We have the following obvious lemma.
admits an action of . Namely,
is an -submodule of .
On the other hand, let us take an arbitrary , then for any , we have
So (as a trace of a nilpotent element.) Thus
also holds.
(2):follows from the general theory.
(3):follows easily from the definition of .
(4):Let us take an arbitrary . For any , we have by using(2)
Thus .
(5): follows clearly from (3).
(6): follows from (4),(5) and Engel's theorem. (We need to note that
holds.) (7): The action of on is equal to zero. So admits an action by , which is semisimple. Now consider the following exact sequence of -modules.
where is the kernel of the restriction map. By a special case of Weyl's theorem on complete reducibility (Lemma 5.51), We see that the sequence splits. (Since we assumed ). This implies that there exists an element such that
Thus gives a splitting of the injection . A Levi subalgebra of is obtained by putting
where is a semisimple (Levi) subalgebra of , and is a solvable (radical) ideal of .
Then from the definition, we is an abelian Lie algebra. It is also easy to verify that is an ideal of . ( is a characteristic ideal of ). We apply the preceding lemma for to obtain a Levi subalgebra of . Then satisfies the following relations.
Since is solvable (and we have assumed ), we see that is strictly smaller than . By induction have a Levi subalgebra . Then it is clear that is a Levi subalgebra of .
ARRAY(0x941b644)