Mathematics › 🎓 University Year 1 › The Eigen-Finder: Eigenvalues and Eigenvectors

The Eigen-Finder: Eigenvalues and Eigenvectors

Most vectors get knocked off course when a matrix acts on them — but a few special directions are only stretched, and those directions reveal what the matrix really does.

University Year 1Calculus II / Linear Algebra

The Eigen-Finder: Eigenvalues and Eigenvectors — illustration

💡

The big idea: A square matrix normally rotates and stretches vectors at the same time, which makes its action hard to picture. But almost every matrix has a handful of privileged directions it leaves pointing the same way, merely scaling them by a factor. Those directions are the eigenvectors and the scale factors are the eigenvalues. Together they are the matrix's fingerprint: they diagonalise it, predict long-term behaviour, and turn a tangle of coupled equations into independent ones.

🎯 By the end, you'll be able to

State the eigenvalue equation Av = λv and explain what it means geometrically
Find eigenvalues by solving the characteristic equation det(A − λI) = 0
Compute an eigenvector for a given eigenvalue by solving (A − λI)v = 0
Use the trace and determinant as quick checks on a pair of eigenvalues
Recognise when a matrix is defective or has complex eigenvalues

📎 You should already know

Matrix–vector multiplication
Determinants of 2×2 matrices
Solving homogeneous linear systems

The directions a matrix leaves alone

When a matrix A multiplies a vector, it usually does two things at once: it rotates the vector to a new direction and stretches it to a new length. Because both happen together, it is hard to say what the transformation “really” does.

But for almost every matrix there are a few special directions that refuse to rotate. A vector pointing along one of these directions comes out pointing exactly the same way — only longer, shorter, or flipped. Those un-rotated directions are the key to understanding the whole transformation.

🔑 Eigenvector and eigenvalue

A non-zero vector v is an eigenvector of A if A v points along the same line as v — that is, A v = λv for some scalar λ. The number λ (lambda) is the matching eigenvalue: the factor by which that direction is stretched (λ > 1), shrunk (0 < λ < 1) or flipped (λ < 0).

\[ A\vec{v} = \lambda\vec{v}, \qquad \vec{v} \neq \vec{0} \]

The eigenvalue equation. Acting with the matrix has the same effect, on this one vector, as simply multiplying by the number λ.

🎮 The Eigen-Finder LIVE

Find the special directions a matrix only stretches, never rotates — the eigenvectors.

How to hunt them down

Rewrite A v = λv as A v − λv = 0, then factor out the vector using the identity matrix I: (A − λI) v = 0. This is a homogeneous system. It has a non-zero solution v only when the matrix A − λI squashes some direction to zero — that is, when it is singular.

A matrix is singular exactly when its determinant is zero. So the eigenvalues are the numbers λ that make the determinant vanish.

\[ \det(A - \lambda I) = 0 \]

The characteristic equation. For an n×n matrix this is a degree-n polynomial in λ whose roots are the eigenvalues.

📝 Worked example: Find the eigenvalues of A = [[2, 1], [1, 2]].

Form A − λI = [[2 − λ, 1], [1, 2 − λ]].
Take the determinant: \( (2-\lambda)^2 - (1)(1) = \lambda^2 - 4\lambda + 3 \).
Factor the characteristic polynomial: \( \lambda^2 - 4\lambda + 3 = (\lambda - 1)(\lambda - 3) \).
Set each factor to zero.

✓ The eigenvalues are λ = 1 and λ = 3.

📝 Worked example: Find an eigenvector of A = [[2, 1], [1, 2]] for each eigenvalue.

For \( \lambda = 3 \): solve (A − 3I)v = 0, i.e. [[−1, 1], [1, −1]]v = 0. Both rows say −x + y = 0, so y = x. A choice is \( \vec{v} = (1, 1) \).
Check: A(1,1) = (2+1, 1+2) = (3, 3) = 3·(1,1). ✓
For \( \lambda = 1 \): solve (A − I)v = 0, i.e. [[1, 1], [1, 1]]v = 0. Both rows say x + y = 0, so y = −x. A choice is \( \vec{v} = (1, -1) \).
Check: A(1,−1) = (2−1, 1−2) = (1, −1) = 1·(1,−1). ✓

✓ Eigenvectors are (1, 1) for λ = 3 and (1, −1) for λ = 1. Any non-zero multiple of each is also an eigenvector.

✨ Two free consistency checks

For any square matrix the eigenvalues obey two shortcuts: their sum equals the trace (the diagonal total) and their product equals the determinant. For our example the trace is 2 + 2 = 4 = 1 + 3, and the determinant is 4 − 1 = 3 = 1 × 3. Both check out, so the eigenvalues are almost certainly right.

\[ \lambda_1 + \lambda_2 = \operatorname{tr}(A), \qquad \lambda_1 \lambda_2 = \det(A) \]

Sum-and-product relations for a 2×2 matrix — a fast sanity check on your eigenvalues.

Why they are worth the trouble

Line up the eigenvectors as the columns of a matrix P. Then P⁻¹AP is diagonal, with the eigenvalues down the diagonal. In eigenvector coordinates the matrix does nothing but scale each axis independently — the messy rotation-plus-stretch becomes pure stretching. This is why eigenvalues govern the long-term behaviour of repeated multiplication, the stability of equilibria, and the natural modes of vibrating systems.

⚠️ Not every matrix cooperates

Two things can go wrong. A shear such as [[1, 1], [0, 1]] has the repeated eigenvalue λ = 1 but only one independent eigenvector — it is defective and cannot be diagonalised. A pure rotation such as [[0, −1], [1, 0]] has no real eigenvectors at all, because it turns every real direction; its eigenvalues are the complex numbers ±i. A real characteristic polynomial with a negative discriminant is the warning sign.

Check your understanding

1. What does it mean for v to be an eigenvector of A?

An eigenvector keeps its direction under A; the matrix only scales it, so Av = λv for the eigenvalue λ.

2. Which equation gives the eigenvalues of A?

(A − λI)v = 0 has a non-zero solution only when A − λI is singular, i.e. det(A − λI) = 0 — the characteristic equation.

3. A 2×2 matrix has trace 5 and determinant 6. What are its eigenvalues?

The eigenvalues sum to the trace (2 + 3 = 5) and multiply to the determinant (2 × 3 = 6), so they are 2 and 3.

4. For A = [[2, 0], [0, 3]], which vector is an eigenvector with eigenvalue 3?

A(0,1) = (0, 3) = 3·(0,1). The vector along the y-axis is scaled by the diagonal entry 3, so its eigenvalue is 3.

5. Why does a 90° rotation matrix have no real eigenvectors?

A rotation turns every real direction to a new one, so no real v satisfies Av = λv; its eigenvalues are complex (±i here).

✅ Key takeaways

An eigenvector is a direction a matrix does not rotate; the eigenvalue λ is the factor it is scaled by, so Av = λv.
Eigenvalues are the roots of the characteristic equation det(A − λI) = 0.
For each eigenvalue, an eigenvector is any non-zero solution of (A − λI)v = 0.
Quick check: the eigenvalues sum to the trace and multiply to the determinant.
Eigenvectors diagonalise a matrix, turning rotation-plus-stretch into pure scaling — but defective matrices and rotations can fail to have a full set of real eigenvectors.