1) Let [itex]v \in T_{\bar p}\mathbb{CP}^n, ||v|| = 1[/itex]. Then: [itex]R(w, v)v = w \forall w \in (\mathbb Cv)^\perp[/itex]

2) Let [itex]v \in T_{\bar p}\mathbb{HP}^n, ||v|| = 1[/itex]. Then: [itex]R(w, v)v = w \forall w \in (v\mathbb H)^\perp[/itex]

Afterwards the following is supposed to be proven:

a) [itex]R(iv, v)v = 4iv[/itex] (in the case of [itex]CP^n[/itex])
b) [itex]R(w, v)v = 4w \forall w \in (\mathbb Rv)^\perp\cap(v \mathbb H)[/itex] (in the case of [itex]HP^n[/itex])

Unfortunately, I don't understand the very beginning of the following proof:

I've been on this since yesterday but I don't see why this is the case. Does it somehow follow from 1)?

In b) it is basically the same thing (I think) but the script is a little bit more elaborate - so maybe this helps. It reads:

Do these two statements immediately follow from 1) and 2)? I mean 1) basically shows:

I might have an idea. At a different point the following theorem is introduced:

Sp(n+1) and U(n+1) are defined as matrices A fulfilling [itex]AA^* = I[/itex].

If the map [itex]B(w) := R(w, v)v[/itex] were in [itex]Sp(n+1), U(n+1)[/itex], then the above theorem might be what the proof is referring to. I figured out that B is a self-adjoint endomorphism, therefore [itex]B = B^*[/itex] (right?). But that doesn't mean [itex]BB^* = I[/itex]. For that to be true, it needs to be [itex]B = B^{-1}[/itex], meaning [itex]R(R(w, v)v, v)v = w[/itex]. I've worked on this for the last couple of hours but I think this is not even true...

Okay, I've got it now. Not that clear imo (at least for me it wasn't) but this is the explanation for anybody who cares:

[itex]R(., v)v[/itex] is a self-adjoint endomorphism. Therefore the tangent space has an orthonormal basis of eigenvectors and all eigenvalues are real. It follows that iv must be an eigenvector.

Then basically the same applies in the case of HPn.