Generated Subgroups, Products, and Isomorphisms

December 1, 2025 · Luciano Muratore

It is well-known that groups can be built up from smaller pieces. A natural question is: given a subset SS of a group GG, what is the smallest subgroup of GG that contains SS?

Generated Subgroups

Let (G,)(G, \cdot) be a group and SGS \subseteq G. The subgroup generated by SS, denoted S\langle S \rangle, is defined as the smallest subgroup of GG containing SS, or equivalently,

S={HGSH}.\langle S \rangle = \bigcap \{H \leq G \mid S \subseteq H\}.

Intuitively: start with SS and close under products and inverses until nothing more can be added without leaving a subgroup. But what does S\langle S \rangle look like concretely?

The answer is that every element of S\langle S \rangle is a finite product of elements of SS and their inverses. More precisely, setting H:={1,s1s2snsiSS1}H := \{1, s_1 s_2 \cdots s_n \mid s_i \in S \cup S^{-1}\}, one can check that HH is a subgroup: 1H1 \in H by the empty product, a product of two words in SS1S \cup S^{-1} is again a word in SS1S \cup S^{-1}, and the inverse of s1sns_1 \cdots s_n is sn1s11s_n^{-1} \cdots s_1^{-1}, which is also in HH. Since HH contains SS and any subgroup containing SS must contain all finite products of its elements and their inverses, we conclude that H=SH = \langle S \rangle.

The Dihedral Group

A beautiful example of a generated subgroup is the dihedral group DnD_n. Consider a regular convex nn-gon in the plane. The dihedral group DnD_n is the group of all isometries of the plane that map the nn-gon to itself, with composition as the operation. It has Dn=2n|D_n| = 2n elements: nn rotations about the center and nn reflections with respect to lines through the center.

Label the vertices 1,2,,n1, 2, \ldots, n in cyclic order, let rr be the counterclockwise rotation by 2πn\frac{2\pi}{n}, and let ss be a reflection through, say, vertex 11 and the center. Then Dn=r,sD_n = \langle r, s \rangle.

Why? The rotations are {1,r,r2,,rn1}=r\{1, r, r^2, \ldots, r^{n-1}\} = \langle r \rangle. For the reflections, consider the conjugates rjsrjr^j s r^{-j} for j{0,,n1}j \in \{0, \ldots, n-1\}. Geometrically, rjr^{-j} rotates the polygon, ss reflects across a fixed axis, and rjr^j rotates back — the net effect is a reflection across the axis obtained by rotating the original axis by j2πnj \cdot \frac{2\pi}{n}. So every reflection appears as such a conjugate, and since DnD_n consists only of rotations and reflections, we get Dn=r,sD_n = \langle r, s \rangle.

Groups of Exponent 2 are Abelian

Here is a small but elegant observation. Let (G,)(G, \cdot) be a group such that g2=1g^2 = 1 for all gGg \in G — that is, every element has order dividing 2. Then GG is abelian.

The key observation is that g2=1g^2 = 1 implies g=g1g = g^{-1} for all gGg \in G. Take any a,bGa, b \in G. By hypothesis, (ab)2=1(ab)^2 = 1, so (ab)1=ab(ab)^{-1} = ab. But also (ab)1=b1a1=ba(ab)^{-1} = b^{-1}a^{-1} = ba, since every element is its own inverse. Therefore ab=baab = ba, and since this holds for all a,bGa, b \in G, the group is abelian.

Direct Products

Given two groups (G,)(G, \ast) and (H,)(H, \star), their direct product G×HG \times H is the set of pairs {(g,h)gG,hH}\{(g, h) \mid g \in G, h \in H\} with componentwise operation:

(g1,h1)(g2,h2)=(g1g2,h1h2).(g_1, h_1) \cdot (g_2, h_2) = (g_1 \ast g_2, h_1 \star h_2).

The intuition is simple: operate in GG on the first component and in HH on the second, independently. One can verify that this makes (G×H,)(G \times H, \cdot) a group: closure and associativity follow from those in GG and HH; the identity is (eG,eH)(e_G, e_H); and the inverse of (g,h)(g, h) is (g1,h1)(g^{-1}, h^{-1}).

