NowhereLog

Abstract Algebra 02 - The Isomorphism Theorems

January 01, 2021

This is the second post of the Introduction to Abstract Algebra series. In the first post, we introduce the basic concepts of groups. In this post, we will introduce four isomorphism theorems. But first of all, we need to define normal groups.

Normal Groups and Quotient Groups

Recall that we define the cosets as the following: Given a group G,HGG, H \subset G is a subgroup, for any gGg\in G, gHgH a left coset of GG, and HgHg is a right coset. We’ve already shown that all the cosets have the same cardinality (by defining an isomorphism). We’ve also shown that the set of left cosets forms a partition of the group. The idea of quotient groups, denoted as G/HG/H is motivated by the following question: can we define a group based on this set, or can we define a proper binary operation :G/H×G/HG/H\circ: G/H \times G/H \to G/H for this set? A perhaps natural binary operation is g1Hg2H=(g1g2)Hg_1H \circ g_2H = (g_1 g_2)H. This definition is very convenient, as it has the identity eHeH, an inverse for each element as gHg1H=eHgH \circ g^{-1}H = eH, and satisfies associativity. But it has one significant problem: it may not be well defined. You may notice that, a coset can be represented in multiple ways as long as g1H=g1Hg_1H = g_1'H. The question is it true that g1Hg2H=g1Hg2Hg1g2H=g1g2Hg_1H \circ g_2H = g_1'H \circ g_2H \equiv g_1g_2 H = g_1'g_2H? This seems hard to achieve, and indeed this is not true for most subgroups HH. This is only true if subgroup HH is normal.

Definition 2.1: Normal Groups

A subgroup HH of group GG is a normal subgroup, denoted as HGH \trianglelefteq G, if gG\forall g \in G, gHg1=H    gH=HggHg^{-1}=H \iff gH=Hg, i.e. left & right cosets coincide.

Definition/Lemma 2.2: Quotient Groups

Given a normal subgroup HH of group GG, the quotient group, demoted as G/HG/H, is the set of cosets of HH in GG, with the binary operation g1Hg2H=(g1g2)Hg_1H \circ g_2H = (g_1 g_2)H for g1,g2Gg_1, g_2 \in G.

ProofProof. We show that given HGH \trianglelefteq G, G/HG/H is indeed a well-defined group. For g1,g1,g2Gg_1, g_1', g_2 \in G, suppose g1H=g1Hg_1H = g_1'H. First observe that g1g2H=g1g2H    g21(g1)1g1g2H=H    g21(g1)1g1g2Hg_1g_2 H = g_1'g_2H \iff g_2^{-1}(g_1')^{-1}g_1g_2 H = H \iff g_2^{-1}(g_1')^{-1}g_1g_2 \in H. Since g1H=g1Hg_1H = g_1'H, (g1)1g1H(g_1')^{-1}g_1 \in H. Thus g21(g1)1g1g2g21Hg2=g21g2H=Hg_2^{-1}(g_1')^{-1}g_1g_2 \in g_2^{-1}Hg_2 = g_2^{-1}g_2H = H. The first equality is due to the normality of HH.

Definition 2.3: Simple Groups

A group GG is simple it has no proper normal subgroups.

Note that every group GG has trivial subgroups: {e},G\{e\}, G. The idea of normal and quotient groups may be intuitive if we make an analogy to integers. If we consider a group GG as an integer, then its normal subgroup is its divisor, and the quotient group is its quotient (and hence the name). Following this analogy, simple groups are like the prime numbers. Just like studying prime numbers is important in number theory as they are the building blocks of the number system, simple groups are like building blocks or fundamental structures. If a group is not simple, we can always “factor” it out into its normal and quotient groups, just like prime factorization.

Following this idea, we can always “project” a group to its quotient group via a homomorphism.

Definition/Lemma 2.4: Projection Homomorphism

If HGH \trianglelefteq G, the mapping π:GG/H\pi: G \to G/H defined as π(g)=gH\pi(g) = gH is a well-defined homomorphism.

ProofProof. The mapping is well-defined since the set of cosets is a partition of GG. It is a homomorphism since g1Hg2H=g1g2Hg_1Hg_2H = g_1g_2H by the normality of HH. Also note that the homomorphism is subjective.

