An eigenvector of a square matrix A is a nonzero vector v that A only stretches, so Av = λv, and the number λ is its eigenvalue. To find them, solve the characteristic equation det(A − λI) = 0 for λ, then solve (A − λI)v = 0 for each λ. Below, six original examples take you from a plain 2×2 through repeated, defective and complex cases to a typical "compute A⁵v" exam question.
What Av = λv is really saying
Most vectors change direction when you multiply them by a matrix. An eigenvector keeps its line: A sends it to a multiple of itself. If λ = 3 the vector triples, if λ = −1 it flips, if λ = 0 it collapses to the zero vector.
Rewrite Av = λv as (A − λI)v = 0. A nonzero solution v exists only when A − λI is singular, which means det(A − λI) = 0. That single fact gives the whole method:
- Form A − λI. Subtract λ from the diagonal entries only.
- Solve det(A − λI) = 0. This polynomial is the characteristic polynomial; its roots are the eigenvalues.
- For each eigenvalue, solve (A − λI)v = 0 by row reduction. The solutions, minus the zero vector, are the eigenvectors. Together with 0 they form the eigenspace.
- Check: multiply A by each eigenvector. You should get λ times the same vector.
If you want someone to watch you run this method on your own problem sets and spot where it slips, our online linear algebra tutoring page explains how sessions with Ajay Vatsyayan work.
Example 1: a 2×2 with two real eigenvalues
Let A = [[4, 1], [2, 3]].
A − λI = [[4 − λ, 1], [2, 3 − λ]]
det(A − λI) = (4 − λ)(3 − λ) − 2 = λ² − 7λ + 10 = (λ − 2)(λ − 5)
So λ = 2 and λ = 5.
λ = 5: A − 5I = [[−1, 1], [2, −2]]. The first row says −x + y = 0, so y = x. Eigenvector (1, 1).
λ = 2: A − 2I = [[2, 1], [2, 1]]. The row says 2x + y = 0, so y = −2x. Eigenvector (1, −2).
Check: A(1, 1) = (5, 5) = 5(1, 1) and A(1, −2) = (2, −4) = 2(1, −2).
Notice that for a 2×2 the two rows of A − λI are always multiples of each other once λ is correct. If they are not, your λ is wrong: go back and recheck the determinant.
Shortcuts and checks worth knowing
For any 2×2, det(A − λI) = λ² − (trace A)λ + det A. Here trace = 7 and det = 10, which gives the same equation instantly. The table below collects the patterns we use most.
| What you notice | What it tells you |
|---|---|
| Matrix is 2×2 | λ² − (trace)λ + det = 0 |
| Matrix is triangular or diagonal | Eigenvalues are the diagonal entries |
| Every row sums to the same number c | c is an eigenvalue, with eigenvector (1, 1, …, 1) |
| det A = 0 | 0 is an eigenvalue |
| Matrix is real and symmetric | All eigenvalues are real; eigenvectors for different eigenvalues are orthogonal |
| Any square matrix | Sum of eigenvalues = trace; product = determinant (counting repeats) |
| You know λ for A | A² has λ², A⁻¹ has 1/λ, A + kI has λ + k, same eigenvectors |
The last two rows are the quickest way to catch an arithmetic slip before you hand anything in.
Example 2: a full 3×3 characteristic polynomial
Let A = [[0, 2, 1], [−1, 3, 1], [−4, 2, 5]].
Expand det(A − λI) along the first row:
det = (−λ)[(3 − λ)(5 − λ) − 2] − 2[(−1)(5 − λ) − (1)(−4)] + 1[(−1)(2) − (3 − λ)(−4)]
The three brackets simplify to λ² − 8λ + 13, then λ − 1, then 10 − 4λ. So
det = −λ³ + 8λ² − 13λ − 2λ + 2 + 10 − 4λ = −λ³ + 8λ² − 19λ + 12
Set it to zero and multiply by −1: λ³ − 8λ² + 19λ − 12 = 0. Any integer root must divide 12, so try 1: 1 − 8 + 19 − 12 = 0. Dividing by (λ − 1) leaves λ² − 7λ + 12 = (λ − 3)(λ − 4).
Eigenvalues: 1, 3, 4. Check: 1 + 3 + 4 = 8 = trace, and 1 × 3 × 4 = 12 = det A.
λ = 1: A − I = [[−1, 2, 1], [−1, 2, 1], [−4, 2, 4]]. Rows 1 and 2 are equal. From −x + 2y + z = 0 and −4x + 2y + 4z = 0, subtracting the first from the second gives −3x + 3z = 0, so z = x, and then y = 0. Eigenvector (1, 0, 1).
