8
$\begingroup$

This is a follow up.

What are the commutative rings $R$, for which given $R$-modules $A$ and $B$, $A \otimes _{\Bbb Z} B = A \otimes _R B$ as abelian groups?


This is true when $R= \Bbb Q$, or $\Bbb Z_m$. We can give an $R$-module structure $A \otimes _{\Bbb Z} B$ satisfying $r (a \otimes b) = ra \otimes b$.

When $R= \Bbb Q$, we get the additional fact that $a \otimes rb=ra \otimes b$. To see this, note that when $r \in \Bbb N$, we have

$$ r a \otimes b = \sum a \otimes b = a \otimes rb $$ by bilinearity - which extends $r$ to $\Bbb Z$ too. When $r=1/m$, $m \in \Bbb Z$, $$ \frac{1}{m}a \otimes b = \frac{1}{m} (a \otimes b) = \frac{1}{m} ( \sum (a \otimes \frac{1}{m} b)) = \frac{1}{m} (ma \otimes \frac{1}{m} b ) = a \otimes \frac{1}{m} b $$ Thus, we have equality for all $r \in \Bbb Q$.


I think generalizing to $\Bbb Q$ is as far as we can get for this naive strategy. I wonder if there exists better method for the classification.

$\endgroup$
  • $\begingroup$ nice question. to make it self-contained, you might edit into it that $A$ and $B$ start life as $R$-modules (it's clear from reading your other post) $\endgroup$ – hunter Nov 28 '18 at 10:41
  • $\begingroup$ A guess: you might have this when $\Bbb Z\to R$ is an epimorphism (as it is in the cases $R = \Bbb Q$ and $R = \Bbb Z/n$). However, I do not have a proof of this (nor am I convinced that even if this is true, that these are all such rings). $\endgroup$ – Stahl Nov 29 '18 at 4:11
3
$\begingroup$

First, there is an ambiguity in your question. The usage of $=$ is ambiguous, but I'll interpret it as meaning that the natural map $\newcommand\tens\otimes\newcommand\ZZ{\mathbb{Z}}A\tens_\ZZ B\to A\tens_R B$ is an isomorphism.

Given this interpretation, we can give a simple criterion which rings with this property must satisfy. Namely the natural map $R\tens_\ZZ R\to R$ must be an isomorphism. This is also sufficient though, since if this is true, then $$A\tens_\ZZ B \simeq (A\tens_R R)\tens_\ZZ (R\tens_R B)\simeq A\tens_R (R\tens_\ZZ R)\tens_R B \simeq A\tens_R R \tens_R B \simeq A\tens_R B,$$ where I'm using $\simeq$ for natural isomorphism.

Thus the question is reduced to the question of for which rings is the natural map $R\tens_\ZZ R \to R$ an isomorphism. This is equivalent to asking for which rings is the diagram $$\newcommand\id{\operatorname{id}} \require{AMScd} \begin{CD} \ZZ @>\iota>> R \\ @V\iota VV @VV\id V \\ R @>\id >> R \end{CD} $$ a pushout diagram (where $\iota$ is the unique map $\ZZ\to R$.

Well, if it is, then for any pair of morphisms $f,g : R\to S$ with $f\circ \iota = g\circ \iota$, then there is a unique map $h : R\to S$ with $f=h\circ \id = g$. Thus $\iota$ is an epimorphism.

Conversely if $\iota$ is an epimorphism, then for any pair of morphisms $f,g : R\to S$ with $f\circ \iota = g\circ \iota$, then $f=g$, so the map $h=f=g : R\to S$ satisfies $f=h\circ\id$ and $g=h\circ\id$. Thus if $\iota$ is epic, this diagram is a pushout.

Hence a commutative ring $R$ has the property that $A\tens_\ZZ B\simeq A\tens_R B$ for all pairs $A$ and $B$ of $R$-modules if and only if the natural map $\ZZ\to R$ is an epimorphism.

Edit

As for what rings $R$ for which the natural map $\ZZ\to R$ is an epimorphism look like, I'm not sure. In general epis in the category of rings are complicated. That said, if I had to guess the answer in this case, my guess would be that these rings would be the subrings of $\Bbb{Q}$ and the rings $\ZZ/n\ZZ$, but that should probably be another question.

$\endgroup$
  • $\begingroup$ I thought this might be the case, but I didn't come up with an argument - very nice! There's a mathoverflow post I saw a bit ago which described or linked to a paper classifying rings such that the map from $\Bbb Z$ is an epi, and its very similar to what you say, but you also need to allow some products or sums of such rings as well. $\endgroup$ – Stahl Dec 15 '18 at 23:26
  • $\begingroup$ @Stahl Thanks, I'll have to go search for that, I'd be very interested in learning more about that. $\endgroup$ – jgon Dec 15 '18 at 23:29
  • 1
    $\begingroup$ Actually it appears to be one of the answers to the question I linked xP $\endgroup$ – jgon Dec 15 '18 at 23:31

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.