6.3-final-2013-strang-b¶
Final Exam, 2013 (strang) · Section 6.3: Symmetric Positive Definite Matrices Source: finalexam_2.pdf, p. 124
Problem¶
9 (6+4 pts.)
(a) Suppose \(v = (v_1, v_2, v_3)\) is a column vector, so \(A = vv^T\) is a symmetric matrix. Show that \(A\) is positive semidefinite, using one of these tests:
-
The eigenvalue test
-
The determinant test
-
The energy test on \(x^T Ax\).
(b) Suppose \(A\) is \(m\) by \(n\) of rank \(r\). What conditions on \(m\) and \(n\) and \(r\) guarantee that \(A^T A\) is positive definite? If those conditions fail, prove that \(A^T A\) will not be positive definite.
Solution
(a) Use the energy test. For every vector \(x\), \[ x^T A x = x^T vv^T x = (v^T x)^T (v^T x) = (v^T x)^2 \ge 0. \] Since the energy \(x^T A x\) is never negative, \(A = vv^T\) is positive semidefinite. (The eigenvalues reflect this: the nonzero eigenvalue of \(vv^T\) is \(v^T v = v_1^2 + v_2^2 + v_3^2 \ge 0\), and the other two eigenvalues are 0.) (b) \(A^T A\) is positive definite exactly when \(r = n\) (which forces \(m \ge n\), since \(r \le \min(m, n)\)). If \(r = n\), then \(A\) has independent columns, \(N(A) = \{0\}\), and for every \(x \neq 0\), \[ x^T A^T A x = (Ax)^T (Ax) = \|Ax\|^2 > 0, \] so \(A^T A\) is positive definite. Conversely, if \(r < n\), there is a nonzero \(v \in N(A)\), and then \[ v^T A^T A v = \|Av\|^2 = 0, \] so \(A^T A\) is not positive definite (it is only positive semidefinite, with an eigenvalue 0).Keywords: positive semidefinite, positive definite, energy test, rank