λ = 3: A − 3I = [[−3, 2, 1], [−1, 0, 1], [−4, 2, 2]]. Row 2 gives z = x; row 1 then gives y = x. Eigenvector (1, 1, 1).
λ = 4: A − 4I = [[−4, 2, 1], [−1, −1, 1], [−4, 2, 1]]. Row 2 gives z = x + y; substituting into row 1 gives −3x + 3y = 0. Eigenvector (1, 1, 2).
Check the last one: A(1, 1, 2) = (4, 4, 8). In hindsight every row of A sums to 3, so the table would have handed you λ = 3 and (1, 1, 1) before any expansion.
Example 3: a repeated eigenvalue with two eigenvectors
Let A = [[5, −3, 3], [3, −1, 3], [3, −3, 5]].
Every row sums to 5, so λ = 5 with eigenvector (1, 1, 1). Now use the checks rather than a cubic. Trace = 9 and det = 20, so the other two eigenvalues add to 9 − 5 = 4 and multiply to 20 ÷ 5 = 4. They solve μ² − 4μ + 4 = 0, so both equal 2.
λ = 2: A − 2I = [[3, −3, 3], [3, −3, 3], [3, −3, 3]]. Rank 1, so one equation, x − y + z = 0, and two free variables. Put y = s and z = t: x = s − t. Eigenspace spanned by (1, 1, 0) and (−1, 0, 1).
The eigenvalue 2 appears twice in the polynomial (algebraic multiplicity 2) and has a two-dimensional eigenspace (geometric multiplicity 2). Because the multiplicities match for every eigenvalue, this matrix has three independent eigenvectors, which is exactly what diagonalisation needs.
Example 4: a repeated eigenvalue with only one eigenvector
Let A = [[5, 1], [−1, 3]]. Trace 8, det 15 + 1 = 16, so λ² − 8λ + 16 = (λ − 4)².
A − 4I = [[1, 1], [−1, −1]], which has rank 1. The only condition is x + y = 0, giving one eigenvector direction, (1, −1).
Algebraic multiplicity 2, geometric multiplicity 1. A matrix like this is called defective. It cannot be diagonalised, and in a differential equations course it is the case that forces the extra t·e^(4t) term.
The rule: geometric multiplicity = n − rank(A − λI), and it is always at least 1 and at most the algebraic multiplicity.
Example 5: complex eigenvalues
Let A = [[1, −2], [1, 3]]. Trace 4, det 3 + 2 = 5, so λ² − 4λ + 5 = 0 and
λ = (4 ± √(16 − 20)) / 2 = 2 ± i
λ = 2 + i: A − (2 + i)I = [[−1 − i, −2], [1, 1 − i]]. The second row is simpler: x + (1 − i)y = 0, so x = (−1 + i)y. Take y = 1: eigenvector (−1 + i, 1).
Check the first row: (−1 − i)(−1 + i) − 2 = (1 + 1) − 2 = 0.
For a real matrix the second eigenvalue is the conjugate, 2 − i, with the conjugate eigenvector (−1 − i, 1). You never need to solve twice. Geometrically, a real 2×2 with complex eigenvalues rotates every direction, which is why no real eigenvector exists.
Example 6: using eigenvectors to compute A⁵v
Return to A = [[4, 1], [2, 3]] from Example 1 and find A⁵v for v = (3, 0) without multiplying A by itself five times.
Write v in terms of the eigenvectors: (3, 0) = a(1, 1) + b(1, −2). Then a + b = 3 and a − 2b = 0, so a = 2 and b = 1.
Each eigenvector just gets multiplied by its eigenvalue each time:
A⁵v = 2 × 5⁵ (1, 1) + 1 × 2⁵ (1, −2) = 6250(1, 1) + 32(1, −2) = (6282, 6186)
This idea, splitting a vector into eigen-directions where A acts simply, is the reason eigenvalues matter in Markov chains, dynamics and data science.
Where students lose marks
- Row reducing A before finding eigenvalues. Row operations change the eigenvalues. Only row reduce A − λI, and only after λ is fixed.
- Subtracting λ from every entry. λI has λ on the diagonal and zeros elsewhere.
- Expanding a 3×3 determinant with a sign slip. The middle cofactor carries a minus sign. Expand along a row or column with zeros if there is one, and confirm with trace and determinant afterwards.
