Skip to content

6.1-final-2003

Final Exam, 2003 · Section 6.1: Introduction to Eigenvalues : Ax = λx Source: finalexam_1.pdf, pp. 105-106

Problem

7. Consider the following matrix \(A\):

\[ A = \begin{bmatrix} 0.5 & 0.4 & 0.2 \\ 0.4 & 0.5 & 0.2 \\ 0.1 & 0.1 & 0.6 \end{bmatrix}. \]

(c) Can you immediately tell one of the eigenvalues of \(A\) (without computing them)? Explain.

(d) Compute the determinant of \(A\).

(e) Find the eigenvalues of \(A\) and the corresponding eigenvectors. (Check your answer.)

(f) Two out of the 3 eigenvectors of \(A\) should be orthogonal to \(\begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix}\). How could you have explained this before computing the eigenvectors?

(g) Write an exact expression for \(A^{100}\).

Solution (c) Yes: \(A\) is a Markov matrix — each *column* sums to 1 (column-stochastic). Therefore \((1,1,1)^T A = (1,1,1)^T\), i.e. \((1,1,1)\) is a left eigenvector of \(A\) (an eigenvector of \(A^T\)) with eigenvalue 1. Since \(\det(A^T - I) = \det(A - I)\), \(1\) is also an eigenvalue of \(A\). (d) Expanding along the first row: \[ \det A = 0.5(0.30 - 0.02) - 0.4(0.24 - 0.02) + 0.2(0.04 - 0.05) = 0.14 - 0.088 - 0.002 = 0.05 = \frac{1}{20}. \] (Consistency check: \(\det A = 1 \cdot \lambda_2 \lambda_3\).) (e) The trace is \(0.5 + 0.5 + 0.6 = 1.6\), so with \(\lambda_1 = 1\): \(\lambda_2 + \lambda_3 = 0.6\) and \(\lambda_2\lambda_3 = 0.05\), giving \(\lambda^2 - 0.6\lambda + 0.05 = 0\), i.e. \(\lambda_2 = 0.5\), \(\lambda_3 = 0.1\). Eigenvectors: for \(\lambda = 1\), solving \((A-I)x = 0\) gives \((2, 2, 1)\); for \(\lambda = 0.5\), \(A - 0.5I\) has rows \((0, 0.4, 0.2)\), \((0.4, 0, 0.2)\), \((0.1, 0.1, 0.1)\), giving \(2y + z = 0\) and \(2x + z = 0\), so \((1, 1, -2)\); for \(\lambda = 0.1\), \(A - 0.1I\) gives \(2x + 2y + z = 0\) and \(x + y + 5z = 0\), so \(z = 0\), \(x + y = 0\), i.e. \((1, -1, 0)\). Check: \[ A\begin{bmatrix} 2 \\ 2 \\ 1 \end{bmatrix} = \begin{bmatrix} 2 \\ 2 \\ 1 \end{bmatrix},\quad A\begin{bmatrix} 1 \\ 1 \\ -2 \end{bmatrix} = \begin{bmatrix} 0.5 \\ 0.5 \\ -1 \end{bmatrix},\quad A\begin{bmatrix} 1 \\ -1 \\ 0 \end{bmatrix} = \begin{bmatrix} 0.1 \\ -0.1 \\ 0 \end{bmatrix}. \] (f) \((1,1,1)\) is a left eigenvector of \(A\) (from (c): the column sums are 1). For a diagonalizable matrix, right eigenvectors for distinct eigenvalues are orthogonal to the corresponding left eigenvectors: if \(w^T A = \lambda w^T\) and \(A v = \mu v\) with \(\lambda \neq \mu\), then \(w^T v = 0\). Here the right eigenvectors for \(\lambda = 0.5\) and \(\lambda = 0.1\) must each be orthogonal to the left eigenvector \((1,1,1)\) for \(\lambda = 1\). (Indeed \((1,1,-2)\cdot(1,1,1) = 0\) and \((1,-1,0)\cdot(1,1,1) = 0\).) (g) \(A = S\Lambda S^{-1}\) with \[ S = \begin{bmatrix} 2 & 1 & 1 \\ 2 & 1 & -1 \\ 1 & -2 & 0 \end{bmatrix},\quad S^{-1} = \frac{1}{10}\begin{bmatrix} 2 & 2 & 2 \\ 1 & 1 & -4 \\ 5 & -5 & 0 \end{bmatrix},\quad \Lambda = \operatorname{diag}(1,\,0.5,\,0.1). \] Then \(A^{100} = S\Lambda^{100}S^{-1} = \lambda_1^{100} v_1 w_1^T + \lambda_2^{100} v_2 w_2^T + \lambda_3^{100} v_3 w_3^T\), where \(w_i^T\) is the \(i\)-th row of \(S^{-1}\). Since \((0.5)^{100} \approx 7.9\times10^{-31}\) and \((0.1)^{100} = 10^{-100}\), both are negligible, and \[ A^{100} = v_1 w_1^T = \frac{1}{5}\begin{bmatrix} 2 \\ 2 \\ 1 \end{bmatrix}\begin{bmatrix} 1 & 1 & 1 \end{bmatrix} = \frac{1}{5}\begin{bmatrix} 2 & 2 & 2 \\ 2 & 2 & 2 \\ 1 & 1 & 1 \end{bmatrix} = \begin{bmatrix} 0.4 & 0.4 & 0.4 \\ 0.4 & 0.4 & 0.4 \\ 0.2 & 0.2 & 0.2 \end{bmatrix}. \] (The \(\lambda_2^{100}\) and \(\lambda_3^{100}\) terms are astronomically small; the columns of \(A^{100}\) all equal the stationary distribution \((2/5,\,2/5,\,1/5)^T\).)

Keywords: Markov matrix, eigenvalues, eigenvectors, matrix powers