Proposition (H1)

Let MM be a [[Exercise A0.III.6.16#definition-cyclic-module|cyclic]] RR-module, with MR/IM \cong R/I for an ideal II (6.16.C2), and let NN be another RR-module. Let QQ be a submodule of NN given by

Q:={nN(aI):an=0}.Q := \{n \in N \mid (\forall a \in I): an = 0\}.

Then

HomR-Mod(M,N)Q.\Hom_{\RMod}(M,N) \cong Q.

Proof

Let π:RR/I\pi : R \to R/I be the quotient map and let aI.a \in I. If φ:MN\varphi: M \to N is a homomorphism, then

aφ(π(1))=φ(aπ(1))=φ(π(a))=0.a \varphi(\pi(1)) = \varphi(a\pi(1)) = \varphi(\pi(a)) = 0.

Now, consider the map v:HomR-Mod(M,N)Qv : \Hom_{\RMod}(M,N) \to Q given by ψψ(π(1)).\psi \mapsto \psi(\pi(1)). It is immediate that it is a module homomorphism. It is injective, since each element of HomR-Mod(M,N)\Hom_{\RMod}(M,N) is determined uniquely by its value on π(1)\pi(1) due to MM being cyclic.

It is surjective. By the universal property of a quotient, whenever a map τ:RN\tau : R \to N sends II to 00, it must factor uniquely through the quotient M,M, giving a homomorphism φn:MN,π(1)n\varphi_n : M \to N, \pi(1) \mapsto n for every value nNn \in N such that for all aI,a \in I, an=0.an = 0.

So vv is an isomorphism.

\blacksquare

NOTE

The same proof shape works without assuming commutativity of R,R, giving an isomorphism of abelian groups HomR-Mod(M,N)Q.\Hom_{\RMod}(M,N) \cong Q.

Proposition (H2)

For all a,bZ>0a,b \in \mathbb{Z}_{>0} we have HomAb(Z/aZ,Z/bZ)Z/gcd(a,b)Z.\Hom_\mathbf{Ab}(\mathbb{Z}/a\mathbb{Z}, \mathbb{Z}/b\mathbb{Z}) \cong \mathbb{Z}/\gcd(a,b)\mathbb{Z}.

Proof

DANGER

This proof is trash but I don’t know any better way. Trying to use H1 leads to even more number chasing.

A homomorphism f:Z/aZZ/bZf : \mathbb{Z}/a\mathbb{Z} \to \mathbb{Z}/b\mathbb{Z} is defined by its action on [1]a.[1]_a.

Suppose the order of [k]b[k]_b divides aa and f([1]a)=[k]bf([1]_a) = [k]_b. Then the map q:ZZ/bZq : \mathbb{Z} \to \mathbb{Z}/b\mathbb{Z} given by q(1)=[k]bq(1) = [k]_b (well-defined by the universal property of Z\mathbb{Z}) has the property q(an)=[ank]b=[0]b.q(an) = [ank]_b = [0]_b. So aZker(q)a\mathbb{Z} \subseteq \ker(q) and hence it factors uniquely (via the universal property of quotients) through Z/aZ.\mathbb{Z}/a\mathbb{Z}.

This shows that HomAb(Z/aZ,Z/bZ)\Hom_\mathbf{Ab}(\mathbb{Z}/a\mathbb{Z}, \mathbb{Z}/b\mathbb{Z}) contains all homomorphisms sending [1]a[1]_a to an element of order dividing a,a, and only those homomorphisms (since for any homomorphism gg, g(x)|g(x)| divides x|x|).

Recall that every subgroup of Z/bZ\mathbb{Z}/b\mathbb{Z} is cyclic and has order that is a factor of b.b. Recall also that for two subgroups A,BZ/bZA, B \leq \mathbb{Z}/b\mathbb{Z} the subgroup A+BA+B is the smallest subgroup containing both AA and B.B.

Now take the sum of all subgroups of Z/bZ\mathbb{Z}/b\mathbb{Z} whose order divides a.a. The order of this sum would be the product of the multiset of prime factors of bb contained within the multiset of prime factors of a.a. But this is, by definition, gcd(a,b).\gcd(a,b).

So there are exactly gcd(a,b)\gcd(a,b) suitable elements of Z/bZ.\mathbb{Z}/b\mathbb{Z}. Let [m]bZ/bZ[m]_b \in \mathbb{Z}/b\mathbb{Z} be an element of order gcd(a,b)\gcd(a,b) (in Ab\mathbf{Ab}, every factor of a finite group’s order corresponds to a subgroup). By definition, [m]b[m]_b generates the cyclic group Z/gcd(a,b)Z.\mathbb{Z}/\gcd(a,b)\mathbb{Z}.

Take fHomAb(Z/aZ,Z/bZ)f \in \Hom_\mathbf{Ab}(\mathbb{Z}/a\mathbb{Z}, \mathbb{Z}/b\mathbb{Z}) defined by f([1]a)=[m]b.f([1]_a) = [m]_b. Note that nfnf sends [1]a[1]_a to n[m]b.n[m]_b. Take any element [u]b[u]_b of Z/gcd(a,b)Z.\mathbb{Z}/\gcd(a,b)\mathbb{Z}. Since [m]b[m]_b generates Z/gcd(a,b)Z,\mathbb{Z}/\gcd(a,b)\mathbb{Z}, [u]b=k[m]b[u]_b = k[m]_b for some kZ.k \in \mathbb{Z}. So kfkf gives the unique homomorphism sending [1]a[1]_a to [u]b.[u]_b. Thus, ff generates the whole group.

Any two cyclic group of identical order are isomorphic.

\blacksquare