1: 

2: \documentclass[10pt]{amsart}

3: 

4: \usepackage{url,graphicx}

5: \usepackage[dvips]{epsfig}

6: \usepackage{latexsym,color}

7: \usepackage{amscd,amssymb,verbatim}

8: 

9: \setlength{\marginparsep}{5pt}

10: 

11: \sloppy

12: \def\pstexInput#1{%

13:   \begin{picture}(0,0)%

14:     \special{psfile=#1.pstex}%

15:   \end{picture}%

16:   \input{#1.pstex_t}%

17:  }

18: 

19: \newcommand{\mc}{\marginpar{*}} 

20: \newcommand{\al}{{\alpha}}

21: \newcommand{\aut}{\text{\rm Aut}\,}

22: 

23: \newcommand{\bbr}{\big{(}\big{)}}

24: \newcommand{\bd}{{\textrm{Bd}\,}}

25: \newcommand{\be}{\begin{enumerate}}

26: \newcommand{\bl}{{\rm{Bl}\,}}

27: \newcommand{\bn}{\cb_n}

28: \newcommand{\bo}{\partial}

29: \newcommand{\bu}{\bullet}

30: 

31: \newcommand{\cat}{\text{\bf Cat}}

32: \newcommand{\ca}{{\mathcal A}}

33: \newcommand{\cb}{{\mathcal B}}

34: \newcommand{\cc}{{\mathcal C}}

35: \newcommand{\cd}{{\mathcal D}}

36: \newcommand{\ce}{{\mathcal E}}

37: \newcommand{\cf}{{\mathcal F}}

38: \newcommand{\cg}{{\mathcal G}}

39: 

40: \newcommand{\chom}{\text{\tt Hom}_{\,0}}

41: \newcommand{\ck}{{\mathcal K}}

42: \newcommand{\cl}{{\mathcal L}}

43: \newcommand{\cliq}{{\text Cliq}\,}

44: \newcommand{\cm}{{\mathcal M}}

45: \newcommand{\cn}{{\mathcal N}}

46: \newcommand{\co}{{\mathcal O}}

47: \newcommand{\colim}{\text{\bf colim}\,}

48: \newcommand{\com}{\complement}

49: \newcommand{\cp}{{\mathcal P}}

50: \newcommand{\crr}{{\mathcal R}}

51: \newcommand{\cs}{{\mathcal S}}

52: \newcommand{\ct}{{\mathcal T}}

53: 

54: \newcommand{\da}{\Delta}

55: \newcommand{\dar}{\downarrow}

56: \newcommand{\dc}{{\mathbb C}}

57: \newcommand{\dr}{{\mathbb R}}

58: \newcommand{\dq}{{\mathbb Q}}

59: \newcommand{\dz}{{\mathbb Z}}

60: 

61: \newcommand{\ee}{\end{enumerate}}

62: 

63: \newcommand{\follows}{\Rightarrow}

64: \newcommand{\fns}{\footnotesize}

65: 

66: \newcommand{\ga}{\Gamma}

67: 

68: \newcommand{\hra}{\hookrightarrow}

69: 

70: \newcommand{\im}{{\text{\rm Im}}\,}

71: \newcommand{\ind}{{\text{\rm Ind}}\,}

72: 

73: \newcommand{\U}{\text{\bf U}}

74: \newcommand{\GL}{\text{\bf GL}}

75: \newcommand{\grp}{\text{\bf Grp}}

76: 

77: \newcommand{\id}{\text{\rm id}}

78: 

79: \newcommand{\la}{\leftarrow}

80: \newcommand{\lam}{\lambda}

81: \newcommand{\Lam}{\Lambda}

82: \newcommand{\lb}{\linebreak}

83: \newcommand{\lk}{\text{lk}\,}

84: \newcommand{\lla}{\longleftarrow}

85: \newcommand{\lmu}{(\lam,\mu)}

86: \newcommand{\Lmu}{(\Lam,\mu)}

87: \newcommand{\lmt}{\longmapsto}

88: \newcommand{\lra}{\longrightarrow}

89: \newcommand{\lr}{\leftrightarrow}

90: 

