Groups, Actions, Quotients, and Covers

A group affects an object only through an action. Elements in the action kernel are invisible on that object, and quotienting by the kernel gives the canonical faithful transformation group with the same realized transformations. Other quotients have different types: $G/H$ is generally only a coset set, $G/N$ is a quotient group only when $N$ is normal, and an orbit quotient is a set or space of equivalence classes.

The same representative-independence test controls covers. An action of $\widetilde G$ descends through a surjective homomorphism $p:\widetilde G\to G$ exactly when every element of $\ker p$ acts trivially. A covering group therefore carries essential global information for that action precisely when its kernel can be detected by the action. The maps $\mathbb R\to U(1)$ and $SU(2)\to SO(3)$ make this criterion concrete.

This page develops that reusable mathematics. Whether an action is an exact symmetry of a QFT requires additional tests on states, observables, dynamics, sectors, correlators, anomalies, and gauge redundancy.

Actions, quotients, and covers

The page defines actions, orbits, stabilizers, kernels, faithful actions, coset and orbit quotients, quotient groups, covering homomorphisms, and descent. It proves the algebraic results it uses and gives only the topology needed to distinguish a cover from an arbitrary quotient. It does not construct universal covers, classify group extensions, develop quotient manifolds, introduce Lie algebras or exponential maps, or classify linear representations and spinors.

A group is not an action

A group is a set $G$ with an associative binary operation $G\times G\to G$, an identity $e$, and an inverse $g^{-1}$ for every $g\in G$. Those axioms do not say what $G$ transforms. An active left action on a set $X$ is a map

$$\begin{aligned} G\times X&\longrightarrow X, & (g,x)&\longmapsto g\cdot x, \\ e\cdot x&=x, & (g_1g_2)\cdot x&=g_1\cdot(g_2\cdot x). \end{aligned}$$

For each $g$, the map $x\mapsto g\cdot x$ is bijective, with inverse given by $g^{-1}$. Hence an action is equivalently a homomorphism

$$\rho:G\longrightarrow \operatorname{Sym}(X), \qquad \rho(g_1g_2)=\rho(g_1)\circ\rho(g_2),$$

where $\operatorname{Sym}(X)$ is the group of bijections of $X$. Conversely, any such homomorphism defines $g\cdot x=\rho(g)(x)$. This equivalence, along with the elementary orbit and stabilizer results below, is developed in Earl 2014, §§9, 10, and 12, PDF.

This page uses active left actions throughout. If $G$ acts on $X$, its induced action on scalar-valued functions is

$$(g\cdot f)(x)=f(g^{-1}\cdot x).$$

The inverse is forced by the left-action law:

$$\begin{aligned} \bigl(g_1\cdot(g_2\cdot f)\bigr)(x) &=(g_2\cdot f)(g_1^{-1}\cdot x)\\ &=f(g_2^{-1}g_1^{-1}\cdot x)\\ &=\bigl((g_1g_2)\cdot f\bigr)(x). \end{aligned}$$

A passive component change or a right action uses a different composition rule and must not be inserted into these formulas without translation. A right action $x\cdot g$ can be converted to a left action by $g\star x=x\cdot g^{-1}$.

Orbits, stabilizers, and the action kernel

For $x\in X$, define its orbit and stabilizer by

$$G\cdot x=\{g\cdot x:g\in G\}, \qquad G_x=\{g\in G:g\cdot x=x\}.$$

The orbits partition $X$: two orbits are either equal or disjoint. The stabilizer $G_x$ is a subgroup, but it need not be normal. The action kernel is different:

$$K=\ker\rho =\{g\in G:g\cdot x=x\ \text{for every }x\in X\} =\bigcap_{x\in X}G_x.$$

Because $K$ is the kernel of a homomorphism, $K\triangleleft G$. Directly, if $k\in K$, then for every $g\in G$ and $x\in X$,

$$(gkg^{-1})\cdot x =g\cdot\bigl(k\cdot(g^{-1}\cdot x)\bigr) =x.$$

Three adjectives answer different questions:

Property	Condition	Meaning
Faithful	$K=\{e\}$	No nonidentity element is invisible everywhere
Free	$G_x=\{e\}$ for every $x$	No nonidentity element fixes even one point
Transitive	$G\cdot x=X$ for one, hence every, $x$	There is one orbit

For nonempty $X$, a free action is faithful, but a faithful action need not be free. The natural action of $S_3$ on $\{1,2,3\}$ is faithful and transitive, yet not free: the stabilizer of $1$ is $\{e,(23)\}$.

The orbit through $x$ has a canonical coset description:

$$\Phi_x:G/G_x\longrightarrow G\cdot x, \qquad \Phi_x(gG_x)=g\cdot x.$$

Indeed,

$$gG_x=hG_x \quad\Longleftrightarrow\quad h^{-1}g\in G_x \quad\Longleftrightarrow\quad g\cdot x=h\cdot x,$$

so $\Phi_x$ is well defined and bijective. If $G$ is finite, this gives the orbit–stabilizer formula

$$|G\cdot x|=[G:G_x]=\frac{|G|}{|G_x|}.$$

The bijection exists without assuming that $G_x$ is normal. It identifies a homogeneous $G$-set, not generally a quotient group.

Three quotient constructions

Let $H\leq G$. The notation $G/H$ first means the set of left cosets $gH$. A multiplication proposed by representatives,

$$(gH)(hH)=(gh)H,$$

is well defined exactly when $H\triangleleft G$. To see necessity, assume the multiplication is well defined. Since $aH=H$ for $a\in H$,

$$(g^{-1}H)(aH)(gH) =(g^{-1}ag)H =(g^{-1}H)(H)(gH) =H,$$

so $g^{-1}ag\in H$. Conversely, normality lets factors from $H$ move past representatives, making the product independent of every choice. This is the quotient-group criterion; see Earl 2014, Proposition 215 and §7, pp. 55–59, PDF.

It is useful to keep the resulting objects typed:

Construction	Elements	Canonical group law?
Orbit $G\cdot x$	Points of $X$ reachable from $x$	No
Orbit quotient $G\backslash X$	All orbits of a left action	No
Coset space $G/H$	Left cosets of any subgroup $H$	Only if $H$ is normal
Quotient group $G/N$	Cosets of a normal subgroup $N$	Yes

Here $G\backslash X$ denotes the orbit space of a left action. The alternate notation $X/G$ is common, especially for right actions, so the action side must be declared.

For the natural $S_3$ action above, $S_3/G_1$ is a three-element coset set that models the orbit of $1$. Since $G_1=\{e,(23)\}$ is not normal, it is not a quotient group. By contrast, $A_3\triangleleft S_3$ and

$$S_3/A_3\cong\mathbb Z_2.$$

There is also a canonical quotient attached to every action. Since $K=\ker\rho$ is normal, Earl 2014, Theorem 234, §8, pp. 61–62, PDF gives

$$G/K\cong\operatorname{im}\rho.$$

The formula

$$(gK)\cdot x=g\cdot x$$

is independent of the representative and defines a faithful action of $G/K$. It has exactly the same transformations and orbits as the original action. What it forgets is the identity of the elements that acted trivially. That forgotten global information can matter when the same abstract group is asked to act on additional objects.

Quotient topology retains continuous invariant data

Suppose now that $G$ is a topological group, $X$ is a topological space, and the action map $G\times X\to X$ is continuous. Let

$$q:X\longrightarrow G\backslash X, \qquad x\longmapsto G\cdot x.$$

The quotient topology declares $U\subseteq G\backslash X$ open exactly when $q^{-1}(U)$ is open in $X$. It is characterized by a factorization property: for any topological space $Y$, an invariant map $F:X\to Y$, satisfying $F(g\cdot x)=F(x)$, defines a unique map

$$\overline F:G\backslash X\longrightarrow Y, \qquad \overline F(G\cdot x)=F(x).$$

The map $\overline F$ is well defined because $F$ is constant on orbits, and it is continuous exactly when $\overline F\circ q=F$ is continuous. Thus the quotient retains precisely the continuous data that cannot distinguish points in the same orbit. The quotient-space construction and its characteristic property are treated in Lee 2011, “New Spaces from Old,” pp. 65–80.

A quotient topology need not be Hausdorff or a manifold. For example, let the multiplicative group $\mathbb R_{>0}$ act on $\mathbb R$ by $\lambda\cdot x=\lambda x$. There are three orbits:

$$(-\infty,0),\qquad \{0\},\qquad (0,\infty).$$

Every saturated open neighborhood of $0$ is all of $\mathbb R$, so the orbit of $0$ cannot be separated from either nonzero orbit in the quotient. Freeness, properness, regularity, and manifold hypotheses must therefore be checked rather than inferred from the notation. General quotient-manifold theorems belong to the later geometry and topology chapters.

Covering groups and the descent criterion

For this page, a covering homomorphism is a surjective continuous homomorphism of Hausdorff topological groups

$$p:\widetilde G\longrightarrow G$$

whose underlying map is a covering map: every point of $G$ has a neighborhood evenly covered by disjoint open sets in $\widetilde G$. Its fibers are cosets of $\ker p$, and its kernel is discrete. If $\widetilde G$ is connected, the kernel is also central: for fixed $k\in\ker p$, the continuous map $\widetilde g\mapsto\widetilde gk\widetilde g^{-1}$ takes values in a discrete set, so it is constant.

