I'm trying to teach myself the rudiments of Category Theory. I have a doubt about the definition of comma categories, more precisely about the morphisms.

Suppose have two functors $A\xrightarrow{S}C\xleftarrow{T}B$, and we construct the comma category $(S\downarrow T)$. A morphism between two objects $(\alpha,\beta,f)$ and $(\alpha',\beta',f')$ in the comma category is a pair $(g,h)$ where $g: \alpha \rightarrow \alpha'$ and $h : \beta \rightarrow \beta'$ are morphisms in $A$ and $B$ respectively, such that the following square commutes:

$\require{AMScd} \begin{CD} S(\alpha) @>{S(g)}>> S(\alpha')\\ @V{f}VV @VV{f'}V \\ T(\beta) @>{T(h)}>> T(\beta') \end{CD}$

My question is: how does $(g,h)$ single out a unique morphism $f'$ in $C$? Couldn't there be a second morphsim $f''$ which also made the square commute? But in that case $(g,h)$ wouldn't have a single target in $(S\downarrow T)$. Where am I going wrong?

  • 1
    $\begingroup$ This question has been posed before and fully answered by David Moews and me: math.stackexchange.com/a/98358/19609 . Unfortunately the title of that question was very vague, so it is not obvious what it is all about, by just browsing the site index. Basically most/all textbooks use a sloppy notation for a morphism in a comma category. The correct one should be a quadruple, as described in the cited reference or in Berci's answer below. $\endgroup$ – magma Dec 29 '13 at 2:08

You're absolutely right in one sense: this phenomenon is analogous to that e.g. the function $x\mapsto x^2$ can be considered as an $\Bbb R\to\Bbb R$ function as well as an $\Bbb R\to\Bbb R^+$ function.

If you insist to separate the targets (and domains) for morphisms, then the proper definition of a morphism in the comma category would simply involve the target (and domain), so, if ${\bf f}=(\alpha,\beta,f)$ and ${\bf f'}=(\alpha',\beta',f')$ then a morphism of the comma category should be precisely defined as the 4-tuple $({\bf f},{\bf f'},g,h)$ with $g,h$ satisfying the commutative square condition.

However, this kind of things is not always handled so strictly, but it rarely would lead to any confusion, as when we write $(g,h):{\bf f}\to{\bf f'}$, the target and domain informations are already indicated...

  • $\begingroup$ Yeah, it's the same way a set theoretic function, as a set of ordered pairs, has no unique codomain, but when you do category theory you assume a definition that does encode the codomain. Same with most morphisms in concrete categories. $\endgroup$ – Malice Vidrine Dec 28 '13 at 16:31

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.