I am attempting to read Atiyah's paper on self-duality in four-dimensional Riemannian geometry, and I came across the following basic example:

Let $S_-$ be the $SU(2)$-bundle of anti-self dual spinors over $S^4$. Then the total space of the projectivised bundle $PS_{-}$ over $S^4$ is $\mathbb{C}P^3$.

I was wondering if someone could give a brief proof/explanation of this fact (after searching the internet I find many references to this fact, but no complete explanations)? I understand that the quaternionic Hopf fibration exhibits $\mathbb{C}P^3$ as an $S^2$-bundle over $S^4$, but I only vaguely see that this bundle is in fact the projectivised ASD spinor bundle. (Basically I have tried to write down transition functions for both for the usual splitting of $S^4 = \mathbb{H}P^1$ into two hemispheres. I was hoping that someone could give a better reason and/or the details of this procedure.)

Thanks in advance, and sorry if this question is basic! Everyone seems to just claim that the total space is obviously $\mathbb{C}P^3$, but to me it is not obvious :(.

(In a similar vein, if someone could describe and explain the twistor space of $\mathbb{C}P^2$, that would also be great!)


2 Answers 2


I am sure there are many ways to see this but here is a quick one. The oriented frame bundle of $\mathbb S^4$ is $SO(4)\to SO(5)\to \mathbb S^4$. The spin cover of this is $Spin(4)\to Spin(5)\to S^4$ which is the same as $Sp(1)\times Sp(1)\to Sp(2)\to \mathbb S^4$. Up to conjugation there is only one $Sp(1)\times Sp(1)$ in $Sp(2)$ so we can think of this just being the canonical diagonal embedding. The half-spin reps $\Delta_\pm$ of $Spin(4)=Sp(1)\times Sp(1)$ are just the quaternionic multiplications on $\mathbb H=\mathbb R^4$ by the first and the second factor respectively. Therefore the unprojectivized $SU(2)$ bundle you are interested in is just $Sp(1)\to Sp(2)/(Sp(1)\times 1)=S^7\to S^4$ which of course is just the Hopf bundle. Taking the quotient of $S^7$ by $S^1\subset Sp(1)$ gives you $\mathbb{CP}^3$.


As a reference for that, I can recommend chapter 3 of 'Geometry of Yang-Mills Fields' by Atiyah.

  • 3
    $\begingroup$ Can you cite page numbers or specific results (which would save time from having to read the entire chapter, potentially)? $\endgroup$
    – Todd Trimble
    Commented Nov 29, 2017 at 18:15
  • $\begingroup$ The chapter is very short and I would recommend reading all of the few pages. $\endgroup$
    – deepfloe
    Commented Dec 25, 2017 at 11:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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