\documentclass[uebung]{../../../lecture} \usepackage[]{gauss} \title{Lineare Algebra I: Übungsblatt 12} \author{Christian Merten} \begin{document} \punkte \vspace{4mm} \begin{aufgabe} Zunächst ist das charakteristische Polynom zu berechnen. \begin{align*} \chi_A(\lambda) = \text{det}(\lambda E_n - A) = \begin{gmatrix}[v] \lambda - 4 & 5 & 4 \\ -6 & \lambda + 7 & 4 \\ -3 & 4 & \lambda -3 \end{gmatrix} = \lambda^{3} - 3 \lambda + 2 = (\lambda - 1)^2(\lambda + 2) .\end{align*} Mit $\chi_A(\lambda) = 0$ folgen die Eigenwerte $\lambda_1 = 1$ und $\lambda_2 = -2$. Die Eigenräume ergeben sich aus den homogenen LGS. Für $\lambda_1 = 1$ folgt \begin{align*} \begin{gmatrix}[p] 1 - 4 & 5 & 4 \\ -6 & 8 & 4 \\ 3 & -4 & -2 \rowops \add{2}{1} \swap{2}{1} \add{1}{0} \add[4]{0}{1} \mult{1}{\frac{1}{3}} \end{gmatrix} \to \begin{gmatrix}[p] 1 & 0 & 2 \\ 0 & 1 & 2 \\ 0 & 0 & 0 \end{gmatrix} &\implies V_{\lambda_1} = \text{Lin}\left[ \begin{pmatrix} 2 \\ 2 \\ -1 \end{pmatrix} \right] \intertext{Für $\lambda_2 = -2$ folgt analog} \begin{gmatrix}[p] -6 & 5 & 4 \\ -6 & 5 & 4 \\ 3 & -4 & -5 \rowops \add[-]{0}{1} \swap{1}{2} \add[2]{1}{0} \add[4]{0}{1} \mult{1}{\frac{1}{3}} \swap{1}{0} \end{gmatrix} \to \begin{gmatrix}[p] 1 & 0 & 1 \\ 0 & 1 & 2 \\ 0 & 0 & 0 \end{gmatrix} &\implies V_{\lambda_2} = \text{Lin}\left[ \begin{pmatrix} 1 \\ 2 \\ -1 \end{pmatrix} \right] .\end{align*} Damit folgt direkt $\mu_{\text{geo}}(\lambda_1) = \text{dim } V_{\lambda_1} = 1 = \text{dim } V_{\lambda_2} = \mu_{\text{geo}}(\lambda_2)$. Damit ist \begin{align*} &\sum_{i=1}^{n} \mu_{\text{geo}}(\lambda_i) = 2 \neq 3 = n \quad \implies A \text{ nicht diagonalisierbar} \intertext{Aber wegen $\mu_{\text{alg}}(\lambda_1) = 2$ und $\mu_{\text{alg}}(\lambda_2) = 1$ folgt} &\sum_{i=1}^{n} \mu_{\text{alg}}(\lambda_i) = 3 = n \implies \quad A \text{ trigonalisierbar} .\end{align*} Ergänze nun $v_1 := \begin{pmatrix} 2 \\ 2 \\ -1 \end{pmatrix} $ und $v_1 := \begin{pmatrix} 1 \\ 2 \\ -1 \end{pmatrix} $ zu einer Basis durch $\begin{pmatrix} 0 \\ 0 \\ 1 \end{pmatrix} $ (l.u. da $0 = 2 - 2$ aber $2 - 1 \neq 0$). Damit folgt direkt \begin{align*} S := M(id)_{\underline{e}}^{\underline{v}} = \begin{gmatrix}[p] 2 & 1 & 0 \\ 2 & 2 & 0 \\ -1 & -1 & 1 \end{gmatrix} .\end{align*} Die Inverse ergibt sich durch Berechnung: \begin{align*} \begin{gmatrix}[p] 2 & 1 & 0 \\ 2 & 2 & 0 \\ -1 & -1 & 1 \end{gmatrix} \begin{gmatrix}[p] 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \rowops \add[-]{1}{0} \add[2]{0}{1} \add{0}{2} \mult{1}{\frac{1}{2}} \add{1}{2} \swap{0}{1} \end{gmatrix} \to \begin{gmatrix}[p] 1 & -\frac{1}{2} & 0 \\ -1 & 1 & 0 \\ 0 & \frac{1}{2} & 1 \end{gmatrix} =: S^{-1} .