MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to compare some homomorphism groups over different base rings, so given a commutative local ring $(A,\mathfrak{m})$ and a finite dimensional Azumaya algebra $R$ over $A$.

If $M$ and $N$ are two left $R$-modules which are finitely generated and torsion free over $A$, is there an $A$-isomorphism $R \otimes_A Hom_R(M,N) \rightarrow Hom_A(M,N)$?

So for example if $M=N=A^n$ and $R=M_n(A)$, then we have $Hom_R(M,N)\cong A$ and $Hom_A(M,N)\cong M_n(A)$ so we have an isomorphism in this case.

What about "non trivial" examples? Or do we need to have put stronger conditions on $M$ and $N$?

$\textbf{New idea}$: I think the Hom-Tensor adjunction won't help here. So next try:

Put $M=R$, this gives $Hom_R(M,N)=Hom_R(R,N)\cong N$ as $A$-modules. So we have $R\otimes_A Hom_R(M,N) \cong R\otimes_A N$. Now because $R$ is Azumaya we have $R\cong Hom_A(R,A)$. So we really have $R\otimes_A Hom_R(M,N) \cong Hom_A(M,N)$.

So we have an isomorphism for $M=R^k$. Now what about if $M$ is a projective $R$-module? Then there is some $R$-module $P$ such that $M\oplus P\cong R^k$. Can we somehow conclude that we have an isomorphism for M in this case?

share|cite|improve this question
up vote 3 down vote accepted

I think this is true. There is a homomorphism $R \otimes_A Hom_R(M,N) \rightarrow Hom_A(M,N)$, given from the obvious $R$-module structure of $Hom_A(M,N)$. To check that this is an isomorphism we can make a faithfully flat extension and split $R$; so we may assume that $R$ is a matrix algebra $M_n(A)$.

Now we use Morita equivalence: tensoring with the $(R-A)$-bimodule $A^n$ gives an equivalence of categories between $A$-modules and $R$-modules. If $M = A^n \otimes_A M'$ and $N = A^n \otimes_A N'$, we have $$ Hom_A(M,N) = Hom_A(A^n \otimes_A M', A^n \otimes_A N') = M_n(A) \otimes_A Hom_A(M',N') = R \otimes_A Hom_R(M,N) $$ and it should be easy to see that this is the desired isomorphism.

Does this work?

share|cite|improve this answer
This looks good. But then it seems to me, that this is in fact true for all $R$-modules $M$ and $N$. Or do we need the finiteness assumptions? Because i think, if $A\rightarrow B$ is the splitting extension, then $Hom_R(M,N)\otimes_A B \cong Hom_{R\otimes_A B}(M\otimes_A B,N\otimes_A B)$ isn't true for all $R$-modules. – TonyS Oct 24 '10 at 17:18
I am pretty sure that being torsion-free is not required; but you are right, finiteness probably is needed in the argument. – Angelo Oct 24 '10 at 17:56
Okay. Thanks a lot. – TonyS Oct 24 '10 at 18:20

May be I am missing something, but the adjoint isomorphism gives:

$$ \text{Hom}_R(M, \text{Hom}_A(R,N)) \cong \text{Hom}_A(R\otimes_R M, N) = \text{Hom}_A( M, N)$$

Since $R$ is Azumaya, $R$ is a projective, hence free $A$-module of finite rank $r>0$. So the left hand side is isomorphic to $ \text{Hom}_R(M,N)^{\oplus r}$, which is also isomorphic (as $A$-modules) to $R\otimes_A \text{Hom}_R(M,N)$.

share|cite|improve this answer
Hold on, perhaps there is a problem: $Hom_A(R,N)$ as an $R$-module, might not be $N^r$?. I have to run now, so I will come back to this... – Hailong Dao Oct 22 '10 at 19:57
Yes, i tried the Hom-Tensor adjunction. This should also give $Hom_R(R\otimes_A M,N)\cong Hom_A(M,N)$.But then i couldn't see why $R\otimes_A M$ is isomorphic to $M^r$ as an $R$-module. The same problem you encountered. – TonyS Oct 23 '10 at 10:06

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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