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 1781–1880 of 10051. Source text is displayed literally.
1781\draw[thin,fill=fwd!45] (3.897,-4.25)--(3.464,-4)--(3.031,-4.25)--(3.031,-4.75)--(3.464,-5)--(3.897,-4.75)--cycle;1782\draw[thin,fill=fwd!45] (4.763,-4.25)--(4.33,-4)--(3.897,-4.25)--(3.897,-4.75)--(4.33,-5)--(4.763,-4.75)--cycle;1783\draw[thin,fill=fwd!45] (5.629,-4.25)--(5.196,-4)--(4.763,-4.25)--(4.763,-4.75)--(5.196,-5)--(5.629,-4.75)--cycle;1784\draw[thin,fill=fwd!45] (6.495,-4.25)--(6.062,-4)--(5.629,-4.25)--(5.629,-4.75)--(6.062,-5)--(6.495,-4.75)--cycle;1785\draw[thin,fill=bwd!18] (7.361,-4.25)--(6.928,-4)--(6.495,-4.25)--(6.495,-4.75)--(6.928,-5)--(7.361,-4.75)--cycle;1786\draw[thin,fill=fwd!45] (8.227,-4.25)--(7.794,-4)--(7.361,-4.25)--(7.361,-4.75)--(7.794,-5)--(8.227,-4.75)--cycle;1787\draw[line width=1.6pt,black] (3.464,-4.5)--(3.031,-3.75)--(2.598,-3)--(3.031,-2.25)--(3.464,-1.5)--(3.897,-0.75)--(4.763,-0.75)--(5.196,0);1788\end{tikzpicture}1789\figcap{A filled Hex board. The dark player joins the top side to the bottom, the light player the left to the right. Here the dark player has a chain, drawn; the other cannot then have one, and one of the two always does.}\label{fig:hex}1790\end{center}1791 1792\noindent The game of Hex gives a third form of the same statement. On a rhombic board of hexagons filled with1793two colours, exactly one of the two players has a chain joining their pair of opposite sides; that it cannot be1794both is the Jordan curve theorem in a simple case, and that it must be one is equivalent to Brouwer's theorem in1795the plane \cite{Gal79}. In $300$ random fillings of the $9\times9$ board there was always exactly one winner, the1796first player in $156$ of them. \chk{S5}1797 1798\section{Dimension}1799 1800The \term{order} of a finite cover is the largest number of its members with a common point, so the nerve of a1801cover of order $k$ has dimension $k-1$. A compact metric space has \term{dimension at most $n$} if it has covers1802by arbitrarily small closed sets of order at most $n+1$ \src{I,\,\S3}; equivalently, if it can be mapped to1803$n$-dimensional nerves with arbitrarily small fibres.1804 1805\begin{theorem}[Lebesgue; Brouwer]\label{thm:dimension}1806\Scl[\src{I,\,\S10, Thms.~13--14}] \emph{(a)} A cover of an $n$-simplex by closed sets, none of which meets all the1807facets, has order at least $n+1$. \emph{(b)} A polyhedron $|K|$ has dimension $\dim K$. In particular $\R^m$ and1808$\R^n$ are not homeomorphic for $m\ne n$.1809\end{theorem}1810\begin{proof}[Idea]1811(a) Assign to each set of the cover a vertex of the simplex opposite a facet it misses; a fine Sperner labelling1812compatible with this assignment has a complete cell, whose vertices lie in $n+1$ different sets, and a limit1813point lies in all of them. (b) The stars of the vertices of a fine subdivision, slightly shrunk, give small1814covers of order $n+1$; by (a) there are none of smaller order.1815\end{proof}1816 1817\begin{center}1818\begin{tikzpicture}[scale=0.62]1819\begin{scope}1820\foreach \j in {0,...,3} \foreach \i in {0,...,3} \draw[thick,fill=fwd!8] (\i,\j) rectangle (\i+1,\j+1);1821\fill[bwd] (2,2) circle (4pt);1822\node at (2,-0.6) {\small square grid: order $4$};1823\end{scope}1824\begin{scope}[xshift=7cm]1825\clip (0,0) rectangle (4,4);1826\foreach \j in {0,...,3} \foreach \i in {-1,...,4} {\pgfmathsetmacro{\o}{0.5*mod(\j,2)} \draw[thick,fill=fwd!8] (\i+\o,\j) rectangle (\i+\o+1,\j+1);}1827\fill[bwd] (2,2) circle (4pt); \fill[bwd] (1.5,1) circle (4pt);1828\end{scope}1829\draw[thick] (7,0) rectangle (11,4);1830\node at (9,-0.6) {\small brick pattern: order $3$};1831\end{tikzpicture}1832\figcap{Two covers of the plane by small closed squares. Four squares meet at each corner of the grid; shifting1833alternate rows leaves only triple points. By \Cref{thm:dimension} no cover of the plane by small sets does1834better: the plane has dimension two.}\label{fig:brick}1835\end{center}1836 1837\noindent \chk{S6} The theorem says that dimension is a property of the nerves that a space admits, and so a1838property of the space as a limit of polyhedra. The same idea gives the source's embedding theorem, that a compact1839metric space of dimension $r$ embeds in $\R^{2r+1}$ \src{I,\,\S3, Thm.~5}: approximate by an $r$-dimensional1840nerve, realize the nerve in general position by \Cref{thm:realization}, and pass to the limit.1841 1842\section*{Notes}1843Sperner's lemma dates from 1928; the signed form is in \cite{Bro90}. The source derives Brouwer's theorem and the1844invariance of dimension from it and returns to fixed points with homological tools in its last section; see1845\Cref{ch:traces}. The equivalence with Hex is Gale's \cite{Gal79}. For the complexity of finding a complete cell,1846and of computing equilibria of games by the same method, see \cite{Pap94} and \cite{GIHC}; consequences for1847learning systems are drawn in \cite{ADTML}.1848 1849\section*{Exercises}1850\begin{exercise}\label{exer:sp1}1851Prove Sperner's lemma in dimension one directly, in its signed form.1852\end{exercise}1853\begin{exercise}\label{exer:door}1854In dimension two, show that the number of complete cells has the parity of the number of boundary edges labelled1855$01$, by counting pairs (cell, edge of the cell labelled $01$) in two ways.1856\end{exercise}1857\begin{exercise}\label{exer:retr}1858Deduce \Cref{thm:noretract}\emph{(b)} from \emph{(a)}.1859\end{exercise}1860\begin{exercise}\label{exer:order}1861A cover of the interval $[0,1]$ by closed intervals of length less than $1$ has order at least two. Give a cover1862of order exactly two, and a cover of the square $[0,1]^2$ by sets of diameter less than $\frac12$ of order three.1863\end{exercise}1864%% ---- end ch/04-sperner.tex ----1865%% ---- begin ch/05-homotopy.tex ----1866\chapter{Homotopy and the Cylinder}\label{ch:homotopy}1867 1868Two maps are homotopic if one can be deformed into the other. The third chapter of the source shows that1869homology cannot tell homotopic maps apart, by one construction: a triangulation of the cylinder $K\times[0,1]$1870and a chain homotopy read off from it. This chapter records that construction, says what kind of structure a1871chain homotopy is, and ends with the map that homology cannot see at all.1872 1873\section{The cylindric construction}1874 1875Order the vertices of $K$ and write $v$ and $v'$ for the bottom and top copies of a vertex in $K\times[0,1]$. By1876\Cref{thm:sset}\emph{(c)} the prism over a simplex $[v_0,\dots,v_r]$ is the union of the $r+1$ simplices1877$[v_0,\dots,v_i,v_i',\dots,v_r']$.1878 1879\begin{center}1880\begin{tikzpicture}[scale=1.0]