Morphisms and comorphisms of sites II Distributors of sites
Olivia Caramello, Axel Osmond · 2025
Read Morphisms and comorphisms of sites II Distributors of sites by Olivia Caramello, Axel Osmond (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-20932.txt.
Introduction
Olivia Caramello and Axel Osmond introduce distributors of sites, a notion that unites two older ideas in category theory, in a 2025 paper. Sites are the stage on which sheaf theory and topos geometry play out. Maps between sites come in two flavors, morphisms and comorphisms. Each flavor captures only part of the picture.
Distributors generalize both at once, with matching notions of flatness and cover preservation. The authors build an adjunction linking distributors to geometric morphisms between sheaf topoi. For well-behaved distributors, that link becomes a full equivalence. They also explore equipment-like traits of the resulting bicategory. Topos theorists gain a cleaner language for comparing sites. The work may simplify future transfer of results between settings.
Catalog record
| CREATOR | Olivia Caramello, Axel Osmond |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 131 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-20932
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Morphisms and comorphisms of sites II Distributors of sites Olivia Caramello, Axel Osmond Abstract We introduce a notion of distributor of sites, involving suited analogs of flatness and cover-preservation, and show that this notion jointly generalizes those of morphism and comorphism of sites. Given two sites, we exhibit an adjunction between the category of distributors of sites between them and the category of geometric morphisms between the associated sheaf topoi; this adjunction restricts to an equivalence between geometric morphisms and continuous distributors of sites. We finally discuss some equipment-like properties of the bicategory of sites and distributors of sites. Introduction This work is the second installment in a program focused on the relations between morphisms and comorphisms of sites, and in particular the problem of mixing them altogether inside a single 2-dimensional structure. In the first part [ 8 ] , we introduced a double category of sites, morphisms and comorphisms, and then we exhibited sheafification as a double functor to the lax quintet double category of topoi. This second part presents a different solution: we show that morphisms and comorphisms can jointly be generalized by a single notion of distributor of sites . Those are distributors that combine appropriate analogs of the usual properties of cover-flatness and cover-preservation involved in the definition of morphisms of sites. A notion of flatness for distributors between categories was first introduced in [ 2 ] ; we describe at definition 2.1.11 a refinement of this notion in the presence of coverages. On the other hand, we introduce at definition 2.1.7 a condition of cover-distribution , expressing a form of joint-preservation of covers along heteromorphisms. This notion happens to be closely related to a notion isolated in [ 10 ] , which was however stated in terms of bidense morphisms. We then explain in proposition 2.2.2 how distributors of sites induce geometric morphisms; however, a notable improvement relative to morphisms or comorphisms is that now, given two sites, any geometric morphism between the associated sheaf topoi can be induced by a distributor between those sites – see proposition 2.2.11 – while one may fail to exhibit a functorial lift if both sites are fixed. This process restricts to an equivalence between geometric morphisms and distributors of sites satisfying a further continuity condition, as stated in theorem 2.2.19 . It should be emphasized that the profunctorial aspects here-examined do not fit in the double-categorical framework investigated in our first part [ 8 ] , where the double-categories of interest were of a very different kind from those usually involved in pro-arrow equipments . Although this second part makes no use of any double-categorical technology, we nevertheless describe in our last section some equipment-like properties of the bicategory of sites together with distributors of sites; while not all axioms of equipments are fulfilled – because of the very dichotomy between morphisms and comorphisms amongst other reasons – it makes sense to expect some similarities with the equipment provided by distributors, for instance well behaved gluing as described at proposition 2.3.5 . A further third part of our program with be devoted to the formal category theory for sites. 