91: \newcommand{\nin}{\noindent}

92: \newcommand{\nl}{\newline}

93: \newcommand{\nlb}{\nolinebreak}

94: \newcommand{\npb}{\nopagebreak}

95: 

96: \newcommand{\ovr}[1]{\overline{#1}}

97: 

98: \newcommand{\pl}{\Pi_\lambda}

99: \newcommand{\pr}{\noindent{\bf Proof. }}

100: 

101: \newcommand{\ra}{\rightarrow}

102: \newcommand{\Ra}{\Rightarrow}

103: \newcommand{\rk}{{\text{rk}}}

104: \newcommand{\rp}{{\mathbb R}{\mathbb P}}

105: 

106: \newcommand{\qd}{$\Box$}

107: 

108: \newcommand{\sd}{\text{sd}}

109: \newcommand{\sgn}{\text{sgn}\,}

110: \newcommand{\sig}{\sigma}

111: \newcommand{\sla}{\Sigma_\lam}

112: \newcommand{\sm}{\setminus}

113: \newcommand{\Span}{\text{\rm span}}

114: \newcommand{\sr}{\stackrel}

115: \newcommand{\st}{\text{\rm St}}

116: \newcommand{\supp}{\text{\rm supp}\,}

117: \newcommand{\susp}{\text{\rm susp}}

118: 

119: \newcommand{\tb}{\tilde\beta}

120: \newcommand{\tbe}{(to be extended) }

121: \newcommand{\tf}{{\tilde f}}

122: 

123: \newcommand{\thom}{\text{\tt Hom}\,}

124: \newcommand{\thomp}{\text{\tt Hom}_+}

125: \newcommand{\ti}{\tilde }

126: \newcommand{\tn}{\text{\tt N}}

127: \newcommand{\tp}{\text{\bf Top}}

128: 

129: \newcommand{\vt}{\vartheta}

130: 

131: \newcommand{\wti}{\widetilde }

132: \newcommand{\wsl}{{\wti\Sigma}_\lam}

133: 

134: \newcommand{\xlm}{X_{\lambda,\mu}}

135: \newcommand{\xLm}{X_{\Lambda,\mu}}

136: 

137: \newcommand{\zz}{{{\mathbb Z}_2}}

138: \newcommand{\zp}{{{\mathbb Z}_p}}

139: 

140: \newtheorem{thm}{Theorem}[section]

141: \newtheorem{df}  [thm]{Definition}

142: \newtheorem{mthm}[thm]{Main Theorem}

143: \newtheorem{lm}  [thm]{Lemma}

144: \newtheorem{crl} [thm]{Corollary}

145: \newtheorem{prop}[thm]{Proposition}

146: \newtheorem{conj}[thm]{Conjecture}

147: \newtheorem{rem} [thm]{Remark}

148: \newtheorem{prob}[thm]{Problem}

149: 

150: \numberwithin{equation}{section}

151: \numberwithin{figure}{section}

152: \numberwithin{table}{section}

153: 

154: \newenvironment{remrm}{\begin{rem} \rm}{\end{rem}}

155: 

156: \begin{document}

157: 

158: \title {Proof of the Lov\'asz Conjecture}

159: 

160: \author{Eric Babson and Dmitry N. Kozlov}

161: 

162: \date{\noindent

163:   \today\\[0.05cm]

164: \hskip15pt 

165: MSC 2000 Classification:

166:   primary 05C15, 

167: secondary 57M15. 

168: \\ Keywords: graphs, chromatic number, graph homomorphisms, 

169: Lov\'asz conjecture, spectral sequences, Stiefel-Whitney

170: characteristic classes, $\thom$ complexes, Kneser conjecture.  }

171: 

172: \address{Department of Mathematics, University of Washington, Seattle,

173:   U.S.A.} \email{babson@math.washington.edu}

174: 

175: \address{Institute of Theoretical Computer Science, Eidgen\"ossische 

176: Technische Hochschule, Z\"urich, Switzerland}

177: \email{dkozlov@inf.ethz.ch}

178: 

179: 

180: 

181: \begin{abstract}

182: To any two graphs $G$ and $H$ one can associate a~cell complex

183: $\thom(G,H)$ by taking all graph multihomorphisms from $G$ to $H$ as

184: cells.

