6
$\begingroup$

I am sorry if this question is too elementary to be posted here, but no experts answer this question when I post it on Math Stackexchange.

Let $\mathfrak{g}=\mathfrak{k}+\mathfrak{p}$ be a Cartan decomposition for a noncompact real simple Lie algebra $\mathfrak{g}$ corresponding to a Cartan involution $\theta$, where $\mathfrak{k}$ is the maximal compact subalgebra of $\mathfrak{g}$. Suppose that $\sigma$ is another involutive automorphism of $\mathfrak{g}$ such that $\sigma\theta=\theta\sigma$. Then $\sigma$ preserves the Cartan decomposition, and $\sigma|_\mathfrak{k}:\mathfrak{k}\rightarrow\mathfrak{k}$. By the classification for symmetric pairs, it seems true that $\sigma|_\mathfrak{k}$ is never the identity map, but how to prove this fact theoretically (instead of case by case)?

In other words, there is no (non-Cartan) involutive automorphism $\sigma$ of a noncompact real simple Lie algebra $\mathfrak{g}$ such that the subalgebra $\mathfrak{g}^\sigma$ of the fixed points under the action of $\sigma$ on $\mathfrak{g}$ contains a maximal compact subalgebra. How to prove it? I shall be grateful if experts here may offer any hint.

$\endgroup$
  • $\begingroup$ Did you delete the question on M.SE? I tried to find it, to link between the two, but I couldn't. $\endgroup$ – David Roberts Jun 4 '20 at 6:49
  • $\begingroup$ @DavidRoberts Yes, I deleted the question on M.SE after I posted the question here. Do I need to undelete it? The link is math.stackexchange.com/questions/3703866/…, but I am not sure whether it appears. $\endgroup$ – Hebe Jun 4 '20 at 7:09
  • $\begingroup$ No, it's ok. If it got no traction over there (like: no answers at all, no substantial comments), then it's better to not have a duplicate and zombie question floating around. The situation would be different if someone had put up a helpful answer but which still wasn't a full answer, or answered a slightly different question. $\endgroup$ – David Roberts Jun 4 '20 at 21:58
  • 1
    $\begingroup$ I guess $\sigma=id$ doesn't count? $\endgroup$ – Torsten Schoeneberg Jun 7 '20 at 17:33
  • 1
    $\begingroup$ @TorstenSchoeneberg I think that in general people do not regard the identity map as an involutive automrphism, but I am not sure. Anyway, here I just suppose that $\sigma$ is not identity on $\mathfrak{g}$. $\endgroup$ – Hebe Jun 8 '20 at 8:52
0
$\begingroup$

Let $\mathfrak g$ be a noncompact simple Lie algebra and let $\mathfrak g=\mathfrak k+\mathfrak p$ be a Cartan decomposition. The simplicity of $\mathfrak g$ implies that the adjoint representation of $\mathfrak k$ on $\mathfrak p$ is irreducible (indeed, if $\mathfrak p_1$ is an $\mathrm{ad}_\mathfrak k$-invariant subspace of $\mathfrak p$, one can show$^\dagger$ that $\mathfrak g_1:=[\mathfrak p_1,\mathfrak p_1]+\mathfrak p_1$ is an ideal of $\mathfrak g$). In particular, $\mathfrak k$ is a maximal (not only maximal compact) subalgebra of $\mathfrak g$. Indeed, if $\mathfrak h$ were a subalgebra containing $\mathfrak k$, then $\mathfrak h =\mathfrak k +\mathfrak h\cap\mathfrak p$ and $\mathfrak h\cap\mathfrak p$ would be an $\mathrm{ad}_{\mathfrak k}$-invariant subspace of $\mathfrak p$.

$\dagger$ Write $\mathfrak p=\mathfrak p_1+\mathfrak p_2$ $\mathrm{ad}_{\mathfrak k}$-invariant decomposition. Let $X_i\in\mathfrak p_i$ and $Y=[X_1,X_2]\in\mathfrak k$. Denote the Cartan-Killing form of $\mathfrak g$ by $B$; it is negative definite on $\mathfrak k$ and positive definite on $\mathfrak p$; in particular, we may assume $B(\mathfrak p_1,\mathfrak p_2)=0$. Now $B(Y,Y)=B(Y,[X_1,X_2])=B([Y,X_1],X_2)\in B(\mathfrak p_1,\mathfrak p_2)=0$, so $Y=0$. We have shown that $[\mathfrak p_1,\mathfrak p_2]=0$. Using Jacobi, one completes the check that $\mathfrak g_1$ is an ideal of $\mathfrak g$.

$\endgroup$
  • $\begingroup$ Thank you for the answer. So you just show that a maximal compact subalgebra in a simple Lie algebra is automatically a maximal subalgebra. Interesting. $\endgroup$ – Hebe Sep 8 '20 at 8:20

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.