4
$\begingroup$

I have been working with something related to Goldman bracket for $G_2$ gauge group. There I have something like "$\text{Tr}(M_{\gamma}O_i)$", where $M_{\gamma}$ is a monodromy which takes value in the group $G_2$ (in the fundamental representation) and $O_i, i=1,2,\ldots,7$" are the 7x7 skew-symmetric matrices that can be obtained from the 8x8 matrix representation of left multiplication by octonionic imaginary units deleting the first row and the first column.

My question is the following: can $M_{\gamma}O_{i}M_{\gamma}^{-1}$ be expressed as a suitable linear combination of the $O_i$'s? The matrices $O_i$'s are given explicitly in my paper: http://arxiv.org/pdf/1310.4519.pdf (eq. (4.3) p-19)

$\endgroup$

1 Answer 1

5
$\begingroup$

Yes. This follows from the fact that $G_2$ acts on the octonions via automorphisms.

Let $e_i$, $i=1,\dots,7$ be a choice of 7 imaginary octonion units and let $1$ denote the identity. Then by definition of the $O_i$, we have that $$ e_i e_j = \sum_k e_k (O_i)_{kj} - \delta_{ij} 1 $$ where the left-hand side is octonion multiplication.

Let $g \in G_2$. Since $G_2$ acts on the octonions via automorphisms, $$ g(e_i) = \sum_j e_j M_{ji}~. $$ The $M_{ji}$ are the entries of the matrix $M_\gamma$ in the question.

Now apply $g \in G_2$ on the first displayed equation. On the LHS we have $$ g(e_i e_j) = g(e_i) g(e_j) = \sum_{k,l} \left(\sum_m e_m (O_k)_{ml} - \delta_{kl} 1\right) M_{ki} M_{lj} = \sum_k (O_k M)_{mj} M_{ki} - \delta_{ij} 1~, $$ where we have used that $M$ is orthogonal, so that $\sum_{k,l} \delta_{kl} M_{ki} M_{lj} = \delta_{ij}$. Similarly, on the RHS we find $$ \sum_{m,k} e_m M_{mk} (O_i)_{kj} - \delta_{ij} 1 $$ using that $g(1) = 1$.

Comparing the two, we find $$ \sum_{k,l,m} e_m (O_k)_{ml} M_{ki} M_{lj} = \sum_{m,k} e_m M_{mk} (O_i)_{kj} $$ which is equivalent to $$ \sum_k (O_k M)_{mj} M_{ki} = (M O_i)_{mj} $$ or, finally, $$ M O_i M^{-1} = \sum_k O_k M_{ki}~. $$

Edit To answer the comment, there is a coordinate-free proof of this fact, and perhaps I should have written that first. What is $M O_i M^{-1}$? It is the matrix of left multiplication by the imaginary octonions (projected to the imaginary octonions) under a $G_2$ change of basis. But $G_2$ is an automorphism of the octonions, hence it preserves the subspace of imaginary octonions. Hence $M O_i M^{-1}$ is the matrix of left multiplication by some imaginary octonion unit, whence some linear combination of the $O_i$, since these form a basis.

$\endgroup$
11
  • $\begingroup$ I think you have RHS and LHS reveresed. Also, it may be worth to mention that you are using $g(e_i e_j) = g(e_i)g(e_j)$. $\endgroup$ Commented Aug 11, 2015 at 15:43
  • $\begingroup$ Indeed. But why do I have LHS and RHS reversed? The action of $g$ on the LHS of the equation is $g(e_i e_j)$... $\endgroup$ Commented Aug 11, 2015 at 15:47
  • $\begingroup$ Sorry, I got confused. Everything seems to be in order. Now I just wonder whether there is a coordinate-free proof. $\endgroup$ Commented Aug 11, 2015 at 16:08
  • $\begingroup$ I've edited with a coordinate-free proof. $\endgroup$ Commented Aug 11, 2015 at 16:52
  • $\begingroup$ I have another quick query. Do these skew-symmetric matrices $O_i$'s live in the complement of the Lie algebra g_2 in so(7)? Can one construct a closed formula for the multiplications of these matrices $O_i$'s (perhaps by introducing a few other 7x7 matrices, if necessary)? $\endgroup$
    – Hasib
    Commented Aug 12, 2015 at 11:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .