Preliminaries

Lemma (L3.1)

Let RR be a commutative ring. Then given any RR-module MM, the RR-module morphism φ:RM\varphi : R \to M is uniquely determined by the image of 11.

Proof

Since φ\varphi is an RR-module morphism, for any rRr \in R we have φ(r)=φ(r1)=rφ(1).\varphi(r) = \varphi(r1) = r \varphi(1). \blacksquare

Lemma (L3.2)

Let RR be a commutative ring. Then given any RR-module MM and an element mMm \in M there exists a morphism φ:RM\varphi : R \to M such that φ(1)=m.\varphi(1) = m.

Proof

Construct a set function φ(r)=rm.\varphi(r) = rm. It is an RR-module homomorphism, since

φ(r1r2)=r1r2m=r1(r2m)=r1φ(r2),\varphi(r_1r_2) = r_1r_2 m = r_1(r_2 m) = r_1 \varphi(r_2),

and

φ(r1+r2)=(r1+r2)m=r1m+r2m=φ(r1)+φ(r2).\varphi(r_1 + r_2) = (r_1 + r_2)m = r_1m + r_2m = \varphi(r_1) + \varphi(r_2).

\blacksquare

Exercise

Proposition

Let AA be any set, let RR be a commutative ring, and let FR(A)F^R(A) be the free RR-module on A.A. Let RAR^{\oplus A} be the direct sum of RR indexed by A.A. Then FR(A)RA.F^R(A) \cong R^{\oplus A}.

Proof

Note that as a direct sum RAR^{\oplus A} comes with an injection for each element of A.A. We can think of it as a map i:ARAi : A \to R^{\oplus A} such that for any a,bAa,b \in A, (i(a))a=1(i(a))_a = 1 and if aba \neq b then (i(a))b=0.(i(a))_b = 0.

Let MM be any RR-module and f:AMf : A \to M be any set-function. For each element aA,a \in A, construct using L3.1 a morphism φa:RM\varphi_a : R \to M such that φa(1)=f(a).\varphi_a(1) = f(a). By L3.2 we know that there is one such morphism for each a.a.

Now, consider the map

φ:RAMgiven byφ=aA(φaπa).\varphi : R^{\oplus A} \to M \quad \text{given by} \quad \varphi=\sum_{a\in A}(\varphi_a \circ \pi_a).

This function is well-defined, since any vRAv \in R^{\oplus A} is non-zero on finitely many indices and hence all but a finite subset of the summands vanish. Is it a homomorphism? For any vRAv \in R^{\oplus A} we have

φ(v)=φm1(vm1)+φm2(vm2)++φmk(vmk),\varphi(v) = \varphi_{m_1}(v_{m_1}) + \varphi_{m_2}(v_{m_2}) + \ldots + \varphi_{m_k}(v_{m_k}),

and so φ\varphi clearly satisfies all requirements.

By construction, φ(i(a))=φa(πa(i(a)))=φa(1)=f(a).\varphi(i(a)) = \varphi_a (\pi_a (i(a))) = \varphi_a(1) = f(a). Since φ\varphi is a copairing of uniquely determined morphisms φa\varphi_a, it is itself uniquely determined.

Thus, we have constructed the morphism φ\varphi required for RAR^{\oplus A} to satisfy the universal property of FR(A).F^R(A). This forces unique isomorphism FR(A)RA.F^R(A) \cong R^{\oplus A}.

\blacksquare