8
$\begingroup$

I assume all rings are commutative with identity. Denote $A':=A/ \sqrt{0}$ for convenience.

Question is simple:

For a given ring $R$ and $R$-algebras $A,B$, does this isomorphism $(A\otimes_R B)'\cong (A'\otimes_R B')'$ hold?

With specific (easy) examples, it seems to be true. How can I show that in general setting?

  • This question arises from the equality $(X_{\text{red}}\times_S Y_{\text{red}})_{\text{red}}\cong (X\times_S Y)_{\text{red}}$ where $X,Y$ are $S-$schemes. In the proof, the author says that equality comes from the induced map $X_{\text{red}}\times_S Y_{\text{red}}\to X\times_S Y$ by the canonical maps $X_{\text{red}}\to X$ and $Y_{\text{red}}\to Y$. But I cannot understand how can we get the isomorphism just from the induced map.
$\endgroup$
  • $\begingroup$ The tensor product is over some 'coefficients', some other ring that makes this two algebras over it, isn't it? In this case, what is that ring? Well, I guess you may want the answer for any coefficients ... Well, you said already below. $\endgroup$ – OR. Jul 30 '13 at 5:51
  • $\begingroup$ @RGB we may assume $A,B$ are $R$-algebra for some ring $R$. $\endgroup$ – User0829 Jul 30 '13 at 5:52
  • 1
    $\begingroup$ Hint: If $I$ is a nil ideal of a ring $C$, then $C' = \left(C/I\right)'$. Try to find a nil ideal $I$ of $A \otimes_R B$ such that $\left( A\otimes_R B \right) / I \cong A' \otimes_R B'$. $\endgroup$ – darij grinberg Jul 30 '13 at 8:14
  • $\begingroup$ @darij I think your assertion is wrong. Take $A=\mathbb{Z}$ and $I=2\mathbb{Z}$. $\endgroup$ – User0829 Jul 30 '13 at 8:43
  • $\begingroup$ @User0829: But $I$ is not a nil ideal! $\endgroup$ – darij grinberg Jul 30 '13 at 8:50
3
$\begingroup$

This can be proved using the following universal property of $A'$: if $f \colon A \to S$ is any $R$-algebra homomorphism where $S$ is reduced, then $f$ factors uniquely as $f = g \circ p$ where $p \colon A \to A/\sqrt{0} = A'$ is the canonical homomorphism. Or more briefly: any $R$-algebra homomorphism from $A$ to a reduced $R$-algebra factors uniquely through $A'$. (I will let you work this out, but feel free to ask in the comments if you need some hints.)

The characterization of $A'$ above makes it unique up to unique isomorphism: If $T$ is any other algebra with a homomorphism $q \colon A \to T$ satisfying the same universal property (with $p$ replaced by $q$), then $T \cong A'$ (via a unique isomorphism that respects $p$ and $q$).

Thus, to prove that $(A \otimes_R B)' \cong (A' \otimes_R B')'$, it is enough to show that $(A' \otimes_R B')'$ equipped with the canonical homomorphism $q \colon A \otimes_R B \to A' \otimes_R B' \to (A' \otimes_R B')'$ satisfies the same universal property as $(A \otimes_R B)'$.

So let $S$ be a reduced ring with an algebra homomorphism $f \colon A \otimes_R B \to S$. The canonical homomorphisms $A,B \to A \otimes_R B$ induce $A,B \to A \otimes_R B \to S$. Since $S$ is reduced, these must factor uniquely through $A'$ and $B'$. This induces $A' \otimes_R B' \to S$. But again, as the latter algebra is reduced (and the former may not be), this factors uniquely through $g \colon (A' \otimes_R B')' \to S$. It's now a small chore to check that this provides a unique factorization $f = gq$.

$\endgroup$
2
$\begingroup$

Maybe @darijgrinberg's suggestions deserve to be an answer as long as they provide a very short one.

One can write $A/I\otimes_RB/J$ as $(A\otimes_RB)/K$, where $K$ is the ideal of $A\otimes_RB$ generated by $a\otimes b$ with $a\in I$ or $b\in J$. In the particular case when $I=N(A)$ and $J=N(B)$ we have $A'\otimes_RB'=(A\otimes_RB)/K$ with $K$ generated by $a\otimes b$ with $a\in A$ or $b\in B$ nilpotent. Then we have $$(A'\otimes_RB')'=\left((A\otimes_RB)/K\right)'=(A\otimes_RB)',$$ where the last equality holds since $K$ is a nil ideal, that is, an ideal with $\sqrt K=\sqrt{(0)}$.

$\endgroup$

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.