Skip to content

5.2-final-2006

Final Exam, 2006 · Section 5.2: Computing and Using Determinants Source: finalexam_1.pdf, p. 141

Problem

2 (6+3+2=11 pts.) (a) Find the inverse of a 3 by 3 upper triangular matrix \(U\), with nonzero entries \(a, b, c, d, e, f\). You could use cofactors and the formula for the inverse. Or possibly Gauss-Jordan elimination.

Find the inverse of \(U = \begin{bmatrix} a & b & c \\ 0 & d & e \\ 0 & 0 & f \end{bmatrix}\).

(b) Suppose the columns of \(U\) are eigenvectors of a matrix \(A\). Show that \(A\) is also upper triangular.

(c) Explain why this \(U\) cannot be the same matrix as the first factor in the Singular Value Decomposition \(A = U\Sigma V^T\).

Solution (a) *By elimination:* (We keep track of the elimination matrix \(E\) on one side, and the product \(EU\) on the other. When \(EU = I\), then \(E = U^{-1}\).) \[ \left[ \begin{array}{ccc|ccc} a & b & c & 1 & 0 & 0 \\ 0 & d & e & 0 & 1 & 0 \\ 0 & 0 & f & 0 & 0 & 1 \end{array} \right] \rightsquigarrow \left[ \begin{array}{ccc|ccc} 1 & b/a & c/a & 1/a & 0 & 0 \\ 0 & 1 & e/d & 0 & 1/d & 0 \\ 0 & 0 & 1 & 0 & 0 & 1/f \end{array} \right] \] \[ \rightsquigarrow \left[ \begin{array}{ccc|ccc} 1 & 0 & 0 & 1/a & -b/ad & (be - cd)/adf \\ 0 & 1 & 0 & 0 & 1/d & -e/df \\ 0 & 0 & 1 & 0 & 0 & 1/f \end{array} \right] = \left[ \begin{array}{cc} I & U^{-1} \end{array} \right] \] *By cofactors:* (Take the minor, then "checkerboard" the signs to get the cofactor matrix, then transpose and divide by \(\det(U) = adf\).) \[ \left[ \begin{array}{ccc} a & b & c \\ 0 & d & e \\ 0 & 0 & f \end{array} \right] \rightsquigarrow \left[ \begin{array}{ccc} df & 0 & 0 \\ bf & af & 0 \\ be - cd & ae & ad \end{array} \right] \rightsquigarrow \left[ \begin{array}{ccc} df & 0 & 0 \\ -bf & af & 0 \\ be - cd & -ae & ad \end{array} \right] \rightsquigarrow \left[ \begin{array}{ccc} df & -bf & be - cd \\ 0 & af & -ae \\ 0 & 0 & ad \end{array} \right] \rightsquigarrow \] \[ \left[ \begin{array}{ccc} 1/a & -b/ad & (be - cd)/adf \\ 0 & 1/d & -e/df \\ 0 & 0 & 1/f \end{array} \right] = U^{-1} \] (b) We have a complete set of eigenvectors for \(A\), so we can diagonalize: \(A = U\Lambda U^{-1}\). We know \(U\) is upper-triangular, and so is the diagonal matrix \(\Lambda\), and we have just shown that \(U^{-1}\) is upper-triangular too. So their product \(A\) is also upper-triangular. (c) The columns are not orthogonal! (For example, the product \(u_1^Tu_2\) of the first two columns is \(ab + 0d + 0 \cdot 0 = ab\), which is nonzero because we're assuming all the entries are nonzero.)

Keywords: matrix inverse, cofactors, triangular matrix, eigenvectors