Usually one distinguishes five classes of categories by size, and there are examples for all of them:

  • finite categories
  • locally finite categories
  • small categories
  • locally small categories
  • large categories

Strictly speaking there are two more classes characterized by sizes:

  • discrete categories: $|\text{Hom}(X,X)|=1, |\text{Hom}(X,Y)|=0$ for $X\neq Y$

  • indiscrete categories: $|\text{Hom}(X,Y)|=1$ for all $X,Y$

I wonder up to which point a more detailed classification of categories by size would make sense and/or for which sizes of categories there are examples resp. no examples.

Let the "size spectrum" of a category $\mathcal{C}$ be the functions $K_1, K_2$ from the class of cardinal numbers to itself (including the - improper - cardinality $\aleph_\infty$ of a proper class) which assign to each cardinality $\kappa$ the size of the class

$$O_1(\kappa) := \lbrace X \in \text{Obj}(\mathcal{C}) : |\text{Hom}(X,X)| = \kappa \rbrace $$ i.e. $K_1(\kappa) = |O_1(\kappa)|$ ("the number of objects with $\kappa$ morphisms to itself"), resp. the size of the class

$$O_2(\kappa) := \lbrace (X,Y) \in \text{Obj}(\mathcal{C})^2 : X \neq Y\ \text{and}\ |\text{Hom}(X,Y)| = \kappa \rbrace $$ i.e. $K_2(\kappa) = |O_2(\kappa)|$ ("the number of pairs of distinct objects with $\kappa$ morphisms from the first to the second").

With this definitions we have:

  • all categories: $K_1(0) = 0$

  • connected categories: $K_2(0) = 0$

  • finite categories: $\Sigma_\kappa K_1(\kappa) < \aleph_0$, $K_1(\kappa) = K_2(\kappa) = 0$ for $\kappa \geq \aleph_0$

  • locally finite categories: $K_1(\kappa) = K_2(\kappa) = 0$ for $\kappa \geq \aleph_0$

  • small categories: $\Sigma_\kappa K_1(\kappa) < \aleph_\infty$, $K_1(\aleph_\infty) = K_2(\aleph_\infty) = 0 $

  • locally small categories: $K_1(\aleph_\infty) = K_2(\aleph_\infty) = 0 $

  • discrete categories: $K_1(\kappa) = 0 $ for $\kappa \neq 1$, $K_2(\kappa) = 0 $ for $\kappa > 0$

  • indiscrete categories: $K_1(\kappa) = K_2(\kappa) = 0 $ for $\kappa \neq 1$

The characterizations have basically the form $K_1(\kappa) = 0$, $K_2(\kappa) = 0$ for $\kappa$ in specific subclasses $\mathbb{K}_1$, $\mathbb{K}_2$ of the cardinal numbers.

I wonder whether there are "interesting" and simply definable subclasses $\mathbb{K}_1$, $\mathbb{K}_2$ of the cardinal numbers such that there are no categories with $K_1(\kappa) = 0$ for $\kappa \in \mathbb{K}_1$ and/or $K_2(\kappa) = 0$ for $\kappa \in \mathbb{K}_2$.


This is an interesting idea, which I can comment almost nothing about in full generality. But regarding your singled-out question at the end, the answer is no-ish. Specifically, if $|S|=\kappa$ and $\cdot$ is a monoid structure on $S$, then there's a category with one object whose self-maps are $S$ with composition $\cdot$, so that $K_1(\kappa)=1$. But can we find such a monoid structure? It's known that there's a group of every cardinality iff the axiom of choice holds, but I don't know analogous facts about monoid structures. The proof for groups relies fundamentally on the freeness of a group's self-action, so the proof itself doesn't generalize. But defining a monoid structure on an arbitrary set feels quite choicy, so maybe the same fact will hold.

Once we have a monoid $S$ of size $\kappa,$ we can also set up a category $\mathcal{C}$ on objects $\star,\square$ with $\mathcal{C}(a,a)\cong S$ for $a$ either object. Then for every $s\in S$ put $sf$ into $\mathcal{C}(\star,\square)$ and define $s_1\circ s_2f\circ s_3=s_1s_2s_3 f$. Note for this we need to pick canonical isomorphisms between $\mathcal{C}(\star,\star)$, $\mathcal{C}(\square,\square),$ and $S$. Let $\mathcal{C}(\square,\star)=\{g\}$ with $s_1\circ g\circ s_2=g.$ Finally define $g\circ sf=s:\star\to\star$ and similarly for $sf\circ g$. Associativity of composition follows from that of $S$, so this finishes the definition of $\mathcal{C}$, which gives an example with $K_2(\kappa)=1$.

We can extend these examples via disjoint union of categories to give $K_1,K_2$ any cardinal or $\aleph_\infty$ whenever there's a monoid structure on $\kappa$. So this doesn't look like a useful angle from which to analyze your functions.

Update: I asked about the existence of monoid structures on arbitrary sets, and Zhen Lin pointed out there's an easy one that doesn't need choice: on a set with at least two elements, pick a $0$ and a $1$ and define the product of any pair of elements neither of which is $1$ to be $0$. So indeed the $K_i$ take on every value at every $\kappa.$

  • $\begingroup$ (+1) Don't forget to update this answer based on Asaf's answer to your question. $\endgroup$ – Cameron Buie Jul 30 '13 at 12:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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