1 Distributors between categories In this section, we recall some generalities about distributors, and how in particular they induce geometric morphisms between presheaf topoi in a covariant way. We devote also a subsection to the lesser known [ 3 ] notion of flat distributors, which have enough property to induce geometric morphisms between corresponding presheaf topoi in a contravariant way. 1.1 Extensions and restrictions Here we recall some basics about extensions and nerves formulas which will be used recurrently in this work. 1.1.1 Nerve and Yoneda extensions. Recall that any functor f:\mathcal{C}\rightarrow\mathcal{D} with \mathcal{C} small and \mathcal{D} locally small induces a nerve functor {\lx@inpgf@ignorespaces{\mathcal{C}}} {\lx@inpgf@ignorespaces{\mathcal{D}}} {\lx@inpgf@ignorespaces{\widehat{\mathcal{C}}}} \scriptstyle{\lx@inpgf@ignorespaces f} \scriptstyle{\lx@inpgf@ignorespaces\text{\char 136\relax}_{\mathcal{C}}} \scriptstyle{\lx@inpgf@ignorespaces\mathcal{D}(f,1)} \scriptstyle{\lx@inpgf@ignorespaces\nu_{f}} sending d in \mathcal{D} to the presheaf \mathcal{D}[f,d]:\mathcal{C}^{\textup{op}}\rightarrow{\bf Set} , which comes equiped together with a 2-cell \nu_{f} whose component at c is the transformation with component at c^{\prime} given as f_{c^{\prime},c}:\mathcal{C}[c^{\prime},c]\rightarrow\mathcal{D}[f(c^{\prime}),f(c)] . On the other hand, if \mathcal{D} is cocomplete, this nerve functor admits a left adjoint given by the left Kan extension along the Yoneda embedding {\lx@inpgf@ignorespaces{\mathcal{C}}} {\lx@inpgf@ignorespaces{\mathcal{D}}} {\lx@inpgf@ignorespaces{\widehat{\mathcal{C}}}} \scriptstyle{\lx@inpgf@ignorespaces f} \scriptstyle{\lx@inpgf@ignorespaces\text{\char 136\relax}_{\mathcal{C}}} \scriptstyle{\lx@inpgf@ignorespaces{\textup{lan}}_{\text{\char 136\relax}_{\mathcal{C}}}f} \scriptstyle{\lx@inpgf@ignorespaces\simeq} Then one can show that the nerve can also be exhibited as a left Kan extension through the relation \mathcal{D}(f,1)={\textup{lan}}_{f}\text{\char 136\relax}_{\mathcal{C}} 1.1.2 Extensions and restrictions along a functor. First recall that any functor f:\mathcal{C}\rightarrow\mathcal{D} induces a triple of adjoints {\lx@inpgf@ignorespaces{\widehat{\mathcal{C}}}} {\lx@inpgf@ignorespaces{\widehat{\mathcal{D}}}} \scriptstyle{\lx@inpgf@ignorespaces\textup{lext}_{f}} \scriptstyle{\lx@inpgf@ignorespaces\textup{rext}_{f}} \scriptstyle{\lx@inpgf@ignorespaces\textup{rest}_{f}} \scriptstyle{\lx@inpgf@ignorespaces\dashv} \scriptstyle{\lx@inpgf@ignorespaces\dashv} where \textup{lext}_{f} (resp. \textup{rext}_{f} ) sends a presheaf X:\mathcal{C}^{\textup{op}}\rightarrow{\bf Set} to its left (resp. right) Kan extension along f^{\textup{op}} as depicted below {\lx@inpgf@ignorespaces{\mathcal{C}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\bf Set}} {\lx@inpgf@ignorespaces{\mathcal{D}^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces X} \scriptstyle{\lx@inpgf@ignorespaces f^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces{\textup{lan}}_{f^{\textup{op}}}X} \scriptstyle{\lx@inpgf@ignorespaces\zeta_{f^{\textup{op}}}} {\lx@inpgf@ignorespaces{\mathcal{C}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\bf Set}} {\lx@inpgf@ignorespaces{\mathcal{D}^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces X} \scriptstyle{\lx@inpgf@ignorespaces f^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces{\textup{ran}}_{f^{\textup{op}}}X} \scriptstyle{\lx@inpgf@ignorespaces\xi_{f^{\textup{op}}}} while the restriction functor \textup{rest}_{f} sends a presheaf Y:\mathcal{D}^{\textup{op}}\rightarrow{\bf Set} to the precomposite {\lx@inpgf@ignorespaces{\mathcal{C}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\mathcal{D}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\bf Set}} \scriptstyle{\lx@inpgf@ignorespaces f^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces\textup{rest}_{f}Y} \scriptstyle{\lx@inpgf@ignorespaces Y} so that in particular for each c in \mathcal{C} one has \textup{rest}_{f}Y(c)=Y(f(c)) Both the extensions and restriction functors can also be constructed formally as follows: for f:\mathcal{C}\rightarrow\mathcal{D} one can construct two possible nerve functors, the left and right nerves, constructed respectively as the composite and the extension {\lx@inpgf@ignorespaces{\mathcal{C}}} {\lx@inpgf@ignorespaces{\mathcal{D}}} {\lx@inpgf@ignorespaces{\widehat{\mathcal{D}}}} \scriptstyle{\lx@inpgf@ignorespaces f} \scriptstyle{\lx@inpgf@ignorespaces\mathcal{D}(1,f)} \scriptstyle{\lx@inpgf@ignorespaces\text{\char 136\relax}_{\mathcal{D}}} {\lx@inpgf@ignorespaces{\mathcal{C}}} {\lx@inpgf@ignorespaces{\mathcal{D}}} {\lx@inpgf@ignorespaces{\widehat{\mathcal{C}}}} \scriptstyle{\lx@inpgf@ignorespaces