Hom functors and a glimpse of Yoneda

Given two objects A and B, Hom(A, B) is simply the set of functions between A and B. From this humble start, things get more interesting quickly.

Hom sets

To make the above definition precise, we need to say what kinds of objects and what kinds of functions we’re talking about. That is, we specify a category C that the object belong to, and the functions are the morphisms of that category [1]. For example, in the context of groups, Hom(A, B) would be the set of group homomorphisms [2] between A and B, but in the context of continuous groups (Lie groups), we would restrict Hom(A, B) to be continuous group homomorphisms.

To emphasize that Hom refers to a set of morphisms in a particular category, sometimes you’ll see the name of the category as a subscript, as in HomC(A, B). Sometimes you’ll see the name of the category as a replacement for Hom as in C(A, B). You may also see the name of the category in square braces as in [C](AB).

Hom functors

So far Hom has been a set, but you’ll also see Hom as a functor. This is where the notation takes a little more interpretation. You may see a capital H with objects as superscripts or subscripts:

\begin{eqnarray*} H^A &=& \mathrm{Hom}(A, -) \\ H_A &=& \mathrm{Hom}(-, A) \end{eqnarray*}

You may also see a lower case h instead. And I’ll use the name of the category instead of “Hom” just to throw that variation in too.

\begin{eqnarray*} h^A &=& {\cal C}(A, -) \\ h_A &=& {\cal C}(-, A) \end{eqnarray*}

I’ll explain what these mean and give a mnemonic for remembering which is which.

Action on objects

The dash in Hom(A, -) and Hom(-, A) means “insert your object here.” That is, Hom(A) takes an object B and maps it to the set of morphisms Hom(AB), and Hom(-, A) takes an object B to Hom(BA).

So Hom(A, -) and Hom(-, A) each take an object in the category C to a set of morphims, i.e. an element in the category Set. But that’s only half of what it takes to be a functor. A functor not only maps objects in one category to objects in another category, it also maps morphisms in one category to morphisms in the other. (And it does so in a way that the interactions between the maps of objects and morphisms interact coherently.)

Action on morphisms

Where do the morphisms come in? We’ve established what HA and HA do to objects, but what do they do to morphisms?

Suppose we have a morphism fXY in the category and a function g in Hom(A, X) to Hom(A, Y)? For each g in Hom(AX), the composition fg is an element of Hom(AY).

Next suppose fYX (note the reversed order) and a function g in Hom(X, A). Now the composition gf is an element of Hom(YA). Note that before we applied f after g, but here we pre-compose with f, i.e. apply f before g.

Aside from the notation, what’s going on is very simple. If you have a function from A to X and a function from X to Y, then the composition of the two is a function from A to Y. Similarly, if you have a function from Y to X and a function from X to A, the composition is a function from Y to A.

Note that HA is a covariant functor, but HA is a contravariant. More on covariant vs contravariant below.

Notation mnemonic

How can you keep HA and HA straight? It’s common to use superscript notation YX to indicate the set of functions from the superscript object X to the base object Y. You may have seen this before even if you don’t think you have.

The notation Y² denotes the product of Y with it self, such as R² for the plane, i.e. pairs of real numbers. A pair of things in Y is a function from a two-element set to Y. You could think of (y1, y2) as the result of mapping the set (1, 2) into Y.

You may also have seen the notation 2X for the power set of X, i.e. the set of all subsets of X. You could think of the power set of X being the set of maps from X to the Boolean values (true, false) where an element x is mapped to true if and only if x is in that particular subset.

The notation using H or h with a superscript A stands for Hom(A, -), i.e. morphisms out of A, which is consistent with the usage described above. And so the other is the other, i.e. a subscript A stands for Hom(-, A), i.e morphisms into A.

(Unfortunately, some authors use the opposite of the convention used here, which blows the whole mnemonic away. But the convention used here is most common.)

Yoneda lemma

We’re close to being able to state one of the most important theorems in category theory, the Yoneda lemma. (Lemmas often turn out to be more useful and better known than the theorem they were first used to prove.) So while we’re in the neighborhood, we’ll take a look.

A corollary of the Yoneda lemma says

\begin{eqnarray*} \mathrm{Hom}(H^A, H^B) &\cong& \mathrm{Hom}(B, A) \\ \mathrm{Hom}(H_A, H_B) &\cong& \mathrm{Hom}(A, B) \\ \end{eqnarray*}

The meaning of “Hom” is different on the left and right because we’re looking at morphisms between different kinds of objects. On the right we have sets of morphisms in our category C as above. The left side takes more explanation.

What kind of things are HA and HB? They are functors from C to Set. The class of functors between two categories forms a category itself. The functors are the objects in this new category, and natural transformations are the morphisms. So Hom in this context is the set of natural transformations between the two functors.

What kind of things are HA and HB? They are contravariant functors from C to Set, and contravariant functors also form a category. However, contemporary category theory doesn’t like to speak of contravariant functors, preferring to only work with covariant functors, and leaving the term “covariant” implicit. So rather than saying HA and HB are contravariant functors on C, most contemporary writers would say they are (covariant) functors on a new category Cop, where “op” stands for opposite. That is, Cop is a category with all the same objects as C, but with all the arrows turned around. Every morphism from A to B in C corresponds to a morphism from B to A in Cop.

Related posts

[1] Morphisms are a generalization of functions. Often morphisms are functions, but they might not be. But still, they have to be things that compose like functions.

[2] The name Hom is a shortened from of “homomorphism.” Different areas of math have different names for structure-preserving functions, and category theory wanted to have one name for them all. It used “Hom” as an allusion to what structure-preserving functions are called in group theory. Similarly, “morphism” is also a shorted form of “homomorphism.” I suppose the goal was to use names reminiscent of group theory, but different enough to remind the reader that the category theory counterparts are distinct.

Incidentally, “homomorphism” comes from the Greek roots meaning “similar” and “shape.” A homomorphism is a function between similar objects (objects in the same category) that preserves structure (shape).