Maximal torus of a positive dimensional compact Lie group is nontrivial?

Let $G$ be a compact Lie group. A torus in $G$ is a subgroup $T \leq G$ isomorphic to $(S^1)^n$ for some $n \geq 0$, where we set $(S^1)^0 = \{*\}$ which is the trivial torus.

A standard theorem asserts the existence of a maximal torus $T$ in $G$, can it be trivial when $G$ has positive dimension? Why not?

• The closure of an abelian subgroup is again an abelian subgroup. Does $G$ have nontrivial connected abelian subgroups? – Daniel Fischer Sep 27 '16 at 10:41
• Sorry, I don't see how that helps. A trivial torus is closed because the Lie group is a Hausdorff space, isn't it? So starting with a trivial torus, I don't see why it has to be contained in a larger one. – hallborrey Sep 27 '16 at 11:00
• Don't start with a torus, start with a nontrivial connected abelian subgroup. And look at this or this. Plus this. – Daniel Fischer Sep 27 '16 at 11:12