16

Let L/K be a (separable?) field extension, let A be a finite dimensional algebra over K, and let M and N be two A-modules. Let $A' = L \otimes_K A$ be the algebra given by extension of scalars, and let $M' = L \otimes_K M$ and $N' = L \otimes_K N$ be the A'-modules given by extension of scalars.

Does $M' \cong N'$ (as A'-modules) imply that $M \cong N$ (as A-modules)?

(This question is obviously related. Note that just as for that question it is easy to see that base extension reflects isomorphisms in the sense that if a map $f: M \rightarrow N$ has the property that $f' : M' \rightarrow N'$ is an isomorphism then f is an isomorphism. This is asking about the more subtle question of whether it reflects the property of being isomorphic.)

I apologize if this is standard (I have a sinking suspicion that I've seen a theorem along these lines before), but I haven't been able to find it. There's a straightforward proof in the semisimple setting, but I have made no progress in the non-semisimple setting.

Community
  • 1
Noah Snyder
  • 27,820

2 Answers2

10

I hope I'm not misunderstanding the question. Here goes:

We'll show that if $M,N$ are finite-dimensional over $K$, then they are isomorphic over $K$.

Think of the linear space $X=\mathrm{Hom}_{A}(M,N)$ as a variety over $K$. Inside $X$ look at the $K$-subvariety $X'$ of maps that are not isomorphisms $M \rightarrow N$. Now $X' \neq X$, because there is an $L$-point of $X$ not in $X'$. Therefore, over an infinite field $K$, there will certainly exist a $K$-point of $X$ that doesnt lie in the proper subvariety $X'$.

If $K$ is finite: $M,N$ are both $K$-forms of the same module $M'$ over $L$. The $L$-automorphisms of $M'$ are a connected group, because they amount to the complement of the hypersurface $X'$ inside the linear space $X$. So its Galois cohomology vanishes, thus the same conclusion.

Edgardo
  • 276
  • I don't think it uses commutativity anywhere (?) We are just identifying $\mathrm{Hom}_A(M,N)$ with the linear subspace of $\mathrm{Hom}_K(M,N)$ which commutes with $A$. – Edgardo Dec 16 '13 at 23:54
  • Sorry, I was confused. – Noah Snyder Dec 17 '13 at 00:13
  • In the infinite field case, you're using the separable hypothesis, right? – Ben Wieland Dec 17 '13 at 00:42
  • I don't think so. It comes down to this: Take a polynomial $f \in K[x_1, \dots, x_n]$. If there exists $(a_1, \dots, a_n) \in L^n$ such that $f(a_1, \dots, a_n) \neq 0$, then also there exists $(b_1, \dots, b_n) \in K^n$ with $f(b_1, \dots, b_n) \neq 0$. The existence of $a_i$ means that $f$ is not identically vanishing. – Edgardo Dec 17 '13 at 00:44
  • I follow the infinite case, very nice! Any chance you could flesh out the last paragraph a little more? Which Galois cohomology group? (I think $H^1$?) Why does vanishing of that Galois cohomology group yield the desired conclusion? Sorry for being slow. – Noah Snyder Dec 17 '13 at 02:51
  • 2
    Sorry, here it is: Modules that become isomorphic to $M$ over the algebraic closure of our finite field $K$ are classified by $H^1(G, \mathrm{Aut}(M'))$, where $M'$ is $M$ base-changed to the algebraic closure, and $G$ is the absolute Galois group of $K$. Now $\mathrm{Aut}(M')$ is the set of $\bar{K}$-points of a connected algebraic $K$-group, namely, the automorphism group of $M$ (considered as a $K$-variety). There is a theorem of Lang and Steinberg that says $H^1$ always vanishes in this setting. – Edgardo Dec 17 '13 at 02:57
  • Would this also work for $A$ finitely generated (as an algebra)? – Dag Oskar Madsen Dec 17 '13 at 12:29
  • @DagOskarMadsen, at least the way I am thinking about it, it is important that everything is finite-dimensional over $K$. I don't know what happens in the general case. Also, it is a bit annoying that one has to treat the cases of K finite and infinite separately. Surely there is a uniform way of proceeding... – Edgardo Dec 17 '13 at 14:48
  • @DagOskarMadsen: Doesn't Julian's answer give a counterexample in the f.g. case? – Noah Snyder Dec 17 '13 at 19:17
  • @NoahSnyder I was thinking $A$ finitely generated, $M$ and $N$ finite dimensional, and wondered if the same proof would work. – Dag Oskar Madsen Dec 17 '13 at 19:21
  • if the finite dimensional representation is initially defined over K, then its image is always a finite dimensional K-algebra. – mhahthhh Nov 20 '22 at 17:49
7

Here's a counterexample to the same statement for infinite dimensional algebras:

Take $K=\mathbb{R}$, $L=\mathbb{C}$, $A=\mathbb{R}[x,y]/(x^2+y^2-1)$. Then $A$ is a Dedekind domain with class group cyclic of order 2, and $A'=A\otimes\mathbb{C}$ is a PID. We can take $M$ and $N$ to be non-isomorphic projective rank 1 modules over $A$, which both necessarily become free after tensoring with $\mathbb{C}$.

Explicitly, we can take $M=A$, $N=(x,y-1)\subset A$.

Noah Snyder
  • 27,820
Julian Rosen
  • 8,941