MIT 18.701 — Lecture 2

Group Examples, Subgroups, Isomorphisms, Homomorphisms

§ Group Theory Definitions (cont.)

Theorem. (Uniqueness of Identity) Any group GG has an identity 1∈G1 \in G. That identity 11, in fact, is unique.

Proof: If GG had two identities 11 and 1′1', then 1=11′=1′1 = 11' = 1', so in fact 1=1′1 = 1'.   ■\blacksquare

Definition (Cyclic Group). The cyclic group of order nn, denoted CnC_n, is the set {0,1,2,…,n−1}\{0, 1, 2, \dots, n - 1\} together with operation “addition modulo nn”.

Definition (Cyclic Group v2). Equivalently, CnC_n is the group {1,g,g2,…,gn−1}\{1, g, g^2, \dots, g^{n - 1}\} together with operation “multiplication, given gn=1g^n = 1”. This representation of CnC_n better shows how CnC_n is generated by a single element gg.

Definition (Direct Product). Given groups GG and HH, the group G×HG \times H is the set {(g,h):g∈G,h∈H}\{(g, h): g \in G, h \in H\} together with the operation (g1,h1)(g2,h2)=(g1g2,h1h2)(g_1, h_1)(g_2, h_2) = (g_1g_2, h_1h_2).

§ Group Examples by Order

We can now classify all groups of small order up to isomorphism.

§ Subgroups

Definition (Subgroup). A group HH is a subgroup of another group GG if (i) H⊆GH \subseteq G, and (ii) the operation of HH is directly inherited from the operation of GG.

Example. The trivial group is a subgroup of any group.

Example. For any groups GG and HH, the group G×HG \times H has subgroups G×{1H}G \times \{1_H\} and {1G}×H\{1_G\} \times H.

Example. The group (R,+)(\mathbb{R}, +) has subgroups (Z,+)(\mathbb{Z}, +), (Q,+)(\mathbb{Q}, +), and (Qπ+Qe,+)(\mathbb{Q}\pi + \mathbb{Q}e, +) (among many others).

Remark. If H1H_1 and H2H_2 are subgroups of GG, then H1∩H2H_1 \cap H_2 is also a subgroup! However, H1∪H2H_1 \cup H_2 is not usually a subgroup; closure is typically violated.

What if we consider subgroups in the context of matrix groups?

Example. The group GLn(R)GL_n(\mathbb{R}) has the following subgroups.

§ Geometric Groups

Groups can capture geometric symmetries, turning geometric problems into algebraic problems!

Definition (Dihedral Groups). For any positive integer n≥3n \geq 3, the group DnD_n consists of all symmetries of a regular nn-gon. Note that ∣Dn∣=2n|D_n| = 2n; there are nn rotational symmetries and nn reflectional symmetries.

More complicatedly, one can consider the group of all symmetries of an icosahedron.

Tip: Learn how to freehand an icosahedron; it's not as hard as it looks!

Get excited to learn more about icosahedra!

§ Isomorphisms

We hand-waved the definition of isomorphism earlier. Let's fix that.

Definition (Isomorphism). Consider groups (G,⋅)(G, \cdot) and (H,⋆)(H, \star). Then an isomorphism is a bijection f:G→Hf: G \to H such that for any g1,g2∈Gg_1, g_2 \in G, we have f(g1⋅g2)=f(g1)⋆f(g2)f(g_1 \cdot g_2) = f(g_1) \star f(g_2). In other words, you can pair the elements of GG and HH so that the products correspond under this pairing.

Definition (Isomorphic). Groups (G,⋅)(G, \cdot) and (H,⋆)(H, \star) are isomorphic if there exists an isomorphism f:G→Hf: G \to H. We use the notation G≅HG \cong H to indicate that GG and HH are isomorphic.

Theorem. (Image of the Identity) Given an isomorphism f:G→Hf: G \to H, we have f(1G)=1Hf(1_G) = 1_H.

Proof: By definition of isomorphism, observe:

f(1G)⋆f(1G)=f(1G⋅1G)   ⟹   f(1G)⋆f(1G)=f(1G)   ⟹  ⋆f(1G)−1 f(1G)=1H.   ■f(1_G) \star f(1_G) = f(1_G \cdot 1_G) ~ \implies ~ f(1_G) \star f(1_G) = f(1_G) ~ \stackrel{\star f(1_G)^{-1}}{\implies} ~ f(1_G) = 1_H. ~~~ \blacksquare

Example. The group of n×nn \times n permutation matrices is isomorphic to SnS_n.

Example. Recall the coincidence 2n=n!2n = n! when n=3n = 3; therefore, ∣D3∣=∣S3∣|D_3| = |S_3|. It turns out, in fact, that D3D_3 and S3S_3 are isomorphic! The isomorphism can be found by interpreting any symmetry of the equilateral triangle as a permutation of its vertices; the nontrivial rotations yield 33-cycles, and reflections yield 22-cycles.

§ Group Generation

Definition (Subgroup Generated by Subset). Let SS be a subset of a group GG. Then the subgroup generated by S\mathbf{S} is the intersection of all subgroups of GG that contain SS.

Equivalently, we can think of the subgroup generated by S={x1,x2,…,xn}S = \{x_1, x_2, \dots, x_n\} as the set of all possible results of multiplying terms of the form xix_i or xi−1x_i^{-1} in some order. (Analogies: span, convex hull, fruit in a blender, …)

Example. If SS is not just a subset of GG, but more strongly a subgroup, then the subgroup generated by SS is just SS.

This lets us more formally define group generation.

Definition (Generation). A group GG is said to be generated by S⊆GS \subseteq G if the subgroup generated by SS is GG.

