MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Hello everyone
thanks to all of you

I have two questions and I hope to get some guide:
1. One of the Lie groups in the Berger's list of holonomy groups of locally irreducible Riemannian manifolds is $Sp(n).Sp(1)\cong Sp(n) \times Sp(1)/(\pm Identity)$, which can be considered as a Lie subgroup of $SO(4n)$. what is the explicit formula for this inclusion?
2. Can be $Sp(n)\times Sp(1).T^1$ a Lie subgroup of $SO(4n)\times T^1$, where $Sp(1).T^1 \cong Sp(1) \times T^1 /(\pm Identity)$?
I think the answer of the second question is negative, but I'm not sure.

share|cite|improve this question
up vote 2 down vote accepted

One way to make the subgroup $\mathrm{Sp}(n)\cdot\mathrm{Sp}(1)\subset\mathrm{SO}(4n)$ explicit is to think of $\mathbb{R}^{4n}$ as $\mathbb{H}^n$, i.e. as columns of quaternions of height $n$, where $\mathbb{H}\simeq\mathbb{R}^4$ is the ring of quaternions. Then $\mathrm{Sp}(n)$ can be thought of as the group of $n$-by-$n$ quaternion matrices $A$ that satisfy $A^\ast A = \mathrm{I}_n$, where $A^\ast$ is the conjugate transpose of $A$. This group then acts on $\mathbb{H}^n$ on the left by the usual matrix multiplication $A\cdot x = Ax$, for $A\in\mathrm{Sp}(n)$ and $x\in\mathbb{H}^n$. This action preserves the positive definite quadratic norm $\|x\|^2= x^\ast x$, so $\mathrm{Sp}(n)$ is exhibited as a subgroup of $\mathrm{SO}(4n)$.

This action is irreducible (it acts transitively on the unit sphere in $\mathbb{H}^n$), and its commuting ring is $\mathbb{H}$, thought of as scalar multiplication on the right: $$ A\cdot (x q) = (A\cdot x) q $$ for $q\in\mathbb{H}$ (which is true because quaternionic multiplication is associative). Remembering that $\mathrm{Sp}(1)$ is the group of unit quaternions, you can now see how $\mathrm{Sp}(n)\times\mathrm{Sp}(1)$ can act on $\mathbb{H}^n$, just use the rule $$ (A,q)\cdot x = A\ x\ \bar q $$ (the conjugation is important, because, otherwise, this won't be a left action). This action is not effective, because $(-A,-q)$ acts the same way as $(A,q)$, so you actually get a faithful representation of $\mathrm{Sp}(n)\cdot\mathrm{Sp}(1)\simeq \bigl(\mathrm{Sp}(n)\times\mathrm{Sp}(1)\bigr)/(\lbrace \pm\mathrm{I}_n,\pm 1\rbrace)$ into $\mathrm{SO}(4n)$.

The commuting ring of this representation is just $\mathbb{R}$, so there is no torus in $\mathrm{SO}(4n)$ that commutes with this action, which answers your second question.

share|cite|improve this answer
Thanks a lot for your helpful comments. I guess there is a misunderstanding in the second question. There, I mean $Sp(n)\times (Sp(1).T^1)$ not $(Sp(n)\times (Sp(1)).T^1$. Could you introduce me a good reference for what you,ve used in your comments, for example "commuting ring of an irreducible representation", please. – Nrd-Math Jan 3 '13 at 11:05
@Nerd-Math: Look up Schur's Lemma; it's basic. Maybe I don't understand what your $\mathrm{Sp}(1).T^1$ means, but I assume that you mean a subgroup of $\mathrm{SO}(4n)$ that commutes with the elements of $\mathrm{Sp}(n)$ and that contains both a copy of $\mathrm{Sp}(1)$ and a separate torus $T^1$ (i.e., a circle). However, the ring of $\mathbb{R}$-linear maps of $\mathbb{H}^n$ to itself that commute with all the elements of $\mathrm{Sp}(n)$ (i.e., its commuting ring) is the linear maps that come from multiplying by elements of $\mathbb{H}$ on the right, so the extra $T^1$ is not possible. – Robert Bryant Jan 3 '13 at 12:45
Pardon me if I am wrong. Your argument implies $Sp(n)\times Sp(1).T^1$ is not a Lie subgroup of $SO(4n)$. Isn't it? whereas the question is about $SO(4n)\times T^1$. – Nrd-Math Jan 3 '13 at 16:55
@Nerd-Math: Oh, sorry. I didn't read the question carefully. The answer is still 'no' because such an inclusion would cause an embedding of $\mathrm{Sp}(n)\times\mathrm{Sp}(1)$ as a subgroup of $\mathrm{SO}(4n)\times T^1$, and this is not possible, since it would have to be a subgroup of $\mathrm{SO}(4n)$, which is clearly impossible. – Robert Bryant Jan 3 '13 at 17:52
I'm sorry for this probably easy question: How does an embedding of $Sp(n)\times Sp(1)$ as a subgroup of $SO(4n)\times T^1$ yield being a subgroup of $SO(4n)$? – user30435 Jan 11 '13 at 11:10

For all $n,m\ge 0$ we have a homomorphism $Sp(n) \times Sp(m)/(\pm \mathrm{Id})\rightarrow SO(4nm)$. This is just the tensor product of the two defining representations.

share|cite|improve this answer
More precisely, it's the 'quaternionic tensor product'. In the usual representation, one has $\mathrm{Sp}(n)\subset\mathrm{SO}(4n)$, so the naïve tensor product would map $\mathrm{Sp}(n)\times\mathrm{Sp}(n)$ into $\mathrm{SO}(4m\cdot 4n)=\mathrm{SO}(16mn)$, which is not what you want. The point is that the commuting ring of $\mathrm{Sp}(n)\subset\mathrm{SO}(4n)$ acting on $\mathbb{R}^{4n}$ is the quaternions $\mathbb{H}$, and one can use this to construct an invariant subspace of dimension $4mn$ in the tensor product $\mathbb{R}^{4m}\otimes \mathbb{R}^{4n}\simeq \mathbb{R}^{16mn}$. – Robert Bryant Jan 2 '13 at 13:00
(correction) Of course, I meant to type $\mathrm{Sp}(m)\times\mathrm{Sp}(n)$ in my comment above instead of $\mathrm{Sp}(n)\times\mathrm{Sp}(n)$. – Robert Bryant Jan 2 '13 at 16:48

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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