On the contravariant functor of preorders and its representability

298 Views Asked by At

$$ \newcommand{op}[1]{\operatorname{#1}} \newcommand{cat}[1]{\op{#1}} \newcommand{hom}[1][2]{\cat{Hom}(#1,#2)} $$

On what at this point is becoming a series, I'm going through an exercise of Emily Riehl's Category Theory in Context,

Exercise 2.4.v. Define a contravariant functor $F: \cat{Set}^{op} \rightarrow \cat{Set}$ that carries a set to the set of preorders on it. What is its category of elements? Is $F$ representable?

I've defined this functor as

$$ \mathfrak{P} : \cat{Set}^{op} \longrightarrow \cat{Set} \\ X \longmapsto \mathfrak{P}(X) \\ (f:X \to Y) \mapsto \mathfrak{P}(f) $$ with $\mathfrak{P}(X) = \{\leq \ \subseteq \ X \times X : \leq \text{ is a preorder on } X\}$ and, given a function $f: X \to Y$,

$$ (x,y) \in \mathfrak{P}(f)(\leq) \iff f(x) \leq f(y) $$

If this is the correct/usual construction of such functor then its category of elements has objects as sets with a fixed preorder $(X, \leq )$ and morphisms as functions that behave as previously described, i.e. $f : (X,\leq) \rightarrow (Y, \leq')$ is a morphism in $\int \mathfrak{P}$ if and only if

$$ x \leq x' \iff f(x) \leq f(x') \quad (\forall x, x' \in X) $$

Now, as is proven in this exercises' chapter, a contravariant functor is representable if and only if its category of elements has a terminal object. In this case, this means that $\mathfrak{P}$ would be representable if and only if there exists a set $X_F$ equipped with a preorder $\preceq$ so that, for any preordered set $(X,\leq)$, there exists a unique function $f : X \rightarrow X_F$ such that

$$ x \leq x' \iff f(x) \preceq f(x') \quad (\forall x,x' \in X) $$

My intuition is that this will not happen, although I am not fully convinced about my (apparent) proof, which goes as follows: suppose that such a final object $(X_F, \preceq)$ exists. Therefore, there should exist a unique function $s : \mathbb{N} \longrightarrow X_F$ that defines the usual order on $\mathbb{N}$ and thus $s(\mathbb{N})$ gives a countably infinite chain on $X_F$. In particular, there exist distinct elements $x \preceq x' \preceq x''$ in $X_F$.

However, this contradicts the unicity of morphisms towards $X_F$ in the case of $T = \{1,2\}$ with the usual order, since the functions

$$ f_1: T \rightarrow X_F \\ 1 \mapsto x \\ 2 \mapsto x'\\ $$

$$ f_2: T \rightarrow X_F \\ 1 \mapsto x' \\ 2 \mapsto x''\\ $$

both verify $y \leq y' \iff f_i(y) \leq f_i(y') \ (\forall y,y' \in T)$.

Have I missed something along the way? I had a strong intuition about $\mathfrak{P}$ being representable when I started thinking about this, but now I am leaning towards a negative answer.

1

There are 1 best solutions below

0
On BEST ANSWER

Yes, that argument looks good. You could just use the order $0<1<2$, instead of going all the way to $\mathbb{N}$.