- Getting a unique solution for (A − λI)v = 0. If row reduction gives only v = 0, the eigenvalue is wrong. Don't write "no eigenvector"; go back.
- Giving the zero vector as an eigenvector. By definition it never is, even though it lies in the eigenspace.
- Marking a scaled answer as wrong. (2, 2) and (1, 1) describe the same eigenvector. When checking against a solution sheet, compare directions.
- Assuming eigenvalues add. The eigenvalues of A + B are not the sums of those of A and B (Practice question 6 shows why).
- Claiming "repeated eigenvalue, so not diagonalisable". Example 3 is a counterexample. Count the eigenspace dimension.
Practice questions
- Find the eigenvalues and eigenvectors of [[6, −2], [2, 1]].
- Find the eigenvalues of [[3, 4, −1], [0, −2, 5], [0, 0, 3]]. How many independent eigenvectors belong to the repeated eigenvalue?
- Find the eigenvalues of [[3, −5], [1, −1]] and an eigenvector for the one with positive imaginary part.
- Using your answer to Question 1, give the eigenvalues of A², A⁻¹ and A − 4I.
- Find the eigenvalues and eigenvectors of [[2, 1, 1], [0, 3, 1], [1, 2, 1]]. (Look at the row sums and the determinant first.)
- True or false: if λ is an eigenvalue of A and μ is an eigenvalue of B, then λ + μ is an eigenvalue of A + B.
Answers.
- λ = 2 with (1, 2); λ = 5 with (2, 1).
- Triangular, so 3, −2, 3. A − 3I has rank 2, so λ = 3 has only one independent eigenvector, (1, 0, 0); the matrix is not diagonalisable.
- λ = 1 ± i; for 1 + i an eigenvector is (2 + i, 1).
- A²: 4 and 25. A⁻¹: ½ and ⅕. A − 4I: −2 and 1.
- Row sums are all 4 and det = 0, so 4 and 0 are eigenvalues; the trace 6 gives the third as 2. Eigenvectors: λ = 4, (1, 1, 1); λ = 2, (3, −1, 1); λ = 0, (−1, −1, 3).
- False. A = [[1, 0], [0, 0]] and B = [[0, 0], [0, 1]] both have eigenvalue 1, but A + B = I has only the eigenvalue 1, never 2.
Where eigenvalues lead next
Eigenvalues are not a one-week topic. They come back in diagonalisation and matrix powers, in systems of differential equations (the eigenvalues of the coefficient matrix decide growth, decay or oscillation), in vibration and stability problems in engineering, in Markov chains, where the long-run state is the eigenvector for λ = 1 (our guide to matrices and Markov chains shows that idea at school level), and in principal component analysis and the singular value decomposition.
If the definitions around this topic (span, basis, eigenspace) still feel slippery, read why linear algebra feels hard once vector spaces arrive. If you are planning which course comes when, see how the US college sequence fits calculus and linear algebra together. The repeated-root and complex-root cases above have exact twins in second order linear differential equations, and Indian B.Tech students meet the same material early, as our piece on moving from JEE to IIT first-year maths explains.
How Maths Bodhi helps
Eigenvalue questions are mostly about method and checks. In sessions we work through your own course problems, ask you to explain each step aloud, and review your attempts so the same slip does not repeat in the exam. We explain and practise; we don't complete graded work. Our university maths support covers the rest of a first linear algebra course too.
Home tuition is available in Gurgaon (Gurugram). Students elsewhere in India and abroad learn with us online. You can book a free demo class.
Written and checked by the Maths Bodhi team (November 2026).
Frequently asked questions
- How do you find the eigenvalues of a 2×2 matrix quickly?
- Solve λ² − (trace)λ + det = 0; for [[4, 1], [2, 3]] that is λ² − 7λ + 10 = 0, so λ = 2 and 5.
- How do you find eigenvectors once you have the eigenvalues?
- For each λ, row reduce A − λI and solve (A − λI)v = 0; any nonzero solution is an eigenvector.
- Can an eigenvalue be zero?
- Yes, exactly when det A = 0; the eigenvectors are then the nonzero vectors in the null space of A.
- Can an eigenvector be the zero vector?
- No, eigenvectors are nonzero by definition, although the zero vector belongs to every eigenspace.
- What is the difference between algebraic and geometric multiplicity?
- Algebraic is how often λ repeats as a root; geometric is the dimension of its eigenspace, n − rank(A − λI).
- Why does a real matrix sometimes have complex eigenvalues?
- When the characteristic polynomial has no real roots, as for a rotation-like matrix; they come in conjugate pairs.