185: 

186: In this paper we prove the Lov\'asz Conjecture which states that

187: \begin{quote}

188: {\it if $\thom(C_{2r+1},G)$ is $k$-connected, then $\chi(G)\geq k+4$,}

189: \end{quote}

190: where $r,k\in\dz$, $r\geq 1$, $k\geq -1$, and $C_{2r+1}$ denotes the

191: cycle with $2r+1$ vertices.

192: 

193: The proof requires analysis of the complexes $\thom(C_{2r+1},K_n)$.

194: For even $n$, the obstructions to graph colorings are provided by the

195: presence of torsion in $H^*(\thom(C_{2r+1},K_n);\dz)$. For odd $n$,

196: the obstructions are expressed as vanishing of certain powers of

197: Stiefel-Whitney characteristic classes of $\thom(C_{2r+1},K_n)$, where

198: the latter are viewed as $\zz$-spaces with the involution induced by

199: the reflection of $C_{2r+1}$.

200: \end{abstract}

201: 

202: \maketitle

203: 

204: 

205: \section{Introduction}\label{sect_intr}

206: 

207: The main idea of this paper is to look for obstructions to graph

208: colorings in the following indirect way: take a~graph, associate to it

209: a~topological space, and then look for obstructions to colorings of

210: the graph by studying the algebraic invariants of this space.

211: 

212: The construction of such a space, which is of interest here, has been

213: suggested by L.\ Lov\'asz.  The obtained complex $\thom(G,H)$ depends

214: on two graph parameters. The algebraic invariants of this space, which

215: we proceed to study, are its cohomology groups, and, when it can be

216: viewed as a~$\zz$-space, its Stiefel-Whitney characteristic classes.

217: 

218: \subsection{The vertex colorings and the category of Graphs.} 

219: $\,$ \vspace{5pt}

220: 

221: All graphs in this paper are undirected. The following definition is

222: a~key in turning the set of all undirected graphs into a~category.

223: 

224: \begin{df}\label{dfgrhom}

225:   For two graphs $G$ and $H$, a {\bf graph homomorphism}

226:   from $G$ to $H$ is a~map $\phi:V(G)\rightarrow V(H)$, such that if

227:   $(x,y)\in E(G)$, then $(\phi(x),\phi(y))\in E(H)$.

228: \end{df}

229: Here, $V(G)$ denotes the set of vertices of $G$, and $E(G)$ denotes

230: the set of its edges.

231: 

232: For a graph $G$ the {\it vertex coloring} is an assignment of colors

233: to vertices such that no two vertices which are connected by an~edge

234: get the same color. The minimal needed number of colors is denoted by

235: $\chi(G)$, and is called the {\it chromatic number} of~$G$.

236: 

237: Deciding whether or not there exists a~graph homomorphism between two

238: graphs is in general at least as difficult as bounding the chromatic

239: numbers of graphs because of the following observation: a vertex

240: coloring of $G$ with $n$ colors is the same as a graph homomorphism

241: from $G$ to the complete graph on $n$ vertices $K_n$. Because of this,

242: one can also think of graph homomorphisms from $G$ to $H$ as vertex

243: colorings of $G$ with colors from $V(H)$ subject to the natural

244: condition.

245: 

246: Since an identity map is a~graph homomorphism, and a~composition of

247: two graph homomorphisms is again a~graph homomorphism, we can consider

248: the category {\bf Graphs} whose objects are all undirected graphs, and

249: morphisms are all the graph homomorphisms.

250: 

251: We denote the set of all graph homomorphisms from $G$ to $H$ by

252: $\chom(G,H)$. Lov\'asz has suggested the following way of turning this

253: set into a~topological space.

254: 

255: \begin{df} \label{dfhom}

256: We define $\thom(G,H)$ to be a~polyhedral complex whose cells are

257: indexed by all functions $\eta:V(G)\rightarrow

258: 2^{V(H)}\setminus\{\emptyset\}$, such that if $(x,y)\in E(G)$, for any

259: $\tilde x\in\eta(x)$ and $\tilde y\in\eta(y)$ we have $(\tilde

260: x,\tilde y)\in E(H)$.

