Indexed text attachment
galois-theory-in-higher-categories.tex
Source for Galois Theory in Higher Categories
View paper · Download original file
Lines 2991–3090 of 4071. Source text is displayed literally.
2991 2992The final chapter of Borceux and Janelidze passes from objects in a2993topos to morphisms between topoi. Their descent groupoid is a kernel2994construction in a two-dimensional category. It records identifications2995between geometric morphisms and thereby differs from the kernel pair2996of a map of objects in a fixed topos. In higher dimensions this2997distinction persists. The kernel of an étale geometric morphism can2998be computed from an internal Čech nerve. An external atlas of a topos2999requires separate information about its overlaps. We first prove3000the representation theorem for étale geometric morphisms, then use3001it to identify both the scope and the limits of the internal3002construction. The classical background is \cite[\S\S7.7--7.10]{bj}.3003 3004\section{The category of geometric morphisms}3005\index{geometric morphism}3006\index{inverse image}3007 3008An $\infty$-topos is a presentable $\infty$-category admitting a3009presentation by an accessible left exact localization of an3010$\infty$-category of presheaves of spaces. Write3011$\mathcal P(C)=\operatorname{Fun}(C^{\mathrm{op}},\mathcal S)$ for the3012presheaves on a small $\infty$-category. A geometric morphism3013$f\colon\mathcal Y\to\mathcal X$ is specified by an inverse image3014$f^*\colon\mathcal X\to\mathcal Y$. It preserves finite limits and3015small colimits, so the adjoint functor theorem supplies its right3016adjoint $f_*$. Thus a morphism of topoi reverses the direction of3017the functor that performs substitution. In particular a localization3018$L\colon\mathcal X\to\mathcal X'$ gives a geometric morphism3019$\mathcal X'\to\mathcal X$ when $L$ is left exact. In the inverse3020image convention, a morphism from $f$ to $g$ in the category of3021geometric morphisms is a natural transformation $f^*\to g^*$.3022Passing to mates reverses this direction if one works with direct3023images.3024 3025For each pair $\mathcal Y,\mathcal X$, put3026\[3027 \operatorname{Geom}(\mathcal Y,\mathcal X)3028 =\operatorname{Fun}^{\mathrm{lex},\mathrm L}3029 (\mathcal X,\mathcal Y).3030\]3031We take a full subcategory of the indicated functor category,3032with all natural transformations between geometric inverse images.3033Composition of functors and their transformations yields the3034$(\infty,2)$-category of geometric morphisms in this convention.3035Its underlying $(\infty,1)$-category, written3036$\mathbf{Top}_{\infty}$, keeps the maximal subgroupoids of the3037hom $\infty$-categories. A comparison saying that two diagrams3038of inverse images are identified in $\mathbf{Top}_{\infty}$ means3039that their comparison transformation is invertible. The3040noninvertible transformations can still be useful. For instance,3041a classifier for a theory can have models as the objects of a hom3042category and homomorphisms as its noninvertible transformations.3043The core records the space of models and their equivalences.3044 3045The terminal $\infty$-topos in $\mathbf{Top}_{\infty}$ is3046$\mathcal S$. For each $\mathcal X$, the inverse image of its structure3047map to $\mathcal S$ sends a space $K$ to3048$\Delta_{\mathcal X}K=\operatorname{colim}_{K}1_{\mathcal X}$. Indeed,3049$\mathcal S$ is the free cocompletion of the terminal category,3050and the requirement to preserve the terminal object specifies its3051image. The resulting functor preserves finite limits by the finite3052limit extension theorem for presheaves3053\cite[Prop.~6.1.5.2]{htt}. Its right adjoint is global sections3054$\operatorname{Map}_{\mathcal X}(1_{\mathcal X},-)$.3055The entire hom category3056$\operatorname{Geom}(\mathcal X,\mathcal S)$ is contractible. This3057argument uses the terminal-object condition on the inverse image,3058so it applies equally when maps of topoi carry higher homotopies.3059 3060An inverse image preserves effective epimorphisms because it3061preserves Čech nerves by left exactness and preserves their3062realizations. It also preserves monomorphisms by the diagonal3063test. An image factorization is an effective epimorphism followed3064by a monomorphism, so an inverse image preserves images as well.3065In particular the family of images of generating covers transported3066by the inverse image remains jointly covering. This elementary3067property is what links geometric morphisms to a site presentation.3068 3069\section{The universal property of a site}3070\index{Grothendieck topology}3071\index{sheafification}3072\label{sec:higher-site-classifier}3073 3074Suppose $C$ has a Grothendieck topology. A sieve $R$ on $c$ is regarded3075as a monomorphism $R\hookrightarrow y(c)$ in $\mathcal P(C)$.3076Sheafification3077$\mathcal P(C)\to\operatorname{Shv}(C;\mathcal S)$ is the3078localization inverting covering sieve inclusions. If a family3079$\{c_i\to c\}$ generates $R$, the image of3080$\coprod_i y(c_i)\to y(c)$ is $R$. The sheaf condition at $R$ for3081$F$ is the equivalence3082\[3083 F(c)\xrightarrow{\sim}3084 \operatorname{Map}_{\mathcal P(C)}(R,F).3085\]3086It imposes a coherent limit condition on the whole sieve. When3087there are pullbacks of the covering arrows, the condition also3088admits a Čech form by replacing $R$ by the realization of the3089Čech nerve of $\coprod_i y(c_i)\to R$. This identification3090takes place already in presheaves, where effective epimorphisms