A Lie algebra $L$ is said to be reductive if for any ideal $\mathfrak{a}$ of $L$, there is an ideal $\mathfrak{b}$ of $L$ such that $L=\mathfrak{a}\oplus\mathfrak{b}$.

It is known that a reductive Lie algebra decomposes as $L = L'\oplus Z(L)$, where $L'$ is the derived subalgebra of $L$ and $Z(L)$ is the center of $L$. Is the converse true?

The answer is "yes" if $L'$ is semisimple; can we deduce that $L'$ is semisimple just from the decomposition $L = L'\oplus Z(L)$?

  • $\begingroup$ Could you provide a reference for defining reductive like this? The definition I am used to is being the direct sum of the radical and the center. $\endgroup$ Jan 30, 2015 at 14:10
  • $\begingroup$ @TobiasKildetoft this is the definition used in Structure and Geometry of Lie Groups, by Hilgert and Neeb, for instance. $\endgroup$
    – essay
    Jan 30, 2015 at 17:37

1 Answer 1


We cannot deduce that. Let $L$ be a finite dimensional semisimple Lie algebra and $V$ be an irreducible representation of $L$ where $1<\dim V<\infty$. Define $L \times V$ a Lie algebra structure by $([X, u], [Y, v])=([X,Y], Xv-Yu)$. Then you can check the followings:

  1. $[L \times V, L \times V]=L \times V$, i.e. $L$ is perfect.
  2. $Z(L \times V)=0$
  3. $V=\{ 0 \} \times V$ is an ideal of $L \times V$. $\nexists$ ideal $I$ of $L \times V$ such that $L \times V = V \oplus I$.

This shows that $L$ is not semisimple (as it is not reductive). We can also prove that $L$ is not semisimple by calculating its radical.

Claim. $Rad(L \times V) = V$.
proof. It suffices to show $Rad(L \times V) \subset V$. If $(X, u) \in Rad(L \times V)$ with $X \neq 0$, we can prove inductively that there exists $0 \neq X_n \in L$ such that $(Y, 0) \subset (Rad(L \times V))^{(n)}$ which contradicts to the solvability, by the following lemma:

Lemma. Let $I$ be a nonzero ideal of a semisimple Lie algebra $L$. Then $[I, I] \neq0$.
proof. By Weyl's theorem, there exists an ideal $J$ of $L$ such that $L=I \oplus J$. If $[I, I]=0$, we have $L=[L,L]=[I,I]+[J,J] \subset [J,J] \subset J$ so contradiction occurs.

In fact, you can deduce $L=[L, L] \oplus Z(L)$ if $[L, L]$ is semisimple. It follows that $L$ is reductive if $[L, L]$ is semisimple. To prove this, view $L$ as a $[L, L]$-module. By Weyl's theorem, $L=[L, L]\oplus M$ for some $[L, L]$-submodule $M$ of $L$. Since $[L, L] \cap Z(L)=0$, it suffices to show the following claim:

Claim. $M \subset Z(L)$.
proof. Let $x \in M$ be given, and suppose there exists $y \in L$ such that $[x, y] \neq 0$. Since $[L, L]$ is semisimple and $0 \neq [x, y] \in [L, L]$, there exists $z \in [L, L]$ such that $[z, [x, y]] \neq 0$. However, Jacobi identity yields $[z, [x, y]] = [x, [z, y]] + [y, [x, z]]$. Here $[x, [z,y]]=0$ since $[z,y] \in [L, L]$ and $L=[L, L] \oplus M$. Similarly, $[y,[x,z]]=[y,0]=0$. Hence $[z, [x,y]]=0$; contradiction.

  • 1
    $\begingroup$ It seems I misunderstood the question, thank you for pointing it out. I deleted my (wrong) answer. $\endgroup$
    – Cronus
    Jun 29, 2020 at 15:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.