Kernel and Image

Now we define the kernel and image of a homomorphism.

Definition 2.5: Kernel and Image

Given a homomorphism ϕ:GH\phi: G \to H,

  1. The kernel of ϕ\phi is Ker(ϕ)={gG:ϕ(g)=eHH}Ker(\phi) = \{g \in G : \phi(g) = e_H \in H\}.
  2. The image of ϕ\phi is Im(ϕ)={ϕ(g)=gG}Im(\phi) = \{\phi(g) = g \in G\}.

Lemma 2.6: Kernel and Image

  1. The kernel of any homomorphism ϕ:GH\phi: G \to H is a normal subgroup of GG.
  2. The image of any homomorphism ϕ:GH\phi: G \to H is a subgroup of HH.

ProofProof. It’s straightforward to check that the kernel and image are indeed subgroups of GG and HH by checking the three axioms or showing that they are closed under multiplication and inverses. Checking that the kernel is a normal subgroup is also straightforward but we will lay out the steps. We want to show that for gGg \in G, gKer(ϕ)g1=Ker(ϕ)gKer(\phi)g^{-1} = Ker(\phi), or for kKer(ϕ),gkg1Ker(ϕ)k \in Ker(\phi), gkg^{-1} \in Ker(\phi). We show this by noting that ϕ(gkg1)=ϕ(g)ϕ(k)ϕ(g1)=ϕ(g)eHϕ(g1)=eH\phi(gkg^{-1}) = \phi(g)\phi(k)\phi(g^{-1}) = \phi(g)e_H\phi(g^{-1}) = e_H.

Corollary 2.7: Kernel of Homomorphisms from Simple Groups

Given a homomorphism ϕ:GH\phi: G \to H where GG is a simple group, then Ker(ϕ)=H or {eH}Ker(\phi) = H \text{ or } \{e_H\}, i.e., ϕ\phi is trivial or injective.

The First Isomorphism Theorem

Theorem 2.8: The First Isomorphism Theorem

Given a homomorphism ϕ:GH\phi: G \to H, G/Ker(ϕ)Im(ϕ)G/Ker(\phi) \cong Im(\phi).

ProofProof. To prove this theorem, all we need to do is to define an isomorphism ψ:G/Ker(ϕ)Im(ϕ)\psi: G/Ker(\phi) \to Im(\phi). The idea is given the homomorphism ϕ\phi and the induced projection homomorphism defined in Lemma 2.4 π:GG/Ker(ϕ)\pi: G \to G/Ker(\phi), we define our ψ\psi so that ϕ=ψπ\phi = \psi \circ \pi, which will complete the proof. The idea is illustrated in Figure 1.

aa02-1

Figure 1. The First Isomorphism Theorem. (Image source: The First Isomorphism Theorem, Intuitively. Nov. 28, 2016. Algebra, Math3ma.)

Formally, let KK denote Ker(ϕ)Ker(\phi). We define ψ:G/KIm(ϕ)\psi: G/K \to Im(\phi) by ψ(gK)=ϕ(g)\psi(gK) = \phi(g). Thus, by definition, ϕ=ψπ\phi = \psi \circ \pi, so we only need to prove that this is a well-defined homomorphism. First, ψ\psi is well-defined since if g1K=g2Kg_1K = g_2K, then

g2g11K    ϕ(g2g11)=ϕ(g2)ϕ(g11)=eH    ϕ(g2)=ϕ(g1)g_2g_1^{-1} \in K \iff \phi(g_2g_1^{-1}) = \phi(g_2) \phi(g_1^{-1}) = e_H \iff \phi(g_2) = \phi(g_1).

Second, ψ\psi is a homomorphism since

ψ(g1Kg2K)=ψ(g1g2K)=ϕ(g1g2)=ϕ(g1)ϕ(g2)=ψ(g1K)ψ(g2K)\psi(g_1Kg_2K) = \psi(g_1g_2K) = \phi(g_1g_2) = \phi(g_1)\phi(g_2) = \psi(g_1K)\psi(g_2K)

Thus, we proved that ψ\psi is an isomorphism.

