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 1281–1380 of 10051. Source text is displayed literally.
1281simplicial approximation of the identity $|\sdv K|\to|K|$.1282\end{proposition}1283 1284\noindent The counts in (b) are $2,6,24,120$. \chk{C5} Statement (c) is how \Cref{ch:complexes} embedded1285complexes in simplicial sets: an unordered complex becomes ordered after one subdivision. For (d), the map1286sending $\sigma$ to its last vertex is a chain map $C_*(\sdv K)\to C_*(K)$ and is onto in top homology for the1287torus, the projective plane and the Klein bottle. \chk{I2} On simplicial sets subdivision has a right adjoint1288$\operatorname{Ex}$, and iterating it gives Kan's functorial replacement of a simplicial set by one in which1289every horn has a filler \cite{Kan57}: the passage from complexes to homotopy types.1290 1291\section{The cone is a contraction}1292 1293\begin{proposition}[Conic construction]\label{prop:cone-op}1294\Scl[\src{I,\,\S8}] On the cone $CK$ with apex $a$ define $D[v_0,\dots,v_r]=(-1)^{r+1}[v_0,\dots,v_r,a]$ for1295simplices not containing $a$, and $D=0$ on the others. Then1296\[ \pd D+D\pd=\id-\varepsilon, \]1297where $\varepsilon$ is zero in positive degrees and sends a $0$-chain $c$ to (sum of coefficients of $c$) times1298$a$.1299\end{proposition}1300 1301\noindent \chk{I3} This is \Cref{prop:cone}(a) with the reason attached: a cone is acyclic because the identity is1302chain homotopic to the projection on a point. Simplices are cones. So any two natural chain maps defined on1303simplices that agree in degree zero are chain homotopic, by building the homotopy one dimension at a time and1304filling each step with $D$. This argument, the method of acyclic models \cite{EiMa53}, is what the source carries1305out by hand for the subdivision operator.1306 1307\begin{theorem}[Invariance under subdivision]\label{thm:sdinv}1308\Scl[\src{I,\,\S11, Thm.~15}] There is a natural chain map $C_*(K)\to C_*(\sdv K)$, sending a simplex to the sum of1309its pieces with their induced orientations. It is a chain homotopy inverse of the map of1310\Cref{prop:sd}\emph{(d)}. Hence $H_*(\sdv K)\cong H_*(K)$.1311\end{theorem}1312 1313\noindent The subdivided projective plane has $31$ vertices, $90$ edges and $60$ triangles, and homology1314$\Z,\Z/2,0$; likewise for the torus and the Klein bottle. \chk{I1}1315 1316\section{The theorem}1317 1318\begin{theorem}[Invariance]\label{thm:invariance}1319\Scl[\src{I,\,\S12, Thms.~16--17}] A continuous map $f\colon|K|\to|L|$ induces homomorphisms1320$f_*\colon H_r(K)\to H_r(L)$, defined by any simplicial approximation on any sufficiently fine subdivision,1321functorially. Homeomorphic polyhedra have isomorphic homology groups.1322\end{theorem}1323 1324\[\begin{tikzcd}[column sep=large]1325\SCx\arrow[r,"|\cdot|"]\arrow[dr,"H_*"'] & \cat{Pol}\arrow[d,dashed,"H_*"]\arrow[r] & \operatorname{Ho}(\cat{Pol})\arrow[dl,dashed]\\1326 & \cat{grAb} &1327\end{tikzcd}\]1328 1329\noindent The theorem is the left dashed arrow: homology of complexes factors through the category $\cat{Pol}$ of1330polyhedra and continuous maps. The right dashed arrow, through the homotopy category, is \Cref{ch:homotopy}. The1331modern route to both introduces the singular complex $\Sing X$, whose $n$-simplices are all continuous maps1332$|\Delta^n|\to X$; $\Sing$ is right adjoint to realization, $H_*(\Sing X)$ is defined for every space with no1333choices, and the content of \Cref{thm:invariance} becomes the statement that the unit $K\to\Sing|K|$ induces an1334isomorphism on homology. The three constructions of this chapter are exactly what that proof uses: approximation1335to make singular simplices simplicial, subdivision to make them small, and the cone to compare.1336 1337\section*{Notes}1338The source's order is: approximation, cone, subdivision, Sperner's lemma, invariance; we have moved Sperner's1339lemma to the next chapter. For acyclic models see \cite{EiMa53}; for $\operatorname{Ex}^\infty$, \cite{Kan57,GoJa99}.1340The estimate of the number of subdivisions is implicit in every proof of \Cref{thm:approx} and is used in1341\cite{ADTML} to bound the size of a piecewise linear network realizing a homotopy class.1342 1343\section*{Exercises}1344\begin{exercise}\label{exer:mesh}1345Show that the diameter of a simplex of the barycentric subdivision of an $r$-simplex of diameter $d$ is at most1346$\frac r{r+1}d$. How many subdivisions of the triangle of \chk{I4} guarantee mesh below $0.01$ by this bound, and1347how many triangles is that?1348\end{exercise}1349\begin{exercise}\label{exer:oct}1350Explain the degree $2$ found for the octagon: list the images of the eight vertices under the rule for $d=3$ and1351count the signed number of times the edge from the last third to the first is covered.1352\end{exercise}1353\begin{exercise}\label{exer:coneop}1354Verify $\pd D+D\pd=\id$ on a $1$-simplex $[v_0,v_1]$ of the base of a cone.1355\end{exercise}1356\begin{exercise}\label{exer:sdf}1357Let $K$ have $f$-vector $(\alpha_0,\alpha_1,\alpha_2)$. Find the $f$-vector of $\sdv K$ and check that the Euler1358characteristic is unchanged. Apply it to the projective plane.1359\end{exercise}1360%% ---- end ch/03-invariance.tex ----1361\part{Maps}1362%% ---- begin ch/04-sperner.tex ----1363\chapter{Sperner, Brouwer, Dimension}\label{ch:sperner}1364 1365In the middle of its invariance proof the source stops for a digression: a combinatorial lemma about labellings1366of a subdivided simplex, and two consequences, the fixed point theorem and the invariance of dimension1367\src{I,\,\S3, \S10}. The lemma is a degree computation, and the consequences are what one gets from knowing that a1368functor does not vanish.1369 1370\section{The lemma}1371 1372Let $K$ be a subdivision of an $n$-simplex $\Delta$ with vertices $e_0,\dots,e_n$. A \term{Sperner labelling}1373assigns to each vertex of $K$ one of the labels $0,\dots,n$ so that a vertex lying in the face spanned by1374$e_{i_0},\dots,e_{i_k}$ gets one of the labels $i_0,\dots,i_k$. A cell of $K$ is \term{complete} if its vertices1375carry all $n+1$ labels.1376 1377\begin{theorem}[Sperner]\label{thm:sperner}1378\Scl[\src{I,\,\S10}] The number of complete cells is odd. Counted with the sign $+1$ if the labelling preserves the1379orientation of the cell and $-1$ if it reverses it, the number of complete cells is exactly $1$.1380\end{theorem}