Am I correct in thinking that the category pGrp, whose objects are pairs $(G,g)$ where $G$ is a group, $g \in G$ and

$$\hom\left((G,g),(H,h) \right) = \{ \varphi: G \rightarrow H \hspace{1mm} \big\vert \hspace{1mm} \varphi(g) = h \},$$

is neither isomorphic to nor equivalent to the regular category Grp?

The counterexample I was thinking of was that $\hom\left((\mathbb{Z},2), (\mathbb{Z},3)\right) = \{ \} $, but that, under the forgetful functor $F$ which essentially forgets the points and leaves the groups and homomorphisms intact, the corresponding hom-set would be $\hom \left( \mathbb{Z},\mathbb{Z} \right) $, an infinite set (since we can have $\varphi(x) = nx $ for any $ n \in \mathbb{Z}$), so $F$ would not be full.

I suppose technically this wouldn't constitute a proof (it would only be a proof that $F$ is not an equivalence), since there are other functors from pGrp to Grp, but is my line of thinking generally correct?


  • $\begingroup$ I don't know enough category theory to tell you whether or not your line of thinking is correct. However, I think for the category $Grp$ the only initial/terminal/null/zero object is the trivial group, whereas for pointed groups we could have that 2-element groups are also terminal objects (if we select the non-identity object as the base point). en.wikipedia.org/wiki/Initial_and_terminal_objects From that it should follow (I imagine) that the two categories cannot be isomorphic. $\endgroup$ – Chill2Macht Jun 18 '16 at 22:42
  • 3
    $\begingroup$ @William If they exist, terminal objects in a category are unique (up to isomorphismus). Here the terminal object is $(\{e\},e)$. $\endgroup$ – Christian Sievers Jun 18 '16 at 23:32
  • $\begingroup$ @ChristianSievers Oh yeah that is a good point, I did learn that terminal objects are "unique up to unique isomorphism". Good thing I didn't post this as an answer then. $\endgroup$ – Chill2Macht Jun 18 '16 at 23:37

You are correct that they are not isomorphic nor equivalent, and your basic idea is along the right lines; you just may want to try and argue with some of the "special" objects in $\mathbf{Grp}$, like the zero object, to ensure that there can be no isomorphism nor equivalence. Consider that in $\mathbf{Grp}$ the group $1$ is both initial and terminal. However, in $\mathbf{Grp}_\ast$, the category of pointed groups, we may want to consider maps of the form $$\phi:(1,1) \to (\lbrace e, g \rbrace, g),$$ where $(1,1)$ is the trivial group with the obvious point identified and $\lbrace e, g \rbrace$ is a group isomorphic to $\mathbb{Z}/2\mathbb{Z}$ with $e$ the identity element. In order for $\phi$ to be a homomorphism in $\mathbf{Grp}_\ast$, $\phi$ must first be a homomorphism of groups (in the usual sense) and secondly must send the identity element of $1$, to the non-identity element of $\lbrace e, g \rbrace$. But then we can easily verify that the assignment $\phi(e) := g$ is not a homomorphism of groups and hence $$ \mathbf{Grp}_\ast((1,1),(\lbrace e, g \rbrace, g)) = \emptyset;$$ we conclude that the trivial group is not initial in $\mathbf{Grp}_\ast$. This shows that in particular the trivial group cannot be a zero object in $\mathbf{Grp}_\ast$ and hence that there can be no equivalence nor isomorphism between the categories.

  • 1
    $\begingroup$ One might suspect that there may be a different zero object. However, $(1,1)$ is obviously a terminal object in $\mathbf{Grp}_*$, but it is not initial, so the category has no zero object. $\endgroup$ – egreg Jun 18 '16 at 23:29
  • $\begingroup$ @egreg If there was a zero object, no hom set could be empty. $\endgroup$ – Christian Sievers Jun 18 '16 at 23:35
  • $\begingroup$ @ChristianSievers Right, but I was commenting on Geoff's argument $\endgroup$ – egreg Jun 19 '16 at 8:48

Geoff's answer is correct, and you are mostly right, too.

However, we have You now agree that $$\hom((\mathbb{Z},2),(\mathbb{Z},3))=\emptyset,$$ and that alone shows that this category is different from the category of groups (because that is connected).

Let me note that I have never seen this category before. With the identity element, groups have a "natural" point. Usually we point categories like the one of sets, of metric or of topological spaces, thereby creating a zero object. In this case, we destroy the zero object. That's really interesting!

By the way, your category is isomorphic (sic!) to the comma category $$(\mathbb{Z}\downarrow\mathcal{Grp})$$

Also note that you deviate from standard terminology. Usually, when you have a category $\mathcal{C}$ with a terminal object $1$, a pointed object of that category is an object $A$ together with a morphism $1\to A$. Morphisms to another such pointed object $1\to A'$ are morphisms $A\to A'$ that make the obvious triangle commutative. That is the category $(1\downarrow\mathcal{C})$. It has a zero object, and if $\mathcal{C}$ alreay had a zero object, both categories are isomorphic.

Your alternative generalization of pointed sets seems to work like this: let $\mathcal{C}$ be a concrete category (a category of structured sets), that is a category with a faithful functor $U:\mathcal{C}\to\mathcal{SET}$. A pointed object is an object $A$ of that category together with an element $a\in U(A)$. A morphism $(A,a)\to(A',a')$ is a morphism $A\to A'$ that satisfies $U(f)(a)=a'$. We have seen that this construction has less nice properties. On the other hand, one may argue that only this construction is of interest if $\mathcal{C}$ already has a zero object. In any case, if $\mathcal{C}$ has a free object $F_1$ over a singleton set, then we get a category that is isomorphic to $(F_1\downarrow\mathcal{C})$.


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.