261: 

262:   The closure of a~cell $\eta$ consists of all cells indexed by

263:   $\ti\eta:V(G)\rightarrow 2^{V(H)}\setminus\{\emptyset\}$, which

264:   satisfy $\ti\eta(v)\subseteq\eta(v)$, for all $v\in V(G)$.

265: \end{df}

266: \nin We think of a cell in $\thom(G,H)$ as a~collection of non-empty

267: lists of vertices of $H$, one for each vertex of $G$, with the

268: condition that any choice of one vertex from each list will yield

269: a~graph homomorphism from $G$ to~$H$. A~geometric realization of

270: $\thom(G,H)$ can be described as follows: number the vertices of $G$

271: with $1,\dots,|V(G)|$, the cell indexed with $\eta:V(G)\rightarrow

272: 2^{V(H)}\setminus\{\emptyset\}$ is realized as a~direct product of

273: simplices $\da^1,\dots,\da^{|V(G)|}$, where $\da^i$ has $|\eta(i)|$

274: vertices and is realized as the standard simplex in ${\mathbb

275:   R}^{|\eta(i)|}$. In particular, the set of vertices of $\thom(G,H)$

276: is precisely $\chom(G,H)$.

277: 

278: The barycentric subdivision of $\thom(G,H)$ is isomorphic as

279: a~simplicial complex to the geometric realization of its face poset.

280: So, alternatively, it could be described by first defining a~poset of

281: all $\eta$ satisfying conditions of Definition~\ref{dfhom}, with

282: $\eta\geq\ti\eta$ iff $\eta(v)\supseteq\ti\eta(v)$, for all $v\in

283: V(G)$, and then taking the geometric realization.

284: 

285:  The $\thom$ complexes are functorial in the following sense:

286:  $\thom(H,-)$ is a covariant, while $\thom(-,H)$ is

287: a~contravariant functor from {\bf Graphs} to {\bf Top}. If

288: $\phi\in\chom(G,G')$, then we shall denote the induced cellular

289: maps as $\phi^H:\thom(H,G)\ra\thom(H,G')$ and

290: $\phi_H:\thom(G',H)\ra\thom(G,H)$.

291: 

292: \subsection{The statement of the Lov\'asz conjecture.} $\,$ \vspace{5pt}

293: 

294: Lov\'asz has stated the~following conjecture, which we prove in this

295: paper.

296: 

297: \begin{thm} {\bf (Lov\'asz Conjecture).} \label{loconj}

298:   {\it Let $G$ be a graph, such that $\thom(C_{2r+1},G)$ is

299:     $k$-connected for some $r,k\in\dz$, $r\geq 1$, $k\geq -1$, then

300:     $\chi(G)\geq k+4$.}

301: \end{thm}

302: 

303: \nin Here $C_{2r+1}$ is a cycle with $2r+1$ vertices:

304: $V(C_{2r+1})=\dz_{2r+1}$,

305: $E(C_{2r+1})=\{(x,x+1),(x+1,x)\,|\,x\in\dz_{2r+1}\}$. 

306: 

307: The motivation for this conjecture stems from the following theorem

308: which Lov\'asz has proved in 1978.

309: 

310: \begin{thm} \label{lothm}

311:   {\rm (Lov\'asz, \cite{Lo}).}  {\it Let $H$ be a graph, such that

312:     $\thom(K_2,H)$ is $k$-connected for some $k\in\dz$, $k\geq -1$,

313:     then $\chi(H)\geq k+3$.}

314: \end{thm}

315: 

316: One corollary of Theorem~\ref{lothm} is the Kneser conjecture from

317: 1955, see~\cite{Knes}.

318: 

319: \begin{rem} The actual theorem from \cite{Lo} is stated using the

320: neighborhood complexes $\cn(H)$. However, it is well known that

321: $\cn(H)$ is homotopy equivalent to $\thom(K_2,H)$ for any graph~$H$,

322: see, e.g., \cite{BK2} for an argument. In fact, these two spaces are

323: known to be simple-homotopy equivalent, see \cite{K5}.

