Indexed text attachment
differential-topology-in-higher-categories.tex
Source for Differential Topology in Higher Categories: Mapping Objects, Derived Intersections, Spaces of Choices, and Cobordism
View paper · Download original file
Lines 1211–1310 of 3851. Source text is displayed literally.
1211rests on a lemma worth isolating.1212 1213\begin{lemma}[Fibrewise transversality]\label{lem:fibre}\Scl1214Let $F\colon V\times M\to N$ be smooth with $F\trans A$, let $W=F^{-1}(A)$ and let $\pi\colon W\to V$ be the1215projection. Then $F_v=F(v,-)$ is transverse to $A$ if and only if $v$ is a regular value of $\pi$.1216\end{lemma}1217 1218\begin{proof}1219Fix $(v,x)\in W$, $z=F(v,x)$, and let $L\colon T_vV\oplus T_xM\to Q=T_zN/T_zA$ be the differential of $F$1220followed by the quotient. By hypothesis $L$ is onto, and $T_{(v,x)}W=\ker L$. Write $L_M$ for the1221restriction of $L$ to $T_xM$; then $F_v\trans A$ at $x$ means that $L_M$ is onto, and $d\pi$ is the1222projection $\ker L\to T_vV$.1223 1224If $L_M$ is onto and $a\in T_vV$, choose $b$ with $L_M(b)=L(a,0)$; then $(a,-b)\in\ker L$ projects to $a$.1225Conversely, if the projection is onto and $q=L(a,b)\in Q$, choose $(a,b')\in\ker L$; then1226$q=L(a,b)-L(a,b')=L_M(b-b')$.1227\end{proof}1228 1229\begin{theorem}[Parametric transversality]\label{thm:parametric}\Scl[Hirsch 3.2.7]1230Let $F\colon V\times M\to N$ be of class $C^r$ with $r>\max\{0,\dim M+\dim A-\dim N\}$ and $F\trans A$. Then1231the set of $v\in V$ with $F_v\trans A$ is residual in $V$, and its complement has measure zero.1232\end{theorem}1233 1234\begin{proof}1235Apply \cref{thm:sard} to $\pi\colon W\to V$, noting $\dim W-\dim V=\dim M+\dim A-\dim N$, and use1236\cref{lem:fibre}.1237\end{proof}1238 1239The lemma has a reading that will recur. The manifold $W$ is the fibre product of the whole family with $A$,1240and it is always a manifold. The individual fibre products $F_v^{-1}(A)$ are its fibres over $V$. So a1241non-transverse member $F_{v_0}^{-1}(A)$ of the family is a singular fibre of a map between manifolds, and1242nearby regular fibres are its replacements. If $F$ is merely a plot of $[M,N]$, with no transversality1243assumed, nothing can be said; the hypothesis $F\trans A$ is what makes a family ``rich enough'' in the sense1244of \cref{rem:nobaire}. It holds in particular when $F$ is a submersion, for instance $F(v,x)=f(x)+v$ with1245$N=V=\R^n$ (\cref{exer:translate}).1246 1247\begin{theorem}[Jet transversality]\label{thm:jettrans}\Scl[Thom; Hirsch 3.2.8, 3.2.9]1248Let $A\subset J^r(M,N)$ be a submanifold and $r<s\le\infty$. The set of $f\in C^s(M,N)$ with $j^rf\trans A$ is1249residual in $C^s_S(M,N)$, and open if $A$ is closed.1250\end{theorem}1251 1252\section{Dimension counts}1253 1254A map transverse to a submanifold of codimension greater than $\dim M$ misses it. So genericity theorems1255become existence theorems as soon as one can compute a codimension.1256 1257\begin{proposition}\label{prop:rank}\Scl1258The $m\times n$ real matrices of rank $\rho$ form a submanifold of $\R^{m\times n}$ of dimension1259$\rho(m+n-\rho)$, that is, of codimension $(m-\rho)(n-\rho)$.1260\end{proposition}1261 1262\begin{corollary}\label{cor:counts}\Scl[Whitney; Hirsch 2.2.12, 2.2.13, 6.1.2]1263Let $\dim M=m$, $\dim N=n$, with $M$ compact.1264\begin{enumerate}[label=(\alph*)]1265\item If $n\ge2m$, immersions are dense in $C^\infty(M,N)$.1266\item If $n\ge2m+1$, embeddings are dense in $C^\infty(M,N)$.1267\item Morse functions are open and dense in $C^\infty(M,\R)$.1268\end{enumerate}1269\end{corollary}1270 1271\begin{proof}1272(a) In $J^1(M,N)$ the jets of rank $\rho<m$ form submanifolds of codimension $(m-\rho)(n-\rho)\ge n-m+1>m$. A1273map whose $1$-jet prolongation is transverse to all of them misses them. (b) For an immersion $f$,1274injectivity is the statement that $f\times f$ restricted to $M\times M\setminus\Delta$ misses $\Delta_N$, of1275codimension $n>2m$; one applies a multijet form of \cref{thm:jettrans}. (c) A function is Morse when $df$,1276as a section of $T^*M$, is transverse to the zero section, a closed submanifold of codimension $m$.1277\end{proof}1278 1279The script checks \cref{prop:rank} by computing the rank of the differential of a local defining map for1280several $(m,n,\rho)$, and checks (b) in a small case by sampling linear projections.1281 1282\section{Clean and transverse}1283 1284Transversality is a sufficient condition for a fibre product in $\Mfd$ to be correct. It is useful to1285separate the two things it provides.1286 1287\begin{definition}\label{def:clean}1288Maps $f\colon M\to N$ and $g\colon P\to N$ intersect \term{cleanly} if $Z=M\times_NP$ is a submanifold of1289$M\times P$ and $TZ=TM\times_{TN}TP$. The \term{excess bundle} is the vector bundle1290$E=\operatorname{coker}(TM\oplus TP\to TN)$ over $Z$, of rank $e=\dim N-\dim M-\dim P+\dim Z$.1291\end{definition}1292 1293\begin{proposition}\label{prop:clean}\Scl1294If $f$ and $g$ intersect cleanly then $Z$ is the fibre product in $\Mfd$ and $T$ preserves it. They are1295transverse if and only if they intersect cleanly with $e=0$. If $M,P\subset N$ are compact submanifolds1296meeting cleanly in $Z$, then for every sufficiently small generic perturbation $M'$ of $M$ the transverse1297intersection $M'\cap P$ is cobordant to the zero set of a generic section of $E$ over $Z$.1298\end{proposition}1299 1300So a clean intersection has the right tangent spaces and the wrong dimension, by $e$, and the excess bundle1301measures exactly what a perturbation will cut away. A submanifold meets itself cleanly with $E$ its normal1302bundle; the diagonal of $M\times M$ meets itself cleanly with $E=TM$. \Cref{ex:three}(b) is not clean. In1303each case there is an object that records the correct answer before any perturbation is made.1304 1305\section{Derived manifolds}1306 1307By \cref{thm:MR} the function algebra of a transverse fibre product is a pushout of $C^\infty$-rings. The1308derived fibre product is defined by the homotopy pushout.1309 1310\begin{definition}\label{def:scr}