Indexed text attachment
differential-topology-in-higher-categories.tex
Source for Differential Topology in Higher Categories
View paper · Download original file
Lines 599–698 of 6119. Source text is displayed literally.
599neighborhood of a closed set $A\subset M$, it can be approximated600in any strong $C^r$ neighborhood by a smooth $g$ agreeing with601$f$ on a possibly smaller neighborhood of $A$. The approximation602can be joined to $f$ by a homotopy fixed near $A$.603\end{proposition}604 605\begin{proof}606For an $\mathbb R^L$-valued map $F$ take a locally finite smooth607partition of unity by functions $\psi_i$ with compact support in608coordinate domains. In each coordinate chart mollify $\psi_i F$609with sufficiently small support radius. The mollified functions610\(\widetilde F_i\) are smooth and supported in slightly larger611compact subsets of the same coordinate domains, chosen locally612finite from the start. Since convolution commutes with derivatives613through order $r$, their differences from $\psi_iF$ approach zero614in $C^r$ in the coordinate chart. Extend by zero and sum.615 616To meet a given strong neighborhood, use locally finite compact617controls on the sum. Each enlarged support meets only finitely618many of these control sets. On each meeting, derivatives in the619control charts are estimated by the chart derivatives using the620chain rule on compact subsets. Choose the smoothing radius for621the $i$th summand so its contribution to each relevant constraint622is at most $2^{-i-1}$ times the allowed error there. The sum623then lies in the prescribed neighborhood. If $F$ is smooth624near $A$, first choose a smooth cutoff $\chi$ with $\chi=1$625near $A$ and with support inside that smooth locus. Replace the626sum $\widetilde F$ by $\chi F+(1-\chi)\widetilde F$.627Multiplication by $1-\chi$ has a bounded derivative estimate628on each compact constraint. Choosing smaller local errors629gives the same approximation with the exact agreement required.630 631For a manifold target use a proper smooth embedding632$i:N\hookrightarrow\mathbb R^L$ and a tubular retraction633$\rho:V\to i(N)$ from an open neighborhood634\cite[Chs.~2, 4]{hirsch}. Apply the Euclidean argument to635$F=i\circ f$. Before smoothing, tighten the compact controls636so that the straight segments between $F$ and its approximation637remain in $V$. This is possible on each compact set since638$F$ maps it into a compact subset of the open set $V$.639The chain rule and compact control also allow the approximation640to be chosen small enough so that $\rho\circ\widetilde F$ meets641the original jet conditions. Composing the straight-line homotopy642with $i^{-1}\rho$ gives the assertion, including the fixed643neighborhood.644\end{proof}645 646Real-analytic approximation has another source. Local647partitions of unity cannot serve as analytic cutoffs, since an648analytic function on a connected domain that vanishes on a649nonempty open set vanishes identically. A compact interval650illustrates a different mechanism. If $f$ is $C^r$ on $[a,b]$,651approximate $f^{(r)}$ uniformly by a polynomial $q$.652For $r\geq1$ integrate $q$ successively $r$ times, choosing the653constants to match $f$ and its lower derivatives at $a$.654The resulting polynomial approximates the $j$th derivative655for $0\le j<r$ with error bounded by656\[657 \|q-f^{(r)}\|_\infty\,\frac{(b-a)^{r-j}}{(r-j)!}.658\]659This follows by integrating the derivative error at each step.660For $r=0$ it is the original polynomial approximation.661On general real-analytic manifolds the global approximation662theorem uses complex analytic methods. Hirsch states the663density of analytic maps in the strong topology for targets664and sources equipped with analytic structures665\cite[Ch.~2, \S5]{hirsch}. Density does not impose exact agreement666with arbitrary prescribed smooth data.667 668\section{The object representing jointly smooth families}669 670Let $\mathcal E$ be the category of sheaves of spaces from671Chapter~\ref{ch:manifolds}. It has an exponential672$[\underline M,\underline N]$, characterized by a natural673equivalence674\[675 \operatorname{Map}(S,[\underline M,\underline N])676 \simeq \operatorname{Map}(S\times\underline M,\underline N).677\]678If $S=\underline U$ is a coordinate probe, the right-hand side679is the discrete space of jointly smooth maps $U\times M\to N$.680To see this, the product is represented by the usual product681manifold, allowing corners for boundary probes, and test-family682reconstruction applies to it. The exponential therefore treats a683family and its evaluation under a single universal property.684Unlike a topology alone, it specifies which parameter tests685have smooth evaluation \cite{baez-hoffnung}.686 687A jointly smooth family $F:U\times M\to N$ is continuous into the688weak $C^r$ topology. For a point $u_0$ and compact set $K$, joint689continuity of the partial derivatives in the $M$ variables and690compactness of $K$ yield one parameter neighborhood for each691required compact control. This proof does not supply infinitely692many controls at once. For $F(s,x)=sx$ on $\mathbb R\times\mathbb R$,693the parameter family is not continuous at $s=0$ into the strong694topology: the strong neighborhood $|g(x)|<1$ on the whole source695contains $F(0,-)$ and no other member. Thus the smooth-family696structure does not implicitly choose the strong topology.697 698When $M$ is compact, familiar infinite-dimensional charts