Symmetric Groups and Alternating Groups

Upon introduction of the first isomorphism theorem, it’s convenient to introduce two kinds of very important groups.

Consider a set S={1,2,3}S = \{1, 2, 3\} which contains 3 elements. We can define a bijective map σ:SS\sigma: S \to S, which is called a permutation. For example, we can send 1 to 2, 2 to 3, 3 to 1. This permutation is denoted as (1,2,3)(1,2,3). We can also send 1 to 3 and leave 2 unchanged, and this permutation is denoted as (1,3)(1,3). This “parenthesized” notation is called a cycle decomposition. There are in total 3!=63! = 6 permutations. For a set with more elements, the cycle decomposition can be more complex. For example, for a set of 66 elements, a permutation can be (1,3)(4,5,6)(1, 3)(4,5,6). It shouldn’t be too hard to check that a k-cycle has an order kk. More generally, a (m-cycle)(n-cycle) has an order lcm(m,n)lcm(m,n). We call a 2-cycle a transposition. A cycle decomposition can always be written as composition of transpositions, for example, (1,2,3)=(1,3)(1,2)(1,2,3) = (1,3)(1,2). In general, a m-cycle (a1,,an)(a_1, \cdots, a_n) is (a1,am)(a1,am1)(a1,a2)(a_1, a_m)(a_1, a_{m-1})\cdots(a_1, a_2).

Symmetric Groups of nn elements SnS_n contains all the permutations permutations on a set of nn elements.

Definition 2.9: Symmetric Groups SnS_n

The symmetric group on n letters, SnS_n, is a group with n!n! elements, where the binary operation is the composition of maps.

It should be straightforward to check that this is indeed a group. It is closed by multiplication and inverse since by definition it contains all the permutations.

Definition/Lemma 2.10: Alternating Groups AnA_n

The alternating group on n letters, AnA_n, is a subgroup of SnS_n, where An={σSn:σ is a product of even number of transpositions}A_n = \{\sigma \in S_n : \sigma \text{ is a product of even number of transpositions}\}.

ProofProof. To justify that AnA_n is indeed a subgroup of SnS_n, first note that the definition is unambiguous. If two products of transpositions are the same permutation, their length must have the same parity. We can first prove that the identity can only be written as the product of an even number of transpositions by induction on the length of the transpositions. This implies that two equivalent products of transpositions must have the same parity. Secondly, permutations that are products of even number of transpositions are closed under multiplication and inverses.

It may be intuitive that the size of AnA_n is always Sn/2|S_n|/2. However, this can be hard to prove if we don’t use techniques from group theory. By applying the first isomorphism theorem, we can prove this rather easily. To do this, we need to define another type of groups CnC_n.

Definition 2.11: Cyclic Groups CnC_n

For nZ1,CnCn \in \mathbb Z_{\geq 1}, C_n \subset \mathbb C, denotes the set of nth roots of unity, i.e., Cn={zC:zn=1}C_n = \{z \in \mathbb C : z^n = 1\}. The cyclic groups is defined as (Cn,×)(C_n, \times), where the binary operation ×\times is the usual multplication for numbers.

You may wonder why CnC_n have the same name as the property “cyclic”. Recall that all the cyclic groups of the same order are isomorphic, e.g., Z/ZnCn\mathbb Z/\mathbb Z_n \cong C_n, so in fact there are only one type of cyclic groups up to isomorphism. We just choose to use one of them for convenience.

Theorem 2.12: [Sn:An]=2[S_n : A_n] = 2

ProofProof. To prove this claim, we want to define the subjective homomorphism ϕ:SnC2\phi: S_n \to C_2 such that An=Ker(ϕ)A_n = Ker(\phi), and then the first isomorphism theorem gives the claim. Recall that C2={1,1}.First,wedefineamappingC_2 = \{1,-1\}. First, we define a mappingA:SnGL(n,Z)A: S_n \to GL(n, \mathbb Z)..GL(n,Z)GL(n,\mathbb Z)isafieldwhichwehaventintroduced,butfornowjustunderstanditasais a field which we haven’t introduced, but for now just understand it as an×nn\times nmatrixoverintegers.Wesketchthedefinitionmatrix over integers. We sketch the definitionAAwithanexamplewithwith an example withn=3n=3.For. Forσ=(1,2,3)\sigma = (1,2,3)$,

A(σ)=(010001100)A(\sigma)= \begin{pmatrix} 0 & 1 & 0 \\ 0 & 0 & 1 \\ 1 & 0 & 0 \end{pmatrix}

We define ϕ:SnC2\phi: S_n \to C_2 as ϕ=detA\phi = det \circ A. Without a proof, we claim that this is indeed a subjective homomorphism.

