Independent scholarship · Open accessThe monthly digest

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 3001–3100 of 3851. Source text is displayed literally.

3001 3002The two earlier volumes of this series ended with an assessment of how much of their higher-categorical3003vocabulary was doing work. The same question is asked here, and the answer is different in both directions:3004more of the vocabulary is essential, and less of the mathematics is touched by it.3005 3006\section{Where the proofs live}3007 3008\begin{principle}[Four inputs]\label{prin:inputs}\Sob3009The theorems of the source cited in this book are derived there from four analytic facts and the topology of3010function spaces: the inverse function theorem; the existence of partitions of unity; the Morse--Sard3011theorem; and the existence, uniqueness and smooth dependence of solutions of ordinary differential3012equations. The first gives \cref{thm:transpull} and tubular neighbourhoods. The second gives approximation,3013collars, and every passage from local to global. The third gives transversality and so every genericity3014statement. The fourth gives the regular interval theorem, isotopy extension, and the handle description of a3015critical point.3016\end{principle}3017 3018None of the four is proved, shortened, or explained by anything in this book. The categorical statements3019begin where the four inputs end. \Cref{thm:CS} characterizes derived manifolds, but that the transverse3020fibre products exist is the inverse function theorem, and that they are dense is Sard's theorem.3021\Cref{thm:concordance} classifies concordance classes, but its proof uses partitions of unity in an3022essential way. \Cref{thm:palais} says that restriction is a fibration, and the lifting is done by3023integrating a vector field. What the language provides is the form of the conclusions, and the form matters3024for what can be done next.3025 3026\section{The source as a book about components}3027 3028\begin{table}[ht]\centering\small3029\begin{tabular}{@{}p{0.36\textwidth}p{0.52\textwidth}@{}}3030\toprule3031theorem of the source & the statement of which it is the shadow on $\pi_0$\\3032\midrule3033homotopic maps pull back isomorphic bundles; classification (4.2.4, 4.3.4) & $\Pi[M,B\GL_k]\simeq\Map(M,B\Or(k))$ (\cref{thm:concordance})\\3034tubular neighbourhoods and collars are unique up to isotopy (4.5.3, 4.6) & the spaces of them are contractible (\cref{thm:tubcontract,thm:collar})\\3035Hopf's theorem (5.1.10) & $S^n\to K(\Z,n)$ is $(n+1)$-connected; Pontryagin--Thom in $M$\\3036invariance of intersection numbers (5.2.1) & derived fibre products have cobordism classes (\cref{thm:spivak})\\3037$\chi$ by vector fields equals $\chi$ by homology (6.3.8) & a trace computed before and after a monoidal functor (\cref{thm:DP})\\3038regular interval theorem (6.2.2) & families of compact manifolds are classified by $\coprod B\Diff$ (\cref{thm:ehresmann})\\3039Thom's theorem (7.2.3) & $B\Cob_d\simeq\Omega^{\infty-1}MTO(d)$ (\cref{thm:GMTW})\\3040isotopy extension (8.1.3) & $\Diff_c(M)\to\Emb(V,M)$ is a fibre bundle (\cref{thm:palais})\\3041embeddings of discs are isotopic (8.3.1) & $\Emb(D^n,M)\simeq\Fr(TM)$ (\cref{thm:discs})\\3042classification of surfaces (9.3.11) & $\mathcal M_2\simeq\coprod B\Diff$, with \cref{thm:EE,thm:MW}\\3043\bottomrule3044\end{tabular}3045\caption{The main theorems of the source as statements about components.}\label{tab:pi0}3046\end{table}3047 3048\Cref{tab:pi0} is the summary of the book. Read down the left column and one has Hirsch's table of contents;3049read down the right and one has a good part of geometric topology since. In every row the left-hand3050statement is what one can prove with one parameter of genericity (a path between two choices) and the3051right-hand statement is what one gets by carrying all parameters. The source's method, described in the3052Preface, already contains the idea: an invariant is independent of choices because two choices are joined by3053a generic path. The higher statement is that the space of choices is contractible because spheres of choices3054bound generic discs.3055 3056\section{What is higher here}3057 3058In the volume on game theory the $\infty$-categorical structure was largely decorative, and in the volume3059on information theory it was concentrated in one place (the bar construction behind entropy). Here it is3060needed in at least four places where no $1$-categorical substitute exists.3061 3062First, derived fibre products. The last row of \cref{tab:four} has no solution in any ordinary category of3063ringed spaces, and \cref{thm:CS} characterizes the $\infty$-category that solves it.3064 3065Second, moduli. The stack $B\GL_k$ and the stack of manifold bundles are not sheaves of sets, and their shapes3066$B\Or(k)$ and $\coprod B\Diff(W)$ are not discrete.3067 3068Third, composition of cobordisms, which is associative only up to the contractible choices of3069\cref{ch:tubular}; the cobordism category is an $\infty$-category in an unavoidable way, and its homotopy type3070is the content of \cref{thm:GMTW}.3071 3072Fourth, embedding spaces, whose description in \cref{thm:weiss,thm:BdBW} is by homotopy limits over3073categories of discs.3074 3075In each case the higher statement was a theorem of someone else. This book contributes arrangement: the3076observation that Hirsch's nine chapters are, in order, the $\pi_0$ of these theories, with the dictionary3077made explicit and the status of each translation marked.3078 3079\section{What the language did not do}3080 3081It proved no theorem of differential topology. It did not make the analytic inputs of \cref{prin:inputs}3082easier. It did not internalize genericity: the Baire property of the strong topology has no formulation in3083terms of the sheaf $[M,N]$ that we know of (\cref{rem:nobaire}), and the density theorems are stated here3084exactly as in the source.3085 3086It is blind, at the level of generality used here, to the phenomena that distinguish the smooth category3087from the topological one. Exotic spheres appear in \cref{thm:spheres} as $\pi_0\Diff(S^n)$ and nowhere else;3088nothing in the categorical formalism predicts them. The same formalism applied to topological or3089piecewise-linear manifolds gives the same tables with different entries, and the differences among the3090entries are the hard part of the subject.3091 3092And several of the statements quoted are much harder than the theorems they generalize. One should not3093come away thinking that \cref{thm:GMTW} is a reformulation of Thom's theorem. It contains Thom's theorem as3094its statement about components, and its proof uses new ideas.3095 3096\section{Directions}3097 3098\begin{problem}[Internal genericity]\label{prob:baire}\Spg3099Find a property of a subobject $A\subset[M,N]$ of the mapping object, stated in terms of plots, that is3100implied by residuality in the strong topology, implies the conclusion of the parametric transversality