5
$\begingroup$

The following problem was subject of examination that was taken place in June. The document is here. Problem 1 states:

The tensor product $\mathbb{Q}\otimes_{\mathbb Z}\mathbb{Q}$ is a vector space over $\mathbb{Q}$ by multiplication in the left factor, i.e. $\lambda(x\otimes y)=(\lambda x)\otimes y$ for $\lambda, x, y\in\mathbb{Q}$. What is the dimension of $\mathbb{Q}\otimes_{\mathbb{Z}}\mathbb{Q}$ as a vector space over $\mathbb{Q}$?

I only know the definition of tensor product for modules (via universal property). How does one go about calculating dimension of such a vector space?

Thanks!

$\endgroup$
  • $\begingroup$ $\mathbb Z$ isn't a field so it doesn't make sense to ask about vector spaces over the integers. Perhaps what was meant was module? $\endgroup$ – kahen Aug 2 '13 at 20:21
  • $\begingroup$ @kahen: sorry it was a typo $\endgroup$ – Prism Aug 2 '13 at 20:21
  • $\begingroup$ As a vector space over $\mathbb{Z}$? Don't you mean as a free module? $\endgroup$ – Dedalus Aug 2 '13 at 20:22
  • $\begingroup$ @Dedalus: I meant $\mathbb{Q}$. Fixed. $\endgroup$ – Prism Aug 2 '13 at 20:22
  • 1
    $\begingroup$ I advocate my answer here as a good discussion of three important properties of tensor products, in particular #3 is relevant to this question. (BenjaLim generalizes it a bit.) $\endgroup$ – anon Aug 3 '13 at 0:17
17
$\begingroup$

Begin by checking

\begin{align}\frac{a}{b}\otimes \frac{c}{d}&=\frac{ad}{bd}\otimes\frac{c}{d}\\ &=\frac{a}{bd}\otimes\frac{cd}{d}\\ &=\frac{a}{bd}\otimes c\\ &=\frac{ac}{bd}\otimes 1 \end{align}

So, $\mathbb{Q}\otimes_{\mathbb{Z}}\mathbb{Q}=\mathbb{Q}$.

Now you can end it.

$\endgroup$
  • 2
    $\begingroup$ Obviously, the more general things going on here is that $\Frac(R)\otimes_R\Frac(R)$ is isomorphic to $\Frac(R)$ as a $\Frac(R)$-module. $\endgroup$ – Alex Youcis Aug 2 '13 at 21:26
  • $\begingroup$ There is a subtle point in this year which perhaps should be pointed out to the OP. The last step is not in general legit for general modules (which don't have $1$ in them, e.g. $\Bbb{2}\Bbb{Z}$ as a $\Bbb{Z} - module). $\endgroup$ – user38268 Aug 3 '13 at 0:18
  • 2
    $\begingroup$ @AlexYoucis: Ah indeed! A very nice presentation is given in Keith Conrad's expository note on tensor products. (Theorem 4.20, and Example 4.21) $\endgroup$ – Prism Aug 3 '13 at 1:58
8
$\begingroup$

I just want to make remark that RGB's answer more generally shows the following. Suppose $M,N$ are $S^{-1}A$ - modules. Then we can form both $M \otimes_{S^{-1}A} N$ and $M \otimes N$. The former is an $S^{-1}A$ - module and the latter an $A$ - module by restriction of scalars. The ultimate point now is this: $M \otimes_A N$ already has the structure of an $S^{-1}A$ - module built into it! Thus $M \otimes_{S^{-1}A} N$ and $M \otimes_A N$ are canonically isomorphic as $S^{-1}A$ modules; this is basically the content of RGB's answer.

If you want to show $\Bbb{Q} \otimes_{\Bbb{Z}} \Bbb{Q} \cong \Bbb{Q}$ using universal properties here is what we do: Let $f : \Bbb{Q} \times \Bbb{Q} \to M$ be a $\Bbb{Z}$ - bilinear map. Consider $\pi : \Bbb{Q} \times \Bbb{Q} \to \Bbb{Q}$ that sends $(a,b)$ to $ab$. Now we have linearity in the first variable because $$\begin{eqnarray*} \pi(na_1 + ma_2,b) &=& (na_1+ma_2)b \\ &=& n(a_1b) + m(a_2b) \\ &=& n\pi(a_1,b) + m\pi(a_2,b)\end{eqnarray*}$$ for any $a_1,a_2,b \in \Bbb{Q}$ and $n,m\in \Bbb{Z}$. By symmetry linearity in the second variable follows and so $\pi$ is bilinear. Let us now define a map $g : \Bbb{Q} \to M$ by $g(a) = f(a,1)$ for any $a \in \Bbb{Q}$. This map $g$ is linear because $f$ is linear in the first variable. Also $g$ is well - defined and furthermore is unique: Any linear map out of $\Bbb{Q}$ to $M$ $\tilde{g}$ such that $\tilde{g} \circ \pi = f$ must necessarily satisfy $$ \tilde{g}(a) = \tilde{g}(\pi(a,1)) = f(a,1).$$

Thus we have shown that $\Bbb{Q}$ satisfies the universal property of $\Bbb{Q} \otimes_{\Bbb{Z}} \Bbb{Q}$ and so the answer to your question comes easily: $\dim_\Bbb{Q} \Bbb{Q} \otimes_{\Bbb{Z}} \Bbb{Q} = 1$.

$\endgroup$
  • 2
    $\begingroup$ Thank you very much for taking your time to compose this detailed answer. I am drawing diagrams to convince myself :) $\endgroup$ – Prism Aug 3 '13 at 1:59
  • $\begingroup$ +1. why are they canonically isomorphic? is there a "functorial" proof? I know that $-\otimes_{A} S^{-1}A\cong S^{-1}$ (functorial isom.) $\endgroup$ – user111072 Jan 26 '14 at 15:39

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.