Principal homogeneous space (G-torsor)

A set or manifold equipped with a free and transitive action of a Lie group, with no preferred origin.
Principal homogeneous space (G-torsor)

Let GG act on a nonempty set XX on the left.

A principal homogeneous space for GG (also called a GG-torsor) is a set XX together with a left action G×XXG \times X \to X, (g,x)gx(g,x)\mapsto g\cdot x, such that:

  1. (Free) If gx=xg\cdot x = x for some xXx\in X, then g=eg=e.
  2. (Transitive) For any x,yXx,y\in X there exists gGg\in G with gx=yg\cdot x = y.

Equivalently, fixing any x0Xx_0\in X, the orbit map

θx0:GX,θx0(g)=gx0 \theta_{x_0}:G\to X,\qquad \theta_{x_0}(g)=g\cdot x_0

is a bijection; different choices of x0x_0 differ by right multiplication in GG, so there is no canonical identification of XX with GG unless a basepoint is chosen.

If XX is a and the action is , then θx0\theta_{x_0} is a GXG\cong X for each choice of x0x_0.

A basic source of torsors in geometry: if PMP\to M is a , then each fiber PxP_x is a (right) GG-torsor under the bundle’s right action.

Examples

  1. The group acting on itself. GG is a GG-torsor under left translation: gh=ghg\cdot h = gh.
  2. Affine spaces. Any affine space AA modeled on a vector space VV is a torsor for the additive Lie group (V,+)(V,+) via translations va:=a+vv\cdot a := a+v.
  3. Fibers of a principal bundle. For a principal bundle π:PM\pi:P\to M, each fiber π1(x)\pi^{-1}(x) is a torsor for GG: for any p,qπ1(x)p,q\in \pi^{-1}(x) there is a unique gGg\in G with pg=qp\cdot g=q.