Chapter 08
Hard

Group Actions

00 · Symbol Glossary

$g\cdot x$G acts on x

Notation for the result of letting group element gg move a point xx in some set XX. Distinct from the group's own operation ghgh — here gg and xx live in different sets.

$\mathrm{Orb}(x)$Orbit of x

Orb(x)={gx:gG}\mathrm{Orb}(x) = \{g\cdot x : g\in G\} — every point reachable from xx by acting with some group element. The set of everywhere xx can be "moved to."

$\mathrm{Stab}(x)$Stabilizer of x

Stab(x)={gG:gx=x}\mathrm{Stab}(x) = \{g\in G : g\cdot x = x\} — every group element that leaves xx fixed in place.

$C_G(x)$Centralizer of x (under conjugation)

CG(x)={gG:gxg1=x}={gG:gx=xg}C_G(x) = \{g\in G : gxg^{-1}=x\} = \{g\in G : gx=xg\} — the stabilizer of xx specifically under the conjugation action. Everything commuting with xx.

$x^G$Conjugacy class of x

xG={gxg1:gG}x^G = \{gxg^{-1} : g\in G\} — the orbit of xx under conjugation. All elements "structurally equivalent" to xx by relabeling via some gg.


01 · What Is a Group Action?

Every example of a group so far has, at heart, been a group of motions: permutations moving points, matrices moving vectors, symmetries moving a polygon's vertices. A group action makes this idea of "GG moves the points of some set XX" completely precise — and, crucially, XX need not have any group structure of its own.

Definition — Group Action

A group GG acts on a set XX if there is a function G×XXG\times X \to X, written (g,x)gx(g,x)\mapsto g\cdot x, satisfying:

A1. ex=xe\cdot x = x for every xXx\in X.

A2. g(hx)=(gh)xg\cdot(h\cdot x) = (gh)\cdot x for every g,hGg,h\in G, xXx\in X.

X is just a set — no group structure required

This is the key generalization beyond homomorphisms: XX can be a completely bare set — the vertices of a polygon, the elements of GG itself, the set of subgroups of GG — with no operation of its own. All the algebraic structure lives entirely in GG; XX merely receives the motion.


02 · Actions Are Permutations, and Some Standard Examples

Theorem 8.1 — Each Group Element Acts as a Bijection

For fixed gGg\in G, the map σg:XX\sigma_g: X\to X, σg(x)=gx\sigma_g(x)=g\cdot x, is a bijection, with inverse σg1\sigma_{g^{-1}}.

Proof

Compute σg1(σg(x))=g1(gx)=(g1g)x=ex=x\sigma_{g^{-1}}(\sigma_g(x)) = g^{-1}\cdot(g\cdot x) = (g^{-1}g)\cdot x = e\cdot x = x (using A2, then A1), and similarly σg(σg1(x))=x\sigma_g(\sigma_{g^{-1}}(x))=x. So σg1\sigma_{g^{-1}} is a two-sided inverse for σg\sigma_g, meaning σg\sigma_g is a bijection.

Example — G acting on itself by left multiplication

X=GX=G, action gx:=gxg\cdot x := gx (the group's own operation). A1: ex=ex=xe\cdot x=ex=x. A2: g(hx)=g(hx)=(gh)x=(gh)xg\cdot(h\cdot x) = g(hx) = (gh)x = (gh)\cdot x, using associativity. Every group acts on itself this way — the very first example, and the source of Cayley's theorem (every group embeds into some SnS_n, since left multiplication realizes GG as bijections of itself).

Example — $D_n$ acting on the vertices of a regular n-gon

X={1,2,,n}X=\{1,2,\ldots,n\} (vertex labels), and each symmetry gDng\in D_n sends vertex xx to wherever that symmetry moves it. A1 (identity motion fixes every vertex) and A2 (composing two symmetries then applying to a vertex matches applying one after the other) both hold by the geometric meaning of composition. This is the action underlying every cycle-notation computation with DnD_n from Chapter 02.

Example — G acting on itself by conjugation

X=GX=G, action gx:=gxg1g\cdot x := gxg^{-1}. A1: ex=exe1=xe\cdot x = exe^{-1}=x. A2: g(hx)=g(hxh1)g1=(gh)x(gh)1=(gh)xg\cdot(h\cdot x) = g(hxh^{-1})g^{-1} = (gh)x(gh)^{-1} = (gh)\cdot x, using (gh)1=h1g1(gh)^{-1}=h^{-1}g^{-1} (Theorem 2.3). This action will drive the rest of the chapter.


