Uniqueness of Isomorphism $M \otimes_A N \cong N \otimes_A M$ (Proposition 2.14 in Atiyah-Macdonald).

commutative-algebratensor-products

Proposition $2.14$ in Atiyah-Macdonald states that there is a unique isomorphism between $M \otimes_A N$ and $N \otimes_A M$ for any two $A$-modules $M$ and $N$. Here is a picture of the relevant passage.

Atiyah-Macdonald 2.14
I'm struggling to figure out why the following is not a counterexample to the claim of uniqueness. Take $A=M=N=\mathbb{Z}$ and consider the two maps
\begin{align*}
\phi_1: \mathbb{Z} \otimes_{\mathbb{Z}} \mathbb{Z} & \to \mathbb{Z} \otimes_{\mathbb{Z}} \mathbb{Z} \\
x \otimes y & \mapsto x \otimes y
\end{align*}

and
\begin{align*}
\phi_2: \mathbb{Z} \otimes_{\mathbb{Z}} \mathbb{Z} & \to \mathbb{Z} \otimes_{\mathbb{Z}} \mathbb{Z} \\
x \otimes y & \mapsto (-x) \otimes y
\end{align*}

As near as I can tell, these constitute a notrivial subgroup of $\mathrm{Aut}_{\mathbb{Z}}(\mathbb{Z} \otimes_{\mathbb{Z}} \mathbb{Z})$ isomorphic to $\mathbb{Z}/2$. Where am I going wrong with my reasoning? Any and all insights are welcome!

Best Answer

This answer is essentially Darij's comment; it exists to remove this question from the unanswered queue.

The isomorphisms i, ii, iii, iv given in this proposition are unique up to the properties a, b, c, d (respectively). As your counterexample shows, there generally are many isomorphisms between $M\otimes N$ and $N\otimes M$.

Related Question