2
$\begingroup$

I have read about the forgetful functor $U$ from $\textsf{Mon}(\mathcal{C})$ to $\mathcal{C}$ (see e.g. this nLab page). I think I have read somewhere that this functor is monoidal, but I cannot find the source for it anymore and now I doubt that I remember this correctly.

I can see that the forgetful functor from $\textsf{Mon}(\mathcal{C})$ to $\mathcal{C}$ would be monoidal if $\textsf{Mon}(\mathcal{C})$ is monoidal. This condition does not seem to be that obvious to me though. I have two questions:

  • Is the category of monoids $\textsf{Mon}(\mathcal{C})$ in a monoidal category $\mathcal{C}$ itself monoidal?
  • Is the category of commutative monoids $\textsf{CMon}(\mathcal{C})$ in a symmetric monoidal category $\mathcal{C}$ itself symmetric monoidal?

Thank you!

$\endgroup$

1 Answer 1

3
$\begingroup$

If $(\mathcal{C},\otimes, I)$ is merely a monoidal category, then $\mathsf{Mon}(\mathcal{C})$ need not be a monoidal category, and if it is it is likely not in any natural way. For instance, taking $\mathcal{C}=\mathrm{Fun}(\mathcal{A},\mathcal{A})$ for a category $\mathcal{A}$, we can define a monoidal structure on $\mathcal{C}$ via composition of functors. If you want the forgetful functor $\mathsf{Mon}(\mathcal{C})\to\mathcal{C}$ to preserve a monoidal structure, you would be asking if, given two monads $T,U$ on $\mathcal{A}$, the composition $TU$ is again a monad. This question has been answered today here, and generally this is false.

If $(\mathcal{C},\otimes, I)$ is a symmetric monoidal category, then $\mathsf{Mon}(\mathcal{C})$ and $\mathsf{CMon}(\mathcal{C})$ are again symmetric monoidal categories, and the forgetful functor is monoidal. The tensor product of monoids is defined as follows: given two monoids $(M,m,e)$, $(N,n,f)$ in $\mathcal{C}$, then we can consider the object $M\otimes N\in\mathcal{C}$, with multiplication law $$ (M\otimes N)\otimes (M\otimes N)\cong(M\otimes M)\otimes(N\otimes N)\xrightarrow{m\otimes n}M\otimes N, $$ and unit $$I\cong I\otimes I\xrightarrow{e\otimes f}M\otimes N.$$ You can check that $M\otimes N$ will be a monoid again, which is commutative if both $M$ and $N$ are. The trivial monoid on $I$ serves as unit object.

If $f\colon (M,m,e)\to(M',m',e')$ is a monoid homomorphism, then $f\otimes N\colon M\otimes N\to M'\otimes N$ will be as well, which is again checking a bunch of diagrams commute. Likewise, the associativity, commutativity and unitality isomorphisms of $\otimes$ in the category $\mathcal{C}$ also lift to monoid isomorphisms if we take the tensor product of monoids. In particular, the coherence axioms just follow from the fact that they hold for $\otimes$ in $\mathcal{C}$.

$\endgroup$
1
  • $\begingroup$ Oh, I see. I thought about the forgetful functor $\textsf{Mon}(\mathcal{C})$ a bit too naïvely. My idea was that moving to a weaker structure surely wouldn't break things, because one just maps monoids $(M, m, e)$ to objects $M$ and homomorphisms $f$ to morphisms $f$, but I realise now that the forgetful functor does not have to be injective (after all you can have multiple monoid structures on $M$). Thanks you for your help and the detailed construction for the symmetric case. $\endgroup$ Commented Feb 29 at 21:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .