8
$\begingroup$

Is every closed connected proper subgroup of $SO(2n+1)$ isomorphic (as a Lie group) to a subgroup of $SO(2n)$? The answer is yes for abelian closed connected subgroups.

$\endgroup$
2
  • 1
    $\begingroup$ It seems to me at this moment that there's no copy of $SO(n,\mathbb R)\times SO(n+1,\mathbb R)$ in $SO(2n+1,\mathbb C)$. Is this sort of thing relevant in the context of your question, or did you intend everything to be complex, or everything to be real, etc.? $\endgroup$ May 1, 2018 at 20:42
  • $\begingroup$ @paulgarrett: I wanted all groups to be real, just the way I asked. $\endgroup$ May 1, 2018 at 20:48

2 Answers 2

12
$\begingroup$

No for $2n+1\ge 7$: $\mathfrak{so}(3)\oplus\mathfrak{so}(2n-2)$ does not embed into $\mathfrak{gl}(2n)$.

Indeed start with $2n+1\ge 9$. Consider a nontrivial irreducible (real) representation of this product, nontrivial on both factors, of dimension $d$. It is isotypic for $\mathfrak{so}(3)$, say of the form $\mathfrak{v}_k^{\oplus m}$ with $\mathfrak{v}_k$ $k$-dimensional. If $k$ is odd, $\mathfrak{v}_k$ is absolutely irreducible, and hence the centralizer is isomorphic to $\mathfrak{gl}_m$, and hence $m\ge 2n-2$, so $$d=km\ge 3m\ge 3(2n-2)=6n-6=(2n+1)+(4n-5)\ge 2n+1.$$ Otherwise $k$ is a multiple of 4 (so $k\ge 4$) and the complexification is a sum of $2m$ $(k/2)$-dimensional irreducible representations, and hence the complexification of the centralizer is $\mathfrak{gl}_{2m}(\mathbf{C})$. So $2m\ge 2n-2$, so $$d=km\ge 4m\ge 4(n-1)=4n-1\ge 2n+1.$$

Such irreducible being discarded, one remains with sums of irreducibles factoring through one quotient, and the smallest faithful one is clearly $(2n+1)$-dimensional.

For $2n+1=7$, one has to show that $\mathfrak{so}(3)\oplus\mathfrak{so}(4)\simeq\mathfrak{so}(3)^3$ does not embed into $\mathfrak{gl}(6)$. This is seen by hand: start with an irreducible of dimension $\le 6$, and suppose that it's nontrivial on the first factor. Then on the first factor, it's either $\mathfrak{v}_3$, $\mathfrak{v}_5$, $\mathfrak{v}_4$ or $\mathfrak{v}_3^{\oplus 2}$. The centralizer is abelian in the first two cases, and is isomorphic to $\mathfrak{so}_3\times\mathbf{R}$ and $\mathfrak{gl}_2$ in the last two cases. This shows that for $\mathfrak{so}(4)$, the only irreducible of dimension $\le 6$ (actually $\le 7$) that is nontrivial on both factors is 4-dimensional, and that there is no irreducible of $\mathfrak{so}(3)^3$ of dimension $\le 6$ that is nontrivial on each factor. If we allow only irreps that are nontrivial on only one factor, one has to add up to 9 to get faithfulness. So one irrep should be nontrivial on 2 factors, hence 4-dimensional. The remaining 2 dimensions are not enough to add anything.

Remains $2n+1=5$. At the level of Lie algebras there's no obstruction. Indeed any subalgebra of the 10-dimensional $\mathfrak{so}(5)$ has codimension $\ge 4$, that is dimension $\le 6$. Hence, if nonabelian it's isomorphic to $\mathfrak{so}(4)$, or $\mathfrak{so}(3)+\mathbf{R}^{k}$ for some $k$. A centralizer argument shows that $k\le 1$, so this indeed embeds into $\mathfrak{so}(4)$. I haven't checked at the group level.

$\endgroup$
3
  • $\begingroup$ A weaker version of my question is: does every Lie subalgebra of $\mathfrak{so}(2n+1)$ is a product of Lie algebras that each embed into $\mathfrak{so}(2n)$? Do you think this is likely to happen? Thank you! $\endgroup$ May 1, 2018 at 21:48
  • 1
    $\begingroup$ @IgorBelegradek or equivalently does every proper simple subalgebra of the former embed in the latter. I guess so; I've just looked into Dynkin's complete works... it's possibly inside but it's really hard to sort things out... $\endgroup$
    – YCor
    May 1, 2018 at 23:27
  • $\begingroup$ As Robert Bryant explains in his answer, what I asked for does not happen. $\endgroup$ May 1, 2018 at 23:46
10
$\begingroup$

As an answer to your weaker version: No. For example, $\mathrm{G}_2$ is a closed proper subgroup of $\mathrm{SO}(7)$, but it is not isomorphic to any subgroup of $\mathrm{SO}(6)$. Of course, this follows from the fact that ${\frak{g}}_2$ has no nontrivial representation of dimension less than $7$.

$\endgroup$
10
  • $\begingroup$ Haha! A crazy fact! :) $\endgroup$ May 1, 2018 at 23:05
  • $\begingroup$ It might me the only exception? I guess the answer is yes for large $n$. $\endgroup$
    – YCor
    May 2, 2018 at 8:31
  • $\begingroup$ @YCor: When one phrases the question as "Which simple Lie algebras over $\mathbb{R}$ of compact type have the property that their nontrivial representation(s) of lowest dimension have odd dimension?", then, yes, the answer is only ${\mathfrak{so}}_{2n+1}$ and ${\mathfrak{g}}_{2}$. $\endgroup$ May 2, 2018 at 11:02
  • $\begingroup$ @RobertBryant you mean, all real forms of these types? $\endgroup$
    – YCor
    May 2, 2018 at 11:05
  • $\begingroup$ @YCor: Well, I would imagine that "simple Lie algebra over $\mathbb{R}$ of compact type" does imply "real form". Since the original question was about closed subgroups of $\mathrm{SO}(2n{+}1)$ which are compact, I feel that it is justified to restrict to simple Lie algebra over $\mathbb{R}$ of compact type, and the 'weaker' question seems to be about simple subalgebras of ${\mathfrak{so}}_{2n+1}$, which are also of this type. Hence my rephrasing of the question in the form that I did. $\endgroup$ May 2, 2018 at 11:12

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.