31 Quotient Groups and Homomorphisms

1

Question

Let φ:GH be a Homomorphism (or Group Morphism) and let E be a Subgroup of H. Prove that φ1(E)G, namely that the preimage or pullback of a subgroup under a homomorphism is a subgroup.
If EH prove that φ1(E)G. Deduce that Ker(φ)G. (see Normal Subgroup)

Proof

(φ1(E)G): Let a,b1φ1(E). We just have to show that ab1φ1(E). By the definition of Fiber (Group Theory), namely φ1(E)={gG|φ(g)E} then there are elements e1,e2E such that φ(a)=e1 and φ(b1)=e2=φ(b)1. Notice that:

φ(ab1)=φ(a)φ(b)1=e1e2G

Thus choose e=e1e2G. Thus we found some element eG such that φ(ab1)=eE, so then ab1φ1(E) as desired.

(EHφ1(E)G): Suppose EH. Already we have φ1(E)G so to show that φ1(E)G we want to show one of the Equivalences for Normal Subgroups, namely that for any gG and aφ1(E) that gag1φ1(E).

Let gG and aφ1(E), so then there is some eE such that φ(a)=e. Construct h=φ(g)H. Notice that since EH then for all hH,eE then heh1E. But we have this since our constructed hH and thus:

heh1E

Thus there is some eE where heh1=e. But look:

heh1Eφ(g)eφ(g1)Eφ(g)φ(a)φ(g1)Eφ(gag1)Egag1φ1(E)

as we desired.

(Ker(φ)G): We already have Ker(φ)G since it is a clear subset and any g1,g21Ker(φ) has it that φ(g1)=φ(g21)=1H so then 1H=φ(g1)φ(g21)=φ(g1g21) so then g1g21Ker(φ) (showing it's a subgroup).

Consider that Ker(φ)=φ1(1H). Thus replacing all the E's with {1H} we trivially know {1H}H so then by our findings then Ker(φ)G.

3

Question

Let A be an Abelian Group and let B be a Subgroup of A. Prove that AB (the Quotient Group) is also abelian. Give an example of a non-abelian group G containing proper Normal Subgroup N such that GN is abelian.

Proof

(Abelian part): Here AB is the set of left-cosets of B in A, or just that for any aA then aB={ab|bB} are all elements of AB.

Let X,YAB be arbitrary. Then X=xB for xA and Y=yB where yA. We want to show that AB is abelian, which is true if XY=YX. This is true using the Coset Product since:

XY=(xB)(yB)=(xy)B=(yx)B(A is abelian xy=yx)=(yB)(xB)=YX

Thus AB is abelian.

(Example): Let G=D8=r,s:r4=1,s2=1,rs=sr3 which is non-abelian. Consider the normal subgroup N={1,r2} which is normal because it is abelian. Notice then that GN is abelian as it is {gn:nN}={gr2i:0i<2}.

4

Question

Prove that in the Quotient Group GN, (gN)α=gαN for all αZ.

Proof

We are assuming, via the definition of quotient group, that NG.

For the base case of our induction notice using Coset Product:

(gN)2=(gN)(gN)=(gg)N=g2N

continuing on gives all αZ+ cases.

Notice for α=0 then (gN)0=N=g0N as N is the identity for the quotient group.

Doing a similar induction on α over Z will work as long as we show the base case that (gN)1=g1N. Notice that:

(gN)1(gN)=N=(g1g)N

by the definition of being the inverse. But what is the inverse of (gN)=gN? Namely notice that:

1N=(g1g)N=(g1N)(gN)

Thus then:

(gN)1(gN)=(g1N)(gN)

and using Cancellation Theorem (right cancellation):

(gN)1=g1N

as we desired.

5

Question

Use the preceding exercise to prove that the Order (Groups) of the element gN in GN (see Quotient Group) is n, where n is the smallest positive integer such that gnN (and gN has infinite order if no such positive integer exists).
Give an example to show that the order of gN in GN may be strictly smaller than the order of g in G.

Proof

(Proof): Say n is the smallest positive integer such that gnN. Then using our previous finding then:

(gN)n=gnN=N

as gnN. Thus the order of gN by definition being in GN must be n.

If such an n didn't initially exist, then nN(gnN). Assume for contradiction that some nN exists such that (gN)n=N (implying that there's a finite order for gN). Then:

(gN)n=gnN

but since gnN then gnNN, which is a contradiction (contradicts the order of gN being n). Thus such an n cannot exist so |gN| is infinite.

(Example): Consider G=S4. Let x be a generator for G, and take N=x2={1,x2}. Now |x2|=2 in G (since (x2)2=x4=1) but x2N=1N has order 1 in GN (since x2N=1N).

