Category Theory
Zulip Server
Archive

You're reading the public-facing archive of the Category Theory Zulip server.
To join the server you need an invite. Anybody can get an invite by contacting Matteo Capucci at name dot surname at gmail dot com.
For all things related to this archive refer to the same person.

Stream: learning: questions

Topic: Do all adjunctions in Sets arise from product ⊣ exponential?

Ignat Insarov (Apr 15 2025 at 22:00):

The adjunction between product and exponential functors, from which the Cartesian closed structure arises, is famous. (Which is to say, the only one I know, sad laugh.) It is also special because it involves two endofunctors — it is not a «free-forgetful» adjunction. Are there any other adjunctions that completely fit into the category of sets and functions like this, and that are not constructed on top of product and exponential, but are entirely independent? Is there any kind of a classification of adjunctions between endofunctors?

John Baez (Apr 15 2025 at 22:21):

Are you familiar with the adjunction between the binary product $\times: \mathsf{Set} \times \mathsf{Set} \to \mathsf{Set}$ and the diagonal functor $\Delta : \mathsf{Set} \to \mathsf{Set} \times \mathsf{Set}$, or the adjunction between the binary coproduct $+ : \mathsf{Set} \times \mathsf{Set} \to \mathsf{Set}$ and the diagonal functor?

Taken together with the one you mentioned, these are nice because we can start with the diagonal functor and then build up addition (binary coproduct), multiplication (binary product) and exponentiation by taking adjoints.

Ignat Insarov (Apr 15 2025 at 22:23):

Yes, I am more or less familiar with them, but they are not endofunctors!

John Baez (Apr 15 2025 at 22:26):

Oh, you need endofunctors?

John Baez (Apr 15 2025 at 22:26):

Okay.

John Baez (Apr 15 2025 at 22:37):

I wonder if I know any "left adjoint monads" on Set. These are endofunctors that happen to be monads and happen to be left adjoints . They have right adjoints that are comonads.

John Baez (Apr 15 2025 at 23:11):

I guess the example you mentioned, $S \times - : \mathsf{Set} \to \mathsf{Set}$, becomes a left adjoint monad when the set $S$ is a monoid, but I'm wondering if there are other examples.

Mike Shulman (Apr 15 2025 at 23:17):

If $F:\rm Set \to Set$ is a left adjoint, then in particular it is cocontinuous. Since Set is the free cocompletion of a point, this means that $F$ is uniquely determined, up to unique isomorphism, by its value on a one-element set. If $F(1) = S$, then the unique cocontinuous extension of this is $F(X) = S\times X$. So the answer is no! There are no other endo-adjunctions of Set.

John Baez (Apr 15 2025 at 23:47):

Darn. I should have seen that.

The good news is that you've now got the complete classification of all adjoint endofunctors on $\mathsf{Set}$, @Ignat Insarov.

Kevin Carlson (Apr 16 2025 at 01:20):

So the category of left adjoint, equivalently right adjoint, endofunctors of Set is not so huge, being just Set itself. But there are vastly more if you loosen up a bit. For instance parametric right adjoints (which preserve just the connected limits, or roughly speaking “all limits except products and the terminal object”) are [[polynomial functors]], which are all the functors of the form $X\mapsto A X^B+C X^D+\cdots$. In particular, they subsume both the left and the right adjoint endofunctors that were listed above!

Ignat Insarov (Apr 16 2025 at 11:02):

I am very proud that I can understand what Mike just said!

Madeleine Birchfield (Apr 16 2025 at 12:37):

It's also true in $(\infty,1)$-category theory for left adjoint endofunctors on $\infty\mathrm{Grpd}$ being the product functor $X \mapsto S \times X$

Jencel Panic (Apr 24 2025 at 07:04):

Mike Shulman said:

If $F:\rm Set \to Set$ is a left adjoint, then in particular it is cocontinuous. Since Set is the free cocompletion of a point, this means that $F$ is uniquely determined, up to unique isomorphism, by its value on a one-element set. If $F(1) = S$, then the unique cocontinuous extension of this is $F(X) = S\times X$. So the answer is no! There are no other endo-adjunctions of Set.

Wow, this comment should be a paper.

Kevin Carlson (Apr 24 2025 at 23:44):

It’s not a new result.

Ignat Insarov (Jun 18 2025 at 03:57):

Kevin Carlson said:

It’s not a new result.

Do you mean this in the sense «this result has been published before»? If you have a reference, please share it with me. And if it is not published, then it would be convenient, for me personally, if Mike published it in some form. It is inconvenient to cite a message from a chat. That this result is privately known to you, your friends, or some other elite categorial luminaries, does not help anyone else in the world.

Mike Shulman (Jun 18 2025 at 05:42):

It's always tricky to draw the "publishable" line. I totally agree that having things in citable form is good. But for better or for worse, I think the mathematical community, as well as individual subject communities within mathematics, also have an interest in maintaining a nontriviality bar for publication.

There probably isn't a published paper proving that $46846281\times 88592984 = 4150251823092504$, but I think most of us would agree that that isn't a result worthy of publication credits. The tricky thing is that nontriviality is in the eye of the beholder. The counter-claim to publication wouldn't be that this result is already known to elite categorical luminaries; rather it would be that any sufficiently experienced category theorist could easily derive this result themselves with a few minutes' thought, so that it is more at the level of a textbook exercise for students. (I say would because I don't necessarily want to make that claim myself -- among other things, I don't think anyone should feel they are an "insufficiently experienced category theorist" just because they wouldn't find it easy to derive this result! (-: )

Of course that isn't any help to those who don't find such a derivation easy themselves and would like to be able to cite the result somewhere! But, for better or for worse, I do think the nontriviality bar for publication of a category theory result must be determined by the community of category theorists. (Ugh, and now that sounds like I'm saying anyone who doesn't find this easy isn't even a category theorist, which was totally not my intent. I guess I'll just leave it, but please nobody get hurt feelings...)

One way we've tried to grapple with this dilemma in the past is to place results of this sort on the nLab, where they can be cited and yet not give the appearance of claiming an ordinary "publication".

Mike Shulman (Jun 18 2025 at 05:44):

Would you be happy to cite an nLab page like [[endo-adjunction of Set]]?

Kevin Carlson (Jun 18 2025 at 06:02):

(deleted as superseded by Arkor below)

Nathanael Arkor (Jun 18 2025 at 06:27):

A reference for this fact is §4.2 of Bunge's thesis Categories of set valued functors.

Mike Shulman (Jun 18 2025 at 14:33):

Hooray! Thanks for finding that.