Let u,v∈ker(f). Then
f(uv−1)=f(u)f(v−1)=f(u)f(v)−1=1⋅1=1.
So ker(f) is a subgroup. Suppose, for contradiction, that ker(f) is not normal. That is, let a∈ker(f) and b∈G such that b−1ab∈ker(f). Then
f(b−1ab)=f(b−1)f(a)f(b)=f(b−1)f(b)=1,
which implies b−1ab∈ker(f). This contradicts our hypothesis.
Let u,v∈im(f). By definition of image, there exist some a,b∈G such that f(a)=u,f(b)=v. Then
f(ab−1)=f(a)f(b−1)=f(a)f(b)−1=uv−1
and f(ab−1) is clearly in the image of f.
Let ∼ be the relation defined as
a∼b⟺f(a)=f(b)
(1) ∼ is an equivalence relation: this is inherited from equality. (2) ∼ is a congruence: suppose a∼b and v∼u. Then
f(av)=f(a)f(v)=f(b)f(u)=f(bu),
so av∼bu as needed. Partition G by ∼ (this constructs G/ker(f) explicitly) and note that any two elements in the same partition are mapped by f to the same element of im(f), so the map φ:G/ker(f)→im(f), sending [g] to f(g), is well defined. Now, let [p],[q]∈G/ker(f) and note that since ∼ is a congruence, [p][q] is well-defined and is equal to [pq]. Moreover
φ([pq])=f(pq)=f(p)f(q)=φ([p])φ([q]),
so φ is a homomorphism. φ is injective by construction, since if φ([a])=φ([b]), then f(a)=f(b) and hence [a]=[b]. It is surjective, since any a∈im(f) has a preimage in G which belongs to one of the equivalence classes. So a=φ([a]). ■
Intuition
A homomorphism gives rise to a congruence on . Since it is a congruence, the quotient is well-defined. By construction, sending any representative of some equivalence class in through covers all of with no overlaps (we collapsed these using ) and no missed spots (by definition of ). This gives an isomorphism.
Theorem B
Let S,N be subgroups of a group G with N normal. Then
The product SN is a subgroup of G.
N is a normal subgroup of SN
The intersection S∩N is a normal subgroup of S
SN/N≅S/(S∩N)
Proof
Let s1,s2∈S, g1,g2∈N and let g=g1g2−1∈N and s=s1s2−1∈S. Then
s1g1(s2g2)−1=s1g1g2−1s2−1=s1gs2−1,
but since N is normal, there exists g′∈N such that gs2−1=s2−1g′ and hence
s1gs2−1=s1s2−1g′=sg′.
so SN passes the subgroup test.
Indeed, conjugating an element of N by any element of G produces an element of N, so in particular conjugating an element of N by an element of SN produces an element of N.
Suppose g∈S and g∈N. Let a∈S and consider a−1ga. Clearly this is an element of S, but also note that since N is a normal subgroup of G, it has to be an element of N as well. Thus, conjugating any element of S∩N by an element of S produces an element of S∩N, so it is a normal subgroup.
Let Z:=S∩N. Let s∈S,n∈N and consider the coset snN∈SN/N. Clearly, snN=sN. Now, if s∈N then sN=N, the converse is also true, since sN=N in particular implies s⋅1=s∈N . Now, take the map f:S→SN/N sending s to sN. Clearly f is surjective, by the earlier argument. Note that f is a group homomorphism, since
f(s1s2)=(s1s2)N=!s1N⋅s2N=f(s1)f(s2),
where the marked equality holds because N is normal. By the first isomorphism theorem, that gives us an isomorphism im(f)=SN/N≅S/ker(f) and this kernel is precisely Z. ■
Intuition
The idea is that given SN, quotienting out N kills all of N and this includes the overlap with S. This can be seen clearly from cosets: if n∈N and s∈S, snN=sN and if s also happens to be in N, then sN=N.
Modules
Theorem B
Proposition
Let N, P be submodules of an R-module M. Then
N+P is a submodule of M;
N∩P is a submodule of P and
NN+P≅N∩PP
Proof
First, observe that N+P is a subgroup of M. Any element of N+P is already an element of M, and since R-action on N+P is inherited from M, it is sufficient to make N+P into a module, provided that closure holds.
Let n∈N and p∈P. For an arbitrary r∈R, consider r(n+p)=rn+rp. Since N is a submodule, rn∈N. Similarly, rp∈P. This demonstrates that rn+rp∈N+P, as required.
Similarly, for N∩P we only need to check closure. Let u∈N∩P and take an arbitrary r∈R. Since u∈N, ru∈N as well. Similarly, ru∈P. So N∩P is a submodule.
Let K=N∩P and consider the map φ:(N+P)→P/K defined as n+p↦p+K. It is clearly a surjective homomorphism. So im(φ)=P/K and ker(φ)=N.
Let’s check that φ is well-defined. Let n1+p1=n2+p2. Note that n1−n2=p1−p2 and since n1−n2∈N and p1−p2∈P, we know that this difference lives in K.