7

Question

Define π:R2R where (x,y)x+y. Prove that π is a surjective Homomorphism (or Group Morphism) and describe the Kernel and fibers of π geometrically.

Proof

(π is surjective): Let rR be arbitrary. Construct the element (r,0)R2
such that π((r,0))=r+0=r. Thus π is arbitrary.

(Homomorphism (or Group Morphism)): Let x,yR2 be arbitrary. Say x=(x1,x2) and y=(y1,y2). Notice that (using the product of vectors to give a vector with corresponding entries added together, and using addition over R):

π(xy)=π(x+y)=π(x1+x2,y1+y2)=x1+x2+y1+y2=(x1+y1)+(x2+y2)=π(x)+π(y)=π(x)π(y)

The kernel Ker(π)={xR2|π(x)=0} would be the set of vectors:

z=(z,z)zR

Since π(z)=z+(z)=0 so clearly zKer(π). If any other vector didn't follow this then it wouldn't be in the kernel. These are the vectors that point in the same direction of (1,1), creating a line that intercepts the origin.

Further for any specific rR then the fiber above r would be:

π1(r)={zR2|π(z)=r}

Notice if we consider any z=(z1,z2) here then if we want π(z)=r then:

z1+z2=rz=(z1,rz1)

Dropping z=z1 then any zR gives a vector z=(z,rz)=z(1,1)+(0,r). Thus, geometrically, any fiber above r would be the lines pointing in the direction (1,) with y-intercept at r.

14ab

Question

Consider the additive Quotient Group QZ:
a. Show that every coset of Z in Q contains exactly one representative qQ in the range 0q<1.
b. Show that every element of QZ has finite order but that there are elements of arbitrarily large order.

Proof

a. We'll just consider left cosets as the right coset case is largely the same. Take any coset of Z in Q, denoted qZ. We want to show that if q1Z=q2Z then q1=q2[0,1).

Suppose q1Z=q2Z. Since the Left (and Right) Cosets form a Partition in G and since q1Z=q2Z then q21q1Z. Then there is some integer mZ where:

m=q21+q1
Note

We are dealing with the additive quotient group, so what looks like multiplication of the elements of Q here is actually their addition.

Going a bit back, realize that if q1,q2[0,1) then they equal some integer plus a fraction that is in that range, and then we'd get rational numbers that are in that range:

q1=n1+q1[0,1),q2=n2+q2m=q1q2=(n1n2)+(q1q2)[0,1)m=q1q2

Thus we know that at the very least we can represent our q's with q1,q2[0,1). For contradiction assume that q1>q2. Notice that m=0 because m=q1q2 which are in the set [0,1) with q1>q2. That means that q1=q2 which is a contradiction, so then q1q2. A similar argument shows q1q2 so then q1=q2, as desired.

b. Let qZQZ. Using (a) then q[0,1) and is unique. Then have q=ab where b0 and a<b. Notice that here since we are dealing with an additive group then our Order (Groups) comes from repeated multiplication, meaning we want to see when nqZ for some integer n.

Notice though that bq=aZ exactly! Thus, then q has order b, which must be some positive natural number.

Since qZ was arbitrary, each coset represented by q has a finite order given by the denominator of q. We can always make this b arbitrarily large.

22

Question

a. Prove that if H and K are Normal Subgroups of a Group G then their intersection HK is also a normal subgroup of G.
b. Prove that the intersection of an arbitrary nonempty collection of normal subgroups of a group is a normal subgroup (you cannot assume the collection is countable).

Proof

See 21 Subgroups#10 for showing how HK (and an arbitrary intersection of them) is also a subgroup.

a. We already know HKG so we need to show that g(HK)=(HK)g for all gG (since then we have Equivalences for Normal Subgroups (1)). Let gG. Notice that for some pg(HK) then:

pg(HK)hHK(p=gh)p=hg(HGKG)hHK(p=hg)p(HK)g

where that middle step comes from the fact that if hH then you use HG and if not then hK so you use KG.

Overall, since p was arbitrary then g(HK)=(HK)g showing normalcy, and thus HKG.

b. Say such a collection A={Hλ:λΛ} is something we are intersecting over. We want to see if:

A=λΛHλ

is a normal subgroup of G. As we showed before it's a subgroup so all we have to show is that for any gG that we have:

g(λΛHλ)=(λΛHλ)g

Now then we have some pg(λΛHλ) iff:

pg(λΛHλ)hλΛHλ(p=gh)p=hghλΛHλ(p=hg)p(λΛHλ)g

where the abelianness, like in (a), comes from the fact that hλΛHλ means there any HλA has HλG, so then since h is in at least one of these Hλ's then HλG and thus p=hg=gh.