A concrete and important example is Z2×Z2\mathbb{Z}_2 \times \mathbb{Z}_2, the Klein group V4V_4. It has four elements {(0,0),(1,0),(0,1),(1,1)}\{(0,0),(1,0),(0,1),(1,1)\}, every non-identity element has order 2, and the group is abelian. Interestingly, V4V_4 appears not only as an abstract product but also inside S4S_4: the set {e,(12)(34),(13)(24),(14)(23)}\{e, (12)(34), (13)(24), (14)(23)\} — the identity and the three bitranspositions — forms a subgroup of S4S_4 isomorphic to Z2×Z2\mathbb{Z}_2 \times \mathbb{Z}_2.

Some Fundamental Isomorphisms

Let me illustrate how direct products and isomorphisms interact through a few examples.

(R×,)(R+,)×(Z2,+)(\mathbb{R}^\times, \cdot) \cong (\mathbb{R}^+, \cdot) \times (\mathbb{Z}_2, +). Every nonzero real can be written as x=sgn(x)xx = \text{sgn}(x) \cdot |x|, and encoding the sign in Z2\mathbb{Z}_2 gives a bijective homomorphism φ(x)=(x,0)\varphi(x) = (|x|, 0) if x>0x > 0 and (x,1)(|x|, 1) if x<0x < 0.

(Z4,+)({1,1,i,i},)(\mathbb{Z}_4, +) \cong (\{1,-1,i,-i\}, \cdot). The map ψ([k]4)=ik\psi([k]_4) = i^k is a homomorphism since ψ(a+b)=ia+b=iaib=ψ(a)ψ(b)\psi(a+b) = i^{a+b} = i^a \cdot i^b = \psi(a)\psi(b), and it is bijective since i0,i1,i2,i3i^0, i^1, i^2, i^3 are distinct and exhaust the target.

(Z6,+)(Z2×Z3,+)(\mathbb{Z}_6, +) \cong (\mathbb{Z}_2 \times \mathbb{Z}_3, +). Define θ([k]6)=([k]2,[k]3)\theta([k]_6) = ([k]_2, [k]_3). This is a homomorphism by the compatibility of reduction modulo 2 and 3. It is injective because θ([k]6)=(0,0)\theta([k]_6) = (0,0) forces kk to be divisible by both 2 and 3, hence by 6. Since both groups have order 6, injectivity implies bijectivity.

The last example is a special case of a more general fact: ZmnZm×Zn\mathbb{Z}_{mn} \cong \mathbb{Z}_m \times \mathbb{Z}_n whenever gcd(m,n)=1\gcd(m, n) = 1. I would let the readers convince themselves of this.

Translations of a Vector Space

Let me close with a charming observation about translations. Let VV be a vector space over a field KK. For any vVv \in V, the translation by vv is the map τv:VV\tau_v : V \to V defined by τv(x)=x+v\tau_v(x) = x + v. The set of all translations T={τvvV}T = \{\tau_v \mid v \in V\}, under composition, forms a group: τvτw=τv+w\tau_v \circ \tau_w = \tau_{v+w}, the identity is τ0\tau_0, and the inverse of τv\tau_v is τv\tau_{-v}.

The map Φ:VT\Phi : V \to T defined by Φ(v)=τv\Phi(v) = \tau_v is an isomorphism. It is a homomorphism since Φ(v+w)=τv+w=τvτw=Φ(v)Φ(w)\Phi(v + w) = \tau_{v+w} = \tau_v \circ \tau_w = \Phi(v) \circ \Phi(w). It is injective because Φ(v)=Φ(w)\Phi(v) = \Phi(w) implies x+v=x+wx + v = x + w for all xx, hence v=wv = w. And it is surjective by definition of TT.

So (V,+)(T,)(V, +) \cong (T, \circ): the additive group of a vector space is nothing more than the group of its own translations. The abstract and the geometric are the same thing, dressed differently.