Standard

Going the other way is even simpler. If you know the coordinates in the new basis and want standard coordinates:

NEW BASIS TO STANDARD

[v]_standard = P × [v]_B

Just multiply by P directly. This works because P × [v]_B = b₁ × c₁ + b₂ × c₂, which is exactly the definition of the vector as a combination of the new basis vectors.

The 2×2 Inverse Formula

For a 2 × 2 matrix, there is a quick formula for the inverse. If P = [[a, b], [c, d]], then:

2×2 MATRIX INVERSE

P⁻¹ = (1 / (ad − bc)) × [[d, −b], [−c, a]]

The number (ad − bc) is called the determinant. If the determinant is zero, the matrix has no inverse and the vectors are not a valid basis.

To illustrate, let's apply this to the basis from the diagram in Section 3: b₁ = (2, 1) and b₂ = (1, 3), so P = [[2, 1], [1, 3]]. The determinant is (2)(3) − (1)(1) = 6 − 1 = 5. Therefore P⁻¹ = (1/5) × [[3, −1], [−1, 2]]. For v = (3, 3) in standard coordinates, the new-basis coordinates are [v]_B = P⁻¹ × (3, 3): first component = (1/5)[(3)(3) + (−1)(3)] = (1/5)(6) = 6/5, second component = (1/5)[(−1)(3) + (2)(3)] = (1/5)(3) = 3/5. We can verify: (6/5)·(2,1) + (3/5)·(1,3) = (12/5, 6/5) + (3/5, 9/5) = (15/5, 15/5) = (3, 3) ✓.

💡 Why Does This Work?

The matrix P encodes a recipe: "combine the new basis vectors with these weights to get a vector in standard coordinates." The inverse P⁻¹ reverses the recipe — it figures out which weights (new coordinates) you need.

SECTION 5

Detailed Breakdown: What the Matrix Does Geometrically

The change-of-basis matrix doesn't just shuffle numbers around. It has a clear geometric meaning. When you multiply by P, you are stretching and rotating the standard grid to match the new basis directions. When you multiply by P⁻¹, you are "un-stretching" and "un-rotating" back.

The standard square grid (left) is transformed into a parallelogram grid (right) defined by the new basis vectors b₁ and b₂. The green dot representing vector v stays in the same position. Multiplying by P⁻¹ finds where v sits on the new (parallelogram) grid.

In the left panel, the square grid is formed by the standard basis vectors e₁ and e₂. In the right panel, the grid is formed by the new basis vectors b₁ and b₂, creating parallelograms instead of squares. The green vector v hasn't moved — its tip is in the same place in both panels. But its "address" on the grid has changed because the grid itself has changed.

This geometric picture explains why the determinant matters. The determinant of P tells you how much the area of each grid cell changes. If the determinant is zero, the parallelogram collapses into a line, and you can't assign unique coordinates — that's why the inverse doesn't exist in that case.

SECTION 6

Worked Example

Let's walk through a complete change-of-basis problem step by step.

📝 Problem Statement

A vector v has coordinates (5, 3) in the standard basis. Find its coordinates in the new basis B = {b₁, b₂} where b₁ = (2, 1) and b₂ = (1, 1).

Finding [v]_B — Coordinates in Basis B

Step 1 — Build the Change-of-Basis Matrix PPlace each new basis vector as a column of a matrix. Since b₁ = (2, 1) and b₂ = (1, 1), we get P = [[2, 1], [1, 1]]. The first column is b₁ and the second column is b₂.

P = [[2, 1], [1, 1]]

Step 2 — Compute the DeterminantFor a 2 × 2 matrix [[a, b], [c, d]], the determinant is ad − bc. So: det(P) = (2)(1) − (1)(1) = 2 − 1 = 1. Since the determinant is not zero, this is a valid basis and the inverse exists.

det(P) = 1

Step 3 — Find P⁻¹Using the 2 × 2 inverse formula: P⁻¹ = (1/det) × [[d, −b], [−c, a]]. Plugging in: P⁻¹ = (1/1) × [[1, −1], [−1, 2]] = [[1, −1], [−1, 2]].

P⁻¹ = [[1, −1], [−1, 2]]

Step 4 — Multiply P⁻¹ × [v]_standardNow multiply: [v]_B = P⁻¹ × (5, 3). First row: (1)(5) + (−1)(3) = 5 − 3 = 2. Second row: (−1)(5) + (2)(3) = −5 + 6 = 1.

[v]_B = (2, 1)

Step 5 — Verify the AnswerCheck: 2 × b₁ + 1 × b₂ = 2 × (2, 1) + 1 × (1, 1) = (4, 2) + (1, 1) = (5, 3). This matches the original vector, so our answer is correct!

