How to Find Eigenvectors of a 3×3 Matrix
Finding eigenvectors of a 3×3 matrix is a fundamental skill in linear algebra that appears in physics, engineering, computer graphics, and data science. Eigenvectors reveal the directions in which a linear transformation acts by simple scaling, and the associated eigenvalues tell you how much scaling occurs. Now, mastering this process not only strengthens your mathematical toolkit but also prepares you for more advanced topics such as diagonalization, stability analysis, and principal component analysis. The following guide walks you through the theory, the step‑by‑step procedure, a worked example, and practical tips to avoid common mistakes Took long enough..
This is the bit that actually matters in practice.
1. Understanding Eigenvalues and Eigenvectors
Before diving into calculations, it helps to clarify what we are looking for Small thing, real impact..
- Eigenvalue (λ) – a scalar such that when the matrix A multiplies a non‑zero vector v, the result is the same vector scaled by λ:
A v = λv. - Eigenvector (v) – the non‑zero vector that satisfies the equation above for a given λ.
For a 3×3 matrix A, there can be up to three distinct eigenvalues (real or complex), each with its own eigenvector (or a whole eigenspace if the eigenvalue is repeated). The eigenvectors are found by solving the homogeneous system
[ (\mathbf{A} - \lambda \mathbf{I})\mathbf{v} = \mathbf{0}, ]
where I is the 3×3 identity matrix. Non‑trivial solutions exist only when the determinant of ((\mathbf{A} - \lambda \mathbf{I})) equals zero, leading to the characteristic polynomial Simple as that..
2. Step‑by‑Step Procedure
Below is a reliable workflow you can follow for any 3×3 matrix.
2.1 Compute the Characteristic Polynomial
- Form the matrix (\mathbf{A} - \lambda \mathbf{I}) by subtracting λ from each diagonal entry of A.
- Calculate its determinant, (\det(\mathbf{A} - \lambda \mathbf{I})).
- Expand the determinant to obtain a cubic polynomial in λ:
[ p(\lambda) = -\lambda^{3} + a_{2}\lambda^{2} + a_{1}\lambda + a_{0}. ]
(The sign convention may vary; the important part is setting the polynomial equal to zero.)
2.2 Find the Eigenvalues
Solve (p(\lambda)=0). For a cubic, you can:
- Use the Rational Root Theorem to test simple integer candidates.
- Apply synthetic division to reduce the polynomial to a quadratic, then solve with the quadratic formula.
- If the roots are not rational, resort to numerical methods (Newton‑Raphson) or a calculator/computer algebra system.
Record each eigenvalue λ₁, λ₂, λ₃ (they may be repeated or complex) That's the whole idea..
2.3 Solve for Eigenvectors
For each eigenvalue λᵢ:
- Build the matrix (\mathbf{M}_i = \mathbf{A} - \lambda_i \mathbf{I}).
- Set up the homogeneous linear system (\mathbf{M}_i \mathbf{v} = \mathbf{0}).
- Perform Gaussian elimination (row‑reduce) to obtain the row‑echelon form.
- Identify the free variables; express the basic variables in terms of them.
- Write the general solution as a span of one or more basis vectors. Each basis vector is an eigenvector associated with λᵢ.
- Optionally, normalize the eigenvector (divide by its length) for convenience.
If an eigenvalue has algebraic multiplicity > 1, you may obtain more than one linearly independent eigenvector; otherwise, you will get a single direction (up to scaling).
2.4 Verify Your Results
Multiply A by each candidate eigenvector and check whether the output equals λ times the same vector. This quick verification catches arithmetic slips.
3. Worked Example
Let’s apply the procedure to the matrix
[ \mathbf{A}= \begin{pmatrix} 4 & 1 & 2\ 0 & 3 & -1\ 0 & 0 & 2 \end{pmatrix}. ]
3.1 Characteristic Polynomial
[ \mathbf{A}-\lambda\mathbf{I}= \begin{pmatrix} 4-\lambda & 1 & 2\ 0 & 3-\lambda & -1\ 0 & 0 & 2-\lambda \end{pmatrix}. ]
Because the matrix is upper‑triangular, its determinant is the product of the diagonal entries:
[ \det(\mathbf{A}-\lambda\mathbf{I}) = (4-\lambda)(3-\lambda)(2-\lambda). ]
Expanding gives
[ p(\lambda)= -\lambda^{3}+9\lambda^{2}-26\lambda+24. ]
3.2 Eigenvalues
Set (p(\lambda)=0):
[ (4-\lambda)(3-\lambda)(2-\lambda)=0 ;\Longrightarrow; \lambda_{1}=4,; \lambda_{2}=3,; \lambda_{3}=2. ]
All eigenvalues are real and distinct.