24

Question

Prove that if NG (see Normal Subgroup) and H is any subgroup of G then NHH.

Proof

Notice that NHH. Notice also if we let g1,g2NH then we can show that g11g2NH since g1,g2H so then by N,H being a subgroup of G then g11,g2NH and thus their product is in NH. Thus NHH.

Now we want to do the Equivalences for Normal Subgroups, so we'll show that h(NH)=(NH)h for all hH. Let hH. Then for any p:

ph(NH)nNH(p=hn)p=nh(nNGng=gngG)nNH(p=nh)p(NH)h

where to clarify for the middle step, because nNG then that's iff ng=gn for all gG. But HG implies that hHG thus hG! Thus this applies and we can make the swap.

27

Question

Let N be a finite subgroup of a group G. Show that gNg1N iff gNg1=N (this is essentially showing one of the equivalences for Equivalences for Normal Subgroups). Deduce that NG(N)={gG|gNg1N} (see Normalizer).

Proof

Clearly the reverse direction is true so well only check the forward direction. Suppose for any gG that gNg1N for some finite NG. We want to show that NgNg1 for that same gG.

Notice that since N is a finite subgroup then |gNg1||N| by gNg1N. Assume for contradiction that instead NgNg1. Then by that same reasoning then |gNg1|<|N|. However, we can create a bijection φ:NgNg1 where ngng1. To show this is a bijection:

Notice using the definition of Normalizer with our equivalence gives the needed deduction.

31

Question

Prove that if HG and N is a Normal Subgroup of H then HNG(N). Deduce that NG(N) is the largest subgroup of G in which N is normal (ie: is the joining of all subgroups H for which NH).

Proof

(HNG(N)): Suppose HG and NH.

First to show HNG(N). Let hH. We want to show that hNG(N)={gG|gNg1=N}, which is true iff hNh1=N. But since NH then by the Equivalences for Normal Subgroups then we automatically get that.

Second, clearly H will be closed under binary operations and inverses by being a subgroup of G, so we don't need to check that.

Notice that N is a normal subgroup of itself, so then NNG(N) (replace H with N from our finding above, then add that N itself is already normal). Since N was arbitrary, then the largest normal N we can choose must be when we use NG(N) as our normal subgroup.

36

Question

Prove that if GZ(G) is a Cyclic Group then G is an Abelian Group.

In proof terms, you'll want to show that if GZ(G) is cyclic with generator xZ(G) then show that every element of G can be written in the form xaz for some integer aZ and some element zZ(G). For a refresher on the notation see Centralizer.

Proof

Let GZ(G) be a quotient group that's also a cyclic group and centralizer Z(G)=CG(G)={gG|gh=hghG}. This quotient group is the set of all left-cosets:

GZ(G)={gZ(G):gG}

Now, let's do the proof. Because GZ(G) is cyclic then it can be generated by one single element xG such that:

GZ(G)={xiZ(G):iZ}

To start showing that G is abelian, let g1,g2G be arbitrary. Notice that g1Z(G),g2Z(G)GZ(G). Thus then i,jZ such that:

g1Z(G)=xiZ(G),g2Z(G)=xjZ(G)

Since g1Z(G)=xiZ(G) then by Left (and Right) Cosets form a Partition in G then we must have xig1Z(G) (and a similar thing for xjg2Z(G). If xig1=z1 and xjg2=z2 then by each z1,z2Z(G) then the z's will commute with any of the g's and x's (since either is in G). Thus:

xig1=z1g1=xiz1:g2=xjz2

Furthermore the z's will commute with each other, as they each are also in G. Thus:

g1g2=(xiz1)(xjz2)=xixjz1z2(xnz1=z1xnnZ)=xjxiz1z2=xjxiz2z1(z1z2=z2z1)=(xjz2)(xiz1)(xnz2=z2xnnZ)=g2g1

41

Question

Let G be a Group. Prove that N=x1y1xy|x,yG is a Normal Subgroup of G and GN is an Abelian Group (here N is called the commutator subgroup of G).

Proof

(N is normal): Clearly already NG. For the normal part, let gG and x1y1xyN (for some nN). Then we have:

g(x1y1xy)g1=gx1(g1g)y1(g1g)x(g1g)yg1=(gx1g1)(gy1g1)(gxg1)(gyg1)N

where by Equivalences for Normal Subgroups then N must be normal. Notice that raising our generator to various powers won't change it being in N as N is closed under these powers.

(GN is abelian): Let zN be arbitrary. Then:

x1y1xy=zxy=yxz

Now notice that:

axyNa=xyz0z0Na=yxzz0ayxN

suggesting that xyN=yxN for any x,yG. Thus for any arbitrary aN,bNGN then:

(ab)N=(ba)Nab=ba

since N is normal.