# Terminal objects of the category of morphisms

I'm reading Basic Category Theory for Computer Scientists by Benjamin C. Pierce and in exercice 1.4.6, he asks what the terminal objects are in $Set^\to$.

Let $C$ be a category.

The category $C^\to$ is defined as a category so that:

• An objet in $C^\to$ is an arrow in $C$
• An arrow in $C^\to$ from $f:A\to B$ to $f':A'\to B'$ is a pair $(a,b)$ so that the following diagram commutes $$\require{AMScd} \begin{CD} A @>{a}>> A'\\ @V{f}VV @V{f'}VV \\ B @>{b}>> B' \end{CD}$$

I think I have solved this exercice (but I'd really appreciate comments on the proofs):

Theorem 1: Terminal objects in $Set^\to$ are isomorphisms whose codomain is a terminal object in $Set$:

Lemma 2: Given a morphism $f:A\to B$ and an isomorphism $f':A'\to B'$, $\exists !(a:A\to A',b:B\to B'), f;b=a;f'$ is equivalent to $\exists !b:B\to B'$.

Proof 2: Since $f'$ is an isomorphism, $\exists !(a:A\to A',b:B\to B'), f;b=a;f'$ is equivalent to $\exists !(a:A\to A',b:B\to B'), f;b;f'^{-1}=a$. We now prove that $\exists !(a:A\to A',b:B\to B'), f;b;f'^{-1}=a$ is equivalent to $\exists !b:B\to B'$.

"$\Longrightarrow$": Suppose $\exists !(a:A\to A',b:B\to B'), f;b;f'^{-1}=a$. The existence of $b:B\to B'$ is trivial. Suppose there were two function $b_1:B\to B'$ and $b_2:B\to B'$. By setting $a_1:=f;b_1;f'^{-1}$ and $a_2:=f;b_2;f'^{-1}$, we contradict the uniqueness hypothesis. So $\exists !b:B\to B'$.

"$\Longleftarrow$": Suppose $\exists !b$. Defining $a:=f;b;f'^{-1}$, we get the existence of $(a:A\to A',b:B\to B')$. The uniqueness of $b$ is given by the hypothesis and the uniqueness of $a$ for a given $b$ is ensured by the equation so $\exists !(a:A\to A',b:B\to B'), f;b=a;f'$.

Proof 1:

"$\Longleftarrow$": Let $f':A'\to B'$ be an isomorphism so that $B'$ is terminal in $Set$. By definition, we have $\forall B \in Set, \exists !b:B\to B'$. This is equivalent to $\forall A,B\in Set, \forall f:A\to B, \exists !b:B\to B'$ which is, by Lemma 1, equivalent to $\forall A,B\in Set, \forall f:A\to B, \exists !(a:A\to A',b:B\to B'), f;b=a;f'$. So $f'$ is terminal in $Set^\to$.

"$\Longrightarrow$": Let $f':A'\to B'$ be a terminal object in $Set^\to$ (where $A'$ and $B'$ are objects of $Set$).

• To prove that $f'$ is injective, suppose it is not. We can find $x,y\in A'$ so that $x\not=y$ but $f(x)=f(y)$. And then the function $$s:\left\{\begin{array}{ll} x\mapsto y\\ y\mapsto x\\ \lambda\mapsto \lambda \text{ for } \lambda\not\in\{x,y\} \end{array}\right.$$ that swaps $x$ and $y$ is a valid candidate for $a$. Since $Id_A$ is a distinct valid candidate, $a$ is not unique which is absurd so $f'$ is injective. $$\require{AMScd} \begin{CD} {A'} @>{a:=Id_A\mid a := s}>> A'\\ @V{f'}VV @V{f'}VV \\ {B'} @>{b}>> B' \end{CD}$$

• To prove that $f'$ is surjective, suppose it is not. We can find $y\in B'$ so that $\forall x\in A', f(x)\not=y$.

