Suppose I have $\mathcal{J} = \mathbf{2}$, which is the discrete category with two objects. Now I have a category $\mathcal{C}$, with:

$$ c, c', c''\\ f: c \to c' \\ g: c \to c'' $$

Now I have a functor $D: \mathbf{1} \to \mathcal{C}^\mathcal{J}$, where $\mathcal{C}^\mathcal{J}$ is the category of diagrams. Additionally, I have a functor $\Delta: \mathcal{C} \to \mathcal{C}^\mathcal{J}$. From this, I want to construct a category of cones $\Delta \downarrow D$. In this category there are the following objects and morphisms:

$$ (c,\ast,h_c:\Delta c \to D\ast) \\ (c',\ast,h_{c'}:\Delta c' \to D\ast) \\ (c'',\ast,h_{c''}:\Delta c'' \to D\ast) \\ (f,1_\ast) \\ (g,1_\ast) $$ Now I want to concentrate on $h_c, h_{c'}, h_{c''}$. They all are morphisms in a functor category, which makes them natural transformations, or, as seen in context, cones. However looking back on $\mathcal{C}$, it turns out that not all of these are complete. E.g. $h_{c'}$ could have $1_{c'}$ as one of its natural transformations, but there is no way to get from $c'$ to $c''$ in $\mathcal{C}$, which makes the cone exist in $\mathcal{C}^\mathcal{J}$, but not in $\mathcal{C}$.

Is this okay? Where did I go wrong? Do I have to leave objects which don't form cones out of the category of cones?


2 Answers 2


The objects of $\Delta \downarrow D$ are all triples $(a,b,h)$ where $a$ is an object of $\mathcal{C}$, $b$ is an object of $\mathbf{1}$, and $h$ is a natural transformation $\Delta(a)\to D(b)$. In particular, such an object depends on a choice of $h$, not just on $a$ and $b$. For any particular choice of $a$ and $b$, there may be no natural transformation $h:\Delta(a)\to D(b)$, or there may be more than one. There is a separate object of $\Delta \downarrow D$ for each such choice.

So it doesn't make sense to refer to $(c,*,h_c)$ as an object until you have chosen a specific definition of $h_c$. If there is not any possible such definition, then there is no such object. If there is more than one such definition, then each one gives you a different object.


As explained in Eric's answer too, you have to check that the natural tranformations ( cones ) exist, before you build the objects in the comma category. For example, with your $C$ category consisting of 3 objects and 2 morphisms, $h_{c'}$ and $h_{c''}$ do not exist because there are no morphisms to form the cones (unless your $D$ is a constant functor ). In short your comma category has just one object (the first one you wrote).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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