Indexed text attachment
actions-of-toposes-on-toposes.tex
Source for Actions of Toposes on Toposes
View paper · Download original file
Lines 341–440 of 1373. Source text is displayed literally.
341action hold up to coherent isomorphism. One can specify its geometric342quotient by a bicategorical codescent property.343A geometric morphism from the quotient to $\mathcal Y$344amounts to $f:\mathcal X\to\mathcal Y$ equipped345with an \emph{invertible} comparison346$f a\simeq f\operatorname{pr}_{\mathcal X}$, subject to347unit and composition. This specifies the quotient348when the indicated universal object exists.349The invertibility convention matters for monoids.350For the trivial action of the additive monoid351$\mathbb N$ on $\mathrm{Set}$, invertible comparisons are352one automorphism of the selected point of $\mathcal Y$,353with its powers. They are equivalently $\mathbb Z$-actions.354The quotient for this convention is the classifying355topos of $\mathbb Z$-sets by the trivial-action case356of the next section. Allowing noninvertible comparisons357instead gives actions of a single endomorphism. The classifying358topos for these data is the category of sets with an endomorphism:359an inverse image into it consists of a left exact cocontinuous360functor to sets and a natural endomorphism, since limits and361colimits of such diagrams are computed pointwise. The forgetful362point of this topos has the natural endomorphism supplied by the363structure map of each set. Its value on the set of natural numbers364with successor is not invertible, so this point with its comparison365gives only a lax cocone.366\par367\section{External actions by geometric automorphisms}368\label{sec:external}369 370Let \(G\) be a small discrete group and \(\mathcal E\) a371Grothendieck \(1\)-topos. An external action of \(G\) is a372pseudofunctor from the one-object groupoid \(BG\) to geometric373morphisms, with value \(\mathcal E\) at the object. Write374\(\rho_g\) for the resulting geometric autoequivalence.375Composition satisfies \(\rho_g\rho_h\simeq\rho_{gh}\), with376the usual associativity and unit coherences. In the direction of377inverse images it is convenient to index by378\(T_g=\rho_{g^{-1}}^*\). Then there are coherent isomorphisms379\(\mu_{g,h}:T_gT_h\to T_{gh}\) and380\(\epsilon:\mathrm{id}\to T_1\).381 382The equivariant category \(\mathcal E^G\) consists of pairs383\((X,\theta)\) where \(\theta_g:T_gX\to X\) are isomorphisms and384\begin{equation}385 \label{eq:ext-cocycle}386 \theta_{gh}\mu_{g,h,X}=\theta_gT_g(\theta_h),387 \qquad \theta_1\epsilon_X=\mathrm{id}_X .388\end{equation}389Morphisms commute with all \(\theta_g\). Thus the underlying390object need only be identified with each translate, and the391identifications themselves are part of the object. A strict392action is the case in which the composition and unit393comparisons are identities.394 395\begin{proposition}[Quotient and fixed-point variance]396\label{prop:ext-quotient}397The category \(\mathcal E^G\) is a Grothendieck topos. Its398forgetful functor \(U:\mathcal E^G\to\mathcal E\) is the399inverse image of a geometric morphism400\(q:\mathcal E\to\mathcal E^G\). With the canonical401isomorphisms \(q\rho_g\simeq q\), this is a pseudocolimit402of the action in geometric direction. In inverse-image403direction it is the pseudolimit of the \(T_g\) in the404\(2\)-category of topoi and left exact left adjoints.405\end{proposition}406 407\begin{proof}408Each \(T_g\) is an equivalence and so preserves all limits409and colimits. The underlying limit or colimit of an410equivariant diagram inherits its equivariance by applying411\(T_g\) to the defining cones. Coherence follows by the412universal property, and \(U\) creates these constructions413and reflects isomorphisms. Disjointness of coproducts,414pullback stability of colimits, and effectiveness of415equivalence relations can therefore be checked in \(\mathcal E\).416For a separating set of objects \(A\) of \(\mathcal E\), use417the equivariant objects418\[419 L(A)=\coprod_{g\in G} T_g A .420\]421The structure maps move summands by the \(\mu_{h,g}\).422A map \(L(A)\to X\) is equivariant precisely when it is423determined by its component \(A\to UX\) at the identity424summand, using \(\epsilon_A\). Its other components are425\(T_gA\xrightarrow{T_g(f)}T_gUX\xrightarrow{\theta_g}UX\).426Since ordinary maps out of the chosen \(A\)'s separate maps427in \(\mathcal E\), the \(L(A)\)'s separate equivariant maps.428Giraud's characterization gives the topos claim429\cite[Ch.~VII]{maclane-moerdijk}. The object430\(\prod_g T_gA\), with factors permuted using the same431coherences, is right adjoint to \(U\). Indeed an equivariant432map to that product is likewise determined by its identity433component \(f:UX\to A\), with \(g\)-component434\(T_g(f)\theta_g^{-1}\). Thus \(U\) is a left exact left adjoint.435 436For a test topos \(\mathcal F\), an inverse image into437\(\mathcal E^G\) is a functor \(S:\mathcal F\to\mathcal E^G\)438preserving finite limits and small colimits. Composing439with \(U\) gives \(F:\mathcal F\to\mathcal E\), together440with natural isomorphisms \(T_gF\simeq F\) obeying