Exercises: Eigenvalues and Diagonalization
exercises, eigenvalues, eigenvectors, diagonalization
1 Scope and Goals
This set trains four things:
- finding eigenvalues and eigenvectors in small examples
- checking whether diagonalization is available
- reading matrix powers through eigen-modes
- preparing for spectral applications and SVD
2 Prerequisites
- matrix multiplication
- basis and independence
- diagonal matrices and powers
3 Warm-Up Problems
Find the eigenvalues of
\[ \begin{bmatrix} 2 & 0 \\ 0 & 3 \end{bmatrix}. \]
Find one eigenvector of
\[ \begin{bmatrix} 2 & 1 \\ 0 & 3 \end{bmatrix} \]
for each eigenvalue.
4 Core Problems
Diagonalize
\[ A = \begin{bmatrix} 4 & 0 \\ 0 & 1 \end{bmatrix} \]
and compute \(A^5\).
Let
\[ P = \begin{bmatrix} 0.8 & 0.2 \\ 0.2 & 0.8 \end{bmatrix}. \]
Find its eigenvalues and explain which mode decays.
Explain why a matrix with \(n\) distinct eigenvalues is diagonalizable.
5 Proof Problems
- Prove that if \(Av = \lambda v\) and \(c \neq 0\), then \(cv\) is also an eigenvector with eigenvalue \(\lambda\).
- Show that if \(A = S \Lambda S^{-1}\), then \(A^k = S \Lambda^k S^{-1}\) for every positive integer \(k\).
6 Computational or Applied Problems
- Open Computation Lab: Matrix Powers and Spectral Modes. Change the mixing parameter and describe how the second eigenvalue controls the rate of convergence.
- In software, compute the eigenvalues of a symmetric matrix and compare them with the singular values of the same matrix when it is positive semidefinite.
7 Hints
- For diagonal matrices, the eigenvalues are already visible.
- For matrix powers, diagonalize first and then power the diagonal entries.
- Distinct eigenvalues give independent eigenvectors.
8 Full Solutions
8.1 Solution 1
The eigenvalues are \(2\) and \(3\).
8.2 Solution 2
For \(\lambda=2\), solve
\[ \begin{bmatrix} 0 & 1 \\ 0 & 1 \end{bmatrix} v = 0, \]
so one eigenvector is \((1,0)^\top\).
For \(\lambda=3\), solve
\[ \begin{bmatrix} -1 & 1 \\ 0 & 0 \end{bmatrix} v = 0, \]
so one eigenvector is \((1,1)^\top\).
8.3 Solution 3
The matrix is already diagonal, so
\[ A = I \begin{bmatrix} 4 & 0 \\ 0 & 1 \end{bmatrix} I^{-1}. \]
Therefore
\[ A^5 = \begin{bmatrix} 4^5 & 0 \\ 0 & 1 \end{bmatrix} = \begin{bmatrix} 1024 & 0 \\ 0 & 1 \end{bmatrix}. \]
8.4 Solution 4
The eigenvalues are \(1\) and \(0.6\).
The consensus mode \((1,1)^\top\) persists, while the disagreement mode \((1,-1)^\top\) decays because \(0.6^k \to 0\).
8.5 Solution 5
Distinct eigenvalues give linearly independent eigenvectors. So an \(n \times n\) matrix with \(n\) distinct eigenvalues has \(n\) independent eigenvectors, which is exactly what diagonalization needs.
9 Common Errors
- treating the zero vector as an eigenvector
- assuming repeated eigenvalues automatically prevent diagonalization
- forgetting that eigenvectors are determined only up to nonzero scaling
10 What To Do Next
Open SVD and Low-Rank Approximation next. That page brings spectral thinking back through the symmetric matrix \(A^\top A\) and turns it into a stable approximation tool.
11 Sources and Further Reading
- MIT 18.06: Lecture 21, Eigenvalues and Eigenvectors -
First pass- official source for computational practice on eigenpairs. Checked2026-04-24. - MIT 18.06SC: Lecture 22, Diagonalization and Powers of A -
Second pass- official source for repeated powers. Checked2026-04-24. - A Comprehensive Survey on Spectral Clustering with Graph Structure Learning -
Paper bridge- useful later for seeing eigen-modes inside graph methods. Checked2026-04-24.