6.3-final-2013-b¶
Final Exam, 2013 · Section 6.3: Symmetric Positive Definite Matrices Source: finalexam_2.pdf, pp. 93-95
Problem¶
4 (23 pts.) The \(8 \times 8\) matrix
is symmetric and satisfies \(A^2 = 8I\). The diagonal elements add to 0.
(a) (3 pts.) What are the eigenvalues of \(A\) and the determinant of \(A\)?
(b) (2 pts.) What is the condition number \(\kappa(A) = \sigma_1(A)/\sigma_8(A)\)? (Hint: \(\kappa \ge 1\) always.)
(c) (3 pts.) Can \(A\) have any Jordan blocks of size greater than 1? Explain briefly.
(d) (2 pts.) What is the rank of the matrix consisting of the first four columns of \(A\)?
(e) (3 pts.) The subspace of \(\mathbb{R}^8\) spanned by the first four columns of \(A\) is the ________ of the last four columns of \(A\). Fill in the blank with the best two words and explain briefly.
(f) (5 pts.) Let \(P\) project onto the first column of \(A\). \(P\) is an \(8 \times 8\) matrix all of whose 64 entries are the same number. This number is ______.
(g) (5 pts.) Projection onto the last seven columns of \(A\) (we are dropping only the first column) gives an \(8 \times 8\) projection matrix whose 8 diagonal entries are the same number. This number is ______.
Solution
(a) Since \(A^2 = 8I\), every eigenvalue satisfies \(\lambda^2 = 8\), so \(\lambda = \pm \sqrt{8}\). The trace of \(A\) is 0, so the eigenvalues sum to 0: there must be four eigenvalues \(\sqrt{8}\) and four eigenvalues \(-\sqrt{8}\). The determinant is the product of the eigenvalues: \[ \det A = (\sqrt{8})^4 (-\sqrt{8})^4 = 8^4 = 4096. \] (b) Since \(A\) is symmetric, the singular values are the absolute values of the eigenvalues: all eight are \(\sqrt{8}\). Therefore \[ \kappa(A) = \frac{\sqrt{8}}{\sqrt{8}} = 1. \] (c) No. Symmetric matrices are always diagonalizable, so every Jordan block has size 1. (d) \(\det A \neq 0\), so all eight columns of \(A\) are linearly independent; in particular the first four columns are independent, and the matrix they form has rank 4. (e) The best words are "orthogonal complement." All columns of \(A\) are pairwise orthogonal (since \(A^T A = 8I\)), so each of the last four columns is orthogonal to each of the first four; the first four span a 4-dimensional space and the last four span another 4-dimensional space, so the two spans are orthogonal complements of each other. (f) The first column of \(A\) is the all-ones vector \(u = (1, \dots, 1)\). The projection onto it is \[ P = \frac{uu^T}{u^T u} = \frac{1}{8} \begin{bmatrix} 1 & \cdots & 1 \\ \vdots & \ddots & \vdots \\ 1 & \cdots & 1 \end{bmatrix}, \] so every entry is the same number \[ \frac{1}{8}. \] (g) The last seven columns are orthogonal to the first column, so their span is the orthogonal complement of \(\mathrm{span}\{u\}\), and the projection onto it is \(I - P\). Its diagonal entries are therefore \[ 1 - \frac{1}{8} = \frac{7}{8}. \]Keywords: symmetric matrix, eigenvalues, condition number, projection matrix