There are three different kinds of finite limits in categories: terminal objects, binary products, and equalizers. In a category $C$, these define functors $1_{C},\times,\mathrm{Eq}\colon\mathrm{Fun}(I,C)\to C$ where $I=\emptyset,\{\bullet\ \ \bullet\}$, and $\{\bullet\rightrightarrows\bullet\}$ respectively.

Monoidal categories generalise the first two in that we now have a functor $1_{C}$ from $\mathcal{C}^{\emptyset}=*$ to $\mathcal{C}$ and a functor $\otimes_C$ from $C^{\{\bullet\ \ \bullet\}}=C$ to $C$, i.e.\ functors \begin{align*} 1_C &\colon * \to \mathcal{C}\\ \otimes_C &\colon \mathcal{C}\times\mathcal{C} \to \mathcal{C} \end{align*} together with associativity and unitality natural isomorphisms satisfying compatibility conditions.

What about equalizers? Has the notion of a category $C$ equipped with a functor $\rm{Eq}\colon\rm{Fun}(\{\bullet\rightrightarrows\bullet\},C)\to C$, a unit functor, and unitality/associativity natural isomorphisms satisfying coherence conditions been studied before? Moreover, are there any examples of such structures "found in nature"?

  • $\begingroup$ How do you formulate associativity for this kind of structure? I think the concept of a 2-monad is relevant here - basically you see monoidal categories as algebras for the 2-monad you have sketched, and want to change that 2-monad so that the "cartesian algebras" are not categories with finite products, but rather categories with binary equalizers (maybe more?). But right now your description has no evident 2-monad structure, at least I don't see it. $\endgroup$ May 20, 2021 at 10:36
  • $\begingroup$ In my brain, the vague idea is that product maps have something to do with the fact the three-element set is a pushout of a span of two-element sets. One can imagine analogues of this for the walking parallel pair of arrows, for instance to get a parallel triple, or to get a three-object version, with two sets of parallel pairs that are composable. Just something to consider! $\endgroup$
    – David Roberts
    May 20, 2021 at 10:48

1 Answer 1


Here is a way to write down what a symmetric monoidal category is starting just from the idea of finite (co)products. We have $$SymMonCat = Fun^\times_{(2,1)}(Span^{fin}_{(2,1)}, Cat_{(2,1)})$$ That is, a symmetric monoidal category is equivalent to the data of a finite-product-preserving $(2,1)$-functor from the $(2,1)$-category $Span^{fin}_{(2,1)}$ of finite sets and spans between them, to the $(2,1)$-category $Cat_{(2,1)}$ of categories.

I can think of at least two ways to understand the significance of $Span^{fin}_{(2,1)}$ from the point of view of finite (co)products.

  1. There is a functor $Span^{fin}_{(2,1)} \to Cat^{\amalg}_{(2,1)}$ (where $Cat^{\amalg}_{(2,1)}$ is the $(2,1)$-category of categories with finite coproducts and finite-coproduct preserving functors). The functor sends $n \mapsto Set^{fin}/n$. This functor is fully faithful. Moreover, $Set^{fin}/n$ is itself the free coproduct completion of the discrete category $n$.

So from this perspective, we have boiled it all down to finite (co)products, but the picture is a bit odd and I'm not sure how to complete the thought from this perspective. Alternatively,

  1. Note that the $(2,2)$-category $Span^{fin}_{(2,2)}$ of finite sets, spans between them, and maps of spans has the following universal property. For any 2-category $\mathcal K$ with finite products, we have that the 2-category $Fun_{(2,2)}^\times(Span^{fin}_{(2,2)}, \mathcal K)$ of all finite-product preserving 2-functors, is equivalent to $FinProd(\mathcal K)$, the 2-category of objects of $\mathcal K$ which internally have finite products.

(2) suggests the following generalization. Let $J$ be a set of small categories, and let $2Cat(J)$ be the category of 2-categories with $J$-limits. I believe that the functor $Comp_J: 2Cat(J) \to 2Cat$ carrying $\mathcal K$ to the 2-category of objects in $\mathcal K$ which internally have $J$-limits, is corepresentable by some $\mathcal J \in 2Cat(J)$. Define a $J$-symmetric monoidal category to be an object of $Fun^J_{(2,1)}(\mathcal J_{(2,1)}, Cat)$, where $\mathcal J_{(2,1)}$ is the maximal sub $(2,1)$-category of the $(2,2)$-category $\mathcal J$, and $Fun^J_{(2,1)}$ means we take $(2,1)$-functors which preserve $J$-limits in the $(2,1)$-categorical sense.

Then in the case where $J$ consists of the finite discrete categories, a $J$-symmetric monoidal category is a usual symmetric monoidal category. When $J$ consists of just the empty category, a $J$-symmetric monoidal category is an $E_0$-category. You could try out other classes of $J$, such as equalizers, but I'm not sure what you get!

  • 1
    $\begingroup$ There's an issue with this answer: in order to say that $K \in \mathcal K$ internally has, say, equalizers, you need $\mathcal K$ to have certain $Cat$-cotensors -- a 2-categorical limit rather than a 1-categorical one. After modifying the above story to account for this, the analogy with symmetric monoidal categories is no longer perfect. $\endgroup$
    – Tim Campion
    Dec 27, 2021 at 19:07

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.