7
$\begingroup$

Let $\mathcal{T}$ be a commutative algebraic theory (for example sets, abelian groups, commutative monoids, but not groups etc.). References include the nlab and Borceux' Handbook of Categorical Algebra 2, section 3.10. Then $\mathsf{Mod}(\mathcal{T})$ is a monoidal category with internal homs.

Question 1. (Answered: Yes) Can we find a property of concrete categories which holds for $\mathsf{Mod}(\mathcal{T})$ if and only if $\mathcal{T}$ is commutative? In other words, does commutativity of an algebraic category not depend on the presentation?

Question 2. (Answered: No) Let $\mathcal{T}$ be a commutative algebraic theory and $C=\mathsf{Mod}(\mathcal{T})$. Assume that $X \in C$ is a Co-$C$-algebra, i.e. we have a factorization of $\hom(X,-) : C \to \mathsf{Set}$ over $C$. Does this have to coincide with the usual factorization? This is well-known to be true in the examples I have mentioned above, for example for every abelian group $A$ there is only one natural abelian group structure on the hom-sets $\mathrm{hom}(A,B)$. This should be all well-known, but I don't know a reference.

$\endgroup$
9
  • 1
    $\begingroup$ To me a "presentation" of an algebraic theory is a choice of operations and axioms, while Lawvere theories and monads embody the extensional essence of an algebraic theory. In any case, I think the answer to question 1 is negative: being a commutative algebraic theory is not stable under Morita equivalence, since e.g. the theory of $k$-modules for a field $k$ is commutative, but the theory of $\textrm{Mat}_n(k)$-modules is not for $n \ge 2$. $\endgroup$
    – Zhen Lin
    Nov 20, 2012 at 8:22
  • $\begingroup$ Yes but these categories are not equivalent as concrete categories. $\endgroup$ Nov 20, 2012 at 8:32
  • $\begingroup$ Indeed, which is why I said Morita equivalence. Once you have fixed a forgetful functor, then the category entirely determines the algebraic theory. (To construct the symmetric monoidal closed structure you only need to know that the monad admits a commutative structure.) $\endgroup$
    – Zhen Lin
    Nov 20, 2012 at 9:29
  • $\begingroup$ Why can't we consider the commutativity condition on the monad $Ran_U U$ as a condition on the concrete category $U: Mod(T) \to Set$? $\endgroup$
    – Todd Trimble
    Nov 20, 2012 at 11:53
  • 1
    $\begingroup$ Um, Martin, as a courtesy I was trying to clarify whether you were expecting something else as regards question 1, since the comment seems to be something of a simple observation. Also, I wanted to think more about your question 2. But, I'm happy to put down my comment as an answer, as long as you're happy. (Later, since I have to dash off to do something.) $\endgroup$
    – Todd Trimble
    Nov 20, 2012 at 13:29

1 Answer 1

6
$\begingroup$

As for question 1: commutativity doesn't depend on the presentation of $T$. If $M = Mod(T)$ and $U: M \to Set$ is the forgetful functor, then commutativity can be formulated as saying that the monad $Ran_U U = U \circ Ran_U 1_M$ is commutative (or monoidal) in the sense of the nLab article here. Perhaps the most interesting aspect of this is that commutativity is a property, not an extra structure on a monad (where the structure of a strength constraint on an endofunctor on $Set$ is canonically given because every such endofunctor is canonically $Set$-enriched). These observations also lift to the enriched setting, provided of course that the functors involved are given as enriched functors (with respect to a base of enrichment $V$).

(Note: $Ran_U 1_M$, which invariably exists, is just the left adjoint $F$ of $U$ if $U$ has a left adjoint. Some related discussion on the codensity monad of a general functor $U: M \to Set$ can be found in this post by Tom Leinster.)

Edit: I had responded to question 2 earlier, but I am now editing that response out as it is superseded by Martin's second comment below, which makes the situation quite clear. Apologies for the noise.

$\endgroup$
2
  • $\begingroup$ Hm, I think that 2) is wrong: Note that 2) is equivalent to the assertion that the enriched Yoneda embedding $C^{op} \to \hom_c(C,C)$ is an equivalence (here $c$ stands for continuous functors). Now take $C=$left $k$-modules for some ring $k$. Then it is well-known (special case of Eilenberg-Watts) that the right hand side equals the category of $k$-bimodules. Even when $k$ is commutative, this is not just $k$-modules. $\endgroup$ Nov 21, 2012 at 7:31
  • $\begingroup$ Martin, you're absolutely right, and that's a very clear way to look at the situation. I'm now going to edit my answer, and delete some comments which are now obsolete. $\endgroup$
    – Todd Trimble
    Nov 21, 2012 at 14:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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