Algebraically, the first isomorphism theorem identifies $G\cong\widetilde G/\ker p$. Topologically, the word cover records the additional fact that $p$ is locally a homeomorphism. A surjective homomorphism with a nondiscrete kernel is not a cover, and the converse from “discrete kernel” to “covering map” is not asserted here without further hypotheses. These distinctions and the connectedness qualification are supported by Lee 2011, “Covering Homomorphisms,” pp. 294–296 and Hall 2015, §§4.7 and 5.8.

The main result does not require topology.

Descent proposition. Let $p:\widetilde G\twoheadrightarrow G$ be a surjective group homomorphism and let $\widetilde\rho:\widetilde G\to\operatorname{Sym}(X)$ be an action. There is a unique action $\rho:G\to\operatorname{Sym}(X)$ such that $\widetilde\rho=\rho\circ p$ if and only if

$$\ker p\subseteq\ker\widetilde\rho.$$

Proof. If the action descends and $k\in\ker p$, then $\widetilde\rho(k)=\rho(e)=\operatorname{id}_X$. Conversely, suppose the covering kernel acts trivially. Define

$$\rho\bigl(p(\widetilde g)\bigr)=\widetilde\rho(\widetilde g).$$

If $p(\widetilde g_1)=p(\widetilde g_2)$, then $\widetilde g_2^{-1}\widetilde g_1\in\ker p$, so $\widetilde\rho(\widetilde g_1)=\widetilde\rho(\widetilde g_2)$. The definition is therefore independent of the lift, and it is immediately a homomorphism. Surjectivity of $p$ makes it unique. $\square$

A linear representation is the special case in which $X$ is a vector space and the transformations are invertible linear maps. Its decomposition and intertwiner theory belong to the later representations page.

The circle: periodicity is descent

The standard covering homomorphism

$$\begin{aligned} p:(\mathbb R,+)&\longrightarrow U(1), & p(\theta)&=e^{i\theta}, & \ker p&=2\pi\mathbb Z \end{aligned}$$

is locally one-to-one on every interval of length less than $2\pi$. For $s\in\mathbb R$, let the covering group act on $\mathbb C$ by

$$\widetilde\rho_s(\theta)z=e^{is\theta}z.$$

It descends to $U(1)$ precisely when

$$e^{is(\theta+2\pi)}=e^{is\theta},$$

or equivalently $e^{2\pi i s}=1$. Thus

$$s\in\mathbb Z, \qquad \rho_s(z)=z^s.$$

These are the integer weights of $U(1)$ in the declared $2\pi$-periodic normalization. This is a global-form statement, not a convention-independent numerical quantization of every physical charge. The classification is cross-checked in Mason n.d., Proposition 2.11.1.

The failure at $s=\tfrac12$ is visible without any classification theorem: $0$ and $2\pi$ name the same element of $U(1)$ but act on a nonzero $z$ with opposite signs. The action exists on $\mathbb R$ and on a circle with $4\pi$ period, but not on the specified $2\pi$-periodic group.

The rotation cover: which actions see the center?

Write the Pauli matrices as

$$\begin{aligned} \sigma_1&= \begin{pmatrix}0&1\\1&0\end{pmatrix}, & \sigma_2&= \begin{pmatrix}0&-i\\i&0\end{pmatrix}, \\ \sigma_3&= \begin{pmatrix}1&0\\0&-1\end{pmatrix}. \end{aligned}$$

Identify $\mathbf v\in\mathbb R^3$ with the traceless Hermitian matrix $X_{\mathbf v}=\mathbf v\cdot\boldsymbol\sigma$. For $U\in SU(2)$, define $R(U)$ by

$$X_{R(U)\mathbf v}=UX_{\mathbf v}U^\dagger.$$

Conjugation makes $R(UV)=R(U)R(V)$. It preserves

$$\frac12\operatorname{tr}(X_{\mathbf v}X_{\mathbf w}) =\mathbf v\cdot\mathbf w$$

and the Pauli commutator

$$[X_{\mathbf v},X_{\mathbf w}] =2iX_{\mathbf v\times\mathbf w},$$

so $R(U)$ preserves both the Euclidean inner product and orientation. Hence $R(U)\in SO(3)$. For a unit vector $\mathbf n$, the axis–angle matrices

$$U(\theta,\mathbf n) = \cos\frac{\theta}{2}\,\mathbf1 -i\sin\frac{\theta}{2}\, \mathbf n\cdot\boldsymbol\sigma$$

reach every spatial rotation, and $U$ and $-U$ give the same rotation. The resulting homomorphism is the standard two-sheeted cover

$$SU(2)\longrightarrow SO(3), \qquad \ker R=\{\pm\mathbf1\}.$$

The matrix construction and its global interpretation are treated in Kosmann-Schwarzbach 2022, “Lie Groups $SU(2)$ and $SO(3)$,” pp. 89–102.

