1
$\begingroup$

I was reading David Spivak's excellent book on category theory, where he uses a category to define a monoid. Another text by Tom Leinster defines monoids in the same way (I've only read through the "Motivation for Topologists"):

A monoid $\mathcal{M}$ is a category with a single object $@$, where the morphisms are the elements of $\mathcal{M}$ and the binary map is given by composition of morphisms. Leinster goes on to describe usage of monoids with this definition in Braid Theory and other parts of topology. I have only ever seen monoids defined in this way where the object $@$ is trivial, and the specific nature of the morphisms is usually quite vague.

Is there an example where the definition of $@$ in a monoid is rigorous and interesting, where $@$ is some algebraic object like a magma, group, or ring, and where the morphisms represent something quite literal outside of their use as objects of $\mathcal{M}$? Please do not give any examples where $@$ is only a set.

$\endgroup$
  • 1
    $\begingroup$ The main point is that the actual nature of @ doesn't matter at all, - the monoid structure lives in arrows and composition. $\endgroup$ – lisyarus Jul 10 '18 at 15:16
  • $\begingroup$ By the Yoneda lemma, you can take @ to be $M$ regarded as a right $M$-set. But @ is just an abstract entity; it's exactly like a vertex in a graph. $\endgroup$ – Qiaochu Yuan Jul 10 '18 at 16:57
2
$\begingroup$

In algebra, a monoid is defined as a pair $(M, *)$ where $M$ is a set and $*$ is a binary operation on $M$ satisfying associativity and having an identity element.

A category with a single object, on the other hand, is given by a triple $(O, M, *)$ where $O$ is a set with a single element (the class of objects), $M$ is a set (the set of morphisms from the single object to itself is the only hom-set) and $*$ is a binary operation on $M$ satisfying associativity and having an identity element (the composition of morphisms turns out to be a binary operation because there is only one hom-set).

Notice that in the definition of a category with a single object there is no relationship between $O$ and $(M, *)$, because the properties that need to hold involve only $(M, *)$. Since those properties are exactly the same as those in the definition of monoid, we can say that a category with a single object consists of a monoid together with a set with a single element.

Therefore, if you want to get a monoid from a category with a single element, you just have to discard the useless set with a single element. Conversely, if you want to get a category with a single element from a monoid, you just have to come up with a set with a single element. Since that element won't play any role in the category, one usually puts a placeholder like $@$ to mean that any object will do.

Two categories with a single element are isomorphic (not just equivalent) whenever the corresponding monoids are isomorphic. Indeed, if $f\colon M \to M'$ is a monoid isomorphism, then the functor $F \colon \mathcal C \to \mathcal C'$ defined by $F(@)=@'$ and $F(m) = f(m)$ for any $m \in M$ (that is $m \colon @ \to @$ as a morphism) is clearly an isomorphism of categories, its inverse $F^{-1} \colon \mathcal C' \to \mathcal C$ being given by $F^{-1}(@') = @$ and $F^{-1}(m')=f^{-1}(m')$ for any $m' \in M'$ (i.e., $m' \colon @' \to @'$).

Now, if you still think that it makes more sense to define a monoid as a category with a single object, and that in order for a monoid defined in this way to be interesting its object should be some algebraic object as well, then you could consider the monoid of endomorphisms of a given algebraic object.

For example, if $(R, +, *)$ is a ring, then you could consider the full subcategory of the category of rings having $(R, +, *)$ as its only object. In this way there is indeed a relationship between the object and the morphisms, since the morphisms are precisely the ring homomorphism from $(R, +, *)$ to itself.

In fact, this construction can be applied to any category, not just a category of algebraic objects (e.g., you could also consider the full subcategory of the category of topological spaces having a given topological space as its only object, and so on).

$\endgroup$

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.