Start with Set2. This has as objects X0→X1. The subobjects are of the form:
f
S0 -> S1
v v
|i |i'
v v
X0 -> X1
g
we can identify i(S0) with a subset T0 of X0, and i′(S1) with a subset T1 of X1.
The diagram commuting implies that g(i(S0))=i′(f(S0)). This means that g(T0)=i′(f(S0)), or that g(T0)∈im(i′)=T1.
Thus, we have that g(T0)⊆T1.
We define the subobject classifier as having values T,▹T,▹∞T, where T is interpreted as "is a subobject" (is true), and ▹ is interpreted as "delay" (ie, will be a subobject in the next timestep).
An element s∈S0⊂X0 will be classified as T.
An element sinX0,s∈X1 will be classified as ▹T, since it lands in X in one timestep.
An element sinX0,s∈X1 will be classified as ▹∞T, since it lands in X after infinite timesteps (ie, never).
We can alternatively think of ▹∞∼▹2, since it takes "two timesteps", but the second timestep is never materialized.
We formally define the subobject classifier as Ω0ω0Ω1, where Ω0≡{T,▹T,▹∞T}, ω1≡{T,▹T}.
The map is force0, T↦T, ▹T↦T, ▹∞T↦▹∞T.
Informally, the map can be said to be given by force≡(T↦T,▹n+1T↦▹nT).
We call it "force" since it forces a layer of delay.
We define the bijection between subobjects (SfX) and classification maps (Xξ[f]Ωas follows: Let i be the least i index such that f(Si)∈Xi. Then have ξ[f]0=▹iT. See that by the square, this determines ξ[f]i for all larger i:
X0 ---Χ[f]0--> Ω0
| |
f0 force0
v v
X1 - Χ[f]1- -> Ο1
[to be determined]