Now apply the descent proposition. Conjugation on the $X_{\mathbf v}$ is insensitive to the kernel and therefore descends to $SO(3)$. The defining linear action $v\mapsto Uv$ on $\mathbb C^2$ does not descend, because $-\mathbf1$ acts as minus the identity. It does induce an $SO(3)$ action on rays in $\mathbb C^2$; equivalently, the operators on $\mathbb C^2$ furnish a projective rather than linear representation of $SO(3)$. That ray action must not be conflated with the linear action on vectors.

This is the prototype for why a covering group can be essential. It is not a construction of Lorentzian Spin groups, a classification of spinors, or a claim that an entire physical theory has one global form.

Controlled QFT bridge: a phase action on a complex field

Let a complex scalar field carry the active internal action

$$(z\cdot\phi)(x)=z^q\phi(x), \qquad z\in U(1), \qquad q\in\mathbb Z.$$

The integer condition is precisely the descent condition just derived. The constant phase cancels in $\phi^*\phi$ and in $\partial_\mu\phi^*\partial^\mu\phi$, so the classical free-field action

$$S[\phi] = \int \mathrm d^4x\, \left( \partial_\mu\phi^*\partial^\mu\phi -m^2\phi^*\phi \right)$$

is invariant. This standard internal-symmetry example is discussed in Tong 2006, §1.3.4.

For $q\neq0$, the kernel on the full configuration space is

$$\mu_{|q|} = \{z\in U(1):z^q=1\}.$$

The zero configuration has stabilizer $U(1)$, while a nonzero configuration has stabilizer $\mu_{|q|}$. Thus the action is transitive on each nonzero phase orbit, is not free on the whole configuration space, and is faithful only when $|q|=1$. Its faithful transformation group is $U(1)/\mu_{|q|}$.

For several fields of nonzero integer charges $q_i$, the common action kernel is $\mu_d$, where

$$d=\gcd(|q_1|,|q_2|,\ldots).$$

The kernel therefore depends on every represented charge sector, not on one selected field. Rescaling the generator changes the numerical labels, so an integer weight must always be reported with the period and normalization.

This calculation proves a well-defined classical action and its kernel. It does not establish that the transformation acts faithfully on all physical operators, survives quantization, is nonanomalous, is unbroken, or should be regarded as global rather than gauge redundancy. Those questions belong to the linked physical treatment.

Common pitfalls

Calling the group its action. The same group can act on many different sets with different kernels. Always name $G$, $X$, and the map $G\to\operatorname{Sym}(X)$.

Equating the stabilizer and the kernel. $G_x$ fixes one point; the action kernel fixes every point. The identity $K=\bigcap_xG_x$ is the quickest check.

Treating faithful, free, and transitive as synonyms. They constrain the global kernel, point stabilizers, and orbit count, respectively. The natural $S_3$ action is faithful and transitive but not free.

Giving every coset space a group law. $G/H$ is a quotient group only when $H$ is normal. The bijection $G/G_x\cong G\cdot x$ does not make an arbitrary orbit into a group.

Inferring a good space from quotient notation. An orbit space can be non-Hausdorff or singular. Topological and smooth conclusions require their own hypotheses.

Confusing pullback with descent. Any $G$-action pulls back along $p:\widetilde G\to G$. A $\widetilde G$-action descends in the other direction only when $\ker p$ acts trivially.

Promoting integer weight to an absolute charge claim. The integer refers to a specified $2\pi$ period and generator normalization. A physical charge spectrum adds dynamical and global input.

Exercises

1. Type the natural action

For the natural action of $S_3$ on $\{1,2,3\}$, find the orbit and stabilizer of $1$, the action kernel, and whether the action is faithful, free, or transitive.

Solution

Every point can be sent to every other point, so $S_3\cdot1=\{1,2,3\}$ and the action is transitive. The permutations fixing $1$ are $e$ and $(23)$, so $G_1=\{e,(23)\}$. Only the identity fixes all three points, hence the kernel is trivial and the action is faithful. Since $G_1$ is nontrivial, the action is not free.

