36
$\begingroup$

I am trying to teach myself category theory and, as a beginner, I am looking for examples that I have a hands-on experience with.

Almost every introductory text in category theory contains following facts.

  1. The class of all posets with isotone maps is a category (called $Pos$).
  2. Every individual poset $P$ is a category, with comparable pairs $x\leq y$ as arrows. This can be called "a categorified poset".

Sometimes, product and coproduct in $Pos$ is characterized and (less frequently) it is pointed out that a Galois connection can be characterized as a pair of adjoint functors of categorified posets. Also, it is sometimes mentioned that products and coproducts in categorified posets are joins $\vee$ and meets $\wedge$, so a categorified poset has products and coproducts iff it is a latttice.

Moreover, some texts contain the fact that the category of finite posets and the category of finite distributive lattices are dually equivalent -- this is extremely useful.

But I cannot find much more. For example, when I tried to search for equalizers and coequalizers in $Pos$, the only source I found is the PhD thesis by Pietro Codara, available here, which is from 2004 and characterizes (co)equalizers in $Pos$.

Does anyone know about some other sources of information about $Pos$ from the category-theoretic viewpoint? Specifically, I am looking for things like

  • useful functors to $Pos$ and from $Pos$,
  • pullbacks, pushouts and other universal constructions in $Pos$,
  • examples of adjoint functors, applications of Yoneda lemma etc.
$\endgroup$
8
  • 13
    $\begingroup$ I would guess that you would learn more by working out what the definitions mean in this simple case, than by looking up answers. $\endgroup$ Mar 16, 2012 at 16:00
  • 2
    $\begingroup$ I don't agree with your definition of a categorified poset. I think this should rather be a $2$-category such that each hom-category is equivalent to the final category or empty (since a poset is a category such that each hom-set is a point or empty), or something similar. The process of categorification means that you "climb up" one abstraction level, not just identifiying something below with something above. $\endgroup$ Mar 16, 2012 at 16:19
  • 14
    $\begingroup$ Martin, I don't think he's using the word "categorified" in that sense -- he just means he's making or considering the poset a category. (-ify meaning generally, "to make") $\endgroup$
    – Todd Trimble
    Mar 16, 2012 at 16:34
  • $\begingroup$ @Martin Brandenburg: Maybe I misused the word "categorification": in the non-categorial mathematical literature a poset is a set (a 0-category) and the definition in my question is a 1-category: the elements become objects and the relations become arrows. It seems to me that your definition climbs one step higher. $\endgroup$ Mar 16, 2012 at 16:47
  • 3
    $\begingroup$ @Todd: Sure - I just wanted to indicate that this terminology does not fit to the usual meaning of categorification. $\endgroup$ Mar 16, 2012 at 17:20

6 Answers 6

24
$\begingroup$

Here are some basic remarks and examples: (Caution. This answer refers to preorders; but many of the remarks also apply to partially ordered sets aka posets)

  • Many concepts of category theory have a nice illustration when applied to preorders; but also the other way round: Many concepts familiar from preorders carry over to categories (for example suprema motivate colimits; see also below).

  • This is partially justified by the following observation: An arbitrary category is a sort of a preorder but where you have to specify in addition a reason why $x \leq y$, in form of an arrow $x \to y$. The axioms for a category tell you: For every $x$ there is a distinguished reason for $x \leq x$, and whenever you have a reason for $x \leq y$ and for $y \leq z$, you also get a reason for $x \leq z$.

  • A preorder is a category such that every diagram commutes.

  • In a preorder, the limit of a diagram is the same as the infimum of the involved objects. Similarly, a colimit is just a supremum. The transition morphisms don't matter.

  • When $f^* : P \to Q$ is a cocontinuous functor between preorders, where $P$ is complete, then $f^*$ has a right adjoint $f_*$; you can write it down explicitly: $f_*(q)$ is the infimum of the $p$ with $f^*(p) \leq q$. This construction motivates the General Adjoint Functor Theorem. In this setting we only have to add the solution set condition, so that the a priori big limit can be replaced by a small one and therefore exists.

  • Let $f : X \to Y$ be a map of sets. Then the preimage functor $\mathcal{P}(Y) \to \mathcal{P}(X)$ between the power sets is right adjoint to image functor $\mathcal{P}(X) \to \mathcal{P}(Y)$. Every cocontinuous monoidal functor $\mathcal{P}(Y) \to \mathcal{P}(X)$ arises this way.

  • The inclusion functor $\mathrm{Pre} \to \mathrm{Cat}$ has a left adjoint: It sends every category to its set of objects with the order $x \leq y$ if there is a morphism $x \to y$. In particular, it preserves all limits. In fact, it creates all limits, and limits in $\mathrm{Cat}$ are constructed "pointwise". Thus, the same is true for limits in $\mathrm{Pre}$ (which one could equally well see directly). For example, the pullback of $f : P \to Q$ and $g : P' \to Q$ is the pullback of sets $P \times_Q P'$ equipped with the order $(a,b) \leq (c,d)$ iff $a \leq c$ and $b \leq d$. If we apply this to difference kernels, we see that $f : P \to Q$ is a monomorphism iff the underlying map of $f$ is injective.

  • The forgetful functor $\mathrm{Pre} \to \mathrm{Set}$ creates coproducts: Take the disjoint union $\coprod_i P_i$ and take the order $a \leq b$ iff $a,b$ lie in the same $P_i$, and with respect to that preorder we have $a \leq_i b$.

