5
$\begingroup$

Typically, when defining the functor category $\mathcal{D}^\mathcal{C}$, where objects are functors $\mathcal{C}\rightarrow\mathcal{D}$ and the morphisms between such objects $F,G$ are the natural transformations $F\rightarrow G$ with the obvious composition and identities, one requires the category $\mathcal{C}$ to be small, i.e. that its objects form a set. $\mathcal{D}$ can be any category.

I've always thought that the reason for the smallness requirement on $\mathcal{C}$ is that "surely you'll get into trouble with the morphisms $F\rightarrow G$ in $\mathcal{D}^\mathcal{C}$ potentially not forming a set otherwise", and I've never given it much thought.

Now, does anyone know of an example which fixes a large category $\mathcal{C}$, a category $\mathcal{D}$ and two functors $F,G:\mathcal{C}\rightarrow\mathcal{D}$ such that that the collection of natural transformations $F\rightarrow G$ does not form a set? I've always thought in the back of my head that this should be easily doable with $\mathcal{C}=\mathcal{D}=\mathrm{Set}$ and some simple functors $F,G:\mathcal{C}\rightarrow\mathcal{D}$ by somehow cooking up at least one natural transformation $F\rightarrow G$ for each object of $\mathcal{C}$, i.e. for each set. Maybe my set theory fu is just weak.

I could of course just be missing something basic here. I'm asking because the books I've checked tend to just state $\mathcal{C}$ must be small. Apologies if this is all trivial!

$\endgroup$

1 Answer 1

10
$\begingroup$

Hi,

let C be a discrete category (i.e. the only morphisms in C are the identity morphisms) and let D be the category consisting of two objects 0 and 1 and (apart from the identity morphisms) two parallel arrows $0\rightrightarrows 1$. Consider the constant functors F with value 0 and G with value 1. A natural transformation $F\to G$ corresponds to the choice of an element in a set of two elements for every object of C so that the conglomerate of all natural transformations $F\to G$ can be identified with the power conglomerate of Ob(C). This is not a set unless Ob(C) is a set.

$\endgroup$
1
  • $\begingroup$ Thanks! I'm glad it was as simple as the back of my head thought it was, although I was obviously a bit off in my actual construction. $\endgroup$
    – gspr
    Sep 17, 2010 at 8:49

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.