\end{align*} Damit ist $S^{-1}AS$ eine obere Dreiecksmatrix. \end{aufgabe} \begin{aufgabe} \begin{enumerate}[(a)] \item Beh.: Die Matrix $A := \begin{pmatrix} 2 & 3 \\ 1 & 0 \end{pmatrix}$ ist eindeutig bestimmt mit \[ \begin{pmatrix} f(n+2) \\ f(n+1) \end{pmatrix} = A \cdot \begin{pmatrix} f(n+1) \\ f(n) \end{pmatrix} .\] \begin{proof} Mit $A = \begin{pmatrix} a & b \\ c & d \end{pmatrix} $ ergibt die Voraussetzung zwei Gleichungen: \begin{align*} f(n+2) &= a f(n+1) + b f(n) \\ f(n+1) &= c f(n+1) + d f(n) .\end{align*} Mit $f(n+2) = 2 f(n+1) + 3 f(n)$ $\forall n \in \N$ folgt direkt $a = 2$, $b = 3$, $c = 1$ und $d = 0$. \end{proof} \item Analog zu Aufgabe 1 folgt \begin{align*} \chi_A(\lambda) = \begin{gmatrix}[v] \lambda - 2 & -3 \\ -1 & \lambda \end{gmatrix} = (\lambda+1)(\lambda -3) = 0 \quad \implies \lambda_1 = -1 \quad \lambda_2 = 3 .\end{align*} Damit ergeben sich die Eigenräume. Aus $\lambda_1 = -1$ folgt \begin{align*} \begin{gmatrix}[p] -3 & -3 \\ -1 & -1 \end{gmatrix} \to \begin{gmatrix}[p] 1 & -3 \\ 0 & 0 \end{gmatrix} \implies V_{\lambda_1} = \text{Lin}\left[ \begin{pmatrix} 1 \\ -1 \end{pmatrix} \right]. \intertext{und aus $\lambda_2 = 3$ folgt} \begin{gmatrix}[p] 1 & -3 \\ -1 & 3 \end{gmatrix} \to \begin{gmatrix}[p] 1 & -3 \\ 0 & 0 \end{gmatrix} \implies V_{\lambda_2} = \text{Lin}\left[ \begin{pmatrix} 3 \\ 1 \end{pmatrix} \right] .\end{align*} \item Die Eigenvektoren aus (b) bilden eine Basis $\underline{v}$. Damit folgt \begin{align*} S := M_{\underline{e}}^{\underline{v}}(id) = \begin{gmatrix}[p] 1 & 3 \\ -1 & 1 \end{gmatrix} .\end{align*} und durch Invertieren von $S$ \begin{align*} S^{-1} = \begin{gmatrix}[p] \frac{1}{4} & -\frac{3}{4} \\ \frac{1}{4} & \frac{1}{4} \end{gmatrix} .\end{align*} Damit hat $S^{-1}AS$ Diagonalgestalt. \item Aus (c) folgt die Diagonalmatrix \begin{align*} D := S^{-1}AS = \begin{gmatrix}[p] -1 & 0 \\ 0 & 3 \end{gmatrix} .\end{align*} Durch die Formel \[ \begin{pmatrix} f(n+2) \\ f(n+1) \end{pmatrix} = A \cdot \begin{pmatrix} f(n+1) \\ f(n) \end{pmatrix} .\] ergibt sich, dass $f(n)$ durch $n-2$-malige Anwendung von $A$ auf die Startwerte $\begin{pmatrix} f(2) \\ f(1) \end{pmatrix} $ entsteht. Zur Berechnung wird die Diagonalmatrix $D$ verwendet. Dafür wird der Startvektor mit $S^{-1}$ in die Basis $\underline{v}$ transformiert und dann $D$ $n-2$ mal angewendet und das Ergebnis mit $S^{-1}$ zurück in die kanonische Basis transformiert. Damit ergibt sich \begin{align*} \begin{pmatrix} f(n) \\ f(n-1) \end{pmatrix} = S \cdot D^{n-2} \cdot S^{-1} \cdot \begin{pmatrix} f(2) \\ f(1) \end{pmatrix} = \begin{gmatrix}[p] 1 & 3 \\ -1 & 1 \end{gmatrix} \cdot \begin{gmatrix}[p] (-1)^{n-2} \left( \frac{f(2)}{4} - \frac{3}{4} f(1) \right) \\ 3^{n-2} \left( \frac{f(2)}{4} + \frac{f(1)}{4} \right)\end{gmatrix} .