4.2-final-2010¶
Final Exam, 2010 · Section 4.2: Projections onto Lines and Subspaces Source: finalexam_2.pdf, pp. 5-6
Problem¶
2. (14 pts.)
Let \(P_1\) be the projection matrix onto the line through \((1, 1, 0)\) and \(P_2\) the projection onto the line through \((0, 1, 1)\).
(a) (4 pts) What are the eigenvalues of \(P_1\)?
(a2) (bonus) (This question is from an earlier version of the exam.) Find an eigenvalue and an eigenvector of \(P_1 + P_2\).
(b) (10 pts) Compute \(P = P_2P_1\). (Careful, the answer is not 0.)
Solution
(a) The eigenvalues of a projection matrix are 1 and 0. The multiplicity of each eigenvalue: the multiplicity of 0 is the dimension of the nullspace, which is \(3 - \text{rank}\). Here \(P_1\) is a projection onto a line, so its column space is made of vectors on that line and has dimension 1. Therefore 0 has multiplicity 2 and 1 has multiplicity 1. (a2) The problem asks only for one eigenvalue and one eigenvector, but since this is for practice, here are all three. Let \(\vec{a}_1 := (1, 1, 0)\) and \(\vec{a}_2 := (0, 1, 1)\). The vectors have the same length \(\sqrt{2}\), and \(\vec{a}_1 \cdot \vec{a}_2 = 1 = 2\cos 60^\circ\), so the angle between them is \(60^\circ\). Geometrically: - \(\vec{v}_1 := \vec{a}_1 \times \vec{a}_2 = (1, -1, 1)\) is an eigenvector of \(P_1 + P_2\) with eigenvalue \(\lambda_1 = 0\): since \(v_1\) is perpendicular to both \((1, 1, 0)\) and \((0, 1, 1)\), it lies in the nullspace of both \(P_1\) and \(P_2\), and hence in the nullspace of \(P_1 + P_2\). - \(\vec{v}_2 := \vec{a}_1 + \vec{a}_2 = (1, 2, 1)\) is an eigenvector with eigenvalue \(\lambda_2 = 2\cos^2(60^\circ/2) = 1 + \cos 60^\circ = 3/2\). - \(\vec{v}_3 := \vec{a}_1 - \vec{a}_2 = (1, 0, -1)\) is an eigenvector with eigenvalue \(\lambda_3 = 2\sin^2(60^\circ/2) = 1 - \cos 60^\circ = 1/2\). The second and third facts can be derived with a bit of trigonometry, or by the usual linear-algebra calculation: \[ P_1 = \frac{1}{1^2 + 1^2 + 0^2} \begin{bmatrix} 1 \\ 1 \\ 0 \end{bmatrix} \begin{bmatrix} 1 & 1 & 0 \end{bmatrix} = \frac{1}{2} \begin{bmatrix} 1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 0 \end{bmatrix}, \qquad P_2 = \frac{1}{2} \begin{bmatrix} 0 & 0 & 0 \\ 0 & 1 & 1 \\ 0 & 1 & 1 \end{bmatrix}, \] so \[ P_1 + P_2 = \frac{1}{2} \begin{bmatrix} 1 & 1 & 0 \\ 1 & 2 & 1 \\ 0 & 1 & 1 \end{bmatrix}. \] You may solve for the roots of \(\det(P_1 + P_2 - \lambda I) = 0\) and find the nullspaces of \(P_1 + P_2 - \lambda_i I\) (for \(i = 1, 2, 3\)) as usual, and get the results described above. (b) Let's just do it directly: \[ P = \frac{1}{4} \begin{bmatrix} 0 & 0 & 0 \\ 0 & 1 & 1 \\ 0 & 1 & 1 \end{bmatrix} \begin{bmatrix} 1 & 1 & 0 \\ 1 & 1 & 0 \\ 0 & 0 & 0 \end{bmatrix} = \frac{1}{4} \begin{bmatrix} 0 & 0 & 0 \\ 1 & 1 & 0 \\ 1 & 1 & 0 \end{bmatrix}. \]Keywords: projection matrix, eigenvalues, eigenvectors