1
$\begingroup$

I'm looking for a source of properties for semisimple Lie algebra elements, specifically finite dimensional classical Lie algebras.

I start with the assumption that I have a complexified Lie algebra $\mathfrak{g}^\mathbb{C}$, with a semisimple element $\Lambda \in \mathfrak{g}^\mathbb{C}$, and that it's conjugate $\bar{\Lambda}$ (under the complex structure of $\mathfrak{g}^\mathbb{C}$) has the property that, $$\left[ \Lambda , \bar{\Lambda}\right] = 0$$ Does this imply that $ker\> ad \> \Lambda = ker\> ad \> \bar{\Lambda}$, I know that in many cases I've worked with, it does, yet is there a reason why in general this is or isn't true?

$\endgroup$
1

1 Answer 1

4
$\begingroup$

First the counterexample: Let $\mathfrak g=\mathfrak{sl}(3,\mathbb C)$ and $$ \Lambda:=\left[ \begin {array}{ccc} 1&-3\,i&0\\3\,i&1&0 \\0&0&-2\end {array} \right] $$ Then $\Lambda$ and $\overline\Lambda$ commute and are semisimple. But only $\Lambda$ commutes with $$ \left[ \begin {array}{ccc} 0&0&i\\0&0&1 \\0&0&0\end {array} \right] $$ Now the explanation: Write $\Lambda=\Lambda_{re}+i\Lambda_{im}$ the decomposition into real and imaginary part. Then $\Lambda_{re},\Lambda_{im}$ span a semisimple commutative subalgebra of $\mathfrak g$. Therefore they lie in a Cartan subalgebra $\mathfrak t$ of $\mathfrak g$. The eigenvalues of $\mathfrak t$ on $\mathfrak g$ are the roots. If $\alpha$ is a root then its conjugate $\overline\alpha$ is a root, as well. So starting with $\mathfrak t$ we are done if we can find $\Lambda\in\mathfrak t$ and a root $\alpha$ with $\alpha(\Lambda)=0$ and $\overline\alpha(\Lambda)\ne0$. This is possible iff $\alpha$ and $\overline\alpha$ are linearly independent iff $\overline\alpha\ne\pm\alpha$. Roots with this property are called complex. In the example above, I chose $$ \left[ \begin {array}{ccc} x&-y&0 \\y&x&0\\0&0&-2 \,x\end {array} \right] $$ The eigenvalues are $x\pm iy$ and $-2x$. The roots are their differences. Then $(x+iy)-(-2x)=3x+iy$ is a complex root. So choose $x=1$ and $y=3i$.

$\endgroup$
1
  • $\begingroup$ If I asked for the elements to be outside of the cartan subalgebra, namely $\Lambda \not\in \mathfrak{t}^\mathbb{C}$ nor the real and imaginary parts in the real Cartan subalgebra, would this change the general picture at all? $\endgroup$
    – Jack Moon
    Aug 14, 2016 at 1:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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