Proposition 2.13: Any subgroup of index 2 is normal.

Given a subgroup HGH \subset G where [G:H]=2[G:H]=2, there are only two left cosets: H,GHH, G-H. For a gGg \in G, if gH=HgH = H, then Hg=HHg = H and gH=HggH=Hg; else if gH=GHgH = G-H, then Hg=GHHg = G-H since HH is closed and gH=HggH=Hg.

Corollary 2.14: AnSnA_n \trianglelefteq S_n

The Second Isomorphism Theorem

Before introducing the second isomorphism theorem, we need some preparations.

Definition 2.15: Normalizer

Given a group GG, a subset HGH \subseteq G, the normalizer of HH in GG, denoted as NG(H)N_G(H), is NG(H)={gG:gHg1=H}N_G(H) = \{g\in G : gHg^{-1} = H\}.

We claim that NG(H)N_G(H) is a subgroup of GG, which should be easy to check by applying definitions. Note that HH is a normal subgroup of NG(H)N_G(H). In fact, the definition of NG(H)N_G(H) can be phrased as the largest subgroup of GG in which HH is normal. Note also that a group KK is normal to GG iff NG(K)=GN_G(K) = G.

Lemma 2.16: Product of Two Subgroups

Given two subgroups H,KGH, K \subset G, we define HK={hk:hH,kK}HK = \{hk : h\in H, k\in K\}.

  1. HKH\cap K is a subgroup of G,H,KG, H, K;
  2. HK=HKHK|HK| = \frac{|H||K|}{|H\cap K|};
  3. HKHK is a subgroup of GG iff HK=KHHK=KH;
  4. If HNG(K)H \subset N_G(K), then HK=KHHK=KH.

ProofProof. The first claim can be shown by checking the definitions. The second claim can be proven by the following steps. First, HK=hH{hK}=K{hK:hH}|HK| = |\cup_{h\in H}\{hK\}| = |K||\{hK : h \in H\}|. Second, for h1,h2Hh_1, h_2 \in H,

h1K=h2K    h21h1K    h21h1HK    h1(HK)=h2(HK)h_1K = h_2K \iff h_2^{-1}h_1 \in K \iff h_2^{-1}h_1 \in H \cap K \iff h_1 (H \cap K) = h_2 (H \cap K)

so {hK:hH}={h(HK):hH}|\{hK : h \in H\}| = |\{h(H \cap K) : h \in H\}|. Third, using the same arguments in the proof of Lagrange’s theorem {h(HK):hH}=HHK|\{h(H \cap K) : h \in H\}| = \frac{|H|}{|H\cap K|}.

The third claim can be shown by working from both directions, and use/check the axioms of groups. For the fourth claim, for all hHh\in H, hKh1=K    hK=KhhKh^{-1} = K \iff hK = Kh. Then HK=hHhK=hHKh=KHHK = \cup_{h\in H} hK = \cup_{h\in H} Kh = KH.

Theorem 2.17: The Second Isomorphism Theorem

Given two subgroups H,KG,HNG(K)H, K \subset G, H \subset N_G(K), then

  1. HKHK is a subgroup of GG;
  2. KHKK \trianglelefteq HK;
  3. HKHH \cap K \trianglelefteq H;
  4. HK/KH/(HK)HK/K \cong H/(H \cap K).

Figure 2. illustrated the theorem.

aa02-2

Figure 2. The Second Isomorphism Theorem. (Image source: p Groups. Intuition about the second isomorphism theorem. Apr. 12, 2016. Stackexchange.)

