Processing math: 0%

Sunday, December 12, 2021

The identity functor

The most basic and fundamental topoi are Sets and Sets^{\to}. These describe the fundamentals of sets and functions respectively. As these are the most important objects of topos theoretic mathematics, it would be nice if the two could be related to one another in a way.

Definition. let Sets be the topos of sets and Sets^{\to} the topos of functions. Then let id : Sets \to Sets^{\to} be the function of categories that maps each set X to its identity function id_X: X \to X with f(x) = x and that maps each morphism of sets f : A \to B to the morphism of functions id_f : id_A \to id_B defined by the ordered pair of functions (f,f) : A^2 \to B^2.

Theorem. id : Sets \to Sets^{\to} is a monofunctor, which makes Sets into a full subcategory of Sets^{\to}.

Proof. (1) let f: A \to B and g : B \to C be morphisms of Sets then their composition is f \circ g : A \to C. The corresponding morphisms of Sets^{\to} are (id_f,id_f) and (id_g,id_g) defined as ordered pairs of functions. Then their composition is defined componentwise to be (id_f \circ id_g, id_f \circ id_g) which can be be refactored as (id_{f \circ g}, id_{f \circ g}). So that id_{f} \circ id_{g} = id_{f \circ g} which makes id a functor.

(2) let id_A : A \to A and id_B : B \to B be the identities of A and B respectively. Suppose that (f,g) is a morphism of functions from id_A to id_B then it must satisfy the commutative diagram which says that id_B \circ f = g \circ id_A which is logically equivalent to f = g. By the fact that f=g, it follows that any morphism of identity functions is of the form (f,f) : id_A \to id_B which can be defined by identity functor id_f which makes Sets a full subcategory of Sets^{\to}. \square

This embeds the topos Sets into the topos Sets^{\to} as the full subcategory Id. It would be interesting, if we could further determine the properties of this embedding and the extent to which it preserves the topos theoretic properties of Sets.

Theorem. Id is closed under taking products and coproducts, but not under subobjects and quotients.

Proof. (1) suppose that id_A: A \to A is an identity function, then it has as a subobject all non-surjections that are taken by reducing the domain and not the codomain. Likewise, given id_A : A \to A we can define a congruence of functions by (=_A, true) which has a constant quotient, rather then an identity quotient. So Id is not closed under subobjects or quotients.

(2) on the other hand suppose that id_A : A \to A and id_B : B \to B are two identity functions. Then id_A \times id_B : A \times B \to A \times B takes f(a,b) to (id_A(a),id_B(b)) which is equal to (a,b) so it is still an identity function. Similarily, the coproduct id_A + id_B : A + B \to A + B takes any a \in A to a and b \in B to b so that it is still an identity function. \square

This completes the process of relating Sets to Sets^{\to}. In the other direction, there are a couple of ways to relate Sets^{\to} back to Sets. Firstly, given any category C with subcategory S then we can define a morphism of topoi Sets^{C} \to Sets^{S} that reduces each set-valued functor to its S components. Using this, we can define input set and output set functors on Sets^{\to}.

A limitation of this approach is that it doesn't make Sets^{\to} into a concrete category, so in order to do that we simply need to use the coproduct construction. This takes any function f : A \to B to its coproduct set A + B constructed from its input and output sets. Together with the input and output set functors, these functors relate Sets{\to} and Sets.

No comments:

Post a Comment