Split-complixificatinon of Lie algebras and Lie groups

122 Views Asked by At

The complexification of the Lie algebras and groups over $\mathbb{C}$ is a well-studied topic. However, I can not find any reference to the similar process for "split-complexification" over $\mathbb{D}$ (say over ${}^2\mathbb{R}$). Obviously, the results from both processes shall be different. (Please do not point me to Clifford algebras, as my interest is precisely in their specific representation via Lie groups).

  1. The general problem:

Let $\mathfrak{h}(\mathbb{R})$ be n-dimensional Lie algebra over the reals with the basis $E$ as $$e_1,e_2,..,e_n$$ with given commutator relations. Now consider the "split-complixification" of the algebra $\mathfrak{h}(\mathbb{D})$. Thus, the basis is doubled by adding $E'$: $$e_{n+1},e_{n+1},..,e_{2n}$$ such as $\quad e_{n+k}=j \:e_{k}\quad,k=0,1...n \quad (j^2=1, j \neq 1)$. The commutator relations for $E'$ appear to be the same as for $E$. Thus, algebra $\mathfrak{h}(\mathbb{D})$ is a new 2n-dimensional algebra. Hence, the new basis can generate a new split-complexified Lie group.

$E'$ with $E$ give $[e_{n+1},e_{1}]\neq 0$,$[e_{n+2},e_{2}]\neq 0 $ etc. Though, in case of "two copies" they would commute.

  1. As an example, the particular problem is Split-biquaternion $\mathbb{H}'(\mathbb{D})$ representation in terms of the Lie groups.

Split-quaternion $\mathbb{H}'(\mathbb{R})$ isomorphic to $Mat(2,\mathbb{R})$ or $SL(2,\mathbb{R}) \cong SU(1,1)$. Therefore, split-biquaternion is split-complixification of $SL(2,\mathbb{R})$ i.e. matrices with entries of split-complex numbers $SL(2,\mathbb{D})$. The split-biquaternions is also $\mathbb{H}\otimes\mathbb{D}=\mathbb{H}\oplus\mathbb{H}$, hence the resulting group is also two copies of $SU(2)$.

Then, looking at the properties of $SL(2,\mathbb{D})$ my guess is that there is a homomorphism $\phi$: $SL(2,\mathbb{D}) \to SO(2,2)$.

As always, thanks for any clue for further study.

1

There are 1 best solutions below

3
On BEST ANSWER

Let $\mathbb{D}=\mathbb{R}[j]/(j^2-1)$ be the split complex numbers. It is well-known that $\mathbb{D}\cong\mathbb{R}\oplus\mathbb{R}$ as rings, with the isomorphism given by linearly extending $(1,0)\leftrightarrow\frac{1}{2}(1+j)$ and $(0,1)\leftrightarrow\frac{1}{2}(1-j)$ (these are the orthogonal primitive idempotents for its Peirce decomposition).

Thus $\mathfrak{g}\otimes_{\mathbb{R}}\mathbb{D}\cong\mathfrak{g}\otimes(\mathbb{R}\oplus\mathbb{R})\cong(\mathfrak{g}\otimes\mathbb{R})\oplus(\mathfrak{g}\otimes\mathbb{R})\cong\mathfrak{g}\oplus\mathfrak{g}$ as real lie algerbas. The isomorphism is actually the same as above: $$(X,Y)\mapsto \frac{1}{2}(1+j)X+\frac{1}{2}(1-j)Y=\frac{1}{2}(X+Y)+\frac{1}{2}(X-Y)j.$$

Let's also examine this claim:

Split-quaternion $\mathbb{H}'(\mathbb{R})$ isomorphic to $Mat(2,\mathbb{R})$ or $SL(2,\mathbb{R}) \cong SU(1,1)$. Therefore, split-biquaternion is split-complixification of $SL(2,\mathbb{R})$ i.e. matrices with entries of split-complex numbers $SL(2,\mathbb{D})$.

First off, let's talk about language. There's a difference between a "normed" algebra and its group of elements of norm $1$. You can say $\mathbb{H}'$ is isomorphic to $M_2(\mathbb{R})$ but not to $SL(2,\mathbb{R})$, the latter isn't even an algebra! So here's the language-fixed version of your statement:

Split-quaternions $\mathbb{H}'(\mathbb{R})$ are isomorphic to $M_2(\mathbb{R})$. Therefore, the split-biquaternions $\mathbb{H}\otimes\mathbb{D}$ are the split-complixification of the split quaternions $\mathbb{H}'$.

But this is not true: the split biquaternions are the split-complexification of the normal quaternions, not of the split quaternions! The split biquaternions are isomorphic to $\mathbb{H}^2$ whereas the split-complexification of the split quaternions are isomorphic to $M_2(\mathbb{R})^2$, which is inequivalent.

That said, there is a $2$-to-$1$ surjective homomorphism $SL_2(\mathbb{D})\to SO(2,2)$. Note that $\mathbb{D}\cong\mathbb{R}^2$ so $SL_2(\mathbb{D})\cong SL_2(\mathbb{R})^2$ which acts on $M_2(\mathbb{R})$ via $(A,B)X:=AXB^{-1}$. This preserves the determinant, which is a quadratic form on $M_2(\mathbb{R})$ of signature $(2,2)$.