6.2-exam3-2007-johnson¶
Quiz 3, 2007 (johnson) · Section 6.2: Diagonalizing a Matrix Source: exam3_1.pdf, pp. 119-121
Problem¶
2 (30 pts.) Suppose we form a sequence of real numbers \(f_k\) defined by the recurrence \(f_{k+1} = f_k - f_{k-1} + f_{k-2}\), starting with the initial conditions \(f_0 = 2\), \(f_1 = 1\) and \(f_2 = 0\).
(a) Define a 3-component vector \(\vec{g}_k = (f_k, f_{k-1}, f_{k-2})^T\) and a \(3 \times 3\) matrix \(A\) so that the recurrence is \(\vec{g}_{k+1} = A\vec{g}_k\).
(b) If you constructed \(A\) correctly, the three eigenvalues should be 1 and \(\pm i\) [I'm giving you these so you don't have to solve a cubic equation], and the latter two eigenvectors should be \((-1, \pm i, 1)^T\). Check that you have these \(\pm i\) eigenvalues and eigenvectors, and find the \(\lambda = 1\) eigenvector.
(c) Give an explicit formula for \(f_k\) for any \(k\). (By "explicit," I mean involving elementary arithmetic and powers of complex numbers only. Formulas involving \(A^k\) are not acceptable.)
(d) Is there any choice of initial conditions that will make \(|f_k|\) diverge as \(k \to \infty\)? Explain.
Solution
(a) This recurrence gives \[ A = \begin{bmatrix} 1 & -1 & 1 \\ 1 & 0 & 0 \\ 0 & 1 & 0 \end{bmatrix}. \] That is, the first row of \(A\) gives \(f_{k+1} = f_k - f_{k-1} + f_{k-2}\), while the second and third rows of \(A\) just give \(f_k = f_k\) and \(f_{k-1} = f_{k-1}\) (copying the first and second rows of \(\vec{g}_k\) to the second and third rows of \(\vec{g}_{k+1}\)). (b) We need to find the nullspace of \(A - \lambda I\), via elimination to obtain row-reduced echelon form. For \(\lambda_1 = 1\): \[ A - I = \begin{bmatrix} 0 & -1 & 1 \\ 1 & -1 & 0 \\ 0 & 1 & -1 \end{bmatrix} \to \begin{bmatrix} 1 & -1 & 0 \\ 0 & -1 & 1 \\ 0 & 1 & -1 \end{bmatrix} \to \begin{bmatrix} 1 & -1 & 0 \\ 0 & 1 & -1 \\ 0 & 0 & 0 \end{bmatrix} \to \begin{bmatrix} 1 & 0 & -1 \\ 0 & 1 & -1 \\ 0 & 0 & 0 \end{bmatrix}, \] for which the nullspace vector is \(\vec{x}_1 = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix}\). To check the provided \(\pm i\) eigenvectors, we just multiply them by \(A\): \[ \begin{bmatrix} 1 & -1 & 1 \\ 1 & 0 & 0 \\ 0 & 1 & 0 \end{bmatrix} \begin{bmatrix} -1 \\ \pm i \\ 1 \end{bmatrix} = \begin{bmatrix} -1 \mp i + 1 \\ -1 \\ \pm i \end{bmatrix} = \begin{bmatrix} \mp i \\ -1 \\ \pm i \end{bmatrix} = \pm i \begin{bmatrix} -1 \\ \pm i \\ 1 \end{bmatrix}. \] (c) We have to expand the initial vector in the eigenvectors (note that the initial vector is \(\vec{g}_2\), not \(\vec{g}_0\), here). By inspection or by solving the linear system \(S\vec{c} = \vec{g}_2\) for the coefficients, we get \(\vec{c} = (1, 1/2, 1/2)^T\): \[ \vec{g}_2 = \begin{bmatrix} 0 \\ 1 \\ 2 \end{bmatrix} = \begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix} + \frac{1}{2} \left[ \begin{bmatrix} -1 \\ i \\ 1 \end{bmatrix} + \begin{bmatrix} -1 \\ -i \\ 1 \end{bmatrix} \right]. \] To get \(\vec{g}_{k+2} = A^k\vec{g}_2\), we just multiply each eigenvector by \(\lambda^k\), and take the third row to get \(f_k\): \[ f_k = 1 + \frac{1}{2} [i^k + (-i)^k] = 1 + \cos(k\pi/2). \] (This is just the sequence 2, 1, 0, 1, 2, 1, 0, 1, 2, 1, ... repeated over and over.) (d) No, because all of the eigenvalues have \(|\lambda| = 1\), hence their powers don't blow up. (However, as one may check, the matrix is not unitary.)Keywords: recurrence, eigenvalues, complex eigenvalues, eigenvectors, matrix powers