6.2-exam3-2008-strang¶
Quiz 3, 2008 (strang) · Section 6.2: Diagonalizing a Matrix Source: exam3_1.pdf, pp. 146-147
Problem¶
2 (30 pts.) The cosine of a matrix is defined by copying the series for \(\cos x\) (which always converges):
(a) Suppose \(Ax = \lambda x\). Show that \(x\) is an eigenvector of \(\cos A\). Find the eigenvalue.
(b) Find the eigenvalues of \(A = \frac{\pi}{2} \begin{bmatrix} 1 & 1 \\ 1 & 1 \end{bmatrix}\). The eigenvectors are \((1, 1)\) and \((1, -1)\). From the eigenvalues and eigenvectors of \(\cos A\), find that matrix \(\cos A\).
(c) The second derivative of the series for \(\cos(At)\) is \(-A^2 \cos(At)\). So \(u(t) = \cos(At)u(0)\) is a short formula for the solution of
Now construct that \(u(t) = \cos(At)u(0)\) by the usual three steps when \(A\) is diagonalizable: \(Ax_1 = \lambda_1x_1, Ax_2 = \lambda_2x_2, Ax_3 = \lambda_3x_3\).
-
Expand \(u(0) = c_1x_1 + c_2x_2 + c_3x_3\) in the eigenvectors.
-
Multiply those eigenvectors by ____, ____, ____.
-
Add up the solution \(u(t) = c_1\)____\(x_1 + c_2\)____\(x_2 + c_3\)____\(x_3\).
Solution
(a) Suppose that \(Ax = \lambda x\). Then \[ \cos(A)x = Ix - \frac{1}{2!}A^2x + \frac{1}{4!}A^4x - \dots = x - \frac{1}{2!}\lambda^2x + \frac{1}{4!}\lambda^4x - \dots \] \[ = \left(1 - \frac{1}{2!}\lambda^2 + \frac{1}{4!}\lambda^4 - \dots\right) x = \cos(\lambda)x. \] So \(x\) is an eigenvector of \(\cos(A)\) with eigenvalue \(\cos(\lambda)\). (b) We know that \((1, 1)\) and \((1, -1)\) are eigenvectors of \(A\). We can find the eigenvalues simply by acting by \(A\): \[ A \begin{bmatrix} 1 \\ 1 \end{bmatrix} = \frac{\pi}{2} \begin{bmatrix} 2 \\ 2 \end{bmatrix} = \pi \begin{bmatrix} 1 \\ 1 \end{bmatrix}, \] so \(A\) has eigenvalue \(\lambda_1 = \pi\). Similarly, \[ A \begin{bmatrix} 1 \\ -1 \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \end{bmatrix}, \] so \(A\) has eigenvalue \(\lambda_2 = 0\). Just as for any other function (\(A^2, e^A, A^{-1}, \dots\)), this means that \(\cos(A)\) has eigenvectors \((1, 1)\) with eigenvalue \(\cos(\pi) = -1\) and \((1, -1)\) with eigenvalue \(\cos(0) = 1\). We can put these into the diagonalization formula to find \(\cos(A)\): \[ \cos(A) = \begin{bmatrix} 1 & 1 \\ 1 & -1 \end{bmatrix} \begin{bmatrix} -1 & 0 \\ 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & 1 \\ 1 & -1 \end{bmatrix}^{-1} = \begin{bmatrix} 0 & -1 \\ -1 & 0 \end{bmatrix}. \] (c) This problem is modeled after what happens for \(e^{At}\). After expanding \(u(0)\), step 2 involves multiplying the eigenvectors by \(\cos(\lambda_1 t)\), \(\cos(\lambda_2 t)\), and \(\cos(\lambda_3 t)\). So the final answer is \[ u(t) = c_1 \cos(\lambda_1 t)x_1 + c_2 \cos(\lambda_2 t)x_2 + c_3 \cos(\lambda_3 t)x_3. \] Some common mistakes were forgetting to include the \(t\), using the function \(e\) instead of \(\cos\), or putting in something entirely different for the coefficients.Keywords: matrix cosine, matrix exponential, eigenvalues, eigenvectors, differential equation