Theorem 16.1.11 (Milnor presentation of \(\omega\)). The canonical square of posets
(defined by sending each element to the corresponding element with the same name)
is a pushout square in \(\Cat\).
Proof of Theorem 16.1.11
It is easy to show that the square (16.7) is a pushout of partially ordered sets:
• The square is tautologically a pushout on cores (a.k.a., underlying sets)
• For any partially ordered set \((S,\leq)\), a set map \(\N\to S\) preserves the order if and only if it preserves the order when restricted to each \(\{n,n+1\} \subseteq \N\).
In other words, if we form the actual pushout
of categories, then we have \(\posetrunc{P}=\omega\) after poset truncation;
it remains to show that \(P\) is already a poset.
For each \(a \in \N\) and \(b\in \N\star \{\infty\}\), denote by \(P[a,b]\) the same pushout but retaining only those coproduct components that are indexed by \(a\leq n< b\) on the left side and \(a\leq n\leq b\) in the top right corner Note that despite the notation suggesting a closed interval, there is no object named \(\infty\) appearing anywhere in \(P[a,\infty]\).. For each finite \(b\), the analogous pushouts formed in posets yield the posets \([a,b]\); hence we have truncation maps \(P[a,b]\to [a,b]\) which are natural with respect to inclusions of intervals.
Observe that for \(b=a\) the pushout (16.10) degenerates to \[ P[a,a]=\{a\}\amalg_\emptyset \emptyset = [0] \] and for \(b=a+1\) the pushout (16.10) degenerates to \[ P[a,a+1]=[a,a+1]\amalg_{\{a,a+1\}}(\{a\}\amalg \{a+1\})=[1]. \]
An elementary rearranging of pushouts and coproducts
shows that for each \(a\leq b<\infty\) and \(b\leq c\leq \infty\)
we have a pushout squares of the following form
In particular:
• Choosing \(c=b+1\) exhibits \(P[a,b+1]\) as a one-point sink attachment of \(P[a,b]\) at the last vertex \(b\). Hence, inductively for all \(a\leq b<\infty\), the category \(P[a,b]\) is a poset, hence \(P[a,b]=[a,b]\).
• Choosing \(a=0\), \(b=1\) and \(c=\infty\) exhibits \(P\) a one-point source attachment of \(P[1,\infty]\). In particular, the inclusion \([0]\to P\) is fully faithful.
• For each \(n\in \N\), choosing \(a=0\), \(b=n\) and \(c=\infty\) yields the pushout
The top map is identified with \(\{0\}\to P[0,\infty]\) by relabeling all coproduct components along the bijection \(\succ^n\colon \N\to \N_{\geq n}\). Since the top map is fully faithful by the previous step, so is the bottom one by Corollary 12.2.12.
If we take cores of the defining pushout (16.10), then the left vertical map becomes an equivalence. Thus it follows from Corollary 13.2.2 that the map \(\N\to P^\simeq\) is a cover.
We can now show that \(P\) is a poset. For any \(x,y\in P\) we have to show that \(P(x,y)\) is a proposition. This statement itself is a proposition, so we may choose \(i,j \in \N\) with \(i\mapsto x\) and \(j\mapsto y\) (because \(\N\to P^\simeq\) is a cover). Now take \(n\coloneq max(i,j)\) so that \(x,y\) are both in the image of the map \([n]\to P\). Since we have shown this map to be fully faithful, we conclude that \(P(x,y)\) is identified with \([n](i,j)\) which is a proposition.