TLDR

Just like linear transformation classified by Jordan canonical form, we use “(generalized) eigenspace”, which is called (generalized) weight space, to analyze representations of Lie algebra. However, for a set of matrices, it is not easy to find a common eigenspace. Luckily, with additional conditions, named solvable and nilpotent, we have some good results.

Two important conclusions will be showed:

  • every solvable Lie algebra has a weight space;
  • for any representation of nilpotent Lie algebra, can be decomposed as direct sum of generalized weight spaces.

Solvable Lie algebra

Eigenspace contained in an ideal is -invariant

Naive try. Suppose , is a subalgebra. Suppose has a weight and is the according weight space. Hope that to be -invariant. That is, for any , and .

For any and , by above assumptions we have

so should be zero. If is an ideal, then should be zero.

Lemma

Suppose is a finite dimensional complex Lie algebra and is an ideal. Let be a weight of and be the corresponding weight space, where is a finite dimensional complex representation. Then is -invariant.

TL; DR: 找到一个下的不变子空间,考虑限制在子空间上的矩阵:它的迹为.

\begin{proof}To show is -invariant, by the above argument it remains to show . For any , and , we can verify

is -invariant.

Note that for any , acts on this space by the matrix

Since is an ideal, . Then the trace of is . Also is trace zero. Thus . \end{proof}

Main result

Proposition

For a finite dimensional complex solvable Lie algebra and its finite dimensional complex representation , this representation has a weight.

\begin{proof}By induction, we only need to prove that if has a weight, then has a weight. Suppose has a weight and the corresponding weight space . For any , by lemma is -invariant. Choose eigenvalue and let be the corresponding eigenspace. Now is a weight space of . Then repeat the procedure, by is finite we know has a weight. \end{proof}

Nilpotent Lie algebra

Proposition

Let be a finite dimensional nilpotent Lie algebra over and be a finite dimensional complex representation. Then:

  • for , the generalized eigenspaces of are -invariant;
  • is the direct sum of generalized weight spaces.

\begin{proof}i) Let be an eigenvalue of and the according generalized weight space. It suffices to show for any and any , , that is,

for some . Note that

and is also nilpotent. So for a big enough , .

ii) Choose a basis of where . Write where each is a generalized eigenspace of . Each is -invariant and hence can be decomposed as the direct sum of generalized eigenspaces of , written as . Continue the process we get where

Define

\begin{aligned} \lambda_{i_1\cdots i_n}:g\to&\mathbb C\\\sum a_jY_j\mapsto&\sum a_jc_{j,i_1\cdots i_n}\end{aligned}.$$ Claim that $E_{i_1\cdots i_n}$ is the generalized weight space associated to $\lambda_{i_1\cdots i_n}$. $g$ is nilpotent yields $g$ is solvable and then there is a basis such that $\rho(Y)|_{E_{i_1\cdots i_n}}$ is an upper triangular matrix for all $Y\in g$. Then

\begin{aligned} Y_j&\to\left[ \begin{matrix} c_{j,i_1\cdots i_n} & & * \ & \ddots & \ & & c_{j,i_1\cdots i_n} \end{matrix} \right]\ Y=\sum a_j Y_j&\to\left[ \begin{matrix} \sum a_jc_{j,i_1\cdots i_n} & & * \ & \ddots & \ & & \sum a_jc_{j,i_1\cdots i_n} \end{matrix} \right] \end{aligned}

So $\rho(\sum a_j Y_j)v_1=(\sum a_jc_j)v_1$, whence $\lambda_{i_1\cdots i_n}$ is a weight. And $[\rho(Y)-\lambda_{i_1\cdots i_n}(Y)]^m=0$ on $E_{i_1\cdots i_n}$ for some $m$. Therefore $E_{i_1\cdots i_n}$ is the generalized $\lambda_{i_1\cdots i_n}$-weight space. `\end{proof}`