ProofProof. For the first claim, we’ve already proven it in Lemma 2.16. For the second claim, note that HNG(K)    KHKH \subset N_G(K) \implies K \trianglelefteq HK. For the third claim, we’ve already in Lemma 2.16 that HKHH \cap K \trianglelefteq H, so HNG(K)    HKHH \subset N_G(K) \implies H \cap K \trianglelefteq H. The first three claims make sure that the fourth claim is well-defined. To prove the fourth claim, we define the homomorphism ϕ:HHK/K\phi: H \to HK/K by ϕ(h)=hK\phi(h) = hK. To complete the proof, check directly that ϕ\phi is a well-defined homomorphism, ϕ\phi is surjective, and that Ker(ϕ)=HKKer(\phi) = H \cap K; then use the first isomorphism theorem to finish the proof.

The Third Isomorphism Theorem

Given two normal subgroups H,KG,HGH, K \subset G, H \subset G, then

  1. K/HG/HK/H \trianglelefteq G/H;
  2. Denote G/HG/H as G\overline G, and K/HK/H as H\overline H, G/KG/K\overline G/ \overline K \cong G/K.

ProofProof. The second claim is a bit like canceling out factors when we are working with numbers. To prove the first claim, we directly apply the definition of normality and check. To prove the second claim, we again define a subjective homomorphism and utilize the first isomorphism theorem. We define the homomorphism ϕ:G/HG/K\phi: G/H \to G/K by ϕ(gH)=gK\phi(gH) = gK. Again, we need to check that 1) ϕ\phi is well-defined; 2) ϕ\phi is a homomorphism; 3) ϕ\phi is subjective; and 4) Ker(ϕ)={gH:ϕ(gH)=K}={gH:gK=K}{gH:gK}=K/HKer(\phi) = \{gH:\phi(gH)=K\} = \{gH:gK=K\} \{gH:g \in K\} = K/H. The first three items can be shown by checking the definitions and applying the premises.

The Lattice Isomorphism Theorem

Given a normal subgroup NGN \trianglelefteq G, there exists a bijection from the set of subgroups of G that contains NN to the set of subgroups A=A/N\overline A = A/N of G/NG/N. Specifically, for all subgroups A,BGA, B \subseteq G such that NA,NBN \subseteq A, N \subseteq B,

  1. ABA \subseteq B iff AB\overline A \subseteq \overline B;
  2. If ABA \subseteq B, then [B:A]=[B:A][B:A] = [\overline B: \overline A];
  3. (AB)\overline{(A \cap B)} = AB\overline A \cap \overline B;
  4. ABA \trianglelefteq B iff AB\overline A \trianglelefteq \overline B;

The bijection is given by some π:GG\pi : G \to \overline G, π1(A)={aA:π(a)=aNA}\pi^{-1}(\overline A) = \{a\in A : \pi(a) = aN \in \overline A\}, i.e. the inverse is the pre-image of π\pi.

This theorem is called the lattice isomorphism theorem, since it allows us to “factor” a group into its subgroups, which is the lattice of the group, as the lattice of D16D_{16} (dihedral groups which we haven’t defined yet) is illustrated in Figure 4.

Example 2.1: Dihedral groups D2nD_{2n}

aa02-4

Figure 3. The Generators of Dihedral groups D2nD_{2n}

Dihedral groups D2nD_{2n} are subgroups of SnS_n which consist of the rigid motions (reflections and rotations) of a regular n-sided polygon or n-gon. It should take some pondering, but there are exactly 2n2n unique motions, further every motion can be generated by the reflection and the 2π/n2\pi/n rotation. Say the reflection is ss and the rotations is rr, then formally,

D2n=r,sr2n=s2=e,rsr=sD_{2n} = \langle r, s \mid r^{2n} = s^2 = e, rsr = s \rangle

aa02-3

Figure 4. The Lattice of D16D_{16}.

References

[1] Siegel, Kyler. MATH GU4041 Introduction to Modern Algebra I. Department of Mathematics, Columbia University. 2019.

[2] Dummit, David and Foote, Richard. Abstract Algebra. Third Edition. John Wiley and Sons, Inc. 2004.


By NowhereMan who goes nowhere.


© 2024, NowhereLog by NowhereMan