2. Diagnose a coset space

Show that $G_1=\{e,(23)\}$ is not normal in $S_3$. Explain why $S_3/G_1$ still models the orbit of $1$ but is not a quotient group.

Solution

Conjugating $(23)$ by $(12)$ gives $(13)\notin G_1$, so $G_1$ is not normal. The representative-independent bijection $gG_1\mapsto g\cdot1$ still identifies the three left cosets with the three points in the orbit. Multiplication of those cosets is not representative-independent, so no quotient-group law is induced.

3. Check the faithful quotient

Let $\rho:G\to\operatorname{Sym}(X)$ have kernel $K$. Prove directly that $(gK)\cdot x=g\cdot x$ is well defined and faithful, and that it has the same orbits as the original action.

Solution

If $gK=hK$, then $h^{-1}g\in K$, so $g\cdot x=h\cdot x$ for every $x$. Thus the quotient action is well defined. If $gK$ fixes every $x$, then $g\in K$, so $gK=K$ and the quotient action is faithful. Its transformations are exactly the maps $\rho(g)$, so its orbit through every $x$ is unchanged.

4. Transfer the descent test

Let $n\in\mathbb Z_{>0}$ and let $p_n:U(1)\to U(1)$ be $p_n(z)=z^n$. The covering group acts on $\mathbb C$ by $z\cdot v=z^m v$, with $m\in\mathbb Z$. For which $m$ does this action descend through $p_n$?

Solution

The covering kernel is $\mu_n=\{\zeta\in U(1):\zeta^n=1\}$. Descent requires $\zeta^m=1$ for every $\zeta\in\mu_n$. A primitive $n$th root satisfies this exactly when $n$ divides $m$. Hence the action descends iff $m\in n\mathbb Z$.

5. Test the rotation cover

Test both $U:v\mapsto Uv$ on $\mathbb C^2$ and $U:X\mapsto UXU^\dagger$ on traceless Hermitian matrices against the kernel $\{\pm\mathbf1\}$ of $SU(2)\to SO(3)$.

Solution

In the defining action, $-\mathbf1$ sends $v$ to $-v$, so the kernel does not act trivially and the action does not descend. Under conjugation, $(-\mathbf1)X(-\mathbf1)^\dagger=X$, so both elements of the covering kernel act trivially and the action does descend.

Where to continue

• Continue to Lie Groups, Lie Algebras, and Exponential and Adjoint Maps for continuous groups, infinitesimal generators, the exponential map, and adjoint actions. That page requires this one.
• Continue to Representations, Intertwiners, Invariants, and Tensor Decomposition for linear actions, intertwiners, invariant tensors, and irreducible decomposition. That page requires this one together with the vector-spaces page.
• The Clifford page only recommends this page. The $SU(2)\to SO(3)$ example is a cover-and-descent prototype, not the later construction of Pin or Spin groups.
• For the physical application to internal symmetry actions and spin covers, continue to What Is a Symmetry of a QFT?. That page adds the exact-theory test and states its additional prerequisites.

References

• Richard Earl, Groups and Group Actions, PDF, Oxford lecture notes, 2014, §§7–10 and 12, pp. 54–72 and 82–83. This open teaching source supports quotient groups, the first isomorphism theorem, actions, orbits, stabilizers, orbit–stabilizer, and the action–homomorphism correspondence.
• Brian C. Hall, Lie Groups, Lie Algebras, and Representations: An Elementary Introduction, second edition, Graduate Texts in Mathematics 222, Springer, 2015, §§4.7 and 5.8. This cross-checks representation descent and the connected-cover qualifications.
• Yvette Kosmann-Schwarzbach, Groups and Symmetries: From Finite Groups to Lie Groups, second edition, Springer, 2022, “General Facts About Groups,” pp. 1–12, and the $SU(2)$/$SO(3)$ chapters, pp. 89–118, develops the group-theoretic foundations and the rotation-cover comparison used here.
• John M. Lee, Introduction to Topological Manifolds, second edition, Graduate Texts in Mathematics 202, Springer, 2011, “New Spaces from Old,” pp. 49–84; “Covering Homomorphisms,” pp. 294–296; and “Quotients by Group Actions,” pp. 311–314, develops the quotient topology, continuous actions, and covering-space results used on this page.
• Jamie Mason (n.d.; accessed August 11, 2026), “Representations of $U(1)$ and Maschke’s theorem”, Durham representation-theory notes, Proposition 2.11.1. This is an open cross-check of the integer-weight statement.
• David Tong (2006), Quantum Field Theory, §1.3.4, “Internal Symmetries”, Cambridge lecture notes. This section develops the bounded complex-scalar phase action; developed symmetry claims remain at the physical continuation.