324: \end{rem}

325: 

326: We note here that Theorem~\ref{loconj} is trivially true for $k=-1$:

327: $\thom(C_{2r+1},G)$ is $(-1)$-connected if and only if it is

328: non-empty, and since there are no homomorphisms from odd cycles to

329: bipartite graphs, we conclude that $\chi(G)\geq 3$. It is also not

330: difficult to show that Theorem~\ref{loconj} holds for $k=0$ by using

331: the winding number. A short argument for a~more general statement can

332: be found in subsection~\ref{ss2.2}.

333: 

334: 

335: 

336: \subsection{Plan of the paper.} $\,$

337: \vspace{5pt}

338: 

339: In Section 2, we formulate the main theorems and describe the general

340: framework of finding obstructions to graph colorings via vanishing of

341: powers of Stiefel-Whitney characteristic classes.

342: 

343: In Section 3, we introduce auxiliary simplicial complexes, which we

344: call $\thomp(-,-)$. For any two graphs $G$ and $H$, there is

345: a~canonical support map $\supp:\thomp(G,H)\ra\Delta_{|V(G)|-1}$, and

346: the preimage of the barycenter is precisely $\thom(G,H)$.  This allows

347: us to set up a~useful spectral sequence, filtering by the preimages of

348: the $i$-skeleta.

349: 

350: In Section 4, we compute the cohomology groups

351: $H^*(\thom(C_{2r+1},K_n);\dz)$ up to dimension $n-2$, and we find the

352: $\zz$-action on these groups. These computations allow us to prove the

353: Lov\'asz conjecture for the case of odd~$k$, $k\geq 1$.

354: 

355: In Section 5, we study a~different spectral sequence, this one

356: converging to $H^*(\thom(C_{2r+1},K_n)/\zz;\zz)$. Understanding

357: certain entries and differentials leads to the proof of the Lov\'asz

358: conjecture for the case of even~$k$ as well.

359: 

360: The results of this paper were announced in \cite{BK1}, where no

361: complete proofs were given. The reader is referred to \cite{IAS} for

362: a~survey on $\thom$ complexes, which also includes a lot of background

363: material which is omitted in this paper.

364: \vspace{5pt}

365: 

366: \nin {\bf Acknowledgments.} The second author acknowledges support by

367: the University of Washington, Seattle, the Swiss National Science

368: Foundation, and the Swedish National Research Council.

369: 

370: 

371: 

372: \section{The idea of the proof of the Lov\'asz conjecture.}

373: 

374: \subsection{Group actions on $\thom$ complexes and Stiefel-Whitney 

375: classes.}

376: $\,$  \vspace{5pt} \label{ss2.1}

377: 

378: Consider an arbitrary CW complex $X$ on which a~finite group $\Gamma$

379: acts freely. By the general theory of principal $\Gamma$-bundles,

380: there exists a~$\Gamma$-equivariant map $\tilde w:X\ra{\bf E}\Gamma$,

381: and the induced map $w:X/\Gamma\ra{\bf B}\Gamma={\bf E}\Gamma/\Gamma$

382: is unique up to homotopy.

383: 

384: Specifying $\Gamma=\zz$, we get a map $\tilde w:X\ra S^{\infty}={\bf

385: E}\zz$, where $\zz$ acts on $S^{\infty}$ by the antipodal map, and the

386: induced map $w:X/\zz\ra{\mathbb R\mathbb P}^\infty={\bf B}\zz$.  We

387: denote the induced $\zz$-algebra homomorphism $H^*({\mathbb R \mathbb

388: P}^\infty;\zz)\ra H^*(X/\zz;\zz)$ by $w^*$. Let $z$ denote the

389: nontrivial cohomology class in $H^1({\mathbb R\mathbb P}^\infty;

390: \zz)$. Then $H^*({\mathbb R\mathbb P}^\infty;\zz) \simeq\zz[z]$ as

391: a~graded $\zz$-algebra, with $z$ having degree~1. We denote the image

392: $w^*(z)\in H^1(X/\zz;\zz)$ by $\varpi_1(X)$. This is the {\it first

393: Stiefel-Whitney class} of the $\zz$-space~X. Clearly,

394: $\varpi_1^k(X)=w^*(z^k)$, since $w^*$ is a~$\zz$-algebra homomorphism.

