6
$\begingroup$

In this question, all rings and algebras are commutative with identity.

Let $R$ be a ring, and let $A$ be an $R$-algebra with an $R$-subalgebra $B$. Suppose that we have an $R$-algebra homomorphism $\phi: B\to R$; then we can form the tensor product $A\otimes_B R$. My question is:

If the structure maps $R\to B$ and $R\to A$ are injective, is the map $R\to A\otimes_B R$ injective as well?

My intuition says Yes: the tensor product $A\otimes_B R$ is a quotient $A / (b-\phi(b): b\in B)$, and since $\phi: B\to R$ is a ring homomorphism preserving elements of $R$, it's hard to see how this ideal could ever contain an element of $R$. But of course that's not enough to go on.

I'm particularly interested in the case where $A = R[x_1,\ldots,x_n]$ and $B = R[x_1,\ldots,x_n]^G$ for some subgroup $G\subseteq S_n$, so if it would help to use the fact that $A$ is a polynomial ring, then by all means please do.

$\endgroup$
  • $\begingroup$ If $R \to A$ is injective then is it not automatic that $R \to B$ is also injective? $\endgroup$ – user38268 Nov 16 '12 at 15:54
  • $\begingroup$ Yes, and of course vice versa. I just didn't want to prejudice anyone by mentioning one and not the other. $\endgroup$ – Owen Biesel Nov 16 '12 at 16:11
3
$\begingroup$

This is false. The ideal $(b - \phi(b))$ in $A$ can contain $1 \in R$ as soon as some $b - \phi(b)$ is a unit in $A$. Here is a concrete counterexample.

Let $R$ be a field and let $B$ be the power series ring $k[[x]]$. Let $A$ be the field of fractions of $B$, namely $k((x))$, and let $\phi : B \to R$ be the $k$-algebra homomorphism which sends $x$ to zero.

Then $A \otimes_R B = A \otimes_R B / xB = A / xA = 0$ since $x$ is invertible in $A$. Hence the map $R \to A \otimes_R B$ is not injective in this case.

In the positive direction, it is for example enough to know that $B$ has a $B$-module complement $C$ inside $A$, since then

$A = B \oplus C \Rightarrow R \hookrightarrow R \oplus (C \otimes_B R) = (B \otimes_B R) \oplus (C \otimes_B R) = A \otimes_B R$.

$\endgroup$
  • $\begingroup$ I had also found the similar example $B=R[x]$, $A = B[x^{-1}]$, and $B\to R$ sending $x\mapsto 0$. Thanks for tying off this loose question! $\endgroup$ – Owen Biesel Jan 13 '14 at 8:42

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.