Relationship between SO(3) and SU(2)

High technology approach

The special orthogonal group SO(3) consists of rotations of the 3D space. It can be defined in several ways, but ultimately we have a group representation

SO(3)GL(R3)

This way, SO(3) corresponds to orthogonal matrices with determinant equal to 1.

The Lie algebra so(3) corresponds, by means of this representation, to antisymmetric matrices

[0cbc0aba0](a,b,c)R3

with generators

x^=[000001010],y^=[001000100],z^=[010100000]

On the other hand, we have the group SU(2), of unitary 2x2 complex matrices with determinant equal to 1. This definition is, in fact, a complex linear representation by itself

SU(2)GL(C2).

The Lie algebra su(2) consists of anti-hermitian matrices (see Hermitian matrix) with null trace, and has a basis

e1=[0ii0];e2=[0110];e3=[i00i]

(by the way, these are the Pauli matrices multiplied by i).
The Lie algebras so(3) and su(2) not only have the same dimension, but they are isomorphic: we can map

x^12e1y^12e2z^12e3

and conserve the commutation relations.

Since SO(3) and SU(2) have the same Lie algebra, they have the same universal covering (in particular SU(2) is a double cover of SO(3)). And since SU(2) is simply connected (because it is diffeomorphic to S3, see section unitary matrix#What is SU 2 topologically), it is the universal covering group of SO(3).

Low level approach

But we can forget about these sophisticated tools, and try to see more directly the relationship between SU(2) and SO(3).The idea follows this way:

zaz+bcz+d

with adbc0 corresponds to

[(abcd)]GL(2,C)/{λI}

with nonzero determinant.

In other words, the complex vector space C2 gives rise to the complex projective line (aka the Riemann sphere). The linear transformations of C2 give rise to Mobius transformations.

But several linear transformations give rise to the same Moebius transformation, so we take equivalence classes. Even if we restrict to the singular matrices with unit determinant, we have a 2-1 relation. That is, we have a 2-1 surjection from SL(2,C) to the group of Moebius, M. That is,

MGL(2,C)/{λI}SL(2,C)/{±I}

so we can express the Moebius transformation (abusing of notation) by

±(abcd)

with adbc=1.

Rzϕ{x=xcosϕysinϕy=xsinϕ+ycosϕz=z

and rotation along the x-axis and angle θ,

Rxθ{x=xy=ycosθzsinθz=ysinθ+zcosθ

Pure algebraic manipulations lead us to the Moebius transformations:

rzϕ=SRzϕS1ζeiϕ2ζ+00ζ+eiϕ2

and

rxθ=SRxθS1ζcosθ2ζ+isinθ2isinθ2ζ+cosθ2

so, in SL(2,C)/{±I}

rzϕ=±(eiϕ200eiϕ2)rxθ=±(cosθ2isinθ2isinθ2cosθ2) R=Rϕ,θ,ψ=(cosϕsinϕ0sinϕcosϕ0001)(1000cosθsinθ0sinθcosθ)(cosψsinψ0sinψcosψ0001)==(cosϕcosψcosθsinϕsinψcosϕsinψcosθsinϕcosψsinϕsinθsinϕcosψ+cosθcosϕsinψsinϕsinψ+cosθcosϕcosψcosϕsinθsinψsinθcosψsinθcosθ)

we get

rϕ,θ,ψ=±(eiϕ200eiϕ2)(cosθ2isinθ2isinθ2cosθ2)(eiψ200eiψ2)=±(cosθ2eiϕ+ψ2isinθ2eiϕψ2isinθ2eiϕψ2cosθ2eiϕ+ψ2)

where the determinant is obviously 1.

±(αββ¯α¯)

with |α|2+|β|2=1, that is, it belong to SU(2)/{±I}. But the converse is also true. Every such pair of matrices in SU(2) comes from a Moebius transformation corresponding to a rotation. To see that, we only have to solve

cosθ2=|α|,sinθ2=|β|,(0θπ)ϕ+ψ2=argα,ψϕ2=argβ

to find the Euler angles.
We can even express the 3x3 rotation matrix in function of the given complexes numbers:

(12(α2β2+α2β2)i2(α2β2+α2+β2)αβα¯β¯i2(α2β2α2+β2)12(α2+β2+α2+β2)i(+αβα¯β¯)αβ¯+α¯βi(αβ¯+α¯β)αα¯ββ¯) π:SU(2)SO(3)

with kernel {±I}. I.e., SU(2) is a double cover of SO(3).