03 · Orbits and Stabilizers

Definition — Orbit and Stabilizer

For xXx\in X: Orb(x)={gx:gG}X\mathrm{Orb}(x) = \{g\cdot x : g\in G\} \subseteq X, and Stab(x)={gG:gx=x}G\mathrm{Stab}(x) = \{g\in G : g\cdot x = x\} \subseteq G.

Theorem 8.2 — Orbits Partition X

The orbits of a group action partition XX: every element of XX lies in exactly one orbit.

Proof

Define xy    y=gxx\sim y \iff y=g\cdot x for some gGg\in G. Reflexive: x=exx=e\cdot x (A1). Symmetric: if y=gxy=g\cdot x, then g1y=g1(gx)=(g1g)x=ex=xg^{-1}\cdot y = g^{-1}\cdot(g\cdot x) = (g^{-1}g)\cdot x = e\cdot x=x (A2, A1), so x=g1yx=g^{-1}\cdot y. Transitive: if y=gxy=g\cdot x and z=hyz=h\cdot y, then z=h(gx)=(hg)xz=h\cdot(g\cdot x) = (hg)\cdot x (A2), so xzx\sim z. By Theorem 1.3, \sim's equivalence classes partition XX — and [x]={y:y=gx some g}=Orb(x)[x]=\{y : y=g\cdot x\text{ some }g\}=\mathrm{Orb}(x) exactly.

Theorem 8.3 — Stabilizers Are Subgroups

Stab(x)G\mathrm{Stab}(x) \le G for every xXx\in X.

Proof

Nonempty: ex=xe\cdot x=x (A1), so eStab(x)e\in\mathrm{Stab}(x). Closure: if g,hStab(x)g,h\in\mathrm{Stab}(x), then (gh)x=g(hx)=gx=x(gh)\cdot x = g\cdot(h\cdot x) = g\cdot x = x (A2, then hh fixes xx, then gg fixes xx), so ghStab(x)gh\in\mathrm{Stab}(x). Inverses: if gx=xg\cdot x=x, apply g1g^{-1} to both sides: g1(gx)=g1xg^{-1}\cdot(g\cdot x) = g^{-1}\cdot x, and the left side simplifies to xx (A2, A1), so x=g1xx=g^{-1}\cdot x, meaning g1Stab(x)g^{-1}\in\mathrm{Stab}(x). By Theorem 3.1, Stab(x)G\mathrm{Stab}(x)\le G.

Example — Orbit and stabilizer under $D_4$ acting on a square's vertices

Fix vertex 11. Orb(1)\mathrm{Orb}(1): every symmetry sends vertex 11 somewhere, and since the square is vertex-transitive (any vertex can be rotated/reflected to any other), Orb(1)={1,2,3,4}\mathrm{Orb}(1) = \{1,2,3,4\}, all four vertices. Stab(1)\mathrm{Stab}(1): the symmetries fixing vertex 11 in place — the identity and the single reflection through vertex 11 and the center. Stab(1)=2|\mathrm{Stab}(1)|=2.

Notice Orb(1)Stab(1)=4×2=8=D4|\mathrm{Orb}(1)|\cdot|\mathrm{Stab}(1)| = 4\times2=8=|D_4| — not a coincidence, as the next theorem shows.


04 · The Orbit-Stabilizer Theorem

Theorem 8.4 — Orbit-Stabilizer Theorem

For any group action of a finite group GG on XX and any xXx\in X:

Orb(x)=[G:Stab(x)]=GStab(x)|\mathrm{Orb}(x)| = [G : \mathrm{Stab}(x)] = \frac{|G|}{|\mathrm{Stab}(x)|}

Proof

Write H=Stab(x)H=\mathrm{Stab}(x). Define ψ:G/HOrb(x)\psi: G/H \to \mathrm{Orb}(x) by ψ(gH)=gx\psi(gH) = g\cdot x.

Well-defined: if gH=gHgH=g'H, then by Theorem 4.1, g1gHg^{-1}g'\in H, so (g1g)x=x(g^{-1}g')\cdot x=x (definition of H=Stab(x)H=\mathrm{Stab}(x)). Applying gg to both sides: g((g1g)x)=gxg\cdot\big((g^{-1}g')\cdot x\big) = g\cdot x, and the left side is (gg1g)x=gx(gg^{-1}g')\cdot x = g'\cdot x by A2. So gx=gxg'\cdot x = g\cdot x: the map doesn't depend on the representative.

