Proposition. Unit quaternions and binary rotation groups [fdip-007G]

The unit quaternions are isomorphic to \(SU(2)\) by \[M(a+bi+cj+dk)= \begin {pmatrix}a+bi&c+di\\-c+di&a-bi\end {pmatrix}.\] Conjugation on the imaginary quaternions defines a surjective homomorphism \[\pi :SU(2)\longrightarrow SO(3),\qquad \pi (M(q))(v)=qvq^{-1},\qquad \ker \pi =\{1,-1\}.\] For the proper rotation groups \(R_T,R_O,R_I\) of the regular tetrahedron, octahedron and icosahedron, define \[2T=\pi ^{-1}(R_T),\qquad 2O=\pi ^{-1}(R_O),\qquad 2I=\pi ^{-1}(R_I).\] Their orders are respectively \(24,48,120\). Each quotient by \(\{1,-1\}\) is its indicated rotation group.

For any of the three rotation groups \(R\), the square records the subgroup and its central double cover. In particular, the Hurwitz-unit model of \(2T\) is this tetrahedral preimage. Its quaternion generators \(i,j,k\) and \((-1+i+j+k)/2\) preserve the 4 imaginary vectors \((1,1,1)\), \((1,-1,-1)\), \((-1,1,-1)\), \((-1,-1,1)\): the first generators change two coordinate signs, and the last cyclically permutes coordinates. Its 24 elements therefore exhaust the preimage. This identifies the group used in the finite spinor gauge example.

TauCeti's Quaternion.unitaryEquivSpecialUnitaryGroup formalizes precisely the group equivalence between unit Hamilton quaternions and \(SU(2)\), using the displayed matrix map. This correspondence does not assert the \(SO(3)\) quotient or the three binary subgroup identifications; those have their separate arguments above. A finite subgroup supplies an available group action, not a physical interpretation of that action.