6.1-exam3-2009-spring¶
Quiz 3, 2009 (spring) · Section 6.1: Introduction to Eigenvalues : Ax = λx Source: exam3_1.pdf, p. 159
Problem¶
Problem 1:
For each part, give as much information as possible about the eigenvalues of the matrix \(A\) described in that part. (Each part describes a different matrix \(A\). \(A\) may be complex.)
(a) The recurrence \(u_{k+1} = Au_k\) has a solution where \(\|u_k\| \to 0\) as \(k \to \infty\) for one initial vector \(u_0\), but also has a solution with \(\|u_k\| \to \infty\) as \(k \to \infty\) for a different choice of the initial vector \(u_0\).
(b) The equation \((A^2 - 4I)x = b\) has no solution for some right-hand side \(b\).
(c) \(A = e^{B^TB}\) for some real matrix \(B\) with full column rank.
(d) \(A = B^TB\) for a \(4 \times 3\) real matrix \(B\), and the matrix \(BB^T\) has eigenvalues \(\lambda = 3, 2, 1, 0\). (Hint: think about the SVD of \(B\).)
Solution
(a) (There was a bug in this problem: in the first condition, we should have required the initial vector \(u_0\) to be nonzero.) The first condition implies that \(A\) has an eigenvalue with absolute value \(|\lambda| < 1\). The second condition implies that either \(A\) has an eigenvalue with absolute value \(|\lambda| > 1\), or \(A\) is defective for 2 eigenvalues \(\lambda\) with \(|\lambda| = 1\). (b) The condition says that \(A^2 - 4I\) is singular. But we know that, if \(\lambda_1, \dots, \lambda_n\) are eigenvalues of \(A\), then the eigenvalues of \(A^2 - 4I\) are \(\lambda_1^2 - 4, \dots, \lambda_n^2 - 4\). The condition \(A^2 - 4I\) being singular says that one of \(\lambda_i^2 - 4\) is zero, and hence \(\lambda_i = 2\) or \(-2\). That is to say \(A\) has an eigenvalue 2 or \(-2\). (c) Since \(B\) has full column rank, the eigenvalues of \(B^TB\) are positive real numbers \(\lambda_i\). Hence, we know \(A = e^{B^TB}\) has eigenvalues \(e^{\lambda_i}\); they are real numbers bigger than 1. (d) Since \(BB^T\) and \(B^TB\) have the same set of *nonzero* eigenvalues, \(B^TB\) must have eigenvalues 3, 2, 1. Moreover, since \(B\) is a \(4 \times 3\) matrix, \(B^TB\) is a \(3 \times 3\) matrix. Hence, 3, 2, 1 are exactly all the eigenvalues.Keywords: eigenvalues, matrix exponential, singular value decomposition, matrix powers