1
$\begingroup$

let $\psi: A \to B$ be an injective $R$ module homomorphism, and it is given that any $f: A \to M$ $\mathbb{Z}$ module homomorphism can be lifted to a $\mathbb{Z}$ module homomorphism $F: B \to M$ s.t $f=F \circ \psi$ $\DeclareMathOperator{\Hom}{Hom}$

To prove that any $f': A \to \Hom_{\mathbb{Z}}(R,M)$ arbitrary $R$ module homomorphism can be lifted to $F': B \to \Hom_{\mathbb{Z}}(R,M)$ $R$ module homomorphism with $f'=F' \circ \psi$

It is an exercise in Dummit Foote now I was trying to solve this(following hint in the book):

Consider $f(a)=f'(a)(1_R)\in M$ then as $f'(a) \in\Hom_{\mathbb{Z}}(R,M)$ we can see that $f: A \to M$ is a $\mathbb{Z}$ module homomorphism then it can be lifted to a $\mathbb{Z}$ module homomorphism $F: B \to M$ s.t $f=F \circ \psi$[ according to the hyp] this means

  1. $F(\psi(a))=f(a)=f'(a)(1_R)$

Now construct $F': B \to \Hom_{\mathbb{Z}}(R,M)$ s.t $F'(b)(r):=F(rb)$.

Claim that $F'$ is the lift of $f'$

Now there are two things to check

  • $F'$ is an $R$ module homomorphism i.e $F'(rb)=rF'(b)$

    Now $rF'(b)(s)=F'(b)(sr)$ for any $s \in R$[By the action of $R$ module on $Hom_{\mathbb{Z}}(R,M)$]

$\Rightarrow rF'(b)(s)=F(srb)=F'(rb)(s)\Rightarrow rF'(b)=F'(rb)$.

So $F'$ is an $R$ module homomorphism.

  • $f'=F' \circ \psi$.

Now here I am having problem. This is clear to me that $(F' \circ \psi(a))(1_R)=F'( \psi(a))(1_R)=F(1_R\psi(a))=F(\psi(a))=f'(a)(1_R)$[from eqn 1]

but I can't understand why $(F' \circ \psi(a))(r)=f'(a)(r)$ I was trying to think in this way that $f'(a)(r)=f'(a)(r1_R)$ but $f'(a)$ is a $\Bbb Z$ module homomorphism so I can't do anything from here. From L.H.S $(F' \circ \psi(a))(r)=F(r\psi(a))$. Now what?

I think I am missing some silly observation but I have to admit that I am lost.

Now one more non trivial question from my point of view that will there be any harm if we consider $f(a):=f'(a)(r_1)$ for some fixed $r_1 \in R$ instead of choosing $f(a)=f'(a)(1_R)$ and follow the same process from there? Please give some detailed explanation or a hint from where I can conclude.

$\endgroup$
  • 1
    $\begingroup$ $f'(ar)(1_R) = (f'(a) \cdot r)(1_R)= f'(a)(r)$. I explain this below. $\endgroup$ – Dean Young Apr 26 at 21:31
1
$\begingroup$

Here I answer why $(F' \circ \psi(a))(r) = f'(a)(r)$. It is (1) and (2) in the centerlined equality below. But before I begin, may I recommend that we start with the following setup instead:

Theorem 1: Let $R$ be a ring, and let $A$ and $B$ be right $R$-modules, and let $\psi : A \rightarrow B$ be an injective $R$-module morphism. Let $M$ be an abelian group, and regard $\text{Hom}_{\mathbb{Z}}( R, M)$ as being a right $R$-module where $(\phi \cdot r)(s) = \phi(rs)$. Suppose that, for every map of abelian groups $f : A \rightarrow M$, there is a map $g : B \rightarrow M$ of abelian groups such that $g \circ \psi = f$. Then, for each map of $R$-modules $f' : A \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$, there is a map $g' : B \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$ of $R$-modules such that $g' \circ \psi = f'$.

The difference is that $A$ and $B$ are right $R$-modules. This modification has the advantage of being consistent with the following heuristic:

Heuristic: In deciding whether a module should be a left, or right module, we have made the simplest and easiest choice when, for each equality involved, the variables occur in the same order in each term term. For instance, in choosing $\text{Hom}_{\mathbb{Z}}(R, M)$ to be a right $R$-module, $r$ always occurs before $s$ in $(\phi \cdot r)(s) = \phi(rs)$.

