6.7 Invariant Direct Sums

这一节考虑direct sum中每一个子空间都是在$T$下invariant的情况，这种情况的好处是：$T$在每一个$W_i$上都可以限制为一个linear operator $T_i$。所以$T$可以看成是$T_1,\dots ,T_k$的direct sum，即$T$的变化结果可以分解为在每个独立的$T_i$在$W_i$上的变化结果。这一$W_i$如何得到，其实Theorem 9已经给出了，即通过投影$E_i$得到，但要保证invariant的附加条件，则需要（且只需要）$T$和每一个$E_i$可交换（Theorem 10)。
对于可对角化的矩阵，Theorem 11说明：如果不同特征值是$c_1,\dots ,c_k$，那么有一个和对角化等价的很复杂的一系列关于projection $E_1,\dots ,E_k$的结果，每一个$E_i$恰好是$c_i$的特征向量空间。这个定理很复杂且很难用语言描述，但其目的之一是使用抽象的只关于operator的代数性计算得到结果。
最后展示了$T=c_1{E}_1+\cdots +c_k{E}_k$这种decomposition的比较好的性质：可以直接套用多项式，即$g(T)=g(c_1)E_1+\cdots +g(c_k)E_k$，也揭示了Lagrange interpolation polynomial和$E_i$的关系。仅使用多项式和operator的性质，可以给Theorem 6一个新的证明（即可对角化和minimal polynomial之间的关系）。

Exercises

1.Let $E$ be a projection of $V$ and let $T$ be a linear operator on $V$. Prove that the range of $E$ is invariant under $T$ if and only if $ETE=TE$. Prove that both the range and null space of $E$ are invariant under $T$ if and only if $ET=TE$.
Solution: If $\text{range }E$ is invariant under $T$, then for any $\alpha \in V,E\alpha \in \text{range }E$, thus $TE\alpha \in \text{range }E$, then $E(TE\alpha )=TE\alpha$ since $E$ is a projection. Conversely, if $ETE=TE$, then for $\beta \in \text{range }E$, there is $\alpha \in V$ such that $\beta =E\alpha$, which means $T\beta =TE\alpha =ETE\alpha$, thus $T\beta$ is in $\text{range }E$.
Now if $ET=TE$, then $ETE=EET=ET$, thus the range of $E$ is invariant under $T$ by what has just proved. Consider the case of null space of $E$. Let $\alpha \in \text{null }E$, then $E\alpha =0$, thus $0=T(E\alpha )=ET\alpha$, so $T\alpha \in \text{null }E$ and $\text{null }E$ is invariant under $T$.
Conversely, suppose $\text{null }E$ and $\text{range }E$ is invariant under $T$, then since $V=\text{null }E\oplus \text{range }E$, for any $\alpha \in V$ we can write $\alpha =\beta +\gamma$, in which $\beta \in \text{null }E$ and $\gamma \in \text{range }E$. Thus $ET\alpha =ET\beta +ET\gamma$, as $\beta \in \text{null }E$ means $T\beta \in \text{null }E$, we have $ET\beta =0$, and $ET\gamma =T\gamma =TE\gamma$, notice that $TE\beta =0$, we have
$$ET\alpha =TE\gamma =TE\gamma +TE\beta =TE\alpha$$

2.Let $T$ be the linear operator on $R^2$, the matrix of which in the standard ordered basis is $\left[\begin{array}{cc}2& 1\\ 0& 2\end{array}\right]$. Let $W_1$ be the subspace of $R^2$ spanned by the vector ${ϵ}_1=(1,0)$.
( a ) Prove that $W_1$ is invariant under $T$.
( b ) Prove that there is no subspace $W_2$ which is invariant under $T$ and which is complementary to $W_1$: $R^2=W_1\oplus W_2$.
Solution:
( a ) From the matrix we know that T ϵ 1 = 2 ϵ 1 ∈ W 1 T\epsilon_1=2\epsilon_1\in W_1 Tϵ1​=2ϵ