3.20 Projector
-
Assume \(A\in\mathbb{C}^{4\times 4}\) is diagonalizable. Find all the diagonal matrix \(B\) in \(\mathbb{C}^{4\times 4}\) that satisfy the following conditions
- \(A\) is similar to \(B\)
- \(m_A(x)\mid x(x^2-1)(x^2+2)\)
- \(\chi_A(x)\in \R[x]\)
- \(x^{2} \nmid \chi_{A}(x)\)
- \(A\) is not invertible
Since \(A\) is diagonalizable, then $A=PDP^{-1}\Rightarrow \det A=\det(PDP^{-1})=\det (P)\det (D)\det (P^{-1})=\det(PP^{-1})\det D=\det D $
If \(A\) is not invertible, then \(0\) is a root of \(\chi_{A}(x)\), that is \(x\mid \chi_A(x)\)
But \(x^2\nmid \chi_A(x)\), then \(0\) is a simple root(multiplicity is \(1\)), in particular \(\chi_A(x)=x\cdot p(x)\) and \(m_{A}(x)=xq(x)\) and \(q(x)\mid p(x)\) and \(q\) has a single root
If \(\chi_A(x)\) has no roots in \(\mathbb{C}\), then the roots of \(q(x)\) are in \(\R\), then \(m_{A}(x)\mid x(x^{2}-1)\)
Then \(m_{A}(x)=x(x-1),x(x+1),x(x^{2}-1)\)
Then \(\chi_A(x)=x(x-1)^3,x(x+1)^3,x(x+1)^2(x-1),x(x+1)(x-1)^2\)
If \(\chi_A(x)\) has root in \(\mathbb{C}\), then \(m_{A}(x)=x(x+\sqrt{2}i)(x-\sqrt{2}i)r(x)\) with \(\deg r(x)\leq 1\)
If \(\deg r(x)=0\), then \(\chi_{A}(x)=kx(x+\sqrt{2}i)^{2}(x-\sqrt{2}i)\) or \(kx(x+\sqrt2i)(x-\sqrt2i)^2\)
But it's contradiction to 3
If \(\deg r(x)=1\), then since 2, \(r(x)\) is a factor of \(x^2-1\), then it can be \((x+1)\) or \((x-1)\)
Therefore, \(B\in\left\lbrace \begin{pmatrix} 0 & & & \\ & 1 & & \\ & & 1 & \\ & & & 1 \end{pmatrix}, \begin{pmatrix} 0 & & & \\ & -1 & & \\ & & -1 & \\ & & & -1 \end{pmatrix}, \begin{pmatrix} 0 & \placeholder{} & \placeholder{} & \placeholder{} \\ \placeholder{} & -1 & \placeholder{} & \placeholder{} \\ \placeholder{} & \placeholder{} & -1 & \placeholder{} \\ \placeholder{} & \placeholder{} & \placeholder{} & 1 \end{pmatrix}, \begin{pmatrix} 0 & \placeholder{} & \placeholder{} & \placeholder{} \\ \placeholder{} & -1 & \placeholder{} & \placeholder{} \\ \placeholder{} & \placeholder{} & 1 & \placeholder{} \\ \placeholder{} & \placeholder{} & \placeholder{} & 1 \end{pmatrix} \begin{pmatrix} 0 & \placeholder{} & \placeholder{} & \placeholder{} \\ \placeholder{} & -\sqrt{2i} & \placeholder{} & \placeholder{} \\ \placeholder{} & \placeholder{} & \sqrt{2i} & \placeholder{} \\ \placeholder{} & \placeholder{} & \placeholder{} & 1 \end{pmatrix}, \begin{pmatrix} 0 & \placeholder{} & \placeholder{} & \placeholder{} \\ \placeholder{} & -\sqrt{2i} & \placeholder{} & \placeholder{} \\ \placeholder{} & \placeholder{} & \sqrt{2i} & \placeholder{} \\ \placeholder{} & \placeholder{} & \placeholder{} & -1 \end{pmatrix}\right\rbrace\)
Definition: \(V\) is a vector space over \(\mathbb{K}\), \(f: V \to V\) is a linear operator, then \(f\) is a projection if \(f^2 = f\).
