6.1-final-2000¶
Final Exam, 2000 · Section 6.1: Introduction to Eigenvalues : Ax = λx Source: finalexam_1.pdf, p. 44
Problem¶
1. (a) (4 points) Explain why every eigenvector of \(A\) is either in the column space \(C(A)\) or the nullspace \(N(A)\) (or explain why this is false).
(b) (4 points) From \(A = S\Lambda S^{-1}\) find the eigenvalue matrix and the eigenvector matrix for \(A^{T}\). How are the eigenvalues of \(A\) and \(A^{T}\) related?
(c) (4 points) Suppose \(Ax = 0\) and \(A^{T}y = 2y\). Deduce that \(x\) is orthogonal to \(y\). You may prove this directly or use the subspace ideas in (a) or the eigenvector matrices in (b). Write a clear answer.
Solution
(a) This is true. If \(\lambda = 0\), then \(Ax = 0\), so \(x\) is in the nullspace \(N(A)\). If \(\lambda \neq 0\), then \(x = Ax/\lambda = \lambda x / \lambda\) is in the column space: \(x = A(x/\lambda)\), a combination of the columns of \(A\). Thus every eigenvector is in \(N(A)\) or in \(C(A)\). (b) Transpose \(A = S\Lambda S^{-1}\) to get \(A^T = (S^{-1})^T \Lambda S^T\). So the eigenvalue matrix of \(A^T\) is the same \(\Lambda\) and the eigenvector matrix of \(A^T\) is \((S^{-1})^T\) (the rows of \(S^{-1}\)). The eigenvalues of \(A\) and \(A^T\) are identical. (c) Direct proof: \(x^T A^T y = (Ax)^T y = 0^T y = 0\). But also \(x^T A^T y = x^T (2y) = 2x^T y\), so \(2x^T y = 0\) and \(x\) is orthogonal to \(y\). Subspace proof: \(x\) is in \(N(A)\) and (since \(y = (A^T y)/2\)) \(y\) is in the column space of \(A^T\), and \(N(A)\) is orthogonal to \(C(A^T)\).Keywords: eigenvectors, column space, nullspace, transpose eigenvalues