3.3 Eigenvectors
For λ = 4:
[ \mathbf{M}_1 = \begin{pmatrix} 0 & 1 & 2\ 0 & -1 & -1\ 0 & 0 & -2 \end{pmatrix} \rightarrow \begin{pmatrix} 0 & 1 & 2\ 0 & 0 & 1\ 0 & 0 & 0 \end{pmatrix} \text{ (after row operations)}. ]
The system yields (v_2 + 2v_3 = 0) and (v_3 = 0) ⇒ (v_2 = 0). (v_1) is free.
Choose (v_1 = 1): eigenvector (\mathbf{v}_1 = (1,0,0)^T).
For λ = 3:
[ \mathbf{M}_2 = \begin{pmatrix} 1 & 1 & 2\ 0 & 0 & -1\ 0 & 0 & -1 \end{pmatrix} \rightarrow \begin{pmatrix} 1 & 1 & 2\ 0 & 0 & 1\ 0 & 0 & 0 \end{pmatrix}. ]
From the second row: (v_3 = 0). First row: (v_1 + v_2 + 2v_3 = 0) ⇒ (v_1 = -v_2).
Let (v_2 = 1): eigenvector (\mathbf{v}_2 = (-1,1,0)^T).
**For λ =
For λ = 2:
[ \mathbf{M}_3 = \begin{pmatrix} 2 & 1 & 2\ 0 & 1 & -1\ 0 & 0 & 0 \end{pmatrix}. ]
Row‑reducing (subtract the second row from the first, then scale) gives
[ \begin{pmatrix} 2 & 0 & 3\ 0 & 1 & -1\ 0 & 0 & 0 \end{pmatrix} ;\xrightarrow{;R_1\leftarrow \tfrac12 R_1;}; \begin{pmatrix} 1 & 0 & \tfrac32\ 0 & 1 & -1\ 0 & 0 & 0 \end{pmatrix}. ]
Hence the system (\mathbf{M}_3\mathbf{v}=0) yields
[ v_1 + \tfrac32 v_3 = 0 \quad\Longrightarrow\quad v_1 = -\tfrac32 v_3, \qquad v_2 - v_3 = 0 \quad\Longrightarrow\quad v_2 = v_3, ]
with (v_3) free. Choosing (v_3 = 2) (to avoid fractions) gives
[ \mathbf{v}_3 = \begin{pmatrix}-3 \ 2 \ 2\end{pmatrix}. ]
Any non‑zero scalar multiple of (\mathbf{v}_3) is also an eigenvector for (\lambda = 2).
Verification (quick check)
[ A\mathbf{v}_1 = \begin{pmatrix} 4 & 1 & 2\ 0 & 3 & -1\ 0 & 0 & 2 \end{pmatrix} !\begin{pmatrix}1\0\0\end{pmatrix}
\begin{pmatrix}4\0\0\end{pmatrix}=4\mathbf{v}_1, ]
[ A\mathbf{v}_2 = \begin{pmatrix} 4 & 1 & 2\ 0 & 3 & -1\ 0 & 0 & 2 \end{pmatrix} !\begin{pmatrix}-1\1\0\end{pmatrix}
\begin{pmatrix}-3\3\0\end{pmatrix}=3\mathbf{v}_2, ]
[ A\mathbf{v}_3 = \begin{pmatrix} 4 & 1 & 2\ 0 & 3 & -1\ 0 & 0 & 2 \end{pmatrix} !\begin{pmatrix}-3\2\2\end{pmatrix}
\begin{pmatrix}-6\6\4\end{pmatrix}=2\mathbf{v}_3. ]
Each product equals the corresponding eigenvalue times the vector, confirming the calculations.
Conclusion
Finding eigenvalues and eigenvectors reduces to solving the characteristic polynomial and then a series of homogeneous linear systems. Here's the thing — for triangular (or block‑triangular) matrices the characteristic polynomial factors immediately from the diagonal entries, often yielding distinct real eigenvalues as in the example. When eigenvalues are repeated, the eigenvector computation may reveal a deficiency, prompting the use of generalized eigenvectors or Jordan chains—a topic beyond the present scope but essential for a full spectral analysis. The verification step, though simple, guards against algebraic slips and builds confidence in the results. With practice, the outlined procedure becomes a reliable toolkit for analyzing linear transformations in any finite‑dimensional vector space Worth keeping that in mind..
Short version: it depends. Long version — keep reading.