Remark: \(f\) is a projection iff \(\forall x \in \text{Im} f, f(x) = x\) . If \(f\) is a projection, then \(V = \text{Ker} f \oplus \text{Im} f\) .
Proposition: \(V\) is a vector space over \(\mathbb{K}\) and \(S, T \leq V\) such that \(V = S \oplus T\) . Then there exists a unique projection \(f: V \to V\) such that \(S = \text{Ker} f\) and \(T = \text{Im} f\) .
-
Find a projection \(f: \mathbb{R}^3 \to \mathbb{R}\) such that \(\text{Ker} f = \text{span} \{ (1, 2, 3) \}\) and \(\text{Im} f = \text{span} \{ (0, 1, 0), (3, 2, 1) \}\).
Solution
\(V, W\) are vector spaces, \(\mathcal{B} = \{v_1, \ldots, v_n\}\) is a basis for \(V\). \(\{w_1, \ldots, w_n\}\) is a set in \(W\). There exists a unique \(f: V \to W\) such that \(f(v_i) = w_i\) for each \(i = 1, \ldots, n\).
\(B = \{ (1, 2, 3), (0, 1, 0), (3, 2, 1) \}\) a basis for \(\mathbb{R}^3\)
Since \(f(1,2,3) = (0,0,0)\), \(f(0,1,0) = (0,1,0)\), \(f(3,2,1) = (3,2,1)\)
Then \([f]_{B}= \begin{bmatrix} 0 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{bmatrix}\)
Since \(f \text{ is a projection iff } f(x) = x \quad \forall x \in \text{Im}(f).\)
Then \([f]_{e} = C(B, E) [f]_{b} C(E,B)^{-1}\), \([f]_e = \begin{bmatrix} 1 & 0 & 3 \\ 2 & 1 & 2 \\ 3 & 0 & 1 \end{bmatrix} \begin{bmatrix} 0 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{bmatrix} \begin{bmatrix} 1 & 0 & 3 \\ 2 & 1 & 2 \\ 3 & 0 & 1 \end{bmatrix}^{-1}\)\(= \begin{bmatrix} 9/8 & 0 & -3/8 \\ 1/4 & 1 & -3/4 \\ 3/8 & 0 & -1/8 \end{bmatrix}\)
Then \(f(x,y,z)=\left(\frac{9x-3z}{8},\quad\frac{x+4y-3z}{4},\quad\frac{3x-z}{8}\right)\)
-
Let \(V\) be a v.s. over K and let \(W_1\) and \(W_2\) be subspace of \(V\) such that \(V=W_1\oplus W_2\), then \(V^*=W_1^\circ\oplus W^\circ _2\)
Remark: \(V^{*}=L\left(V,\mathbb{K}\right)\) and \(W^{\circ}=\left\lbrace f:V\to K:f\left(w\right)=0,\forall w\in W\right\rbrace\)
Theorem
Consider \(f_i:V\to V\) such that \(f_i\) is a projection on \(W_i\) along \(W_j\)
We want \(F_{i}:V^{*}\to V^*\), define the map is \((g:V\to K)\mapsto (F_{i}(g):V\to K)\)
\(v\mapsto F_i(g)(v)=g(f_i(v))\)
-
\((F_i)^2=F_i\)
\(F_{i}^{2}(g)(v)=F_{i}(F_{i}(g))(v)=F_i(g)(f_i(v))=g(f_i(f_i(v)))=g(f_i(v))=F_i(g)(v)\) 2. \(F_iF_j=0\) if \(i\neq j\)
\((F_{i}F_{j})(g)(v)=F_{i}(F_{j}(g))(v)=F_{j}(g)(f_{i}(v))=g(f_{j}f_{i}(v))=g(0)=0\) 3. \(id_{V^*}=F_{1}+F_{2}\)
\((F_{1}+F_{2})(g)(v)=F_{1}(g)(v)+F_{2}(g)(v)=g\left(f_{1}\left(v\right)\right)+g\left (f_{2}\left(v\right)\right)=g\left(\left(f_{1}+f_{2}\right)\left(v\right)\right)= g\left(id_{v}\left(v\right)\right)=g\left(v\right)\) 4. \(ImF_i=W_j^\circ\)
\(\subseteq\)) Assume \(h\in ImF_{i}\), then \(\exists g\in V^{*}:F_{i}\left(g\right)=h\)
Let \(v\in W_{j},h(v)=F_{i}(g)(v)=g\left(f_{i}\left(v\right)\right)=g\left(0\right)=0\), thus \(h(v)=0,\forall v\in W_j\)
Thus \(h\in W_j^\circ\)
\(\supseteq\))Let \(g\in W_j^\circ\), then \(g(w_j)=0,\forall w_j\in W_j\). Now if \(v\in V\), \(\exists!w_1\in W_1\) and \(w_2\in W_2\) such that \(v=w_1+w_2\), \(g(v)=g(w_{i}+w_{j})=g\left(w_{i}\right)\)
\(F_{i}(g)(v)=g(f_{i}(w_{1}+w_{2}))=g(w_{i})=g\left(v\right)\), thus \(F_i(g)(v)=g(v),\forall v\in V\), then \(F_i(g)=g\)
Thus \(V^{*}=W_{1}^{\circ}\oplus W_{2}^{\circ}\)
-
-
Definition: Let \(V\) be a finite-dimensional vector space over \(K\). Let \(f: V \to V\) be a linear operator.