Proof of Theorem 1: take $f' : A \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$. $f'$ induces a $\mathbb{Z}$-linear $f : A \rightarrow M$ sending $a$ to $f'(a)(1_R)$ i.e $f(a)=f'(a)(1_R)$. $f$ induces a lift $F : B \rightarrow M$ by assumption. $F$ induces an $R$-linear map $F'$ sending $n$ to the map $R \rightarrow M$ sending $r$ to $F(nr)$. That is, $F'(n)(r) = F(nr)$. To see that $F'$ is a lift of $f'$, take $a \in A$ and $r \in R$. Then, $$(F' ( \psi(a)))(r) = F(\psi(a) r) = F(\psi(ar)) = f(ar) = f'(ar)(1_R) \stackrel{(1)}{=} (f'(a) \cdot r)(1_R) \stackrel{(2)}{=} f'(a)(r)$$ To see (1), note that $f' : A \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$ is an $R$-linear map of right $R$-modules. To see (2), note that the right $R$-module structure on $\text{Hom}_{\mathbb{Z}}(R, M)$ is given by $(\phi \cdot r)(s) = \phi(rs)$.


In this extra section, I explain how to go back and forth between right $R$-module maps $$ N \rightarrow \text{Hom}_{\mathbb{Z}} (R, M) $$ and $\mathbb{Z}$-module maps $$ N \rightarrow M$$

Take a ring $R$ Let $N$ be a right $R$-module. Let $M$ be a left $R$-module. $\text{Hom}_{\mathbb{Z}} (R, M)$ is a right $R$-module. We set $ \phi \cdot r : R \rightarrow M$ to be the map sending $s$ to $\phi(rs)$. Then $$(\phi \cdot (rs))(t) = \phi(rst) = (\phi \cdot r)(st) = ((\phi \cdot r )\cdot s) (t)$$

1) For the first direction, given a map $f : N \rightarrow \text{Hom}_{\mathbb{Z}} (R, M)$, we seek to define a map $g : N \rightarrow M$ of abelian groups. Set $g(n) = f(n)(1_R)$. Then $$g(n +m)= f(n+m)(1_R) = (f(n) + f(m))(1_R) = f(n)(1_R) + f(m)(1_R) = g(n) + g(m)$$

2) For the other direction, given a map $g : N \rightarrow M$ of $\mathbb{Z}$-modules, we seek to define a map $f : N \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$ of right $R$-modules. Set $f(n)(r) = g(nr)$. We check that $f : N \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$ is a map of right $R$-modules. Take $r \in R$ and $n \in N$. Take $s \in R$. Then $$f(nr)(s) = g(nrs) = f(n)(rs) = (f(n) \cdot r) (s)$$ So $f(nr) = f(n) \cdot r$. Note $f(n) \cdot r$ was defined above- we had to make the right choice as to whether $\text{Hom}_{\mathbb{Z}}(R, M)$ was a left or a right $R$-module.

Now to check that these operations are inverse. Take $f : N \rightarrow \text{Hom}_{\mathbb{Z}}(R, M)$ a map of right $R$-modules, and put $g : N \rightarrow M$ the map sending $n$ to $f(n)(1_R)$. Put $f_2$ the map sending $n$ to the map sending $r$ to $g(nr)$. Then $f_2 (n) = f(n)$ for each $n$. Indeed, for each $n \in N$ and each $r \in R$, $$ f_2(n)(r) = g(nr) = f(nr)(1_R) = (f(n) \cdot r) (1_R) = f(n)(r 1_R) = f(n)(r)$$ notice how in all these terms, $n$ always occurs before $r$. That is the heuristic that we have followed to make sure all the parities match up. So $f = f_2$.

Next, take a map of abelian groups $g : N \rightarrow M$. Set $f$ to be the map sending $n$ to the map $f(n) : R \rightarrow M$ of abelian groups sending $r$ to $g(nr)$. Then set $g_2 (n) : N \rightarrow M$ to be the map of abelian groups sending $n$ to $f(n)(1_R)$. Then $$g_2 (n) = f(n)(1_R) = g(n1_R) = g(n)$$

This establishes the desired correspondence.

