5.2-exam3-2001-spring¶
Midterm Exam 3, 2001 (spring) · Section 5.2: Computing and Using Determinants Source: exam3_1.pdf, pp. 59, 63-64
Problem¶
- (40pts.) Let \(A_n\) be the \(n \times n\) tridiagonal matrix
\[
A_n = \begin{bmatrix} 1 & -a & 0 & 0 & \cdots & 0 \\ -a & 1 & -a & 0 & \cdots & 0 \\ 0 & -a & 1 & -a & \cdots & 0 \\ \vdots & & \ddots & \ddots & \ddots & \vdots \\ 0 & 0 & \cdots & -a & 1 & -a \\ 0 & 0 & \cdots & 0 & -a & 1 \end{bmatrix}.
\]
(a) Show for \(n \geq 3\) that
\[
\det(A_n) = \det(A_{n-1}) - a^2 \cdot \det(A_{n-2}). \quad (1)
\]
(b) Show that eq.(1) can equivalently be written as \(x_n = B x_{n-1}\), where
\[
x_n = \begin{bmatrix} \det(A_n) \\ \det(A_{n-1}) \end{bmatrix} \quad \text{and} \quad B = \begin{bmatrix} 1 & -a^2 \\ 1 & 0 \end{bmatrix}.
\]
(c) For \(a^2 = \frac{3}{16}\), find an expression for \(\det(A_n)\) for any \(n\). (Hint: One method starts by writing \(B\) in the form \(B = S\Lambda S^{-1}\), where \(\Lambda\) is a diagonal matrix.)
Solution
(a) Let us expand the determinant of \(A_n\) along the first column: \[ \det(A_n) = 1 \cdot \det \begin{bmatrix} 1 & -a & 0 & \cdots & 0 \\ -a & 1 & -a & \cdots & 0 \\ \vdots & \ddots & \ddots & \ddots & \vdots \\ 0 & \cdots & -a & 1 & -a \\ 0 & \cdots & 0 & -a & 1 \end{bmatrix} + a \det \begin{bmatrix} -a & 0 & 0 & \cdots & 0 \\ -a & 1 & -a & \cdots & 0 \\ \vdots & \ddots & \ddots & \ddots & \vdots \\ 0 & \cdots & -a & 1 & -a \\ 0 & \cdots & 0 & -a & 1 \end{bmatrix} \] \[ = \det(A_{n-1}) - a^2 \det \begin{bmatrix} 1 & -a & 0 & \cdots \\ -a & 1 & -a & \cdots \\ \vdots & \ddots & \ddots & \ddots \\ 0 & \cdots & -a & 1 \end{bmatrix} \] \[ = \det(A_{n-1}) - a^2 \det(A_{n-2}). \] Here, we have expanded the determinant in the second term of the first line along the first row. (b) \[ x_n = \begin{bmatrix} \det(A_n) \\ \det(A_{n-1}) \end{bmatrix} = \begin{bmatrix} \det(A_{n-1}) - a^2 \det(A_{n-2}) \\ \det(A_{n-1}) \end{bmatrix} \] \[ = \begin{bmatrix} 1 & -a^2 \\ 1 & 0 \end{bmatrix} \begin{bmatrix} \det(A_{n-1}) \\ \det(A_{n-2}) \end{bmatrix} = \begin{bmatrix} 1 & -a^2 \\ 1 & 0 \end{bmatrix} x_{n-1}. \] (c) The answer is given by \(x_n = B^{n-2}x_2\). To determine \(x_2\), we need \(\det(A_1) = 1\), and \[ \det(A_2) = \det \begin{bmatrix} 1 & -a \\ -a & 1 \end{bmatrix} = 1 - a^2 = 1 - \frac{3}{16} = \frac{13}{16}. \] To find \(B^{n-2}\), we first need to diagonalise \(B\). Eigenvalues are given by \[ \det \begin{bmatrix} 1-\lambda & -\frac{3}{16} \\ 1 & -\lambda \end{bmatrix} = 0 \] \[ \Leftrightarrow \lambda^2 - \lambda + \frac{3}{16} = 0 \] \[ \Leftrightarrow \lambda_1 = \frac{3}{4}, \lambda_2 = \frac{1}{4}. \] The eigenvector corresponding to \(\lambda_1 = \frac{3}{4}\) is given by \(\begin{bmatrix} 1/4 & -3/16 \\ 1 & -3/4 \end{bmatrix} v_1 = 0\), so that \(v_1 = \begin{bmatrix} 3 \\ 4 \end{bmatrix}\). The eigenvector corresponding to \(\lambda_2 = \frac{1}{4}\) is given by \(\begin{bmatrix} 3/4 & -3/16 \\ 1 & -1/4 \end{bmatrix} v_2 = 0\), and hence \(v_2 = \begin{bmatrix} 1 \\ 4 \end{bmatrix}\). So \(B = S\Lambda S^{-1}\), where \[ S = \begin{bmatrix} 3 & 1 \\ 4 & 4 \end{bmatrix}, \qquad \Lambda = \begin{bmatrix} 3/4 & 0 \\ 0 & 1/4 \end{bmatrix}, \qquad S^{-1} = \frac{1}{8} \begin{bmatrix} 4 & -1 \\ -4 & 3 \end{bmatrix}. \] Hence \[ x_n = S\Lambda^{n-2}S^{-1}x_2 = \frac{1}{8} \begin{bmatrix} \frac{27}{4} (\frac{3}{4})^{n-2} - \frac{1}{4} (\frac{1}{4})^{n-2} \\ 9 (\frac{3}{4})^{n-2} - (\frac{1}{4})^{n-2} \end{bmatrix} \] and \[ \det(A_n) = \frac{2}{4^{n+1}} (3^{n+1} - 1). \]Keywords: determinant, tridiagonal matrix, recurrence, diagonalization