Injective: if ψ(gH)=ψ(gH)\psi(gH)=\psi(g'H), i.e. gx=gxg\cdot x = g'\cdot x, apply g1g^{-1}: x=g1(gx)=(g1g)xx = g^{-1}\cdot(g'\cdot x) = (g^{-1}g')\cdot x, so g1gStab(x)=Hg^{-1}g'\in\mathrm{Stab}(x)=H, giving gH=gHgH=g'H by Theorem 4.1.

Surjective: every element of Orb(x)\mathrm{Orb}(x) has the form gxg\cdot x for some gg, which is ψ(gH)\psi(gH).

ψ\psi is a bijection between G/HG/H (as a set of cosets — not requiring HH normal here, since we only need a bijection of sets, not a group isomorphism) and Orb(x)\mathrm{Orb}(x). So Orb(x)=G/H=[G:H]|\mathrm{Orb}(x)| = |G/H| = [G:H], and by Lagrange's Theorem, [G:H]=G/H[G:H]=|G|/|H|.

Common mistake — Stab(x) need not be normal

Unlike Chapter 05's quotient groups, ψ\psi here is only a bijection of sets, not a group isomorphism — Orb(x)\mathrm{Orb}(x) generally has no group structure at all (it's just a subset of XX, and XX was never assumed to be a group). G/Stab(x)G/\mathrm{Stab}(x) is used here purely as a counting device, valid whether or not Stab(x)\mathrm{Stab}(x) happens to be normal. If Stab(x)\mathrm{Stab}(x) is not normal, you cannot treat G/Stab(x)G/\mathrm{Stab}(x) as a quotient group — only the orbit–stabilizer bijection for counting is valid.


05 · Conjugation and the Class Equation

Applying the general machinery of Sections 03–04 to the specific conjugation action from Section 02 gives one of the most powerful counting tools in finite group theory.

Definition — Conjugacy Class and Centralizer

Under the conjugation action, Orb(x)=xG={gxg1:gG}\mathrm{Orb}(x) = x^G = \{gxg^{-1}:g\in G\} is the conjugacy class of xx, and Stab(x)=CG(x)={gG:gxg1=x}\mathrm{Stab}(x) = C_G(x) = \{g\in G : gxg^{-1}=x\} is the centralizer of xx.

Example — Central elements form their own conjugacy classes

If xZ(G)x\in Z(G), then gxg1=xgxg^{-1}=x for every gg (that's exactly what commuting with everything means), so xG={x}x^G = \{x\} — a conjugacy class of size 1. Conversely, if xG={x}x^G=\{x\}, then gxg1=xgxg^{-1}=x for every gg, meaning xZ(G)x\in Z(G). Conjugacy classes of size 1 are exactly the elements of Z(G)Z(G).

Theorem 8.5 — The Class Equation

For a finite group GG with conjugacy classes {e}=x0G,\{e\}=x_0^G,\ldots (writing out the size-1 classes from Z(G)Z(G) separately) and noncentral class representatives x1,,xkx_1,\ldots,x_k:

G=Z(G)+i=1k[G:CG(xi)]|G| = |Z(G)| + \sum_{i=1}^{k} [G:C_G(x_i)]

Proof

By Theorem 8.2, conjugacy classes partition GG, so G|G| is the sum of all class sizes. By Theorem 8.4 (Orbit-Stabilizer applied to conjugation), each class xGx^G has size [G:CG(x)][G:C_G(x)]. Separating the size-1 classes (exactly the elements of Z(G)Z(G), by the Example above) from the rest gives G=Z(G)size-1 classes+i=1k[G:CG(xi)]noncentral classes|G| = \underbrace{|Z(G)|}_{\text{size-1 classes}} + \sum_{i=1}^k \underbrace{[G:C_G(x_i)]}_{\text{noncentral classes}}.

