Independent scholarship · Open accessThe monthly digest

Indexed text attachment

differential-topology-in-higher-categories.tex

Source for Differential Topology in Higher Categories

View paper · Download original file

Lines 999–1098 of 6119. Source text is displayed literally.

999and work with \(f:U\to\mathbb R^n\) for \(U\subseteq\mathbb R^m\)1000open. The case \(n=0\) has no critical points. For \(n\geq1\),1001induct on \(m\), with the case \(m=0\) giving at most a single1002image point on each chart. Write \(C\) for the critical set1003and \(C_j\) for the subset on which every partial derivative1004of every component of \(f\), of orders \(1\) through \(j\),1005vanishes. Thus \(C_1\subseteq C\).1006 1007\emph{Positive derivative rank.}1008Near a point of \(C\setminus C_1\) there is a nonzero1009first partial of some component. After permuting coordinates,1010use that component in place of the first source coordinate.1011The inverse function theorem makes this a local coordinate1012change and the new expression for \(f\) has the form1013\[1014       f(t,u)=(t,F(t,u)).1015\]1016If \(n=1\), there are no critical points in this chart.1017Otherwise the derivative has rank1018\(1+\operatorname{rank}d_u F(t,u)\). At each \(t\), the1019corresponding section of the image of the critical set1020lies among the critical values of \(u\mapsto F(t,u)\).1021It is null by induction. To apply Fubini, exhaust the1022chart by countably many compact subsets and intersect the1023critical set with each one. Since criticality is a closed1024condition, the images of these intersections are compact,1025and their slices are measurable. Fubini gives measure zero1026on each compact image. Countably many such charts cover1027\(C\setminus C_1\).1028 1029\emph{A first nonvanishing higher derivative.}1030Near any \(x\in C_j\setminus C_{j+1}\), for \(j\geq1\),1031some scalar partial derivative \(g\) of order \(j\)1032has nonzero differential. It vanishes on \(C_j\), so1033the portion of \(C_j\) in this neighborhood lies in the1034smooth hypersurface \(g^{-1}(0)\) through \(x\). At every1035point of \(C_j\) there, \(df=0\), and therefore the1036restriction of \(f\) to the hypersurface is critical at each1037of these points. Its source dimension is \(m-1\), so the1038induction hypothesis gives a null image of this part of1039\(C_j\). A countable collection of hypersurface charts1040covers \(C_j\setminus C_{j+1}\).1041 1042\emph{High-order vanishing.}1043Choose \(j\) with \((j+1)n>m\).1044On a closed coordinate box \(K\subset U\), fix a bound for all1045derivatives of order \(j+1\). Subdivide the box into1046\(O(\epsilon^{-m})\) pieces of diameter at most a constant1047times \(\epsilon\). For each piece meeting \(C_j\), expand at1048one such point. Taylor's formula on the segment inside \(K\)1049bounds the oscillation from that point by \(O(\epsilon^{j+1})\).1050Thus \(f(C_j\cap K)\) is covered by1051\(O(\epsilon^{-m})\) balls of radius \(O(\epsilon^{j+1})\).1052Their total \(n\)-volume is \(O(\epsilon^{(j+1)n-m})\),1053which tends to zero. A countable box cover handles all of1054\(C_j\). The three cases exhaust \(C\).1055\end{proof}1056 1057This smooth proof follows the induction underlying1058\cite[Section 3]{milnor-differentiable}. Hirsch also treats finite1059differentiability. For \(C^r\) maps, the sufficient threshold is1060\(r>\max\{m-n,0\}\) \cite[Chapter 3]{hirsch}. In the smooth argument1061above, choosing a high order of vanishing helps obtain the1062last covering estimate. The sharper finite-differentiability threshold uses a different1063argument. Sard's conclusion1064concerns the image of the critical set. The set itself can1065contain open regions, as for any map constant on an open set.1066 1067\section{Boundary strata and openness}1068 1069Suppose \(M\) has boundary and \(N\) has none. If a value is regular1070for both \(f|_{\operatorname{int}M}\) and \(f|_{\partial M}\),1071the fiber is a manifold with boundary, with boundary equal to1072the fiber of the second restriction. At a boundary point write1073a source chart as \((u,t)\) with \(t\geq0\). The surjectivity on1074the boundary stratum lets us choose the invertible minor in1075the \(u\)-directions. The coordinate change in the regular1076preimage proof then leaves \(t\) unchanged. Apply Sard's1077theorem to the two strata and take a union of null sets to1078see that almost every value has this property. Testing only1079the ambient tangent space at a boundary point is insufficient.1080For the inclusion of \([0,\infty)\) in \(\mathbb R\), the1081ambient derivative at \(0\) is surjective, but the boundary1082restriction is critical there.1083 1084For a proper \(f\) without boundary, the critical set is closed1085and its image is closed, so the regular values are open as1086well as dense. Properness cannot be discarded in this statement.1087For a connected example, let \(f:\mathbb R\to\mathbb R\) be1088\(f(x)=e^{-x}(2+\sin(2x))\). It is strictly positive, so1089\(0\) is a regular value with empty fiber. At \(x=j\pi\)1090the derivative1091\[1092 f'(x)=e^{-x}\bigl(2\cos(2x)-2-\sin(2x)\bigr)1093\]1094vanishes, and the resulting critical values \(2e^{-j\pi}\)1095approach zero as \(j\to\infty\). This illustrates why openness1096statements for families typically include compactness or1097control of the behavior at infinity.1098