6.2-exam3-2005-b¶
Quiz 3, 2005 · Section 6.2: Diagonalizing a Matrix Source: exam3_1.pdf, pp. 81, 88-89
Problem¶
2 (40 pts.) This 4 by 4 matrix \(H\) is a Hadamard matrix:
\[
H = \begin{bmatrix} 1 & 1 & 1 & 1 \\ 1 & -1 & 1 & -1 \\ 1 & 1 & -1 & -1 \\ 1 & -1 & -1 & 1 \end{bmatrix} \qquad \text{Key Properties: } H^T = H \text{ and } H^2 = 4I
\]
(a) Figure out the eigenvalues of \(H\). Explain your reasoning.
(b) Figure out \(H^{-1}\) and the determinant of \(H\). Explain your reasoning.
(c) This matrix \(S\) contains three eigenvectors of \(H\). Find a 4th eigenvector \(x_4\) and explain your reasoning:
\[
S = \begin{bmatrix} 1 & 1 & 0 \\ 1 & 0 & -1 \\ 1 & 0 & 1 \\ -1 & 1 & 0 \end{bmatrix}
\]
(d) Find the solution to \(du/dt = Hu\) given that \(u(0) =\) third column of \(S\).
Solution
(a) Suppose \(Hv = \lambda v\) for some non-zero vector \(v\). Then \(H^2v = \lambda^2v = (4I)v = 4v\), so \(\lambda^2 = 4\), and thus every eigenvalue of \(H\) is equal to either 2 or \(-2\). The trace of \(H\) is 0, hence the sum of the eigenvalues of \(H\) is 0. We conclude that \(H\) has eigenvalues \(\lambda = 2, 2, -2, -2\). (b) From \(H^2 = 4I\) we obtain \[ H^{-1} = \frac{1}{4}H. \] The determinant of a matrix is the product of its eigenvalues: \[ \det H = 2 \cdot 2 \cdot (-2) \cdot (-2) = 16. \] (c) The first two eigenvectors correspond to \(\lambda = 2\), so the missing eigenvector corresponds to \(\lambda = -2\). Denote the unknown eigenvector \(v_4\) by \(\begin{bmatrix} a \\ b \\ c \\ d \end{bmatrix}\). Then \[ (H + 2I)v_4 = \begin{bmatrix} 3 & 1 & 1 & 1 \\ 1 & 1 & 1 & -1 \\ 1 & 1 & 1 & -1 \\ 1 & -1 & -1 & 3 \end{bmatrix} \begin{bmatrix} a \\ b \\ c \\ d \end{bmatrix} = \begin{bmatrix} 3a + b + c + d \\ a + b + c - d \\ a + b + c - d \\ a - b - c + 3d \end{bmatrix} = 0. \] The third component of \((H + 2I)v_4\) is equal to the second, and the fourth is the sum of the first two, hence we can choose \(v_4\) to be any vector satisfying \(3a + b + c + d = 0\) and \(a + b + c - d = 0\) which is not a multiple of the third eigenvector \((0, -1, 1, 0)\). For example, we can choose \[ v_4 = \begin{bmatrix} 1 \\ -1 \\ -1 \\ -1 \end{bmatrix}. \] Note that since \(H\) is symmetric and the three given eigenvectors are pairwise orthogonal, any non-zero vector perpendicular to them is automatically a fourth eigenvector (and \(v_4\) above is in fact such a vector). On the other hand, \(v_4\) doesn't have to be orthogonal to the three given eigenvectors: we could have chosen any vector \(c_1(0, -1, 1, 0) + c_2(1, -1, -1, -1)\) with \(c_2 \neq 0\). (d) Let \(v_3\) be the third column of \(S\). It is an eigenvector corresponding to \(\lambda_3 = -2\), so \(u = e^{-2t}v_3\) is a solution to \(du/dt = Hu\), and in fact it gives \(u(0) = v_3\), so it is the desired solution.Keywords: Hadamard matrix, eigenvalues, determinant, inverse matrix, differential equation