Orbit Spaces, Stabilizers, and Stratified Quotients

A gauge-orbit quotient is generally stratified by stabilizer type. On a stratum where the stabilizer is conjugate to a fixed subgroup $H$, orbit dimensions and local quotient models are controlled; where the stabilizer jumps, the quotient acquires singular or orbifold-like behavior. Treating a reducible connection as a free-action point erases isotropy and gives a false tangent dimension.

Required background. Gauge configuration groupoids supplies the action and stabilizer. Group actions, quotients, and covers supplies orbit–stabilizer geometry. Gauge orbits, Gauss constraints, and stabilizers supplies the physical reduction. Helpful background. Smooth manifolds and tangent tensors supplies tangent and normal spaces.

Orbit type and the local quotient model

For a proper smooth action of a finite-dimensional Lie group $K$ on a manifold $X$, the stabilizer $K_x$ is compact and a slice $S_x$ transverse to $K\cdot x$ gives a neighborhood modeled by

$$K\times_{K_x}S_x.$$

After quotienting by $K$, the local model is $S_x/K_x$, not simply $S_x$. If the action is free, $K_x=\{e\}$ and the quotient is locally a manifold with

$$\dim(X/K)=\dim X-\dim K.$$

With nontrivial isotropy, the correct orbit formula is $\dim(K\cdot x)=\dim K-\dim K_x$, and the residual $K_x$-action on the slice can make its quotient singular. Palais proves the slice theorem for proper actions in Palais 1961, §2.3, pp. 305–310. The theorem is local and does not imply that one slice meets every orbit once.

For connections, fix the Sobolev setting $\mathcal A_k(P)$ and $\mathcal G_{k+1}(P)$ with $k>d/2+1$. The stabilizer of $A$ is isomorphic to the centralizer of its holonomy group; its Lie algebra is $\ker d_A$. Connections with conjugate stabilizers form an orbit-type subset. Under the standard compactness and Sobolev hypotheses, these subsets fit into a stratification of $\mathcal A_k/\mathcal G_{k+1}$; the pointed gauge group is useful because it removes constant isotropy before the remaining compact action is analyzed. The slice and stratification results, with their precise function spaces, are developed in Kondracki and Rogulski 1986, §§3.1–4.4, pp. 30–60.

An irreducible connection for a semisimple group has stabilizer equal to the finite center after the appropriate convention; a reducible connection has a larger centralizer. Thus “irreducible” does not always mean literally free. One may divide by the center, or use a based gauge group, but that choice must be stated before applying a free-action theorem.

SU(2) conjugation as a complete model

The first application is the stabilizer analysis used in Gauge Orbits, Gauss Constraints, and Stabilizers. Model a single holonomy by $U\in SU(2)$ and let $SU(2)$ act by conjugation. Every element can be written

$$U=\cos\theta\,\mathbf 1+i\sin\theta\,\widehat{\mathbf n}\cdot\boldsymbol\sigma, \qquad 0\leq\theta\leq\pi.$$

For $0<\theta<\pi$, the stabilizer is the maximal torus $U(1)$ generated by $\widehat{\mathbf n}\cdot\boldsymbol\sigma$. At $U=\pm\mathbf 1$, every group element commutes with $U$, so the stabilizer jumps to all of $SU(2)$. Conjugation can rotate $\widehat{\mathbf n}$ but cannot change $\theta$, and the coarse quotient is the closed interval $[0,\pi]$. Equivalently, it is $T/W$, where the Weyl group acts on the torus by $\theta\mapsto-\theta$; its fixed points become the two endpoints.

This example exposes all essential mechanisms without analysis. The generic orbit has dimension $3-1=2$, while the central orbits have dimension $3-3=0$. A formula that subtracts $\dim SU(2)=3$ everywhere gives the wrong answer at every point: the action is never free, and it is maximally nonfree at the endpoints. Goldman’s representation-space analysis likewise identifies tangent cocycles, orbit coboundaries, centralizers, and singular strata in Goldman 1984, §1.2, pp. 202–205.

Tangent spaces and the failure test

At a connection $A$, the infinitesimal orbit is $\operatorname{im}d_A\subset\Omega^1(M,\operatorname{ad}P)$. A Coulomb normal space is $\ker d_A^*$, but their direct-sum interpretation uses an elliptic decomposition with declared boundary conditions. The stabilizer modes $H_A^0=\ker d_A$ do not live in the tangent orbit; they record infinitesimal transformations that act trivially. When $H_A^0$ jumps, so can the rank of the orbit map, and no single manifold chart covers both neighboring orbit types.

The independent check in the $SU(2)$ model is the trace: $\operatorname{tr}U=2\cos\theta$ is conjugation invariant and separates the interval. Its derivative vanishes at $\theta=0,\pi$, exactly where the quotient coordinate ceases to behave like a regular coordinate inherited from the circle.

For the adversarial test, apply the free-action quotient theorem at $U=\mathbf 1$. It predicts a three-dimensional orbit and removes three tangent directions, yet the actual orbit is one point because the stabilizer is $SU(2)$. The contradiction identifies the missing hypothesis. Conversely, observing a singular coarse quotient does not alone determine a unique stabilizer: different group actions can produce the same underlying topological space.

Exercises

Compute the stabilizers and orbit dimensions for the conjugation action of $SU(2)$ on itself, and explain why the endpoints of $T/W$ are special.

Solution

For $0<\theta<\pi$, $U$ has two distinct eigenvalues, so a commuting matrix preserves its eigenspaces; inside $SU(2)$ these matrices form $U(1)$. The orbit dimension is $3-1=2$. At $U=\pm\mathbf1$, commutation is automatic, the stabilizer is $SU(2)$, and the orbit dimension is zero. The Weyl reflection has fixed points at those two central elements, so the interval endpoints retain enhanced isotropy and cannot be treated as ordinary free-quotient points.

References

• Goldman, William M. “The Symplectic Nature of Fundamental Groups of Surfaces.” Advances in Mathematics 54 (1984): 200–225. DOI; Open PDF.
• Kondracki, Witold, and Jan S. Rogulski. On the Stratification of the Orbit Space for the Action of Automorphisms on Connections. Dissertationes Mathematicae 250. Warsaw: Polish Scientific Publishers, 1986. Repository record and PDF.
• Palais, Richard S. “On the Existence of Slices for Actions of Non-Compact Lie Groups.” Annals of Mathematics 73, no. 2 (1961): 295–323. DOI.