$\DeclareMathOperator\Spin{Spin}$The seven-sphere can be written as the reductive space $S^7=\Spin(7)/G_2$. Has the decomposition $\Spin(7)=G_2\times S^7$ been calculated somewhere; maybe in terms of Cayley numbers?
-
6$\begingroup$ What makes you think that $Spin(7)=G_2\times S^7$? Think of the Hopf fibration $S^2=S^3/S^1$: would you then conclude that $S^3=S^2\times S^1$? $\endgroup$– André HenriquesCommented May 19, 2014 at 2:09
-
$\begingroup$ @Andre: A result in Helgason says that such a decomposition exists for reductive spaces; at least locally. In the case of $S^2$ we have $SO(3)=SO(2)\times S^2$. $\endgroup$– Oliver JonesCommented May 19, 2014 at 2:27
-
1$\begingroup$ But that's wrong, Oliver. $SO(3)$ is not the product of $SO(2)$ and $S^2$. The homotopy groups don't work out. It's a non-trivial fibre bundle, $SO(2) \to SO(3) \to S^2$ $\endgroup$– Ryan BudneyCommented May 19, 2014 at 2:29
-
4$\begingroup$ The statement $SO(3) = SO(2) \times S^2$ isn't a local statement. What kind of "local" are you talking about? If you mean in the sense of fibre bundles, yes there is a fibre bundle, but it is not trivial (as your notation presumes). $\endgroup$– Ryan BudneyCommented May 19, 2014 at 2:30
-
1$\begingroup$ You're also implicitly assuming that there is a preferred way of trivializing the bundle $Spin(7)\to G_2\to S^7$ in a neighborhood of a given point $p\in S^7$. My guess is that there are many ways of locally trivializing that bundle, and that they are all pretty ugly to write down. $\endgroup$– André HenriquesCommented May 19, 2014 at 4:19
2 Answers
This is too long for a comment. (But I have now added an answer to the original question.)
As André and Ryan have hinted at in the comments, what you have is a principal fibre bundle $G_2 \to \operatorname{Spin}(7) \stackrel{\pi}{\to} S^7$. Fibre bundles are locally trivial, so of course around every point of the 7-sphere there is a neighbourhood $U$ so that $\pi^{-1}U \cong U \times G_2$. The diffeomorphism depends on a choice of section: a way to assign to every point in $U$ a unique element of $\operatorname{Spin}(7)$.
What I know how to describe in terms of octonions (i.e., Cayley numbers) is the above fibre bundle, and maybe this helps you.
Let $\mathbb{O}$ denote the Cayley numbers. They form an 8-dimensional vector space with basis $e_1,\dots,e_8$, where $e_1,\dots,e_7$ are imaginary units and $e_8 = 1$. Let $L_i$ denote left multiplication by the imaginary unit $e_i$, for $i=1,\dots, 7$. The $L_i$ are endomorphisms of $\mathbb{O}$ which obey the Clifford relations $$ L_i \circ L_j + L_j \circ L_i = \begin{cases} - 2\operatorname{id} & i=j \cr 0 & i\neq j \end{cases}$$ whence they define a linear representation of the Clifford algebra $C\ell(7)$ on $\mathbb{O}$. (You could also use right multiplication and this would give the other inequivalent Clifford module of $C\ell(7)$. Both Clifford modules turn out to be equivalent under the spin group.)
The spin group $\operatorname{Spin}(7)$ naturally lives inside $C\ell(7)$, whence we also have a linear representation of $\operatorname{Spin}(7)$ on $\mathbb{O}$. This is nothing but the spinor representation $\operatorname{Spin}(7) \to \operatorname{SO}(\mathbb{O})$. The orbit of $1 \in \mathbb{O}$ under $\operatorname{Spin}(7)$ is the sphere of unit octonions, which we can identify with $S^7$. The stabiliser of $1$ in $\operatorname{Spin}(7)$ is precisely a $G_2$ subgroup. This then gives the principal bundle $$G_2 \to \operatorname{Spin}(7) \stackrel{\pi}{\to} S^7$$ with $\pi(g) = g \cdot 1$.
Added
Here is how to construct a local section of $S^7 \to \operatorname{Spin}(7)$ near $1 \in S^7 \subset\mathbb{O}$ using octonions.
Under the identification of $S^7$ with the unit-norm octonions, the tangent space $T_1 S^7$ is identified with the imaginary octonions $\operatorname{Im}(\mathbb{O})$. Let $\varphi$ denote the $G_2$-invariant 3-form on $\operatorname{Im}(\mathbb{O})$. The image of the map $\operatorname{Im}(\mathbb{O}) \to \Lambda^2\operatorname{Im}(\mathbb{O})$ which sends $\xi$ to $\iota_\xi \varphi$ is a $G_2$-invariant complement of $\mathfrak{g}_2 \subset \mathfrak{spin}(7)$, and we can exponentiate them to elements of $\operatorname{Spin}(7)$ near the identity. Acting on $1 \in S^7$ we coordinatise a neighbourhood of $1$.
