Indexed text attachment
galois-theory-in-higher-categories.tex
Source for Galois Theory in Higher Categories
View paper · Download original file
Lines 3091–3190 of 4071. Source text is displayed literally.
3091and colimits are calculated pointwise.3092 3093\begin{theorem}[Site description of inverse images]3094\label{thm:geo-site-inverse}3095Let $C$ be a small $\infty$-category with finite limits and a3096Grothendieck topology. Let3097$j\colon C\to\operatorname{Shv}(C;\mathcal S)$ be Yoneda followed3098by sheafification. For every $\infty$-topos $\mathcal Y$,3099precomposition by $j$ gives an equivalence of3100$\operatorname{Geom}(\mathcal Y,\operatorname{Shv}(C;\mathcal S))$3101with the full subcategory of3102$\operatorname{Fun}(C,\mathcal Y)$ on functors $A$ satisfying3103the following conditions.3104\begin{enumerate}3105\item The functor $A$ preserves finite limits.3106\item For every family $\{c_i\to c\}$ generating a covering sieve,3107the map $\coprod_i A(c_i)\to A(c)$ is an effective epimorphism.3108\end{enumerate}3109\end{theorem}3110 3111\begin{proof}3112The free cocompletion property identifies $A$ with its3113colimit-preserving extension3114$\overline A\colon\mathcal P(C)\to\mathcal Y$. Under this3115identification the extension is left exact precisely when $A$ is3116left exact. Necessity uses the left exactness of Yoneda. For3117sufficiency apply the finite limit extension theorem3118\cite[Prop.~6.1.5.2]{htt} to the finitely complete domain $C$.3119 3120The extension then takes the image factorization3121\[3122 \coprod_i y(c_i)\longrightarrow R\longrightarrow y(c)3123\]3124to an image factorization. Thus the condition on the generating3125family is equivalent to $\overline A$ inverting the covering sieve.3126By the universal property of localization, $\overline A$ factors3127if and only if all these inclusions are inverted. The descended3128functor is left exact since it is the restriction of $\overline A$3129along the right adjoint inclusion of sheaves, which preserves limits.3130It preserves colimits by the localization property or by taking3131the corresponding right adjoint on presheaves, whose image lies in3132sheaves. These constructions are equivalences of full functor3133subcategories and preserve natural transformations. This proves3134the equivalence including its morphisms3135\cite[Prop.~6.2.3.20]{htt}.3136\end{proof}3137 3138For a presentation by an arbitrary accessible left exact3139localization $L\colon\mathcal P(C)\to\mathcal X$, the first part3140of the argument survives without any topology on $C$. Restriction3141to $C$ identifies inverse images with those functors for which3142$\overline A$ is left exact and sends the $L$-equivalences to3143equivalences. When $L$ is displayed by a set of generators,3144testing those generators suffices since $\overline A$3145preserves colimits. An arbitrary localization can impose3146relations beyond those specified by covering sieves. Thus the3147general inverse image theorem starts with the actual localization,3148which may include a condition at a map between presheaves that3149is not a sieve inclusion.3150 3151For an ordinary site $(C,J)$ this construction provides a3152canonical lift of its ordinary topos. The $0$-truncated objects3153of $\operatorname{Shv}(C;\mathcal S)$ are exactly its sheaves3154of sets. To see this, evaluation on the sheafified representables3155tests whether a sheaf takes $0$-truncated values. Conversely,3156mapping into a sheaf taking such values can be tested on3157a colimit of sheafified representables and then gives a limit of3158$0$-truncated spaces. The canonical lift depends only on the3159ordinary topos. In the terminology of higher topos theory it3160is the $1$-localic lift, and inverse images from an3161$\infty$-topos into it are detected by the corresponding functors3162on $0$-truncated objects3163\cite[Prop.~6.4.5.7]{htt}. One may therefore extend an ordinary3164geometric morphism between Grothendieck topoi to the3165$1$-localic lifts. Recovering an arbitrary $\infty$-topos from its3166$0$-truncated objects is a stronger assertion and generally fails.3167 3168\section{Étale maps and dependent substitution}3169\index{etale geometric morphism@étale geometric morphism}3170\index{Beck--Chevalley comparison}3171 3172For a map $u\colon U\to V$ in an $\infty$-topos $\mathcal X$,3173base change has adjoints on both sides3174\[3175 u_!\dashv u^*\dashv u_*3176 \qquad3177 \left(3178 u^*\colon\mathcal X_{/V}\longrightarrow\mathcal X_{/U}3179 \right).3180\]3181The first adjoint composes an arrow to $U$ with $u$.3182The second exists because pullback preserves colimits. Pullback3183preserves finite limits as well. The resulting map of topoi3184$\mathcal X_{/U}\to\mathcal X_{/V}$ is étale. In particular3185$\pi_U\colon\mathcal X_{/U}\to\mathcal X$ denotes the map obtained3186from $U\to1_{\mathcal X}$. More generally a geometric morphism3187to $\mathcal X$ is called étale when it is equivalent over3188$\mathcal X$ to some $\pi_U$. In this use of the word $U$3189may have higher homotopy in its fibers.3190