\end{align*} Da nur $f(n)$ benötigt wird, ergibt sich so $f(n)$ durch die obere Zeile des Vektors, also \begin{align*} f(n) &= (-1)^{n-2} \left( \frac{f(2)}{4} - \frac{3}{4}f(1) \right) + 3 \cdot 3^{n-2} \left( \frac{f(2)}{4} + \frac{f(1)}{4} \right) \\ &= \frac{3 f(1)}{4} \left( (-1)^{n+1} + 3^{n-2}\right) + \frac{f(2)}{4} \left( (-1)^{n} + 3^{n-1} \right) .\end{align*} \end{enumerate} \end{aufgabe} \begin{aufgabe} \begin{enumerate}[(a)] \item Beh.: $\gamma$ ist eine symmetrische, nicht-ausgeartete Bilinearform \begin{proof} Seien $f, g, h \in K[X]_{\le n}$ und $\lambda \in K$ bel. Zz.: $\gamma$ ist symmetrisch. Da $K[X]$ kommutativer Ring folgt \begin{align*} \gamma(f,g) = \left( \int (f \cdot g) dx \right)(1_K) = \left( \int (g \cdot f) dx \right)(1_K) = \gamma(g, f) .\end{align*} Zz.: $\gamma$ ist bilinear. \begin{align*} \gamma(\lambda f + h, g) &= \left( \int ((\lambda f + h) \cdot g) dx \right) (1_K) \\ &= \left( \int ( \lambda f \cdot g + h \cdot g)dx \right)(1_K) \\ &= \lambda \left( \int \left( f \cdot g \right) \right) + \left( \int \left( h \cdot g) \right) dx \right)(1_K) \\ &= \lambda \gamma(f, g) + \gamma(h, g) .\end{align*} Wegen der Symmetrie von $\gamma$ folgt die Linearität auch im zweiten Argument. Zz.: $\gamma$ ist nicht ausgeartet. Sei $\mathcal{B}$ die kanonische Basis des $K[X]_{\le n}$. Dann ist die Fundamentalmatrix von $\gamma$ bezügl. $\mathcal{B}$ offensichtlich invertierbar. Damit ist $\gamma$ nicht ausgeartet. \end{proof} \item Für $K = \Q$ und $n = 3$ muss $\gamma$ für alle Kombinationen der Basisvektoren $(1, x, x^2, x^{3})$ berechnet werden. Also beispielsweise für $x$ und $x^2$ folgt \begin{align*} \gamma(x, x^2) = \left( \int (x \cdot x^2) dx \right)(1_K) = \left( \int x^{3} dx \right)(1_K) = \frac{1}{3} .\end{align*} Damit folgt analog für die restlichen Kombinationen die Fundamentalmatrix \begin{align*} G = \begin{gmatrix}[p] 1 & \frac{1}{2} & \frac{1}{3} & \frac{1}{4} \\ \frac{1}{2} & \frac{1}{3} & \frac{1}{4} & \frac{1}{5} \\ \frac{1}{3} & \frac{1}{4} & \frac{1}{5} & \frac{1}{6} \\ \frac{1}{4} & \frac{1}{5} & \frac{1}{6} & \frac{1}{7} \end{gmatrix} .\end{align*} \item Um eine Orthogonalbasis zu bestimmen, müsste $G$ aus (c) diagonalisiert werden. Dieser Prozess ist allerdings für eine $4\times 4$ Matrix mit diesen Einträgen langwierig. Deswegen bestimme die Orthogonalbasis induktiv. Für $n = 2$ ist sofort ersichtlich, dass $\{1, 1 - 2x\} $ eine Orthogonalbasis bilden, da \begin{align*} \gamma(1, 1-2x) = \left( \int (1-2x)dx \right)(1_K) = 1 - 1 = 0 .