395: We will be mainly interested in the {\it height} of the

396: Stiefel-Whitney class, i.e., largest $k$, such that $\varpi_1^k(X)\neq

397: 0$; it was called cohomology co-index in~\cite{CF}.

398: 

399: Turning to graphs, let $G$ be a graph with $\zz$-action given by

400: $\phi:G\ra G$, $\phi\in\chom(G,G)$, such that $\phi$ flips an edge,

401: that is, there exist $a,b\in V(G)$, $a\neq b$, $(a,b)\in E(G)$, such

402: that $\phi(a)=b$ (which implies $\phi(b)=a$). For any graph $H$ we

403: have the induced $\zz$-action $\phi_H:\thom(G,H)\ra\thom(G,H)$. In

404: case $H$ has no loops, it follows from the fact that $\phi$ flips an

405: edge that this $\zz$-action is free.

406: 

407: Indeed, since $\phi_H$ is a~cellular map, if it fixes a~point from

408: some cell $\eta:V(G)\ra 2^{V(H)}\sm\{\emptyset\}$, then it maps $\eta$

409: onto itself. By definition, $\phi$ maps $\eta$ to $\eta\circ\phi$, so

410: this means that $\eta=\eta\circ\phi$. In particular,

411: $\eta(a)=\eta\circ\phi(a)=\eta(b)$. Since $\eta(a)\neq\emptyset$, we

412: can take $v\in V(H)$, such that $v\in\eta(a)$. Now, $(a,b)\in E(G)$,

413: but $(v,v)\notin E(H)$, since $H$ has no loops, which contradicts the

414: fact that $\eta\in\thom(G,H)$.

415: 

416: Therefore, in this situation, $\thom(G,-)$ is a~covariant functor from

417: the induced subcategory of {\bf Graphs}, consisting of all loopfree

418: graphs, to {\bf $\zz$-spaces} (the category whose objects are

419: $\zz$-spaces and morphisms are $\zz$-maps).

420: 

421: We order $V(C_{2r+1})$ by identifying it with $[1,2r+1]$ by the map

422: $q:\dz\ra\dz_{2r+1}$, taking $x\mapsto[x]_{2r+1}$. With this notation

423: $\zz$ acts on $C_{2r+1}$ by mapping $[x]_{2r+1}$ to $[-x]_{2r+1}$, for

424: $x\in V(C_{2r+1})$. Let $\gamma\in\chom(C_{2r+1},C_{2r+1})$ denote the

425: corresponding graph homomorphism. This action has a~fixed point

426: $2r+1$, and it flips one edge $(r,r+1)$.

427: 

428: Furthermore, let $\zz$ act on $K_m$ for $m\geq 2$, by swapping the

429: vertices 1 and 2 and fixing the vertices $3,\dots,m$; here, $K_m$ is

430: the graph defined by $V(K_m)=[1,m]$, $E(K_m)=\{(x,y)\,|\,x,y\in V(K_m),

431: x\neq y\}$. Since in both cases the graph homomorphism flips an~edge,

432: they induce free $\zz$-actions on $\thom(C_{2r+1},G)$ and

433: $\thom(K_m,G)$, for an arbitrary graph $G$ without loops.

434: 

435: \subsection{Non-vanishing of powers of Stiefel-Whitney classes as

436:   obstructions to graph colorings.} $\,$ \vspace{5pt}

437: \label{ss2.2}

438: 

439: 

440: The connection between the non-nullity of the powers of

441: Stiefel-Whitney characteristic classes and the lower bounds for graph

442: colorings is provided by the following general observation.

