Is every spin group $Spin(n,R)$ over the reals contained in some metaplectic group $Mp(m,R)$ for some $m$ in such a way that the spin representation is obtained by restriction of the metaplectic representation?
-
1$\begingroup$ For those interested, the metaplectic group is the connected double cover of the symplectic group $Sp(m,\mathbf{R})$ (acting on $\mathbf{R}^{2m}$ or possibly $\mathbf{R}^m$ assuming $m$ even, according to the conventions) en.wikipedia.org/wiki/Metaplectic_group $\endgroup$– YCorCommented May 21, 2015 at 15:08
-
3$\begingroup$ The answer is yes, because the spin group, as a linear group embeds into some $GL_k(\mathbf{R})$, which itself embeds into $Sp(2k,\mathbf{R})$; then by simple connectedness of Spin, the latter lifts to an embedding of Spin into the metaplectic group. $\endgroup$– YCorCommented May 21, 2015 at 15:10
-
$\begingroup$ @YCor I guess the question asked implicitly for a morphism $SO(n)\to Sp(m)$ such that the pullback of the diagram $SO(n)\to Sp(m)\leftarrow Mp(m)$ is isomorphic to $Spin(m)$. $\endgroup$– few_repsCommented May 21, 2015 at 22:13
-
1$\begingroup$ I don't see how you guess this, but we'll see if the question is edited. $\endgroup$– YCorCommented May 21, 2015 at 22:37
-
$\begingroup$ @YCor: I made my question more precise. $\endgroup$– Arnold NeumaierCommented May 22, 2015 at 8:42
2 Answers
Let $\alpha : 1\to K \to Spin(m)\to SO(m)\to 1$ be the universal central extension, so that $K=\mathbf Z$ if $m= 2$ and $K=\mathbf Z/2$ if $m\geq 3$.
Let $\beta : 1\to \mathbf Z \to \tilde{Sp}(n)\to {Sp}(n)\to 1$ be the universal central extension.
Let $\gamma : 1\to \mathbf Z/2 \to {Mp}(n)\to {Sp}(n)\to 1$ be the metaplectic central extension.
I interpret the question as : can one find a morphism $f:SO(m)\to Sp(n)$ such that the pullback $f^*(\gamma)$ of $\gamma$ along $f$ be isomorphic to $\alpha$ ?
It seems that the answer is always no. Indeed,
for $m\geq 3$, $f^*(\gamma)$ is the image of $f^*(\beta)$, which is an extension of $SO$ by $\mathbf Z$ and hence is trivial.
for $m=2$, it is impossible since $\mathbf Z\neq \mathbf Z/2$. Nevertheless, in this case, if $f$ is the inclusion $SO(2)\to SL(2)$, then $f^*(\beta)$ is isomorphic to $\alpha$.
The fact is that because $Sp(n, \mathbb{R})$ is diffeomorphic to the product of the unitary group $U(n)$ and an Euclidean space. So, the fundamental group of $Sp(n, \mathbb{R})$ is $\mathbb{Z}$. So $Sp(n, \mathbb{R})$ has unique double covering, which we denote by $Mp(n, \mathbb{R})$ and is called Metaplectic group. Note also that, the metaplectic group $Mp(2,\mathbb{R})$ is not a matrix group, so metaplectic group is a little bit complicate.
To have a better picture of metaplectic group we give a general definition for it. Let $(V, \omega)$ be a symplectic vector space with $dimV=2n$ over $\mathbb{F}$ (here $\mathbb{F}$ is a nonarchimedean local field of characteristic 0 and residual characteristic $p$) with associated symplectic group $Sp(V)$. The group $Sp(V)$ has a unique two-fold central extension $Mp(V)$ which is called the metaplectic group: $$0\to \{\pm 1\}\to Mp(V)\to Sp(V)\to 0.$$ So, we can write $Mp(V)=Sp(V)\oplus \{\pm 1\}$ with group law given by $$(g_1, \epsilon_1).(g_2, \epsilon_2) = \left(g_1g_2,\epsilon_1\epsilon_2c(g_1, g_2)\right)$$ for some 2-cocycle $c$ on $Sp(V)$ valued in $\{\pm 1\}$.
Now, note that we have the natural embedding $GL(n,\mathbb{R})\hookrightarrow Sp(n,\mathbb{R})$ given by
$$A\mapsto \left[\matrix{ A&0\cr0&A^{*-1}}\right]$$ where $A^*$ is the transpose of $A$.
So, $GL(n,\mathbb{R})$ can be viewed as subgroup in $Sp(2n, \mathbb{R})$ as the subgroup that preserves the standard Lagrangian submanifold $\mathbb{R}^n\hookrightarrow \mathbb{R}^{2n}$. Restriction of the metaplectic group extension along this inclusion defines the metalinear group $Ml(n,\mathbb{R})$.
\begin{array}{lll} ML(n,\mathbb{R}) & \rightarrow &Mp(2n,\mathbb{R})\\ \downarrow && \downarrow \\ GL(n,\mathbb{R}) & \rightarrow & Sp(2n,\mathbb{R}) \end{array}
For the definition of metalinear group, the quotient $ML(n,\mathbb{C}):=({\mathbb{C}\times SL(n,\mathbb{C})})/{2\mathbb{Z}}$ is called as complex metalinear group of dimension $n$. The elements of $ML(n,\mathbb{C})$ can be written as following forms $$\overline{(m,B)}=\left\{\left(m+\frac{4\pi ik}{n},e^{-\frac{4\pi ik}{n}}B\right):k\in {\mathbb{Z}} , \right\}$$
where $B\in SL(n,\mathbb{C}) $
So, we will have a covering map
$$\rho :ML(n,\mathbb{C})\to GL(n,\mathbb{C}),$$ $$\overline{(m,B)}\mapsto e^mB$$
See my expose in Lille, 2014
-
$\begingroup$ Your post has no introduction and no conclusion, it's hard to guess what you claim to prove (and actually I don't understand Arnold's question in its edited [22 May 2015] form; I only understand its interpretation by @few_reps hoping that it matches the desired question - which is probably the case since his/her answer was accepted). $\endgroup$– YCorCommented Jun 8, 2017 at 9:10