Skip to content

6.3-final-2013

Final Exam, 2013 · Section 6.3: Symmetric Positive Definite Matrices Source: finalexam_2.pdf, pp. 89-90

Problem

2 (20 pts.) The matrix

\[ A = \begin{bmatrix} z & z & z \\ z & z+c & z-c \\ z & z-c & z+c \end{bmatrix}. \]

(a) (4 pts.) Under what conditions on \(z\) and \(c\) would \(A\) be positive semidefinite?

(b) (4 pts.) Under what conditions on \(z\) and \(c\) would \(A\) be Markov?

(c) (4 pts.) Under what conditions on \(z\) and \(c\) would the first column of \(A\) be a free column?

(d) (4 pts.) Under what conditions on \(z\) and \(c\) does \(A\) have rank \(r = 2\)?

(e) (4 pts.) Assuming \(A\) has rank 2, for which \(b\) in \(\mathbb{R}^3\) does the equation \(Ax = b\) have a solution?

Solution Elimination on \(A\): subtract row 1 from rows 2 and 3, then add the new row 2 to the new row 3: \[ A \Rightarrow \begin{bmatrix} z & z & z \\ 0 & c & -c \\ 0 & -c & c \end{bmatrix} \Rightarrow \begin{bmatrix} z & z & z \\ 0 & c & -c \\ 0 & 0 & 0 \end{bmatrix}. \] The pivots are \(z\) and \(c\). (a) \(A\) is symmetric, so it is positive semidefinite exactly when both pivots are nonnegative: \[ z \ge 0 \quad \text{and} \quad c \ge 0. \] (b) A Markov matrix has columns summing to 1 with all entries nonnegative. All three columns sum to \(3z\), so \(z = 1/3\). The entries of the last two columns are \(z + c\) and \(z - c\), so we need \[ z + c \ge 0 \quad \text{and} \quad z - c \ge 0, \] i.e. with \(z = 1/3\), \[ -\frac{1}{3} \le c \le \frac{1}{3}. \] (c) A column is free when it is a linear combination of the columns before it. For the first column there are no columns before it, so this happens exactly when the column is zero: \[ z = 0. \] (d) As the elimination shows, the rank is always at most 2. The rank is exactly 2 when both pivots are nonzero: \[ z \neq 0 \quad \text{and} \quad c \neq 0. \] (If \(z = 0\) or \(c = 0\) the rank drops to 1, since all columns become multiples of one vector.) (e) Apply the same elimination to the augmented system. The last equation becomes \[ 0 = b_2 + b_3 - 2b_1, \] so (when \(\mathrm{rank}(A) = 2\)) the equation \(Ax = b\) has a solution if and only if \[ b_2 + b_3 = 2b_1. \]

Keywords: positive semidefinite, Markov matrix, rank, elimination