443: 

444: \begin{thm} \label{thmmain}

445: Let $G$ be a graph without loops, and let $T$ be a~graph with

446: $\zz$-action which flips some edge in $T$. If, for some integers

447: $k\geq 0$, $m\geq 1$, we have $\varpi_1^k(\thom(T,G))\neq 0$, and

448: $\varpi_1^{k}(\thom(T,K_m))=0$, then $\chi(G)\geq m+1$.

449: \end{thm}

450: \pr We have already shown that, under the assumptions of the theorem,

451: $\thom(T,H)$ is a~$\zz$-space for any loopfree graph $H$.  Assume now

452: that the graph $G$ is $m$-colorable, i.e., there exists a~homomorphism

453: $\phi:G\ra K_m$. It induces a~$\zz$-map

454: $\phi^T:\thom(T,G)\ra\thom(T,K_m)$. Since the Stiefel-Whitney classes

455: are functorial and $\varpi_1^{k}(\thom(T,K_m))=0$, the existence of

456: the $\zz$-map $\phi^{T}$ implies that $\varpi_1^{k}(\thom(T,G))=0$,

457: which is a~contradiction to the assumption of the theorem. \qed

458: 

459: \begin{rem} \label{rem:conn}

460: If a~$\zz$-space $X$ is $k$-connected, then there exists a~$\zz$-map

461: $\phi:S_a^{k+1}\ra X$, in particular, $\varpi_1^{k+1}(X)\neq 0$.

462: \end{rem}

463: \pr To construct $\phi$, subdivide $S_a^{k+1}$ simplicially as a join

464: of $k+2$ copies of $S^0$, and then define $\phi$ on the join of the

465: first $i$ factors, starting with $i=1$, and increasing $i$ by $1$ at

466: the time. To define $\phi$ on the first factor $\{a,b\}$, simply map

467: $a$ to an arbitrary point $x\in X$, and then map $b$ to $\gamma(x)$,

468: where $\gamma$ is the free involution of~$X$. Assume $\phi$ is defined

469: on $Y$ - the join of the first $i$ factors. Extend $\phi$ to

470: $Y*\{a,b\}$ by extending it first to $Y*\{a\}$, which we can do, since

471: $X$ is $k$-connected, and then extending $\phi$ to the second

472: hemisphere $Y*\{b\}$, by applying the involution~$\gamma$.

473: 

474: Since the Stiefel-Whitney classes are functorial, we have

475: $\phi^*(\varpi_1^{k+1}(X))=\varpi_1^{k+1}(S_a^{k+1})$, and the latter is

476: clearly nontrivial.\qed \vspace{5pt}

477: 

478: Let $T$ be any graph and consider the following equation

479: \begin{equation}\label{eq:ccvan}

480: \varpi_1^{n-\chi(T)+1}(\thom(T,K_n))=0, \text{ for all }

481: n\geq \chi(T)-1.

482: \end{equation}

483: 

484: \begin{thm} \label{thmmain1} $\,$

485: \begin{enumerate}

486: \item[(a)] The equation \eqref{eq:ccvan} is true for $T=K_m$, $m\geq 2$.

487: \item[(b)] The equation \eqref{eq:ccvan} is true for $T=C_{2r+1}$,

488: $r\geq 1$, and odd~$n$.

489: \end{enumerate}

490: \end{thm}

491: \pr The case $T=K_m$ is \cite[Theorem 1.6]{BK2} and has been proved

492: there. The case $T=C_{2r+1}$ will be proved in the

493: Section~\ref{sect5}. \qed

494: 

495: \begin{rem}

496: For a fixed value of $n$, if the equation \eqref{eq:ccvan} is true for

497: $T=C_{2r+1}$, then it is true for any $T=C_{2\ti r+1}$, if $r\geq\ti r$.

