This question is extremely related to this other question. In fact, a positive answer here directly implies a positive answer there. However, since it is a mathematically different question I decided to ask it separately.

The Freudenthal-Tits Magic Square construction relating exceptional Lie algebras to algebras of infinitesimal isometries of certain "projective planes" over the tensor product of two division algebras

$$\mathfrak{isom}((\mathbb{A}\otimes\mathbb{B})\mathbb{P}^2) = \mathfrak{der}(\mathbb{A}) \oplus \mathfrak{der}(J_3(\mathbb{B})) \oplus (\mathbb{A}_0\otimes J_3(\mathbb{B})_0)$$

can be generalized to projective spaces of arbitrary dimension $n$ for the division rings (and $n\leq2$ for the octonions), yielding

$$\mathfrak{isom}((\mathbb{A}\otimes\mathbb{B})\mathbb{P}^n) = \mathfrak{der}(\mathbb{A}) \oplus \mathfrak{der}(J_{n+1}(\mathbb{B})) \oplus (\mathbb{A}_0\otimes J_{n+1}(\mathbb{B})_0)$$

In particular the right side still makes sense for $n=0$:

$$\mathfrak{isom}((\mathbb{A}\otimes\mathbb{B})\mathbb{P}^0) = \mathfrak{der}(\mathbb{A}) \oplus \mathfrak{der}(\mathbb{B})$$

However, $(\mathbb{A}\otimes\mathbb{B})\mathbb{P}^0$ is just a point.

In his notes on octonions, John Baez says

We can think of $\mathbb{HP}^n$ as the unit sphere in $\mathbb{H}^{n+1}$ with points $x$ and $\alpha x$ identified whenever $\alpha$ is a unit quaternion, and as before, $\mathbb{HP}^n$ inherits a Riemannian metric. The group $\mathrm{Sp}(n+1)$ acts as isometries of $\mathbb{HP}^n$, but this action comes from right multiplication, so

$$\mathrm{Isom} (\mathbb{HP}^n) \cong \mathrm{Sp}(n+1)/\{ \pm 1 \}$$ since not $\mathrm{Sp}(1)$ but only its center $\{\pm 1\}$ acts trivially on $\mathbb{HP}^n$ by right multiplication. At the Lie algebra level, this gives $$\mathfrak{isom}(\mathbb{HP}^n) \cong \mathfrak{sp}(n+1)$$

Does this argument work when $n=0$? How can a point have a nontrivial isometry group?

  • $\begingroup$ The best answer, I think, is that Baez does not offer a mathematical proof, hence, one cannot talk about "his argument"; the claim he makes indeed fails for $n=0$. $\endgroup$ – Moishe Kohan Aug 23 '16 at 9:52

No, a point cannot have a non-trivial isometry group. There's only a single map $\{*\}\to \{*\}$, the identity (this holds in the category of sets, and hence clearly also in the metric space category).

edit: removed second part of post; was not relevant to answer.

  • $\begingroup$ Note that his definition of $HP^n$ is different from yours. He starts with the unit sphere in $H^{n+1}$. If one follows Baez' definition then indeed $HP^0$ is a single point. Of course, the group of bijections of a singleton is still trivial. $\endgroup$ – Moishe Kohan Aug 23 '16 at 9:51
  • $\begingroup$ @studiosus You're right, I missed that. Removed that part from the answer. $\endgroup$ – user2520938 Aug 23 '16 at 10:09

Concerning the quaternion example, there is a fundamental difference between $\mathbb{H}$ and $\mathbb{H}^n$ with $n >1$.

When $n=1$, then every $\mathbb{H}$-linear homotheties $\mathbb{H} \rightarrow \mathbb{H}$ preserve lines (in fact there is only one line). When $n >1$, only the $\mathbb{H}$-linear homotheties $\mathbb{H}^n \rightarrow \mathbb{H}^n$ with real coefficient preserve lines. (Note : if a scalar $\lambda \in \mathbb{H}$ acts on a vector $x$ by $\lambda x$, then a $\mathbb{H}$-linear homothetie is a map $x \mapsto x.a$ where $a \in \mathbb{H}$ is the coefficient. Note that $x \mapsto a.x$ is not linear).

Thus in Baez's notes, when $n=0$, the line $$ \mathrm{Isom}(\mathbb{H}P^n) \simeq \mathrm{Sp}(n+1)/\{±1\}$$ should be replaced by $$ \mathrm{Isom}(\mathbb{H}P^0) \simeq \mathrm{Sp}(1)/\{ \text{units}\}$$ which is trivial.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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