Following the Wiki's definition, given a semisimple Lie algebra $\mathfrak g$, a toral subalgebra is a subalgebra whose elements have semisimple adjoint operators (I'm not quite sure whether one defines toral subalgebras for more general cases, the wiki isn't very clear).
It is mentioned there, as well as in this other question, that when $\mathfrak g$ is defined over an algebraically closed field, any toral subalgebra is abelian. A sketch of the argument as I understand it being:
- given toral $\mathfrak t\le \mathfrak g$, and $x\in\mathfrak t$, if $\operatorname{ad}(x)\neq 0$ then it being a diagonalisable operator in an algebraically closed field, there's $y\neq0$ such that $[x,y]=\lambda y$ for some $\lambda\neq0$.
- thus $\operatorname{ad}(y)^2x=0$. But a semisimple (ie diagonalisable) matrix $A$ such that $A^2x=0$ must also satisfy $Ax=0$, which is a contradiction. So $[x,y]=0$ for all $x,y\in\mathfrak t$.
What are some examples of toral subalgebras (in non-algebraically-closed fields) that are not abelian?