#### [SOLVED] What is the value of $[S^3/G] \in \pi_3(Sphere)$ for a finite subgroup $G \subset SU(2)$?

Let $$G\subset \mathrm{SU}(2)$$ be a finite group. (These are famously classified through the McKay correspondence.) The Lie group framing of $$\mathrm{SU}(2) = S^3$$ descends to the quotient manifold $$S^3 / G$$, at least after getting some "left"s and "right"s in the correct places.

Every framed $$k$$-manifold determines a class in the $$k$$th stable homotopy group of spheres $$\pi_k(\text{Sphere})$$. For example, $$\mathrm{SU}(2)$$ with its Lie group framing provides a generator of $$\pi_3(\text{Sphere}) = \mathbb{Z}/24$$.

What are the values of the homotopy-group classes determined by the framed 3-manifolds $$S^3/G$$? How do these values relate to other Lie theoretic data like the rank or (dual) Coxeter number of the ADE Dynkin diagram?

#### @Alex Suciu 2019-01-12 03:28:18

The answer can be found in Theorem 2.1 from a paper of José Seade and Brian Steer (Complex singularities and the framed cobordism class of compact quotients of $$3$$-dimensional Lie groups by discrete subgroups, Comment. Math. Helv. 65 (1990), no. 3, 349–374, available here): If $$G$$ is the cyclic group of order $$r$$, then $$S^3/G$$ represents $$r$$ times a generator of $$\pi_3^s=\mathbb{Z}/24\mathbb{Z}$$, whereas if $$G$$ is the $$\langle p,q,r\rangle$$ triangle group, then $$S^3/G$$ represents $$(p+q+r-1)$$ times a generator of $$\pi_3^s$$.

#### @Theo Johnson-Freyd 2019-01-12 15:41:38

If I am not mistaken, this number $r$ or $(p+q+r-1)$ is the number of complex irreps of $G$, which in terms of the ADE classification is one more than the rank of the compact Dynkin diagram.

#### @Alex Suciu 2019-01-12 20:38:11

Quick reality check: Take $G=I^*$ (the binary icosahedral group of order 120), so that $S^3/G=\Sigma(2,3,5)$ is the Poincaré homology sphere. Then the corresponding Dynkin diagram is $E_8$, and $2+3+5-1=9=8+1$. Yep, that works.

#### @Theo Johnson-Freyd 2019-01-13 00:30:26

Incidentally, the reason I wanted to say things in terms of the Dynkin diagram is that there is a moonshine-type construction that takes in certain Dynkin data and produces certain modular objects, and which might, in some yet-to-be understood way, factor through some geometric objects e.g. 3-manifolds. I had hoped to go (Dynkin data) -> (S^3/G) -> (modular object), but this seems not to work, or at least I need to think more.