I need to prove that $(R^{ \oplus A_1 })^{ \oplus A_2 } \cong R^{ \oplus (A_1 \times A_2)}$ as $R$-modules.

Recall that

if $N$ is an $R$-module and $A$ is a set, then we can construct $R$-module $N^{ \oplus A} = \{ \alpha: A \to N \ | \ \alpha(a) \neq 0$ for only finitely many $a \in A \}$.

An $R$-module structure is defined componentwise: if $r \in R, \alpha, \beta \in N^{ \oplus A}$, then

$(\alpha+\beta)(a) = \alpha(a) + \beta(a)$

$0_{N^{ \oplus A}} = 0: A \to N$ such that $\forall a \in A \ \ 0(a) = 0$

$(- \alpha)(a) = -\alpha(a)$

$(r\alpha)(a) = r(\alpha(a))$

Now we have two such $R$-modules. Since $R^{\oplus (A_1 \times A_2)}$ is the free $R$-module on a set $A_1 \times A_2$ , every function $\alpha \in R^{ \oplus (A_1 \times A_2) }$ can be written as a finite sum $\sum\nolimits_{x \in A_1, y \in A_2} r_{x,y}j_{x,y}$ such that $r_{x,y} = \alpha(x,y)$ and $j_{x,y}(x',y') = 1$ if $x' = x$ and $y' = y$ and $j_{x,y}(x',y') = 0$ is $x' \neq x$ or $y' \neq y$.

So we need to construct an $R$-module isomorphism $\phi: R^{ \oplus (A_1 \times A_2)} \to (R^{ \oplus A_1 })^{ \oplus A_2 }$.

Since $\phi$ must be an $R$-module homomorphism,

$\phi( \sum\nolimits_{x \in A_1, y \in A_2} r_{x,y}j_{x,y} ) = \sum\nolimits_{x \in A_1, y \in A_2} r_{x,y}\phi(j_{x,y})$

So we it suffices to define an image of each $j_{x,y}$. Any ideas?

Assumed background: basic properties of rings and modules, categories and universal properties( such as the one of a free $R$-module or a categorical product ) are introduced, but functors, limits and colimits are not. So this is not a duplicate of this question.


Let $f \in R^{\oplus(A_{1} \times A_{2})}$. Define a map $\varphi_{f} \colon A_{2} \to R^{\oplus A_{1}}$ by $\alpha \mapsto h_{f}$, where $h_{f} \in R^{\oplus A_{1}}$ is the map given by $h_{f}(\beta) = f(\beta, \alpha)$. Now, define a map $\Phi \colon R^{\oplus(A_{1} \times A_{2})} \to (R^{\oplus A_{1}})^{\oplus A_{2}}$ by $\Phi(f) = \varphi_{f}$. You should check three things:

$1$) $\Phi$ is well-defined, i.e $\phi_{f}$ indeed belongs to $(R^{\oplus A_{1}})^{\oplus A_{2}}$ (part of this will involve checking that $h_{f}$ belongs to $R^{\oplus A_{1}}$)

$2$) $\Phi$ is an $R$-module homomorphism

$3$) $\Phi$ is bijective

which establishes the claim. This is a bit tedious, so you can see the advantage of the categorical point of view.


How about checking that $X:=(R^{ \oplus A_1 })^{ \oplus A_2 }$ satisfies the universal property of the free $R$-module on $A_1 \times A_2$? Define $i:A_1\times A_2 \to X$ by $i(a_1,a_2) = \alpha_{a_1,a_2}$, where the mapping $(\alpha_{a_1,a_2}(b_2))(b_1) := 1$ if $a_i = b_i$ and vanishes otherwise. Note that this lets us decompose any $\alpha \in X$ as a unique $R$-linear combination, $\alpha = \sum_{a_1 \in A_1, \, a_2 \in A_2} (\alpha(a_2))(a_1)\alpha_{a_1,a_2}$.

Now given an arbitrary map $f:A_1 \times A_2 \to M$ into a $R$-module $M$, its only possible extension $\overline{f}:X \to M$ (that is, $R$-module homomorphism such that $f = \overline{f}i$) must be given by $$\overline{f}(\alpha) = \overline{f}\left( \sum_{a_1 \in A_1, \, a_2 \in A_2} (\alpha(a_2))(a_1)\alpha_{a_1,a_2} \right) = \sum_{a_1 \in A_1, \, a_2 \in A_2} (\alpha(a_2))(a_1) f(a_1,a_2).$$

This proves the universal property. As any two objects satisfying the universal property must be isomorphic, we are done.

  • $\begingroup$ How do we know that defining $(\alpha_{a_1,a_2}(b_2))(b_1) := 1$ if $a_i=b_i$ lets us decompose $\alpha \in X$ as a $R$-linear combination $\alpha = \sum_{a_1 \in A_1, \, a_2 \in A_2} (\alpha(a_2))(a_1)\alpha_{a_1,a_2}$? $\endgroup$ – Al Jebr Jun 20 '19 at 21:34
  • $\begingroup$ @AlJebr In order for the maps on both sides to be equal, by definition they must yield the same value whenever you evaluate them at any $b_2 \in \oplus A_2$. This now yields maps $\oplus A_1 \to R$, which again to be equal must yield the same value when evaluated at any $b_1 \in \oplus A_1$. And this is the case, since both sides then yield $(\alpha(b_2))(b_1)$. $\endgroup$ – Alex Provost Jun 21 '19 at 2:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.