\end{align*} Für $n = 3$ ergibt sich dann aus den Gleichungen $\gamma(1 - 2x, a + bx +cx^2) = 0$ und $\gamma(1, a + bx + cx^2) = 0$ folgendes Gleichungssystem: \begin{align*} - b -c &= 0 \\ a + \frac{b}{2} + \frac{c}{3} &= 0 .\end{align*} Durch Wahl von $a = -1$ ergibt sich damit $b = 6$ und $c = -6$. Für $n = 4$ ergeben sich analog zu $n = 3$ die Gleichungen \begin{align*} -10b - 10c - 9d &= 0 \\ d + \frac{2}{3} c &= 0 \\ a + \frac{b}{2} + \frac{c}{3} + \frac{d}{4} &= 0 .\end{align*} Durch Wahl von $d = 10$ ergibt sich damit $a = -\frac{1}{2}$, $b = 6$, $c = -15$. Damit ist \begin{align*} \{1, 1-2x, -1 + 6x -6x^2, -\frac{1}{2} + 6x - 15 x^2 + 10x^3\} .\end{align*} offensichtlich l.u. und eine Orthogonalbasis des $Q[X]_{\le 3}$ bezüglich $\gamma$. \end{enumerate} \end{aufgabe} \begin{aufgabe} Es gilt nach VL: $M_{\underline{e}}^{\underline{e}}(f) = \left(M_{\underline{e^{*}}}^{\underline{e^{*}}}(f^{*})\right)^{t}$. Damit folgt \begin{align*} \chi_f(t) &= \text{det}(t E_n - M_{\underline{e}}^{\underline{e}}(f)) \\ &= \text{det}\left[\left(t E_n - M_{\underline{e}}^{\underline{e}}(f)\right)^{t}\right] \\ &= \text{det}\left[(t E_n)^{t} - \left(M_{\underline{e}}^{\underline{e}}(f)\right)^{t}\right] \\ &= \text{det}\left[t E_n - M_{\underline{e^{*}}}^{\underline{e^{*}}}(f^{*})\right] \\ &= \chi_{f^{*}}(t) .\end{align*} Damit folgt: $\lambda \in K$ Eigenwert von $f$ $\iff$ $\lambda$ Nullstelle von $\chi_f(\lambda)$ $\iff$ $\lambda$ Nullstelle von $\chi_{f^{*}}(\lambda)$ $\iff$ $\lambda$ Eigenwert von $f^{*}$. Sei nun $\lambda \in K$ ein Eigenwert von $f$ bzw. $f^{*}$. Dann bezeichne $V_{\lambda}$ den Eigenraum von $\lambda$ bezüglich $f$ und $V^{*}_{\lambda}$ den Eigenraum bezüglich $f^{*}$. Da $V$ e.d. gilt $\text{dim } V = \text{dim } V^{*}$. \begin{align*} \text{dim } V_{\lambda} &= \qquad\text{dim } \text{ker}(\lambda E_n - M_{\underline{e}}^{\underline{e}}(f)) \\ &\stackrel{V \text{ e.d.}}{=} \qquad \text{dim } V - \text{Rg}(\lambda E_n - M_{\underline{e}}^{\underline{e}}(f)) \\ &\stackrel{\text{SR = ZR}}{=} \qquad \text{dim } V - \text{Rg}\left[ \left( \lambda E_n - M_{\underline{e}}^{\underline{e}}(f) \right)^{t} \right] \\ &= \qquad \text{dim } V^{*} - \text{Rg}\left[ \lambda E_n - \left(M_{\underline{e}}^{\underline{e}}(f) \right)^{t} \right] \\ &\stackrel{\text{dim } V = \text{dim } V^{*}}{=} \qquad \text{dim } V^{*} - \text{Rg}\left( \lambda E_n - M_{\underline{e^{*}}}^{\underline{e^{*}}}(f^{*}) \right) \\ &\stackrel{V^{*} \text{ e.d.}}{=} \qquad \text{dim } \text{ker}(\lambda E_n - M_{\underline{e^{*}}}^{\underline{e^{*}}}(f^{*})) \\ &= \qquad \text{dim } V^{*}_{\lambda} .\end{align*} \end{aufgabe} \end{document}