Step-by-step — The class equation for $S_3$
1
List conjugacy classes of S3S_3: {e}\{e\} (size 1), the three transpositions {(12),(13),(23)}\{(1\,2),(1\,3),(2\,3)\} (one class, since conjugation permutes cycle shape, and all transpositions share the same shape), the two 3-cycles {(123),(132)}\{(1\,2\,3),(1\,3\,2)\} (one class).
2
Check sizes sum to S3=6|S_3|=6: 1+3+2=61+3+2=6. ✓
3
Identify Z(S3)Z(S_3): only the size-1 class contributes, so Z(S3)={e}Z(S_3)=\{e\}, size 1 — consistent with S3S_3 being nonabelian (Chapter 05's fact that Z(G)=GZ(G)=G iff abelian).
4
Write the class equation: 6=1+3+26 = 1 + 3 + 2, i.e. G=Z(G)+[G:CG((12))]+[G:CG((123))]|G|=|Z(G)| + [G:C_G((1\,2))] + [G:C_G((1\,2\,3))], giving [G:CG((12))]=3[G:C_G((1\,2))]=3 and [G:CG((123))]=2[G:C_G((1\,2\,3))]=2.
5
Back out the centralizer sizes: CG((12))=6/3=2|C_G((1\,2))| = 6/3=2 and CG((123))=6/2=3|C_G((1\,2\,3))|=6/2=3 — matching direct computation ((12)(1\,2) commutes only with itself and ee; (123)(1\,2\,3) commutes with all of {e,(123),(132)}\{e,(1\,2\,3),(1\,3\,2)\}).

06 · Application: Cauchy's Theorem

Lagrange's Theorem (Chapter 04) is one-directional: subgroup orders divide G|G|, but not every divisor is achieved (recall A4A_4, order 12, with no subgroup of order 6). Remarkably, for divisors that are prime, the converse always holds — and the class equation is exactly the tool needed to prove it.

Theorem 8.6 — Cauchy's Theorem

If GG is a finite group and pp is a prime dividing G|G|, then GG has an element of order pp.

Proof

By strong induction on G|G|.

Case 1: GG is abelian. Pick any aea\neq e in GG, and let N=aN=\langle a\rangle. If pp divides a=N|a|=|N|, then aN/pa^{|N|/p} has order exactly pp (a direct check: raising it to the pp-th power gives aN=ea^{|N|}=e, and no smaller positive power works, by minimality of N/p|N|/p among divisors), and we're done.

Otherwise pNp\nmid|N|. Since GG is abelian, NGN\trianglelefteq G (Chapter 05), so G/NG/N is a group of order G/N<G|G|/|N| < |G| (strictly smaller, since N>1|N|>1). Since G=NG/N|G|=|N|\cdot|G/N| and pp divides G|G| but not N|N|, pp must divide G/N|G/N| (a prime dividing a product but not one factor divides the other). By the induction hypothesis, G/NG/N has an element yNyN of order pp. Let m=ym=|y| in GG. Then (yN)m=ymN=eN(yN)^m = y^mN = eN, so p=yNp=|yN| divides mm (order of an element divides any exponent returning it to the identity — the same fact used in Theorem 7.2's proof). Then ym/py^{m/p} has order pp in GG.

Case 2: GG is nonabelian. If pp divides Z(G)|Z(G)|, then since Z(G)Z(G) is abelian and smaller than or equal to GG (with Z(G)GZ(G)\neq G since GG is nonabelian, but we can still apply Case 1 directly to the abelian group Z(G)Z(G) regardless of its size), Z(G)Z(G) has an element of order pp by Case 1 — and that element also lives in GG.

Otherwise pZ(G)p\nmid|Z(G)|. By the class equation (Theorem 8.5), G=Z(G)+i[G:CG(xi)]|G| = |Z(G)| + \sum_i [G:C_G(x_i)]. If pp divided every term [G:CG(xi)][G:C_G(x_i)] in the sum, it would divide the whole sum i[G:CG(xi)]=GZ(G)\sum_i[G:C_G(x_i)] = |G|-|Z(G)|; combined with pGp\mid|G|, this would force pZ(G)p\mid|Z(G)| too — contradicting our assumption. So some noncentral class representative xix_i has p[G:CG(xi)]p \nmid [G:C_G(x_i)]. Since G=[G:CG(xi)]CG(xi)|G| = [G:C_G(x_i)]\cdot|C_G(x_i)| and pp divides G|G| but not the index factor, pp must divide CG(xi)|C_G(x_i)|. Since xix_i is noncentral, CG(xi)GC_G(x_i)\neq G, so CG(xi)<G|C_G(x_i)|<|G| — a strictly smaller group. By the induction hypothesis, CG(xi)C_G(x_i) has an element of order pp, which is also an element of GG.

A genuine partial converse to Lagrange

Cauchy's Theorem says: for a prime divisor pp of G|G|, an element (equivalently, by Theorem 3.4, a cyclic subgroup) of order pp is guaranteed. Chapter 09's Sylow theorems extend this from single primes to full prime powers — guaranteeing subgroups of order pkp^k whenever pkp^k divides G|G| — using the exact same class-equation machinery built here, one level more refined.