Theorem. (Generation by a Single Element) If GG is a group generated by a single element gg, then GG is either a cyclic group Cn={1,g1,g2,…,gn−1}C_n = \{1, g^1, g^2, \dots, g^{n - 1}\}, or GG is (Z,+)(\mathbb{Z}, +).

Proof: By definition of subgroup generation, GG must contain every element in the set {…,g−2,g−1,1,g1,g2,… }\{\dots, g^{-2}, g^{-1}, 1, g^1, g^2, \dots\}. If all of these elements are distinct, then GG is just (Z,+)(\mathbb{Z}, +).

Suppose some two elements are not distinct, then. This lets us choose distinct integers aa and bb such that ga=gbg^a = g^b, which implies ga−b=1g^{a - b} = 1. If we choose (a,b)(a, b) so that n=∣a−b∣n = |a - b| is as small as possible, then G≅CnG \cong C_n.   ■\blacksquare

Unfortunately, the set of groups generated by two elements is much less tractable.

Theorem. (Subgroups of (Z,+)\mathbf{(\mathbb{Z}, +)}) The only subgroups HH of (Z,+)(\mathbb{Z}, +) are of the form (dZ,+)(d\mathbb{Z}, +) for some integer dd.

Proof: One case is H={0}H = \{0\}, corresponding to d=0d = 0. So let's assume H≠{0}H \neq \{0\}, which lets us take dd to be the smallest positive integer in HH.

Claim. For any n∈Hn \in H, nn must be a multiple of dd.

Proof: Suppose the contrary. Divide nn by dd to yield qq as a quotient and rr as a remainder. It must follow that n=dq+rn = dq + r, where 0<r<d0 < r < d.

By closure of HH, however, observe that n−d−d⋯−d⏟q times=rn \underbrace{- d - d \dots - d}_{q \text{ times}} = r must also be in HH. But 0<r<d0 < r < d, contradicting the assumption that dd was the smallest positive integer in HH.   □\square

The above claim immediately implies H=dZH = d\mathbb{Z}, as desired.   ■\blacksquare

§ Homomorphisms

What if isomorphisms didn't have to be bijective?

Definition (Homomorphism). A homomorphism between groups (G,⋅)(G,\cdot) and (H,⋆)(H, \star) is a function f:G→Hf: G \to H such that for any g1,g2∈Gg_1, g_2 \in G, we have f(g1⋅g2)=f(g1)⋆f(g2)f(g_1 \cdot g_2) = f(g_1) \star f(g_2).

Example. All isomorphisms are homomorphisms.

Example. Let (G,⋅)(G, \cdot) and (H,⋆)(H, \star) be literally any groups. Then f:G→Hf: G \to H defined by f(g)=1Hf(g) = 1_H for all g∈Gg \in G is a homomorphism: the trivial homomorphism.

Example. If HH is a subgroup of GG, then the inclusion (“do-nothing”) function f:H→Gf: H \to G is a homomorphism.

Example. There is a homomorphism f:(C,+)→(C×,⋅)f: (\mathbb{C}, +) \to (\mathbb{C}^{\times}, \cdot) (where C×:={z∈C:z≠0}\mathbb{C}^{\times} := \{z \in \mathbb{C}: z \neq 0 \}) defined by f(z)=ezf(z) = e^z. The condition f(z1+z2)=f(z1)⋅f(z2)f(z_1 + z_2) = f(z_1) \cdot f(z_2) is equivalent to the identity ez1+z2=ez1⋅ez2e^{z_1 + z_2} = e^{z_1} \cdot e^{z_2}.

Example. Take G=(Z,+)G = (\mathbb{Z}, +) and H=CnH = C_n. Then f:G→Hf: G \to H defined by f(k)=“the remainder of k modulo n”f(k) = \text{``the remainder of } k \text{ modulo } n\text{''} is a homomorphism. This is a preview of a quotient homomorphism.

Example. Take G=GLn(R)G = GL_n(\mathbb{R}) and H=(R×,⋅)H = (\mathbb{R}^{\times}, \cdot) (where R×:={r∈R:r≠0}\mathbb{R}^{\times} := \{r \in \mathbb{R}: r \neq 0 \}). Then f:G→Hf: G \to H defined by f(M)=det⁡Mf(M) = \det M is a homomorphism.

That last example lets us prove a very funny theorem.

Theorem. (Determinant of Permutation Matrices) Let MM be an n×nn \times n permutation matrix. Then det⁡M=±1\det M = \pm 1.

Proof: Let ff denote the aforementioned homomorphism f:GLn(R)→(R×,⋅)f: GL_n(\mathbb{R}) \to (\mathbb{R}^{\times}, \cdot). Then:

f(M)⋅f(M−1)=f(MM−1)=f(In)=1f(M) \cdot f(M^{-1}) = f(MM^{-1}) = f(I_n) = 1

Recall f(M)=det⁡Mf(M) = \det M and f(M−1)=det⁡M−1f(M^{-1}) = \det M^{-1} must be integers since MM and M−1M^{-1} are both permutation matrices. Yet the product of these integers is 11, so the only possibility is that both f(M)f(M) and f(M−1)f(M^{-1}) are ±1\pm 1, as desired.   ■\blacksquare

Taking into account the fact that SnS_n is isomorphic to the group of n×nn \times n permutation matrices, we may define:

Definition (Permutation Sign). Consider the natural isomorphism f:Sn→group of permutation matricesf: S_n \to \text{group of permutation matrices}. Then we say that the sign of a permutation π∈Sn\pi \in S_n is det⁡f(π)\det f(\pi). (Note that the sign is always ±1\pm 1.)