$\endgroup$
  • $\begingroup$ I think it would $g:B \to M$ in the highlighted portion. $\endgroup$ – Shadow Apr 26 at 23:22
  • $\begingroup$ There were some typos e.g $($ was missing and $A$ was written instead of $B$ in the first highlighted section and also I suggested a different notation $f'$ and $g'$ in that part. It is a detailed nice answer. Thanks a lot. $\endgroup$ – Shadow Apr 26 at 23:37
  • $\begingroup$ You're welcome, Shadow. Did the errors get fixed? $\endgroup$ – Dean Young Apr 27 at 3:01
  • 1
    $\begingroup$ Yes, thanks a lot. :) $\endgroup$ – Shadow Apr 27 at 23:00
0
$\begingroup$

This is really just the tensor-hom adjunction. We have $$ \mathrm{Hom}_R(A,\mathrm{Hom}_{\mathbb Z}(R,M)) \cong \mathrm{Hom}_{\mathbb Z}(R\otimes_RA,M) \cong \mathrm{Hom}_{\mathbb Z}(A,M). $$ Explicitly, we can identify $R$-linear maps $A\to\mathrm{Hom}_{\mathbb Z}(R,M)$ with $\mathbb Z$-linear maps $A\to M$, by sending $f'$ to $f$ such that $f(a)=f'(a)(1_R)$, and sending $f$ to $f'$ such that $f'(a)(r)=f(ra)$.

Note that the $R$-action on $\mathrm{Hom}_{\mathbb Z}(R,M)$ is given by $r\theta\colon s\mapsto \theta(sr)$, so given $f$, the map $f'$ such that $f'(a)(r)=f(ra)$ really is $R$-linear: $$ f'(ra)(s) = f(sra) = f'(a)(sr) = (rf'(a))(s). $$

Following this through, we start with an $R$-linear map $f'\colon A\to\mathrm{Hom}_{\mathbb Z}(R,M)$, obtain a $\mathbb Z$-linear map $f\colon A\to M$, lift to a $\mathbb Z$-linear map $F\colon B\to M$ such that $F\psi=f$, and obtain the corresponding $R$-linear map $F'\colon B\to\mathrm{Hom}_{\mathbb Z}(R,M)$.

We just need to check that $F'\psi=f'$. By construction we have $$ F'(\psi(a))(r) = F(r\psi(a)) = F(\psi(ra)) = f(ra) = f'(a)(r). $$ This holds for all $r\in R$, so $F'\psi(a)=f'(a)$. This holds for all $a\in A$, so $F'\psi=f'$.

$\endgroup$
  • $\begingroup$ From $f(a)=f'(a)(1_R)$ why is $f(ra)=f'(a)(r)$? $\endgroup$ – Shadow Apr 26 at 18:37
  • $\begingroup$ In your notation, $A$ and $B$ are left $R$-modules. Therefore $\text{Hom}_{\mathbb{Z}}(R, M)$ must be a left $R$-module for this to make sense. Therefore $R$ in $\text{Hom}_{\mathbb{Z}}(R, M)$ must be a right $R$-module for this to make sense. Therefore the correct multiplication of $R$ on $\text{Hom}_{\mathbb{Z}}(R, M)$ sends $(r, \phi )$ to the map $R \rightarrow M$ sending $s$ to $\phi(sr)$, not $\phi(rs)$. Of course, it's not a problem if we take $R$ to be commutative here :) $\endgroup$ – Dean Young Apr 26 at 20:36
  • $\begingroup$ In the original post we are considering left $R$-modules, so I continued with this. I then wrote explicitly that the left $R$-action on $\mathrm{Hom}_{\mathbb Z}(R,M)$ is given by $r\theta\colon s\mapsto \theta(sr)$. $\endgroup$ – Andrew Hubery Apr 26 at 23:18
  • $\begingroup$ My bad- this should work. So this suggests that there are two forms of the hom-tensor adjunction- one where $\text{Hom}_R (S, -) : R \text{-mod} \rightarrow S \text{-mod}$ is hom as right $R$-modules, and another where it is hom as left $R$-modules. $\endgroup$ – Dean Young Apr 27 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.