Independent scholarship · Open accessThe monthly digest

Indexed text attachment

algebraic-and-differential-topology-in-higher-categories.tex

Source for Algebraic and Differential Topology in Higher Categories

View paper · Download original file

Lines 301–400 of 10051. Source text is displayed literally.

301companion script and drew the figures. The bibliography was assembled from memory, except for the entries marked as302checked, and should be verified before it is relied on.}303\clearpage304 305\chapter*{Preface}306The third volume of Pontryagin's \emph{Selected Works} \cite{Pon86} contains two short monographs. The first,307\emph{Foundations of Algebraic Topology}, defines the homology groups of a polyhedron, proves that they do not308depend on the triangulation, and ends with the fixed point theorem of Lefschetz and Hopf. The second,309\emph{Smooth Manifolds and Their Applications in Homotopy Theory}, computes the first homotopy groups of310spheres that are not given by homology, by representing a map of spheres as a manifold with a field of normal311frames. They have no notation in common.312 313This volume reads them as one theory with two coefficient rings.314 315\begin{itemize}[leftmargin=1.4em]316\item \emph{Chains are a free module.} The chain complex of a space is the free module over the integers on its317homotopy type, and homology is forced by that (\Cref{thm:leftadjoint}). The constructions of the first monograph,318the polyhedron, subdivision, the cone, the cylinder, are a colimit, a functor, a contraction and a product319(\Cref{ch:complexes,ch:invariance,ch:homotopy}).320\item \emph{A fixed point theorem is a statement about traces.} The Lefschetz number is the trace of a321dualizable object, and Sperner's lemma is a degree computed by Stokes' theorem322(\Cref{thm:sperner,thm:doldpuppe}).323\item \emph{Framed manifolds are a free module too.} Pontryagin's construction identifies framed manifolds up to324cobordism with the homotopy of the sphere spectrum, which is the free commutative group, in the homotopical325sense, on one point (\Cref{thm:pontryagin,thm:sphere}).326\item \emph{The first three groups.} A signed count of points, the parity of the turns of a frame along a327circle, and the Arf invariant of a quadratic form on a surface (\Cref{thm:hopfdeg,thm:first,thm:second}).328\item \emph{The comparison.} The first monograph is the second tensored with the integers; what is lost is329everything in positive degrees (\Cref{thm:tworings}).330\end{itemize}331 332Each qualitative statement has a computed companion: homology with torsion by Smith normal form, signed counts333of Sperner cells, Lefschetz numbers of all automorphisms of a small torus, linking numbers of Hopf fibres by334Gauss's integral, lifts of loops of rotations to quaternions, counts of quadratic forms by Arf invariant. The335companion script performs $63$ checks. There are thirty-one figures, about half of them computed.336 337Nothing here is a new theorem, and the long proofs are indicated, not given. What is higher, and how much, is338assessed in \Cref{ch:coda}: of the volumes of this series, this is the one whose central object, the sphere339spectrum, cannot be stated without higher categories. A companion paper \cite{ADTML} applies the material to340machine learning.341 342The volume belongs to a series with \cite{GIHC}, \cite{ESCHC}, \cite{DTHC}, \cite{EGHC} and \cite{PDEHC}. It343overlaps with \cite{DTHC}, which treats the Pontryagin--Thom construction for general tangential structures;344here the framed case is taken slowly and all the way to the first computations.345 346\tableofcontents347%% ---- end front.tex ----348\mainmatter349\part{Complexes}350%% ---- begin ch/01-complexes.tex ----351\chapter{Complexes as Presheaves}\label{ch:complexes}352 353The first monograph of the source builds everything from one kind of object, the finite simplicial complex, and354one kind of map, the simplicial map. This chapter says what category that is, what is wrong with it, and what the355standard repair is. The polyhedron of a complex turns out to be a colimit, the nerve of a cover a limit of sorts,356and the awkwardness of products the reason simplicial sets were invented.357 358\section{Simplices, complexes, polyhedra}359 360\begin{definition}\label{def:complex}361An \term{abstract simplicial complex} $K$ is a finite set $V$ of \emph{vertices} with a family of non-empty362subsets, the \emph{simplices}, closed under passing to non-empty subsets and containing every singleton. A363simplex with $r+1$ vertices has dimension $r$. A \term{simplicial map} $K\to L$ is a map of vertex sets taking364simplices to simplices; it need not be injective on a simplex. This is the category $\SCx$.365\end{definition}366 367\noindent A simplex $\sigma$ has a geometric realization $|\sigma|$, the convex hull of its vertices placed in368general position, and a face $\tau\subseteq\sigma$ gives an inclusion $|\tau|\subseteq|\sigma|$. The source defines369a complex as a set of simplices in a Euclidean space that meet properly, and its polyhedron as their union370\src{I,\,\S2}. Categorically:371 372\begin{proposition}[The polyhedron is a colimit]\label{prop:colim}373\Srf[\cite{Spa66}] Regard $K$ as a poset under inclusion. Then374\[ |K|\;=\;\colim_{\sigma\in K}|\sigma| \]375in topological spaces, and $K\mapsto|K|$ is a functor $\SCx\to\Top$: a simplicial map is realized by the map that is376affine on each simplex.377\end{proposition}378 379\noindent The colimit says exactly what ``meeting properly'' says: two simplices are glued along their common380face and nowhere else. That the abstract colimit can be drawn in a Euclidean space of small dimension is the first381theorem of the source.382 383\begin{theorem}[Realization]\label{thm:realization}384\Scl[\src{I,\,\S1--2, Thms.~1--3}] A complex of dimension $r$ has a realization by straight simplices in385$\R^{2r+1}$: place the vertices in general position, so that every $2r+2$ of them are affinely independent.386\end{theorem}387\begin{proof}388Two simplices have at most $2r+2$ vertices between them. If these are affinely independent, the convex hulls of389two subsets of them meet in the convex hull of the common part.390\end{proof}391 392\noindent Points on the moment curve $t\mapsto(t,t^2,\dots,t^{n})$ are in general position in $\R^n$, by the393Vandermonde determinant; for seven points and $n=5$ all seven six-point subsets are affinely independent, so394every $2$-complex on seven vertices, the torus of \Cref{fig:torus7} included, sits in $\R^5$. \chk{C2} The number395$2r+1$ returns in \Cref{ch:position} as Whitney's embedding dimension, for the same reason: two $r$-dimensional396things in a space of dimension $2r+1$ do not meet in general.397 398\section{What is wrong with complexes: products}399 400The category $\SCx$ has products, but they are the wrong ones. A vertex of $K\times L$ is a pair of vertices, and a