07 · Exercises

EXERCISE 8.1

Check both action axioms directly: does the identity permutation fix everything, and does composing two permutations first vs. applying them in sequence give the same result?

X={1,,n}X=\{1,\ldots,n\}, action σk:=σ(k)\sigma\cdot k := \sigma(k). A1: the identity permutation ee satisfies e(k)=ke(k)=k for every kk, so ek=ke\cdot k=k. A2: σ(τk)=σ(τ(k))=(στ)(k)=(στ)k\sigma\cdot(\tau\cdot k) = \sigma(\tau(k)) = (\sigma\circ\tau)(k) = (\sigma\tau)\cdot k, matching the composition convention from Chapter 02. Both axioms hold: SnS_n acts on {1,,n}\{1,\ldots,n\} by direct evaluation.

Verify that SnS_n acts on X={1,,n}X=\{1,\ldots,n\} via σk:=σ(k)\sigma\cdot k := \sigma(k), by checking both action axioms.

EXERCISE 8.2

Use the Orbit-Stabilizer Theorem: you're given G=12|G|=12 and Orb(x)=4|\mathrm{Orb}(x)|=4; solve for Stab(x)|\mathrm{Stab}(x)|.

By Theorem 8.4, Orb(x)=G/Stab(x)|\mathrm{Orb}(x)| = |G|/|\mathrm{Stab}(x)|, so 4=12/Stab(x)4 = 12/|\mathrm{Stab}(x)|, giving Stab(x)=3|\mathrm{Stab}(x)| = 3.

A group of order 12 acts on a set, and some point xx has an orbit of size 4. What is Stab(x)|\mathrm{Stab}(x)|?

EXERCISE 8.3

Use the fact from Section 05: conjugacy classes of size 1 correspond exactly to central elements.

By the Example in Section 05, xG={x}x^G=\{x\} (a size-1 conjugacy class) if and only if xZ(G)x\in Z(G). If GG is abelian, Z(G)=GZ(G)=G (Chapter 05, Exercise 5.4), so every element's conjugacy class has size 1 — conjugation does nothing at all, consistent with gxg1=xgxg^{-1}=x for all g,xg,x when GG is abelian.

If GG is abelian, what can you say about the size of every conjugacy class? Justify using the Section 05 fact about Z(G)Z(G).

EXERCISE 8.4

A4=12=22×3|A_4|=12=2^2\times3. Apply Cauchy's Theorem for each prime dividing 12.

12=22×312 = 2^2\times3, so the prime divisors are 22 and 33. By Theorem 8.6, A4A_4 must have an element of order 22 and an element of order 33 — but Cauchy's Theorem says nothing about elements of order 44 (since 44 is not itself prime, even though it divides 1212). This is consistent with A4A_4's well-known structure: it has elements of order 1,2,31,2,3 but none of order 44 or 66, showing Cauchy's Theorem is genuinely restricted to prime orders, not all divisors.

A4=12|A_4|=12. Using Cauchy's Theorem, what element orders are guaranteed to exist in A4A_4? What does the theorem not guarantee?


08 · Chapter Summary

ConceptStatement
Group actionG×XXG\times X\to X with ex=xe\cdot x=x and g(hx)=(gh)xg\cdot(h\cdot x)=(gh)\cdot x
Actions are bijectionsEach gg acts as a bijection of XX (Thm 8.1)
OrbitOrb(x)={gx:gG}\mathrm{Orb}(x)=\{g\cdot x:g\in G\}; orbits partition XX (Thm 8.2)
StabilizerStab(x)={g:gx=x}\mathrm{Stab}(x)=\{g:g\cdot x=x\}; always a subgroup (Thm 8.3)
Orbit-Stabilizer TheoremOrb(x)=[G:Stab(x)]\lvert\mathrm{Orb}(x)\rvert = [G:\mathrm{Stab}(x)] (Thm 8.4)
Conjugation actiongx=gxg1g\cdot x = gxg^{-1}; orbits = conjugacy classes; stabilizer = centralizer
Class equationG=Z(G)+[G:CG(xi)]\lvert G\rvert = \lvert Z(G)\rvert + \sum[G:C_G(x_i)] (Thm 8.5)
Cauchy's Theorempp prime, pG    Gp\mid\lvert G\rvert \implies G has an element of order pp (Thm 8.6)

Next: Chapter 09 — Sylow Theorems sharpens Cauchy's Theorem from single primes to full prime-power divisors, using the same orbit-counting and class-equation techniques developed here to describe the subgroup structure of any finite group almost completely.