Given any small category $\mathcal{C}$, the Yoneda embedding $y:\mathcal{C} \to \mathbf{Set}^{\mathcal{C}^{\operatorname{op}}}$ is well-known to represent the free cocompletion of $\mathcal{C}$. That is, the functor $F \mapsto F \circ y$ gives an equivalence of categories between cocontinuous functors from $\mathbf{Set}^{\mathcal{C}^{\operatorname{op}}}$ to $\mathcal{D}$ and functors from $\mathcal{C}$ to $\mathcal{D}$ for any cocomplete category $\mathcal{D}$.

But there is also an alternative construction which seems more intuitive and obvious.

Namely, one could define the free cocompletion $\hat{\mathcal{C}}$ as follows:

  • The objects of $\hat{\mathcal{C}}$ are the small diagrams in $\mathcal{C}$ (i.e., small categories $I$ with a functor $D:I \to \mathcal{C}$).
  • The set of morphisms from $(I, D)$ to $(J, E)$ in $\hat{\mathcal{C}}$ is given by a "limit of colimits" in the category of sets. To be precise, it is $\operatorname{lim}_{i \in I} \operatorname{colim}_{j \in J} \operatorname{Hom}_{\mathcal{C}}(D(i), E(j))$.
  • Composition and identities in $\hat{\mathcal{C}}$ are defined using the ones in $\mathcal{C}$.

The functor $\mathcal{C} \to \hat{\mathcal{C}}$ would then map each object $X$ to the corresponding "singleton" diagram $(1, X)$ (where $1$ is the category with only a single object and its identity morphism and $X$ is treated as a functor $1 \to \mathcal{C}$ in the obvious way, also denoted $X$ by abuse of notation). Also, each object $(I, D)$ of $\hat{\mathcal{C}}$ should formally represent the colimit of $D$.

The cocontinuous extension of any functor $F:\mathcal{C} \to \mathcal{D}$ would then map $(I, D)$ to $\operatorname{colim}_{i \in I} F(D(i))$.

So, is there any source on category theory that mentions the above alternative construction of the free cocompletion?


1 Answer 1


This is essentially the topic of Beurier–Pastor–Guitart's Presentations of clusters and strict free-cocompletions. They define a category $\mathrm{Clu}(\mathscr C)$ for any category ($\mathscr C$ is not necessarily small), and prove (Theorem 4.4) that it exhibits the free strict cocompletion of $\mathscr C$. The morphisms from a diagram $P$ to a diagram $Q$ in $\mathscr C$ is defined to be the set of clusters from $P$ to $Q$ (Definition 3.1): they prove this is equivalent to the limit–colimit formula you give in Proposition 3.11.


You must log in to answer this question.

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