More explicitly, in case this is useful, the components $\varphi_{ijk}$ of $\varphi$ relative to the basis $e_i$ above are given by octonion multiplication as follows: $$ e_i e_j = -\delta_{ij} 1 + \sum_{k=1}^7 \varphi_{ijk} e_k $$ Then the point of $S^7$ with coordinates $\xi = \sum_{i=1}^7 \xi_i e_i \in \operatorname{Im}(\mathbb{O})$, is given by $$ \exp\left(\tfrac12 \sum_{i,j,k=1}^7\xi_i \varphi_{ijk} L_j\circ L_k\right) \cdot 1 $$ where the $L_j$ endomorphisms were defined above.
-
1$\begingroup$ José's clear and simple explanation also shows that Spin(7) can be written as a group of 8-by-8 matrices, not 7-by-7. As for a "polar decomposition", perhaps you (Oliver) should start by showing us what you mean for a simpler example such as the standard Hopf fibration $S^2 = S^3/S^1$. $\endgroup$ Commented May 19, 2014 at 4:55
-
3$\begingroup$ David: OliverJones only cares about what happens locally. And locally, $Spin(7)$ and $SO(7)$ are the same, $\endgroup$ Commented May 19, 2014 at 5:13
-
1$\begingroup$ @OliverJones: The situation with hyperbolic space is slightly different. First, $SO(3,1)$ is not connected, so what you have written is not strictly speaking hyperbolic space, but the full two-sheeted hyperboloid in Minkowski spacetime. It may be better to write $H^3 = SL(2,\mathbb{C})/SU(2)$. It is indeed the case that $SL(2,\mathbb{C})$ is (analytically, even) diffeomorphic to $SU(2) \times H^3$, but this is because $SU(2)$ is the maximal compact subgroup of $SL(2,\mathbb{C})$. That's not the case in the example you are interested in. $\endgroup$ Commented May 19, 2014 at 8:46
-
1$\begingroup$ (cont'd) Having said that, this example is different than the Hopf fibration is that whereas it is clear that $S^3$ and $S^2 \times S^1$ do not have the same de Rham cohomology, say, $Spin(7)$ and $G_2 \times S^7$ do. $\endgroup$ Commented May 19, 2014 at 8:48
-
2$\begingroup$ What you can do for $G/K$, where $K$ is compact, is to fix a reductive decomposition $\mathfrak g = \mathfrak k + \mathfrak p$, namely, $[\mathfrak k,\mathfrak p]\subset\mathfrak p$. Then you can use the exponential map of $G$ restricted to $\mathfrak p$, fix a basis of $\mathfrak p$, and use canonical coordinates of the first kind. This will give you local coordinates around the base point in $Spin_7/G_2$. In principle you can view $Spin_7$ as real matrices of order $8$, $G_2$ as a subgroup of matrices, and write a product decomposition in a nbhd of 1. $\endgroup$ Commented May 19, 2014 at 11:58
The bundle $G_2 \to Spin(7) \to S^7$ is nontrivial, indeed. The formula above $$g_{S^7} = \exp\left\{ \frac 12 \xi_j \phi_{jkl} \Gamma_k \Gamma_l \right \} \qquad \qquad (1)$$ ($\Gamma_j$ are the matrices $8 \times 8$ satisfying the Clifford algebra) is a local section, but not a global section of this bundle. It can be proven in quite explicit "physical" way. I do so in 2102.07415 [hep-th]. The digest of the proof is the following.
(1) is a section because
(i) Seven generators $T_j = \phi_{jkl} \Gamma_k \Gamma_l$ are orthogonal to 14 generators of $G_2$ and
(ii) (1) parameterize the whole $S^7$: the point $\xi_j = 0$ is its north pole with $g= \mathbb{1}$ and, if $\|\xi_j\| = \pi$, one obtains $g = - \mathbb{1}$, and this is the south pole.
However, (1) is not a global section by the following not quite trivial reason:
A subgroup $G_2 \subset Spin(7)$ are the matrices that do not transform a particular 8-component real spinor $\psi_0$. Consider the action of the matrices (1) on this spinor. One can show that $g \psi_0 = \cos(3\alpha) \psi_0 + \sin(3\alpha) \psi_1$, where $\alpha = \|\xi_j\|$ and $\psi_1$ is some other spinor. If $\alpha = 0, \pm 2\pi/3$, $\psi_0$ is left intact. Thus, many elements of (1) belong to one and the same fiber of the bundle and hence (1) cannot be a section.