f} \scriptstyle{\lx@inpgf@ignorespaces\text{\char 136\relax}_{\mathcal{C}}} \scriptstyle{\lx@inpgf@ignorespaces\mathcal{D}(f,1)} \scriptstyle{\lx@inpgf@ignorespaces n_{f}} where \mathcal{D}(1,f) sends c to the presheaf \mathcal{D}(1,f(c))=\text{\char 136\relax}_{f(c)} , while \mathcal{D}(f,1) sends d in \mathcal{D} to the presheaf \mathcal{D}(f,d):\mathcal{C}^{\textup{op}}\rightarrow{\bf Set} , which is actually the left Kan extension \mathcal{D}(f,1)={\textup{lan}}_{f}\text{\char 136\relax}_{\mathcal{D}} . But now one can compute either the left Kan extension {\textup{lan}}_{\text{\char 136\relax}_{\mathcal{C}}}D(1,f) on one side, while one can compute the left Kan extension of \mathcal{D}(f,1) one the other side. Proposition 1.1.3 . For any functor f:\mathcal{C}\rightarrow\mathcal{D} the nerve satisfies the following identities \displaystyle\textup{rest}_{f} \displaystyle={\textup{lan}}_{\text{\char 136\relax}_{\mathcal{D}}}\mathcal{D}(f,1) \displaystyle=\widehat{\mathcal{D}}(\text{\char 136\relax}_{\mathcal{D}}f,1) \displaystyle={\textup{lan}}_{\mathcal{D}(1,f)}\text{\char 136\relax}_{\mathcal{C}} while the left and right extensions can be computed as the extensions \textup{lext}_{f}={\textup{lan}}_{\text{\char 136\relax}_{\mathcal{C}}}\mathcal{D}(1,f)\hskip 28.45274pt\textup{rext}_{f}={\textup{ran}}_{\text{\char 136\relax}_{\mathcal{C}}}\mathcal{D}(1,f) Proof. The first two expressions of the restriction functor are standards; for the third expression, observe that \mathcal{C} is cocomplete, so using that \text{\char 136\relax}_{D}f=\mathcal{D}(1,f) and applying the nerve expression of 1.1.1 as an extension gives: \widehat{\mathcal{D}}(\mathcal{D}(1,f),1)={\textup{lan}}_{\mathcal{D}(1,f)}\text{\char 136\relax}_{\mathcal{C}} For the extensions, observe that {\textup{lan}}_{\mathcal{D}(1,f)}\text{\char 136\relax}_{\mathcal{C}} is right adjoint to the left extension {\textup{lan}}_{\text{\char 136\relax}_{\mathcal{C}}}\mathcal{D}(1,f) ; hence by uniqueness of the left adjoints, we have the expression of the left extension. ∎ 1.1.4 Functoriality. Extensions and restrictions are functorial relative to composition: \textup{lext}_{g}\textup{lext}_{f}=\textup{lext}_{gf}\hskip 28.45274pt\textup{rest}_{gf}=\textup{rest}_{f}\textup{rest}_{g}\hskip 28.45274pt\textup{rext}_{g}\textup{rext}_{f}=\textup{rext}_{gf} Suppose now one has a globular 2-cell {\lx@inpgf@ignorespaces{\mathcal{C}}} {\lx@inpgf@ignorespaces{\mathcal{D}}} \scriptstyle{\lx@inpgf@ignorespaces f} \scriptstyle{\lx@inpgf@ignorespaces g} \scriptstyle{\lx@inpgf@ignorespaces\phi} Then, the corresponding restriction functors are related as follows: \phi induces in a contravariant way a 2-cell between the nerves \mathcal{D}(\phi,1):\mathcal{D}(g,1)\Rightarrow\mathcal{D}(f,1) which in turn induces a 2-cell \textup{rest}_{g}\Rightarrow\textup{rest}_{f} whose component at a presheaf Y:\mathcal{D}^{\textup{op}}\rightarrow{\bf Set} is the whiskering X*\phi^{\textup{op}}:Xg^{\textup{op}}\Rightarrow Xf^{\textup{op}} . This 2-cell has a mate \textup{lext}_{\phi}:\textup{lext}_{f}\Rightarrow\textup{lext}_{g} (resp. \textup{rext}_{\phi}:\textup{rext}_{f}\Rightarrow\textup{rext}_{g} ), whose component at a presheaf X is the universal 2-cell {\textup{lan}}_{f^{\textup{op}}}X\Rightarrow{\textup{lan}}_{g^{\textup{op}}}X (resp. {\textup{ran}}_{f^{\textup{op}}}X\Rightarrow{\textup{ran}}_{g^{\textup{op}}}X ) obtained from the universal property of the Kan extension from the composite 2-cell {\lx@inpgf@ignorespaces{\mathcal{C}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\bf Set}} {\lx@inpgf@ignorespaces{\mathcal{D}^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces X} \scriptstyle{\lx@inpgf@ignorespaces f^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces g^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces{\textup{lan}}_{g^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces\zeta_{g}} \scriptstyle{\lx@inpgf@ignorespaces\phi^{\textup{op}}} {\lx@inpgf@ignorespaces{\mathcal{C}^{\textup{op}}}} {\lx@inpgf@ignorespaces{\bf Set}} {\lx@inpgf@ignorespaces{\mathcal{D}^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces X} \scriptstyle{\lx@inpgf@ignorespaces g^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces f^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces{\textup{ran}}_{f^{\textup{op}}}} \scriptstyle{\lx@inpgf@ignorespaces\phi^{\textup{op}}} \scriptstyle{\lx@inpgf@ignorespaces\xi_{f}} There is another way to mix morphisms and comorphisms, this time as two instances of a same notion of 1-cells. It also relates to the