Let \(U\) be a universe and let \(C\) be a category that has finite limits and \(U\)-small colimits. We say \(U\)-small colimits in \(C\) satisfy descent if, for any \(U\)-small category \(I\), any functors \(X,Y:{I}^{\triangleright }\rightarrow C\), and any natural transformation \(p:X\Rightarrow Y\), if \(Y\) is a colimit diagram and if the restriction of \(p\) to \(I\) is cartesian (that is, all naturality squares are pullbacks), then \(X\) is a colimit diagram if and only if \(p\) is cartesian.
[0000] Logoses and toposes
We review the \(2\)-category of toposes. Following [Anel--Joyal--2021-0000], we first define the \(2\)-category of logoses and then define the \(2\)-category of toposes to be its opposite.
Let \(U\) be a universe. We say a category \(C\) is lex \(U\)-cocomplete if the following conditions are satisfied.
- -\(C\) has finite limits.
- -\(C\) has \(U\)-small colimits.
- -\(U\)-small colimits in \(C\) satisfy descent.
Let \(U\) be a universe and let \(C\) and \(D\) be lex \(U\)-cocomplete categories. We say a functor \(F:C\rightarrow D\) is a morphism of lex \(U\)-cocomplete categories if the following conditions are satisfied.
- -\(F\) commutes with finite limits.
- -\(F\) commutes with \(U\)-small colimits.
Let \(U\) be a universe and let \(C\) and \(D\) be lex \(U\)-cocomplete universes. We define \(\mathord {\textnormal {\textsf {LexCocomp}}}_{U}(C,D)\) to be the full subcategory of the category of functors from \(C\) to \(D\) spanned by the morphisms of lex \(U\)-cocomplete categories.
Let \(U\) be a universe and let \(V\) be a universe strictly greater than \(U\). We define \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\) to be the locally full subcategory of the \(2\)-category of \(V\)-small categories whose objects are the \(V\)-small lex \(U\)-cocomplete categories and morphisms are the morphisms of lex \(U\)-cocomplete categories.
Let \(U\) be a universe.
- -Let \(\kappa \) be a \(U\)-small sound doctrine. We say a category \(C\) is \((U,\kappa )\)-presentable if there exist a \(U\)-small category \(A\), a reflective full subcategory \(L\) of the category of \(U\)-small presheaves on \(A\) closed under \(U\)-small \(\kappa \)-filtered colimits, and an equivalence \(C\simeq L\).
- -We say a category \(C\) is \(U\)-presentable if it is \((U,\kappa )\)-presentable for some \(U\)-small sound doctrine \(\kappa \).
Let \(U\) be a universe and let \(V\) be a universe strictly greater than \(U\). Then \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\) is \((V,U)\)-presentable.
Hint
The theory of lex \(U\)-cocomplete categories is “algebraic”, so this should be straightforward.
Logoses are usually defined as lex cocomplete categories that have small presentations. Having small presentations can be replaced by compactness.
Let \(U\) be a universe and let \(V\) be a universe strictly greater than \(U\). Then a \(V\)-small lex \(U\)-cocomplete category \(C\) is \(U\)-presentable if and only if it is \(U\)-compact in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\).
Hint
The “only if” part follows from the following facts/observations.
- -If \(C\) is a \(U\)-presentable lex \(U\)-cocomplete category, then it is a lex localization of the category of \(U\)-small presheaves on a \(U\)-small lex category.
- -A lex localization of a \(U\)-presentable lex \(U\)-cocomplete category is a coinverter in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\).
- -The category of \(U\)-small presheaves on a \(U\)-small lex category is \(U\)-compact in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\).
For the “if” part, observe that the \(U\)-compact objects in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\) are constructible under \(U\)-small colimits from the free lex \(U\)-cocomplete category over one object. Then the proof proceeds by induction.
- -The free lex \(U\)-cocomplete category over one object is the category of \(U\)-small presheaves on the free lex category over one object. Thus, it is \(U\)-presentable.
- -\(U\)-presentable lex \(U\)-cocomplete categories are closed under \(U\)-small colimits in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\). This is proved by tracking the construction of colimits of \(U\)-presentable lex \(U\)-cocomplete categories given in Section 6.3.4 of [Lurie--2009-0000]. The key observation is that colimits of \(U\)-presentable lex \(U\)-cocomplete categories are decomposed as colimits of \(U\)-small lex categories and lex localizations both of which are colimits in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\).
Let \(U\) be a universe. The \(2\)-category \(\mathord {\textnormal {\textsf {Logos}}}(U)\) of \(U\)-logoses is defined to be the smallest full subcategory of \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\) spanned by the \(U\)-compact objects, where \(V\) is some universe strictly greater than \(U\). Note that the choice of \(V\) does not matter.
Let \(U\) be a universe. The \(2\)-category \(\mathord {\textnormal {\textsf {Topos}}}(U)\) of \(U\)-toposes is defined to be \({(\mathord {\textnormal {\textsf {Logos}}}(U))}^{\mathord {\textnormal {\textsf {op}}}}\).
Let \(U\) be a universe. We define \(\mathord {\textnormal {\textsf {Topos}}}^{(1)}(U)\) to be the \(1\)-categorical core of \(\mathord {\textnormal {\textsf {Topos}}}(U)\).
Let \(U\) be a universe. We refer to the canonical equivalence \({(\mathord {\textnormal {\textsf {Topos}}}(U))}^{\mathord {\textnormal {\textsf {op}}}}\simeq \mathord {\textnormal {\textsf {Logos}}}(U)\) as \(\mathord {\textnormal {\textsf {Sh}}}\).
Let \(U\) be a universe. The action of \(\mathord {\textnormal {\textsf {Sh}}}:{(\mathord {\textnormal {\textsf {Topos}}}(U))}^{\mathord {\textnormal {\textsf {op}}}}\simeq \mathord {\textnormal {\textsf {Logos}}}(U)\) on morphisms is also denoted by \(f\mapsto {f}^{*}\). We call \({f}^{*}\) the inverse image of a morphism \(f\) of \(U\)-toposes.
Let \(U\) be a universe and let \(V\) be a universe strictly greater than \(U\). Then \(\mathord {\textnormal {\textsf {Logos}}}(U)\) is closed in \(\mathord {\textnormal {\textsf {LexCocomp}}}(U,V)\) under \(U\)-small colimits.
Proof
This is because \(U\)-compact objects in a \((V,U)\)-presentable category are closed under \(U\)-small colimits.
Let \(U\) be a universe. Then \(\mathord {\textnormal {\textsf {Topos}}}(U)\) has finite limits.
Proof
By [0007].
Let \(U\) be a universe. Then \(\mathord {\textnormal {\textsf {Topos}}}(U)\) has \(U\)-small products.
Proof
By [0007].