Computing Eigenvalues with the Characteristic Polynomial

This article provides a direct explanation of how the characteristic polynomial is used to find the eigenvalues of a square matrix. By translating a matrix equation into a scalar polynomial equation using determinants, mathematicians can systematically solve for eigenvalues. Below, you will find the theoretical foundation of this process, a step-by-step guide to the computation, and a practical example demonstrating the method in action.

The Fundamental Equation

To understand how the characteristic polynomial works, we must start with the definition of an eigenvalue. For a given \(n \times n\) square matrix \(A\), a scalar \(\lambda\) (lambda) is called an eigenvalue if there exists a non-zero vector \(v\) (the eigenvector) such that:

\(Av = \lambda v\)

This equation states that multiplying the matrix \(A\) by the vector \(v\) has the exact same effect as scaling the vector \(v\) by the number \(\lambda\).

To solve for \(\lambda\), we must first algebraically rearrange the equation to bring all terms to one side. Because we cannot directly subtract a scalar (\(\lambda\)) from a matrix (\(A\)), we multiply \(\lambda\) by the identity matrix \(I\) of the same dimension. This gives us:

\(Av - \lambda Iv = 0\)

Factoring out the vector \(v\) yields the homogeneous system of linear equations:

\((A - \lambda I)v = 0\)

The Role of the Determinant

By definition, an eigenvector \(v\) cannot be the zero vector. We are only interested in "non-trivial" solutions to the equation \((A - \lambda I)v = 0\).

In linear algebra, a system of the form \(Mv = 0\) has a non-zero solution for \(v\) if and only if the matrix \(M\) is singular (non-invertible). A matrix is non-invertible if and only if its determinant is equal to zero.

Applying this rule to our rearranged eigenvalue equation, the matrix \((A - \lambda I)\) must have a determinant of zero for \(\lambda\) to be an eigenvalue. This gives us the characteristic equation:

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

Forming the Characteristic Polynomial

The expression \(\det(A - \lambda I)\) is what we call the characteristic polynomial of the matrix \(A\), often denoted as \(p(\lambda)\).

When you compute the determinant of the matrix \((A - \lambda I)\), the matrix structure disappears, leaving behind a single algebraic polynomial where \(\lambda\) is the variable. For an \(n \times n\) matrix, the characteristic polynomial will always be of degree \(n\). For example, a \(2 \times 2\) matrix produces a quadratic polynomial, and a \(3 \times 3\) matrix produces a cubic polynomial.

Because \(\det(A - \lambda I)\) must equal zero, the eigenvalues of the matrix \(A\) are simply the roots (or zeroes) of the characteristic polynomial.

Step-by-Step Computation Process

To compute the eigenvalues of any square matrix \(A\), follow these four steps:

  1. Subtract \(\lambda\) from the main diagonal: Construct the matrix \((A - \lambda I)\) by subtracting the variable \(\lambda\) from the diagonal entries of \(A\), leaving all other entries unchanged.
  2. Compute the determinant: Calculate the determinant of the resulting \((A - \lambda I)\) matrix. Expand and simplify the algebraic terms to write the characteristic polynomial in standard form.
  3. Set the polynomial to zero: Form the characteristic equation by setting your simplified polynomial equal to zero.
  4. Solve for the roots: Factor the polynomial or use formulas (like the quadratic formula) to solve for \(\lambda\). Every solution for \(\lambda\) is an eigenvalue of the matrix \(A\).

Practical Example

Let us find the eigenvalues for the following \(2 \times 2\) matrix \(A\):

Row 1: \([4, 1]\)
Row 2: \([2, 3]\)

Step 1: Construct \((A - \lambda I)\)
Subtract \(\lambda\) from the diagonal elements (4 and 3):

Row 1: \([4 - \lambda, 1]\)
Row 2: \([2, 3 - \lambda]\)

Step 2: Compute the determinant to find the characteristic polynomial
For a \(2 \times 2\) matrix, the determinant is found using the formula \((ad - bc)\):

\(p(\lambda) = (4 - \lambda)(3 - \lambda) - (1)(2)\)

Expand the terms:

\(p(\lambda) = 12 - 4\lambda - 3\lambda + \lambda^2 - 2\)
\(p(\lambda) = \lambda^2 - 7\lambda + 10\)

This resulting expression, \(\lambda^2 - 7\lambda + 10\), is the characteristic polynomial.

Step 3: Set the polynomial to zero
\(\lambda^2 - 7\lambda + 10 = 0\)

Step 4: Solve for the roots
Factor the quadratic equation:

\((\lambda - 5)(\lambda - 2) = 0\)

Solving for \(\lambda\) gives the roots \(\lambda = 5\) and \(\lambda = 2\). Therefore, the eigenvalues of the matrix \(A\) are 5 and 2.

Limitations for Larger Matrices

While the characteristic polynomial is the definitive tool for finding eigenvalues of \(2 \times 2\), \(3 \times 3\), and simple \(4 \times 4\) matrices, it becomes impractical for very large matrices. According to the Abel-Ruffini theorem, there is no general algebraic formula to find the exact roots of polynomials of degree 5 or higher. Consequently, for matrices of size \(5 \times 5\) and larger, mathematicians and computer scientists bypass the characteristic polynomial and instead use iterative numerical algorithms, such as the QR algorithm, to approximate the eigenvalues.