✓ Verified: 2·b₁ + 1·b₂ = (5, 3) = v

✓ CHECK YOUR WORK

Always verify by plugging your new coordinates back in. Multiply each new coordinate by its basis vector and add the results. You should get back the original vector in standard coordinates.

SECTION 7

Strengths & Limitations

Change of basis is a powerful technique, but it is not always necessary. Understanding when to use it — and when to stick with the standard basis — is part of developing good mathematical judgment.

Strengths and limitations of change-of-basis techniques

Aspect	Strengths	Limitations
Problem Simplification	A clever basis can make matrices diagonal, turning hard problems into easy ones.	Finding the right basis can itself be a hard problem, especially in higher dimensions.
Computation	Once P⁻¹ is found, converting many vectors is quick (just matrix multiplication).	Computing the inverse of large matrices is computationally expensive.
Geometric Insight	Reveals hidden structure, such as principal directions of stretching or rotation.	The new coordinates may be less intuitive than the standard x-y system.
Applications	Essential in computer graphics, quantum mechanics, data science (PCA), and signal processing.	Only applies within linear (vector space) settings. Doesn't directly handle curved or nonlinear systems.

🔑 WHEN TO CHANGE BASIS

Think of change of basis like choosing the right tool from a toolbox. A wrench is great for bolts but useless for screws. Similarly, switching to a new basis is powerful when the problem has a natural direction — like the axis of rotation or the direction of maximum stretch — that doesn't line up with the standard axes.

SECTION 8

Connection to Advanced Topics

Change of basis is not just a standalone trick. It connects deeply to several advanced ideas in linear algebra. Once you understand how to switch between bases, you unlock the door to powerful concepts like diagonalization, eigenvalues and eigenvectors, and similarity transformations.

How change of basis connects to advanced linear algebra

This Lesson	Advanced Extension
Change-of-basis matrix P converts coordinates between two bases.	Similarity transformation: If a linear transformation T has matrix A in the standard basis, then its matrix in basis B is P⁻¹AP. This is a "change of basis for transformations."
Some bases make coordinates simpler.	Diagonalization: If you choose the eigenvectors as your new basis, the transformation matrix becomes diagonal — the simplest possible form.
The determinant of P tells if a basis is valid.	Invertible Matrix Theorem: A matrix is invertible if and only if its columns form a basis. This connects change of basis to many other equivalent conditions.
Works in 2-D and 3-D.	General n-dimensional spaces: The same formula [v]_B = P⁻¹[v]_standard works in any finite-dimensional vector space, including function spaces used in Fourier analysis.

In college-level linear algebra and beyond, nearly every major topic involves a change of basis in some form. Mastering this concept now gives you a strong foundation for all of these future ideas. If you continue in math, engineering, or data science, you will see change of basis again and again.

SECTION 9

Practice Problems

Try these five problems to test your understanding. They start easy and get progressively more challenging. Full answers are provided — try each problem on your own before reading the solution!

PROBLEM 1 — CONCEPTUAL

In your own words, explain: when you perform a change of basis, does the vector itself move? What actually changes?

PROBLEM 2 — BASIC CALCULATION

Given the new basis B = {b₁, b₂} where b₁ = (1, 0) and b₂ = (0, 2), find the coordinates of v = (3, 8) in basis B.

PROBLEM 3 — INTERMEDIATE

Let B = {b₁, b₂} where b₁ = (3, 1) and b₂ = (1, 2). A vector has coordinates [v]_B = (2, −1) in basis B. What are its standard coordinates?

PROBLEM 4 — APPLIED

A robot moves on a tilted surface. Its local axes are b₁ = (1, 1) (forward-right) and b₂ = (−1, 1) (forward-left). If the robot reports it moved (3, 2) in its own coordinate system, what is the displacement in standard map coordinates?

PROBLEM 5 — CRITICAL THINKING

Suppose someone claims a new basis B = {(2, 4), (1, 2)} for ℝ². Explain why this is not actually a valid basis, and connect your reasoning to the change-of-basis matrix.

SUMMARY

Summary

A change of basis lets you re-describe a vector using a different set of reference directions. Every basis is a collection of linearly independent vectors that span the space. The vector itself never moves — only its coordinates change. To convert coordinates from the standard basis to a new basis B, you build the change-of-basis matrix P (whose columns are the new basis vectors) and compute [v]_B = P⁻¹ × [v]_standard. To go the other way, use [v]_standard = P × [v]_B.

The determinant of P must be nonzero for the basis to be valid. Geometrically, changing basis stretches and rotates the coordinate grid while the vector stays fixed. This technique is the foundation for advanced topics like diagonalization and similarity transformations, and it has real-world applications in computer graphics, data science, and physics. Always verify your answer by checking that the new coordinates, combined with the new basis vectors, reconstruct the original vector.