When performing the "opposite" operation in a category, can you reverse arrows that are part of an object's definition?

48 Views Asked by At

I'm reading Emily Riehl's Category Theory in Context and pondering Exercise 1.2.i, where the author writes "Defining $\mathcal{C}/c$ to be $(c/\mathcal{C}^{op})^{op}$," in an effort to exhibit duality between the slice and co-slice categories. If you allow yourself complete freedom to just "reverse all the arrows" in the underlying commutative diagrams of the underlying category $\mathcal{C}$, this makes sense, but I get stuck when I try to step through the construction.

So, we start with $\mathcal{C}$, and we have objects $x,y,c$ and arrows $f:c\rightarrow x$, $g:c\rightarrow y$ and $h:x\rightarrow y$ such that $g=hf$.

Moving to $\mathcal{C}^{op}$, we have the same objects $x,y,c$, and now we have arrows that are the reverse of all the arrows in $\mathcal{C}$. But, the construction of $c/\mathcal{C}^{op}$ still requires that we consider, in this categories, arrows $r^{op}: c\rightarrow x$, $s^{op}:c\rightarrow y$, $t^{op}:x\rightarrow y$. Importantly, $c$ is still the domain of these arrows (and presumably for these arrows to exist, there must have been arrows $r,s,t$ in $\mathcal{C}$ such that $r:x\rightarrow c$, $s:y\rightarrow c$, and $t:y\rightarrow x$.

In $c/\mathcal{C}^{op}$, a morphism is $t^{op}(r^{op})=s^{op}$, and the objects here are $(r^{op}:c\rightarrow x)$ and $(s^{op}:c\rightarrow y)$. So, when you pass to the opposite cateogry $(c/\mathcal{C}^{op})^{op}$, can you "reach in" to the objects and change their arrow directions (i.e., mutate the objects so that ($r^{op}:c\rightarrow x$) becomes ($r:x\rightarrow c)$)? That's the only way I can get this framework to logically work.

1

There are 1 best solutions below

2
On BEST ANSWER

Let's drop the informal terminology of "reaching in", "mutating", "changing arrow directions" and so on and treat this exercise somewhat more formally; this will hopefully clear up the confusion. Namely, let's construct an isomorphism (that is, an invertible functor) $F\colon (c / \mathcal{C}^\mathrm{op})^\mathrm{op} \to \mathcal{C} / c$.

What are the objects of $(c / \mathcal{C}^\mathrm{op})^\mathrm{op}$? They're the same as the objects of $c / \mathcal{C}^\mathrm{op}$, i.e. morphisms of the form $r^\mathrm{op}\colon c \to x$ where $r\colon x \to c$ is a morphism. This gives us an obvious choice for what $F$ should be on objects: we'll define $F(r^\mathrm{op}) = r$, which by definition is an object of $\mathcal{C} / c$.

As for morphisms, what's a morphism in $(c / \mathcal{C}^\mathrm{op})^\mathrm{op}$? Well, it's the ${{-}}^\mathrm{op}$ of a morphism in $(c / \mathcal{C}^\mathrm{op})$, and a morphism in $(c / \mathcal{C}^\mathrm{op})$ from $r^\mathrm{op}\colon c \to x$ to $l^\mathrm{op}\colon c \to y$ is a morphism $h^\mathrm{op}\colon x \to y$ in $\mathcal{C}^\mathrm{op}$ such that $l^\mathrm{op} = h^\mathrm{op} \circ r^\mathrm{op}$, so you could write it as $(h^\mathrm{op})^\mathrm{op}\colon y \to x$. Again it should not come as a surprise that $F$ will be defined on morphisms as $F((h^\mathrm{op})^\mathrm{op}) = h$.

Of course one now needs to check that this actually defines a functor and find an inverse; this I leave as exercise :)