I want to prove that on a semisimple Lie algebra $\mathfrak{g}$ over ${\bf R}$:

$\mathfrak{g}$ is compact if and only if the Killing form is strictly negative definite.

Here the Lie algebra is compact if Int$(\mathfrak{g})$, which is a Lie group of $\{ ad_X | X\in \mathfrak{g}\}$, is compact.

I am reading a Helgason's book : The proof is short and I cannot catch the strategy.

The proof : If Killing form $B$ is strictly negative definitie then $O(B)$ is a set of all linear transformations which leave $B$ invariant. Then $O(B)$ is compact. I cannot understand.

Thank you in advance.


The key is to look at just the vector space structure of $\mathfrak g$ and then use some nice topological properties.

So for now, think of $\mathfrak g$ as just a real vector space, say $\mathbb R^n$, so that the Killing form is essentially an inner product. Since the Killing form is negative definite, there is a an element $A \in GL_n(\mathbb R)$ with all negative eigenvalues such that $B(x,y) = x^T A y$. The collection of all linear transformations which preserve B will be this more generalized orthogonal group $O(B)$, which is closed and bounded by precisely the same arguments as those showing that $O(n)$ is compact (just no longer using the Euclidean inner product).

Now $B$ is adjoint invariant and all Lie algebra automorphisms of $\mathfrak g$ must preserve the commutator (and hence the adjoints) we get that $\operatorname{Aut}(\mathfrak g) \subseteq O(B)$. Furthermore, $\operatorname{Int}(\mathfrak g)$ is a closed subgroup of $\operatorname{Aut}(\mathfrak g)$ and so is also a closed subspace of $O(B)$ and hence compact. Thus the Lie algebra is also compact as required.

  • $\begingroup$ We do not need the negativity in the proof right? $\endgroup$ – S. D. ZHU Feb 23 '20 at 4:26

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.