5
$\begingroup$

I am trying to to prove that the inclusion $\Delta \to \mathbf{Pos}$ preserves colimits, where $\mathbf{Pos}$ is the category of partially ordered sets and monotone map, and $\Delta$ is the full subcategory restricted to finite non-empty totally ordered sets (aka the simplex category).

If it helps, I think I have a proof that $\mathbf{TOrd} \to \mathbf{Pos}$ preserves colimits, where $\mathbf{TOrd}$ is the category of totally ordered sets, so it would suffice to show that $\Delta \to \mathbf{TOrd}$ preserves colimits.

$\endgroup$
3
  • 5
    $\begingroup$ $\Delta$ doesn't have too many colimits, does it? $\endgroup$ Commented Mar 25 at 23:30
  • 3
    $\begingroup$ No, it doesn't have most colimits (e.g. no coproducts) but it does have some. Actually once you show that $\Delta \to \mathbf{Pos}$ preserves colimits, you get that a diagram in $\Delta$ has a colimit iff its colimit in $\mathbf{Pos}$ is totally ordered, so it's easy to check which diagrams in $\Delta$ have colimits. $\endgroup$ Commented Mar 25 at 23:53
  • $\begingroup$ There’s also some yoga about most of the pushouts which exists in $\Delta$ being absolute — they are preserved by any function whatsoever. For one thing, the pushout of any epimorphism along any epimorphism exists and is absolute (i.e. $\Delta$ is an elegant Reedy category). For another, many of the pushouts of monomorphisms which exist are absolute — I believe this fact is used in Cisinski’s Homotopical Algebra book $\endgroup$ Commented Mar 27 at 3:53

1 Answer 1

3
$\begingroup$

I’ll show that $\Delta \to \mathbf{TOrd}$ preserves colimits. Let $F: J \to \Delta$ be a diagram that admits colimit $L$ with morphisms $\phi_j: F(j) \to L$ for each $j \in J$. I claim that $(L, \{\phi_j\}_{j \in J})$ is also the colimit of $F$ regarded as a functor to $\mathbf{TOrd}$. To that end, let $T$ be a totally ordered set and $\psi_j: F(j) \to T$ be a monotone map for each $j \in J$, s.t. $\psi_{j_2} \circ F(f) = \psi_{j_1}$ whenever $f: j_1 \to j_2$ is a morphism in $J$. For each nonempty finite subset $\tau = \{t_1, \cdots, t_n\}$ of $T$, where $t_1 < \cdots < t_n$, we define the collapsing morphism $c_\tau: T \to \tau$ by,

$$c_\tau(t) = \begin{cases} t_n &, \, \mathrm{if} \, t \geq t_n\\ t_i &, \, \mathrm{if} \, t_i \leq t < t_{i+1} \, \mathrm{for} \, 1 \leq i < n\\ t_1 &, \, \mathrm{if} \, t < t_1 \end{cases}$$

Since $\tau \in \Delta$, there exists a unique morphism $u_\tau: L \to \tau$ s.t. $u_\tau \circ \phi_j = c_\tau \circ \psi_j$ for each $j \in J$.

Now, I claim that $\cup_{j \in J} \mathrm{range}(\psi_j)$ is finite. In fact, its cardinality is bounded by $|L|$. Assume to the contrary, then we may let $\tau \subset \cup_{j \in J} \mathrm{range}(\psi_j)$ be a finite subset of cardinality larger than $|L|$. But then by definition of $c_\tau$, we see that $\cup_{j \in J} \mathrm{range}(c_\tau \circ \psi_j)$ contains $\tau$. As $u_\tau \circ \phi_j = c_\tau \circ \psi_j$, this implies $\tau \subset \mathrm{range}(u_\tau)$. But $\tau$ has cardinality larger than $|L|$ whereas $u_\tau$ has domain $L$, which is a contradiction.

Now, let $\tau = \cup_{j \in J} \mathrm{range}(\psi_j)$, which is finite. Then regarding $u_\tau$ as a map with codomain $T$, we easily see that $u_\tau \circ \phi_j = \psi_j$ for all $j \in J$. Assume another map $u: L \to T$ has the same property. We observe that $\cup_{j \in J} \mathrm{range}(\phi_j) = L$. Indeed, we easily see that $\cup_{j \in J} \mathrm{range}(\phi_j)$, together with $\{\phi_j\}_{j \in J}$, is a colimit of $F$ in $\Delta$ by using similar methods as the collapsing morphisms defined above. So, the uniqueness of colimits ensures that $\cup_{j \in J} \mathrm{range}(\phi_j) = L$. But this means,

$$\tau = \cup_{j \in J} \mathrm{range}(\psi_j) = \cup_{j \in J} \mathrm{range}(u \circ \phi_j) = \mathrm{range}(u)$$

Now, by the property of colimits, and again because $\tau \in \Delta$, we see that $c_\tau \circ u_\tau = c_\tau \circ u$. But both $u_\tau$ and $u$ have ranges in $\tau$, so this implies $u_\tau = u$, concluding the proof that $L$ is the colimit of $F$ in $\mathbf{TOrd}$.

$\endgroup$
3
  • 1
    $\begingroup$ I’d say the proof that $\mathbf{TOrd} \to \mathbf{Pos}$ preserves colimits is pretty similar, where one can use the completion of partial orders to total orders to show that $\cup_{j \in J} \mathrm{range}(\psi_j)$ must be totally ordered if a colimit in $\mathbf{TOrd}$ exists. $\endgroup$
    – David Gao
    Commented Mar 26 at 2:43
  • 1
    $\begingroup$ A technical point: I glossed over the possibility that $\cup_{j \in J} \mathrm{range}(\psi_j) = \varnothing$ in my proof. But since $\Delta$ does not contain the empty set, this could only happen when $F$ is the empty diagram. But $\Delta$ has no initial object, so in that case $F$ does not admit a colimit in $\Delta$ in the first place. $\endgroup$
    – David Gao
    Commented Mar 26 at 2:54
  • $\begingroup$ This makes sense, thank you! Yeah the proof for $\mathbf{TOrd} \to \mathbf{Pos}$ uses the fact that every partial order has a total order extension. $\endgroup$ Commented Mar 26 at 11:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .