Indexed text attachment
differential-topology-in-higher-categories.tex
Source for Differential Topology in Higher Categories
View paper · Download original file
Lines 179–278 of 6119. Source text is displayed literally.
179has each $U_i$ as an open chart. To see this, the saturate of an180open subset of $U_i$ is a union of open sets obtained by the181transition diffeomorphisms. The overlap identities ensure that182the quotient map is injective on each $U_i$. When the quotient183is Hausdorff and second countable, these charts give a smooth184manifold, with the manifold dimension and boundary behavior185specified on each chart.186 187Hausdorffness is a condition on this gluing. Take two copies of188the real line and identify them on $\mathbb R\setminus\{0\}$.189The resulting space has two origins. A neighborhood of either190contains the image of some punctured interval, so two neighborhoods191of the different origins cannot be disjoint. The coordinate changes192are smooth, but the result is outside the category of Hausdorff193manifolds. A setting for smooth families can still retain the194gluing diagram and its compatibility data.195 196\section{Smooth families and full faithfulness}197 198Fix a small site $\mathsf{U}$ of coordinate domains, with open covers199as its coverings. For boundary arguments we allow relatively open200subsets of quadrants $[0,\infty)^k\times\mathbb R^{n-k}$ with the201same convention of local smooth extension. Morphisms are all smooth202maps between the domains. Including quadrants allows finite products203of tests. The manifolds in a particular argument need only use the204boundary or boundaryless charts required there. Let $\mathcal E$205denote the $\infty$-category of sheaves of spaces on this site.206 207Associate to a manifold $M$ the discrete-valued functor208\[209 \underline M(U)=C^\infty(U,M),\qquad U\in\mathsf{U}.210\]211A map $U\to M$ can be recovered from compatible restrictions to212an open cover of $U$. Its assembled function is continuous because213continuity is local, and it is smooth because every point has a214smooth local expression. This proves the sheaf condition on215$\underline M$. If a test $U$ has coordinates $(u_1,\dots,u_d)$, a216section of $\underline M(U)$ is a smooth $d$-parameter family,217possibly with parameters on the faces.218 219\begin{proposition}[Reconstruction from test families]220\label{prop:smooth-yoneda}221The association $M\mapsto\underline M$ is fully faithful.222A natural transformation $\underline M\to\underline N$223is composition with a unique smooth map $M\to N$. As maps into224$\underline N$ in $\mathcal E$, these form a discrete mapping space.225\end{proposition}226 227\begin{proof}228Let $a_i:U_i\to M$ be open chart maps from coordinate domains229whose images cover $M$. A natural transformation $\tau$ assigns230the smooth map $\tau_{U_i}(a_i):U_i\to N$ to each of these families.231On an overlap, use an open subdomain of $U_i$ and its coordinate232change to $U_j$. Naturality identifies the two restrictions.233The maps thus glue to a smooth $f:M\to N$. Given any234$u:U\to M$, restrict its domain to open pieces on which it factors235through one of the $a_i$. Naturality on each piece shows that236$\tau_U(u)$ restricts to $f\circ u$. The sheaf condition then gives237$\tau_U(u)=f\circ u$. This also proves uniqueness. The representable238smooth objects under consideration are $0$-truncated sheaves.239By definition all mapping spaces into a $0$-truncated object are240$0$-truncated, giving the last assertion.241\end{proof}242 243The value of $\underline M$ at the point remembers only its244underlying set. The extra naturality requirement is substantial.245For instance, $x\mapsto x^3$ is bijective on points, but the246identity plot $u\mapsto u$ of its target has no smooth lift on247a neighborhood of $0$. The inverse is therefore absent even as248a local map of smooth sheaves. Test families detect a singularity249of a set-theoretic bijection.250 251\section{Atlases and interval homotopies}252 253An atlas takes a universal form in the category of sheaves.254Set $A=\coprod_i\underline U_i$ in $\mathcal E$ and let255$A\to\underline M$ be induced by the chart maps. Every test-section256of $\underline M$ locally lifts to $A$, by pulling back the257coordinate cover. In this sense the atlas is an effective258epimorphism, meaning that $\underline M$ is reconstructed as259the colimit of its \v{C}ech nerve. The first stages are260\[261 \begin{tikzcd}[column sep=large]262 A\times_{\underline M} A263 \arrow[r,shift left=.55ex,"\mathrm{pr}_1"]264 \arrow[r,shift right=.55ex,"\mathrm{pr}_2"'] &265 A\arrow[r] & \underline M .266 \end{tikzcd}267\]268The iterated fiber products of $A$ in this nerve record all269multiple overlaps. The reconstruction can be checked by testing270against a sheaf $F$ of spaces. A map $\underline M\to F$ is271given on plots into $M$ by sections of $F$. A compatible system272on the nerve restricts, for each plot, to compatible sections on273its pulled-back open cover, with the higher homotopies supplied274on repeated intersections. Sheaf descent glues these into a section275over the plot. This gluing is compatible with changes of plot and276has a contractible space of choices once the descent data are fixed.277Equivalently, mapping out of $\underline M$ gives the limit of the278mapping spaces out of the nerve.