  • The construction of coequalizers seems to be more delicate; see this SE discussion.

I don't have a reference for all these observations, but they are easy. A general reference for basic category-theoretic constructions (and it surely says something about preorders and posets) is the book "Abstract and Concrete Categories - The Joy of Cats" by Adamek, Herrlich, Strecker which you can find online.

EDIT: Here is something not so basic: Sefi Ladkani studied the notion of derived equivalent posets. Two posets $X,Y$ are called (universally) derived equivalent if for some specific (every) abelian category $\mathcal{A}$ the diagram categories $\mathcal{A}^X$, $\mathcal{A}^Y$ are derived equivalent.

$\endgroup$
6
  • 1
    $\begingroup$ Though I think I've made the same error as P. May: My answer talks about preorders. Sorry about that ... $\endgroup$ Mar 16, 2012 at 17:24
  • 5
    $\begingroup$ @Martin: morally, I don't think you've made an error. If the OP wants to learn this stuff, one of the things they should spend time understanding is why it's OK to treat posets and preorders as essentially the same thing. $\endgroup$ Mar 16, 2012 at 17:41
  • 3
    $\begingroup$ The sole purpose of this comment is to contain the word "quasiorder" (an alternative name for preorders) -- greetings to the all-seeing eye of google. $\endgroup$ Mar 17, 2012 at 9:22
  • 1
    $\begingroup$ Sorry, this may be ordinary categorical folklore, but could you mention briefly why or whether it is more natural to view x<=y in the preorder as an arrow x->y in the category? (i.e. as opposed to the opposite direction :P ) $\endgroup$ May 8, 2012 at 13:17
  • 1
    $\begingroup$ There are lots of reasons showing up in "practice", for example when dealing with simplicial objects. Then $0 \leq 1$, you want to have a map $X_0 \to X_1$, and not vice versa. [But of course this is just notation] $\endgroup$ May 8, 2012 at 15:41
38
$\begingroup$

Here is a fact that should be much more widely known than it is. The category of posets is isomorphic (not just equivalent) to the category of $T_0$ Alexandrof spaces. A topological space is said to be Alexandrof if arbitrary (not just finite) intersections of open sets are open. For example, every finite topological space is an Alexandrov space. Finite spaces are fascinating and the connection with algebraic topology is very close: finite spaces have associated finite simplicial complexes. Don't just look categorically: that takes the fun out of it! There are notes on my web page and there is a book by Barmak.

$\endgroup$
2
  • 9
    $\begingroup$ Do you mean preorders rather than posets, or do you require the spaces to be $T_0$, or am I missing something? $\endgroup$ Mar 16, 2012 at 16:14
  • 8
    $\begingroup$ Right you are, I mean T_0 Alexandrov spaces, or else I should say preorders. $\endgroup$
    – Peter May
    Mar 16, 2012 at 21:35
14
$\begingroup$

Probably not what you are looking for, but:

There is this recent paper by George Raptis:

Homotopy theory of posets, 2010

discussing about model category structures on the category of posets.

$\endgroup$
1
12
$\begingroup$

In a paper of Doubilet, Rota and Stanley , they see the incidence algebra of a poset as a contravariant functor from some subcategory of $Pos$ to the category of algebras over a field $K$. For instance, you may uniquely recover the poset from its incidence algebra.

I think this is quite interesting cos they then show how to arrive at some types of generating functions familiar from combinatorics.

$\endgroup$
1
9
$\begingroup$

Abstract simplicial complexes happen to be posets, and to every abstract simplicial complex, one may associate a topological space, its geometric realization. This is as functor $\mathrm{Pos} \rightarrow \mathrm{Top}$. (asc's are defined by the property that if $A \in \Delta$ and $B\subseteq A$, then $B \in \Delta$)

Also, you might want to take a look into the theory of Bruhat--Tits buildings. Basically, one associates to a simple algebraic group a certain simplicial complex $\Delta(G)$. However, I have one tried to figure out if that association is a functor without any success since given up.

Both example's seem to be more geometric/topological/algebraic than really category theoretical (viz. more focused on the objects than the functor), but maybe still instructive.

$\endgroup$
2
  • $\begingroup$ Looking back at your post, you ask for useful functors ... maybe the first example is a bit too trivial then. $\endgroup$
    – Malte
    Mar 16, 2012 at 16:11
  • 3
    $\begingroup$ This is an extremely useful functor, when composed with other functors on the left. In combinatorics, one frequently takes an object (let us say a graph) constructs a poset (say a poset of certain sets of vertices) cuts off the bounds (to have a non-contractible space) and applies your Pos->Top functor. In some cases, the resulting space is a wedge of spheres and the dimension and/or the number of spheres express some properties of the original object. $\endgroup$ Mar 16, 2012 at 16:57
3
$\begingroup$

There is an example of a model category structure on a preorder of families of sets, defined as follows: $X\leq Y$ iff every element $x\in X$ is bounded $x\leq y_x$ by some element of $y_x\in Y$; families $X$ and $Y$ are weakly equivalent iff $X\leq Y$ and for every $y\in Y$ there is $x\in X$ such that $y$ and $x$ differ by finitely many elements; cofibrant objects are families of countable sets.

The example is also quite easy, however, it is set theoretic in nature. You may find the definitions in

A homotopy theory for set theory, I

A homotopy approach to set theory

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.