Computing Eigenvalues

Eigenvalues are special scalars that represent the factors by which eigenvectors are scaled during a linear transformation. Understanding how to compute them is fundamental to analyzing the behavior of matrices in various applications, from principal component analysis to quantum mechanics. This lesson focuses on the algebraic method: deriving and solving the characteristic equation for a given matrix.

The Characteristic Equation

The core idea behind finding eigenvalues stems from the definition $Av = \lambda v$ , where $A$ is a matrix, $v$ is an eigenvector, and $\lambda$ is an eigenvalue. Rearranging this equation gives $Av - \lambda v = 0$ . To combine $A$ and $\lambda$ , we introduce the identity matrix $I$ , allowing us to write $Av - \lambda Iv = 0$ , which simplifies to $(A - \lambda I)v = 0$ . For this homogeneous system to have non-trivial solutions (i.e., $v \neq 0$ ), the matrix $(A - \lambda I)$ must be singular, meaning its determinant must be zero.

📐 Characteristic Equation

The characteristic equation for a square matrix $A$ is given by:

det(A - \lambda I) = 0

Here, $det$ denotes the determinant, $A$ is the matrix, $\lambda$ is the eigenvalue (a scalar variable), and $I$ is the identity matrix of the same dimension as $A$ . Solving this polynomial equation for $\lambda$ yields the eigenvalues.

Finding Eigenvalues for 2x2 Matrices

For a 2x2 matrix, the characteristic equation simplifies to a quadratic polynomial, which is straightforward to solve. We first construct the matrix $(A - \lambda I)$ , then compute its determinant. Setting this determinant to zero gives us a quadratic equation in terms of $\lambda$ . The roots of this quadratic equation are the eigenvalues of the matrix.

pythonComputing Eigenvalues for a 2x2 Matrix

Check Your Understanding

What is the degree of the characteristic polynomial for an

n \times n

matrix?

Real and Distinct Eigenvalues

The simplest case occurs when the characteristic polynomial yields distinct real roots. Each unique real root corresponds to a unique eigenvalue, and for each eigenvalue, there will be at least one corresponding eigenvector. These eigenvalues represent distinct scaling factors along different directions in the vector space, making the transformation behavior clear.

pythonMatrix with Distinct Real Eigenvalues

Repeated Eigenvalues (Algebraic Multiplicity)

Sometimes, the characteristic polynomial has repeated roots. If an eigenvalue $\lambda_0$ appears $k$ times as a root of the characteristic polynomial, it has an algebraic multiplicity of $k$ . This means the matrix scales along certain directions by the same factor multiple times. While the algebraic multiplicity tells us how many times an eigenvalue is a root, it does not directly tell us how many linearly independent eigenvectors correspond to it (that's the geometric multiplicity, a separate concept).

pythonMatrix with Repeated Eigenvalues

Check Your Understanding

If a

3 \times 3

matrix has eigenvalues

[5, 5, 2]

, what is the algebraic multiplicity of

\lambda = 5

Complex Eigenvalues

Not all matrices have real eigenvalues. For real matrices, if complex eigenvalues exist, they always appear in conjugate pairs. This means if $a + bi$ is an eigenvalue, then its conjugate $a - bi$ must also be an eigenvalue. Complex eigenvalues often arise in transformations that involve rotation, where vectors are rotated rather than simply scaled along their original direction.

pythonMatrix with Complex Eigenvalues

Generalizing to Larger Matrices

The process of computing eigenvalues by solving $det(A - \lambda I) = 0$ extends to $n \times n$ matrices. For a $3 \times 3$ matrix, the characteristic equation will be a cubic polynomial. For larger matrices, the characteristic polynomial will be of degree $n$ . While theoretically the same, finding the roots of high-degree polynomials analytically becomes increasingly complex and often impossible for $n \geq 5$ using general formulas.

⚠️ Computational Complexity

For matrices larger than $3 \times 3$ , solving the characteristic polynomial by hand is impractical. In numerical linear algebra, specialized iterative algorithms (like the QR algorithm) are used to approximate eigenvalues efficiently. These methods avoid direct polynomial root-finding altogether.

pythonEigenvalues for a 3x3 Matrix

Try It Yourself

Modify the matrix H below so that it has a repeated eigenvalue of 3. You should only change the values in the matrix, not the np.linalg.eigvals call.

python

Types of Eigenvalues

Type	Characteristic Polynomial Roots	Geometric Interpretation	Example Matrix
Real and Distinct	Unique real roots	Clear, distinct scaling along specific directions	$\begin{pmatrix} 4 & 2 \\ 1 & 3 \end{pmatrix}$
Repeated (Real)	Real roots with algebraic multiplicity > 1	Multiple scaling factors are identical; can lead to fewer independent eigenvectors	$\begin{pmatrix} 2 & 1 \\ 0 & 2 \end{pmatrix}$
Complex (Conjugate Pairs)	Complex conjugate roots	Involves rotation and scaling; no real eigenvectors	$\begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix}$

Eigenvalues can be real and distinct, repeated, or complex, each indicating different transformation behaviors.

Key Takeaways

Eigenvalues are found by solving the characteristic equation: $det(A - \lambda I) = 0$ .
For an $n \times n$ matrix, the characteristic equation is an $n$ -th degree polynomial in $\lambda$ .
Real and distinct eigenvalues indicate unique scaling factors along different eigenvector directions.
Repeated eigenvalues have an algebraic multiplicity greater than one, meaning the same scaling factor applies multiple times.
Complex eigenvalues for real matrices always appear in conjugate pairs and often signify rotational components in the transformation.
While manual computation is feasible for $2 \times 2$ and $3 \times 3$ matrices, numerical methods are essential for larger matrices.

← All lessons in Linear Algebra: Eigenvalues