• If $B'$ contains an element $z\in B'$ so that $z\not= y$, the function $$p:\left\{\begin{array}{ll} y\mapsto z\\ \lambda\mapsto \lambda \text{ for } \lambda\not=y \end{array}\right.$$ that sends all elements to themselves except $y$ which is sent to $z$ is a valid candidate for $b$. Since $Id_B$ is a distinct valid candidate, $b$ is not unique which is absurd.
• If $B'=\{y\}$, $\forall x\in A', f(x)\not = y$ has to be vacuously true so $A'=\emptyset$. Setting $A:=\{y\}$, $B:=\{y\}$ and $f:=Id_{\{y\}}$, we don't have the existence of $a$ which is absurd. $$\require{AMScd} \begin{CD} {A:=\{y\}} @>{a}>> A'=\emptyset\\ @V{f:=Id_{\{y\}}}VV @V{f'}VV \\ {B:=\{y\}} @>{b}>> B'=\{y\} \end{CD}$$
• Since both cases are absurd, $f'$ is surjective.
• Since $f'$ is injective and surjective, it is bijective and is therefore an isomorphism in $Set$.

• For any object $X$ of $Set$, we can take $A:=X$, $B:=X$ and $f:=Id_X$. $$\require{AMScd} \begin{CD} A:=X @>{a}>> A'\\ @V{f:=Id_X}VV @V{f'}VV \\ B:=X @>{b}>> B' \end{CD}$$ We have $\exists !(a:X\to A',b:X\to B'), f;b=a;f'$ which is, by Lemma 1, equivalent to $\exists !b:X\to B'$. So $B'$ is terminal in $Set$.

I tried to generalize this to all categories. Lemma 1 seems to be general and doesn't need any modification. In Theorem 1, the part that becomes hard is proving that $f'$ is an isomorphism. I assume I should prove that it's both monic and epic and then find a right-inverse (or a left-inverse) to prove it's an isomorphism but I wasn't able to do any of that.

So my question is: Is there a generalisation of Theorem 1 for all categories?

• If yes, then I'd like to know the exact statement, and maybe a hint on the direction that I should take (including waiting to know more stuff if need be) but no proof (unless it's a hidden proof or a linked proof or any proof that I won't see until I want to).

• If not, then I'd like to know if there is a generalisation for some categories (I assume there must be something to be done with algebra categories but I haven't looked into it yet). I'd also like to see an example where it fails (hidden or linked and with juste a hint in the question if it's an example one can find).

Cool, ty :) - Can you give me a hint on how to prove it? My problem (for monic and epic instead o finjective and surjective) is that even though I have my $u\not=v$ so that $u;f'=v;f'$, I don't know how to ensure that $a:=u$ (or $v$) makes the diagram commute... – xavierm02 Aug 7 '14 at 22:52
If $C$ is any category with a terminal object $t$ and $I$ is a category, then it is an easy exercise ($\approx$ 2 lines) that $\Delta(t)$ (the constant functor with value $t$) is a terminal object in the functor category $C^I$. Any other terminal object of $C^I$ is isomorphic to this one. Now apply this to $I=\{0 < 1\}$.
Thank you for your answer. I haven't read the part about functors yet so I don't understand the details but I can kind of see how it'll work. I'm not sure what $\{0<1\}$ is though. The poset category with 2 objets and 3 morphisms? – xavierm02 Aug 8 '14 at 16:11
Yes, every poset is regarded as a category. Even if you don't know the functor category, you should learn that every two terminal objects are isomorphic (in a unique way), and that every object isomorphic to a terminal object is also a terminal object. Thus, if you have found one terminal object, you have found all of them. In the case of $C^{\to}$, observe that the morphism $id_t : t \to t$ is terminal. It follows that the terminal objects are the isomorphisms $x \to y$, where $x,y$ are terminal. – Martin Brandenburg Aug 8 '14 at 16:27