I'm not sure if this question is too elementary for MO; nevertheless, I have seen many helpful discussions surrounding this topic here. I'm interested in studying the adjunction $\operatorname{Spec}\dashv\Gamma$ for schemes through Isbell duality. As I understand it, Isbell duality relates appropriate "algebraic" and "geometric" categories via the following construction (see 1).

Let $\mathcal{A}$ be a (coextensive by 1, so $\mathcal{A}^\text{op}$ is extensive) category, and $\mathcal{B}$ a (Cartesian, to form ring objects) category. If there exists a dualizing object $\mathbb{A}$ for $\mathcal{A}$ and $\mathcal{B}$, consider the functors $$\underline{\operatorname{hom}}_{\mathcal{B}}(-,\mathbb{A}):\mathcal{B}\longrightarrow\mathcal{A}^\text{op}$$ and $$\underline{\operatorname{hom}}_{\mathcal{A}}(-,\mathbb{A}):\mathcal{A}^\text{op}\longrightarrow\mathcal{B}.$$ For "nice enough" categories, we can realize an adjunction $\underline{\operatorname{hom}}_{\mathcal{B}}(-,\mathbb{A})\dashv\underline{\operatorname{hom}}_{\mathcal{A}}(-,\mathbb{A})$ by taking the unit $\eta$ and counit $\epsilon$ as evaluation. This construction induces an equivalence of categories between those $X$ for which $\eta_X$ is an isomorphism and $A$ for which $\epsilon_A$ is an isomorphism.

This construction (pulling from 2) allows us to recover Gelfand duality $C^*\mathsf{Alg}_{\text{com}}^\text{op}\leftrightarrows\mathsf{Top}_\text{cpt}$, the equivalence of affine varieties over $k=\overline{k}$ and finitely generated integral domains (over $k$), and Stone duality, among others.

My question stems from the following realization. In each of these cases, the dualizing object in $\mathcal{B}$ is realized as the internalization of an "$\mathcal{A}$-type object" to $\mathcal{B}$. Indeed, Gelfand duality stems from considering the $\mathbb{C}$-algebra object $\mathbb{A}=\mathbb{C}$ in $\mathsf{Top}$, Stone duality comes from regarding the Sierpinski space $\{0,1\}$ as a frame, and our "affine-geometric duality" follows from regarding $\mathbb{A}^1_k$ as a ring object in $\mathsf{Aff}_k$.

However, in the category of schemes, there does not seem to be a "natural" commutative ring to consider dual (that is, to represent the functor $\mathsf{CRing}^\text{op}\longrightarrow \mathsf{Sch}$) to the affine line $\mathbb{A}^1=\operatorname{Spec}(\mathbb{Z}[x])$, which of course we can equip with a ring object structure (see 3 for more details).

Thus, is there a way that we can see the adjunction $\operatorname{Spec}\dashv\Gamma$ through Isbell duality, without appealing to dualizing objects? If possible, I'd love to see how this duality is expressed for (complex) analytic spaces. Thank you for any help or clarification.


1: Why is there a duality between spaces and commutative algebras?

2: Theme of Isbell duality

3: Dualizing object in the duality between commutative rings and affine schemes

  • 4
    $\begingroup$ Spec is multi-adjoint to $\hom(-,\mathbb A^1)$. There is a general notion of Diers spectrum that works in many other interesting cases. $\endgroup$ Jul 24 at 6:38
  • $\begingroup$ @მამუკაჯიბლაძე: What categories are you viewing those functors as going from/to, to view them as a multi-adjunction? $\endgroup$ Jul 24 at 7:20
  • $\begingroup$ @მამუკაჯიბლაძე: Interesting! I'll take some time to read about this very topic. $\endgroup$ Jul 24 at 17:09
  • $\begingroup$ @PeterLeFanuLumsdaine I meant the forgetful functor from local commutative rings and local homomorphisms to commutative rings and homomorphisms. It has a multiadjoint which assigns to a ring $A$ the collection of localizations $(A\to A_{\mathfrak p})_{{\mathfrak p}\in\operatorname{Spec}(A)}$. It is multiadjoint since any homomorphism $A\to L$ to a local ring $L$ factors uniquely through some local homomorphism $A_{\mathfrak p}\to L$ for a unique $\mathfrak p$ $\endgroup$ Jul 25 at 6:37
  • $\begingroup$ But to be honest, to include all schemes the only approach that I am aware of goes all the way to ringed toposes vs (local ring)ed toposes and invokes the theory of Cole spectra $\endgroup$ Jul 25 at 6:42

1 Answer 1


Stone Spaces by Peter Johnstone (CUP 1982) is about just this subject.

It is an exceptionally well written book. It would be better that you read it than that I make any attempt to summarise it for you.

By "exceptionally" I mean in contrast to the usual way in which research level pure maths books are written, namely by making it clear before the end of the first page that you are not welcome as a reader unless you are a grad student of the author.

Stone Spaces was written to sell locale theory (thought Peter would never use such a vulgar word) to the kind of mathematicians who define "filters" or "ideals" in their structures and then put a topology on the set of them. He demonstrates with several examples how you get the locale directly from the algebra and that Choice is only needed to define "points" of it, but that these are not really needed at all.

  • 1
    $\begingroup$ Thank you for the reference (I'm excited to get my hands on a copy). In the meantime, if you don't mind, could you briefly explain how locales show up in a "geometric" context? Taking a look at nLab, we do get a full subcategory-inclusion by $\operatorname{Sh}(-):\mathsf{Loc}\hookrightarrow\mathsf{Topos}$, so is there any way of connecting this inclusion to ring objects associated to (localic, or possibly arbitrary) topoi? I'm not too familiar with topos theory, so I really appreciate any help. $\endgroup$ Jul 24 at 17:36
  • $\begingroup$ @LiminalSpace: I'm sorry but the topics that you mention were the ones that I learned least thoroughly from the book and they have been swapped out of my skull. Assuming that your Maths Stack Exch questions represent your PhD topic, you should at least get your university library to order the book. You won't regret it. $\endgroup$ Jul 24 at 19:48

Your Answer

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

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