We say that \(f\) is nilpotent of index \(r\) if \(m_{r}\left(x\right)=x^{r}\). That is, \(f^r = 0\) but \(f^{i}\neq0\) if \(i < r\).
Prove that if \(f\) is nilpotent of index \(r\) and \(\dim V = n\), then \(\{0\}\subsetneq\text{Ker}\,f\subsetneq\text{Ker}\,f^{2}\subsetneq\text{Ker}\,f^{3} \subsetneq\ldots\subsetneq\text{Ker}\,f^{r-1}\subsetneq\text{Ker}\,f^{r}=V\)
\(\text{Ker} f^r = V\) since \(f^r = 0 \implies f^r(v) = 0 \ \forall v\)
\(\text{Ker } f^{i-1} \subsetneq \text{Ker } f^i\)
\(\subseteq\)) \(v \in \text{Ker }f^{i-1}\Rightarrow f^{i}(v)=0 \Rightarrow f(0) = f(f^{i-1}(v)) = f^{i}(v)\)
\(\therefore v\in \text{Ker }f^{i}\)\(\supseteq\)) By def \(m_{f}(x)=x^{r}\Rightarrow f^{r}=0\) and \(f^{r-1} \neq 0\) \(\Rightarrow \exists v \in \text{Ker }f^{r} \setminus \text{Ker }f^{r-1}\)
Now, \(f(v) \in \text{Ker }f^{r-1}\) and \(f(v) \notin \text{Ker }f^{r-2}\)
\(f(v) \in \text{Ker }f^{r-i}\setminus \text{Ker }f^{r-i-1} \text{ for }i = 0, \ldots, r-1\)
\(\dim V = n\)
-
If \(f\) is nilpotent of index \(n\), there exist \(f\)-inv. subspaces \(S_i \subseteq V\) with dim\(S_i = i\) \(\forall i = 0, \ldots, n\)
\(\{0\}\subsetneq\text{Ker }f\subsetneq\text{Ker }f^{2}\subsetneq\ldots\subsetneq\text{Ker }f^{n-2}\subsetneq\text{Ker }f^{n-1}\subsetneq V=\text{Ker }f^{n}\)
\(S_n = V = \text{span} \{ v, fv, \ldots, f^{n-1}(v) \}\)
\(S_{n-1}=\text{span}\{fv,f^{2}(v),\ldots,f^{n-1}(v)\}\)
\(S_{n-2}=\text{span}\{f^{2}(v),f^{3}(v),\ldots,f^{n-1}(v)\}\)
\(\vdots\)
\(S_{2} = \text{span}\{ f^{n-2}(v), f^{n-1}(v) \}\)
\(S_1 = \text{span} \{ f^{n-1}(v) \}\)
\(S_0 = \{ 0 \}\)
\(f(S_n) = S_{n-1} \subseteq S_n\)
\(f(S_{n-1}) = S_{n-2} \subseteq S_{n-1}\)
\(\vdots\)
\(f(S_i) = S_{i-1} \subseteq S_i \quad \text{\checkmark}\)