Visual Tools
Calculators
Tables
Mathematical Keyboard
Converters
Other Tools

How to Find Eigenvectors

An eigenvector for λ\lambda is a nonzero solution of (A−λI)v=0(A - \lambda I)\mathbf{v} = \mathbf{0}. Once the eigenvalues are known, each one gives a homogeneous system; row reduce it and read a solution from the free variable.

Recognize the task

You have the eigenvalues, or can find them, and the question asks for eigenvectors, eigen-directions or invariant lines: "find an eigenvector for λ=6\lambda = 6", "find all eigenvectors of (5412)\begin{pmatrix} 5 & 4 \\ 1 & 2 \end{pmatrix}". Test: A−λIA - \lambda I must be singular; if row reduction leaves only v=0\mathbf{v} = \mathbf{0}, the λ\lambda is wrong.
The procedure is Finding Eigenvectors After Finding Eigenvalues; the eigenvectors for one λ\lambda, together with 0\mathbf{0}, form its eigenspace.

Ways to do it

Row reduce A − λIany size
(A−λI)v=0(A - \lambda I)\mathbf{v} = \mathbf{0}
2×2 shortcut2×2, b ≠ 0
A=(abcd) ⇒ v=(bλ−a)A = \begin{pmatrix} a & b \\ c & d \end{pmatrix} \ \Rightarrow \ \mathbf{v} = \begin{pmatrix} b \\ \lambda - a \end{pmatrix}
The first row of A−λIA - \lambda I is (a−λ, b)(a - \lambda, \ b), and (b, λ−a)(b, \ \lambda - a) makes it zero; the second row follows because det⁡(A−λI)=0\det(A - \lambda I) = 0.
3×3 cross product3×3, one free variable
v=r1×r2\mathbf{v} = \mathbf{r}_1 \times \mathbf{r}_2
When A−λIA - \lambda I has rank 22, the cross product of two independent rows is perpendicular to every row, so it solves the system.

Steps

StepOn A=(5412)A = \begin{pmatrix} 5 & 4 \\ 1 & 2 \end{pmatrix}, λ=6\lambda = 6
1Get the eigenvalues first. Use finding eigenvalues; then work on one λ\lambda at a time.λ=6\lambda = 6 and λ=1\lambda = 1
2Form A−λIA - \lambda I. Subtract λ\lambda from the diagonal only.A−6I=(−141−4)A - 6I = \begin{pmatrix} -1 & 4 \\ 1 & -4 \end{pmatrix}
3Row reduce. At least one row must become zero. If none does, recheck λ\lambda.(1−400)\begin{pmatrix} 1 & -4 \\ 0 & 0 \end{pmatrix}
4Solve with a free variable. Set each free variable to a parameter and solve for the pivot variables; pick a value that clears fractions.v1=4v2v_1 = 4v_2; v2=1v_2 = 1 gives v=(4,1)\mathbf{v} = (4, 1)
5Check Av=λvA\mathbf{v} = \lambda\mathbf{v}. One matrix-vector product catches almost every slip. Repeat for the other eigenvalues.A(4,1)=(24,6)=6(4,1)A(4, 1) = (24, 6) = 6(4, 1) ✓; for λ=1\lambda = 1: (1,−1)(1, -1)

Worked examples

Example 1 · a symmetric matrix

2×2
A=(1331)A = \begin{pmatrix} 1 & 3 \\ 3 & 1 \end{pmatrix}, λ=4\lambda = 4: A−4I=(−333−3)A - 4I = \begin{pmatrix} -3 & 3 \\ 3 & -3 \end{pmatrix}
subtract 44 on the diagonal
−3v1+3v2=0-3v_1 + 3v_2 = 0, so v=(1,1)\mathbf{v} = (1, 1)
second row is a multiple
λ=−2\lambda = -2: A+2I=(3333)A + 2I = \begin{pmatrix} 3 & 3 \\ 3 & 3 \end{pmatrix}, so v=(1,−1)\mathbf{v} = (1, -1)
check: A(1,−1)=(−2,2)A(1, -1) = (-2, 2)

Example 2 · one eigenvalue of a 3×3

3×3
A=(401−210−201)A = \begin{pmatrix} 4 & 0 & 1 \\ -2 & 1 & 0 \\ -2 & 0 & 1 \end{pmatrix}, λ=2\lambda = 2
eigenvalues are 1,2,31, 2, 3
A−2I=(201−2−10−20−1)A - 2I = \begin{pmatrix} 2 & 0 & 1 \\ -2 & -1 & 0 \\ -2 & 0 & -1 \end{pmatrix}: 2v1+v3=02v_1 + v_3 = 0, −2v1−v2=0-2v_1 - v_2 = 0
row 3 repeats row 1
v1=1v_1 = 1: v=(1,−2,−2)\mathbf{v} = (1, -2, -2); Av=(2,−4,−4)A\mathbf{v} = (2, -4, -4) ✓
one free variable

Example 3 · the other two by cross product

3×3 shortcut
λ=1\lambda = 1: A−IA - I has rows (3,0,1)(3, 0, 1), (−2,0,0)(-2, 0, 0); their cross product is (0,−2,0)(0, -2, 0)
so v=(0,1,0)\mathbf{v} = (0, 1, 0)
λ=3\lambda = 3: (1,0,1)×(−2,−2,0)=(2,−2,−2)(1, 0, 1) \times (-2, -2, 0) = (2, -2, -2)
rows 1 and 2 of A−3IA - 3I
v=(−1,1,1)\mathbf{v} = (-1, 1, 1); Av=(−3,3,3)=3vA\mathbf{v} = (-3, 3, 3) = 3\mathbf{v} ✓
scaled by −1/2-1/2
Step through the row reduction in the eigenvalues and eigenvectors visualizer.

See it happen

Subtracting λI\lambda I turns the question "which vectors does AA only stretch by λ\lambda?" into "which vectors does A−λIA - \lambda I send to zero?" That second set is a null space, a line or plane through the origin, so eigenvectors never come alone: every nonzero multiple of an eigenvector is another one. Row reduction finds the line; choosing v2=1v_2 = 1 just picks one point on it. See those lines in the eigenvectors 2D visualizer.

Common mistakes

Taking a row as the eigenvector

✗ the reduced row is (1,−4)(1, -4), so v=(1,−4)\mathbf{v} = (1, -4)
✓ solve the row equation v1−4v2=0v_1 - 4v_2 = 0: v=(4,1)\mathbf{v} = (4, 1)
A(1,−4)=(−11,−7)A(1, -4) = (-11, -7), which is not a multiple of (1,−4)(1, -4).

Accepting the zero vector

✗ v=0\mathbf{v} = \mathbf{0} solves (A−λI)v=0(A - \lambda I)\mathbf{v} = \mathbf{0}
✓ eigenvectors are nonzero by definition; give a nonzero solution

Subtracting λ from every entry

✗ A−6=(−1−2−5−4)A - 6 = \begin{pmatrix} -1 & -2 \\ -5 & -4 \end{pmatrix}
✓ A−6I=(−141−4)A - 6I = \begin{pmatrix} -1 & 4 \\ 1 & -4 \end{pmatrix}

Thinking a different answer is wrong

✗ the book says (8,2)(8, 2), so (4,1)(4, 1) must be wrong
✓ any nonzero multiple is an eigenvector: (8,2)=2(4,1)(8, 2) = 2(4, 1)