498: \end{rem}

499: 

500: \pr If $r\geq\ti r$, there exists a~graph homomorphism

501: $\phi:C_{2r+1}\ra C_{2\ti r+1}$ which respects the $\zz$-action.  This

502: induces a~$\zz$-map 

503: $$\phi_{K_n}:H^*(\thom(C_{2r+1},K_n))\ra H^*(\thom(C_{2\ti r+1},

504: K_n)),$$ 

505: yielding 

506: $$\ti\phi_{K_n}:H^*(\thom(C_{2r+1},K_n)/\zz;\zz)\ra H^*(\thom(C_{2\ti

507: r+1},K_n)/\zz;\zz).$$ Clearly,

508: $\ti\phi_{K_n}(\varpi_1(\thom(C_{2r+1},K_n)))= \varpi_1(\thom(C_{2\ti

509: r+1},K_n))$. In particular, $\varpi_1^i(\thom(C_{2r+1},K_n))=0$,

510: implies $\varpi_1^i(\thom(C_{2\ti r+1}, K_n))=0$.  \qed \vspace{5pt}

511: 

512: Note that for $T=C_{2r+1}$ and $n=2$, the equation \eqref{eq:ccvan} is

513: obvious, since $\thom(C_{2r+1},K_2)=\emptyset$. We give a~quick

514: argument for the next case $n=3$. One can see by inspection that the

515: connected components of $\thom(C_{2r+1},K_3)$ can be indexed by the

516: winding numbers $\alpha$. These numbers must be odd, so $\alpha=\pm

517: 1,\pm 3,\dots,\pm(2s+1)$, where

518: \[

519: s=

520: \begin{cases}

521: (r-1)/3, & \text{ if } r\equiv 1 \mod 3,\\

522: \left\lfloor(r-2)/3\right\rfloor, & \text{ otherwise,}

523: \end{cases}

524: \] 

525: in particular $s\geq 0$. Let $\phi:\thom(C_{2r+1},K_3)\ra\{\pm 1,\pm

526: 3, \dots,\pm(2s+1)\}$ map each point $x\in\thom(C_{2r+1},K_3)$ to the

527: point on the real line, indexing the connected component

528: of~$x$. Clearly, $\phi$ is a $\zz$-map.  Since $H^1(\{\pm 1,\pm 3,

529: \dots,\pm(2s+1)\}/\zz;\zz)=0$, the functoriality of the characteristic

530: classes implies $\varpi_1(\thom(C_{2r+1},K_3))=0$.

531: 

532: \begin{conj} \label{conj:sw}

533: The equation \eqref{eq:ccvan} is true for $T=C_{2r+1}$, $r\geq 1$, and

534: all~$n$.

535: \end{conj}

536: 

537: %\vspace{5pt}

538: 

539: \subsection{Completing the sketch of the proof of the Lov\'asz Conjecture.} 

540: $\,$ \label{ss2.3} 

541: 

542: Consider one of the two maps $\iota:K_2\ra C_{2r+1}$ mapping the edge

543: to the $\zz$-invariant edge of $C_{2r+1}$. Clearly, $\iota$ is

544: $\zz$-equivariant. Since $\thom(-,H)$ is a~contravariant functor,

545: $\iota$ induces a~map of $\zz$-spaces

546: $\iota_{K_n}:\thom(C_{2r+1},K_n)\ra \thom(K_2,K_n)$, which in turn

547: induces a~$\dz$-algebra homomorphism

548: $\iota_{K_n}^*:H^*(\thom(K_2,K_n);\dz)\ra H^*(\thom(C_{2r+1},K_n);\dz)$.

549: 

550: \begin{thm}\label{thm:even_n}

551: Assume $n$ is even, then $2\cdot\iota_{K_n}^*$ is a $0$-map.

552: \end{thm}

553: 

554: Theorem \ref{thm:even_n} is proved in Section~\ref{sect4}. The results

555: of this paper were announced in \cite{BK1}, and the preprint of this

556: paper has been available since February 2004. In the summer 2005 an

557: alternative proof of Theorem~\ref{thm:even_n} appeared in the preprint

558: \cite{Z1}, and a~proof of Conjecture~\ref{conj:sw} was announced by

559: C.~Schultz.

560: 

561: 

562: 

563: %\newpage

564: 

565: \vspace{5pt}

566: 

567: \noindent

568: {\bf Proof of Theorem~\ref{loconj} (Lov\'asz Conjecture).}

569: 

570: \noindent

571: The case $k=-1$ is trivial, so take $k\geq 0$. Assume first that $k$

572: is even. By the Remark~\ref{rem:conn}, we have

573: $\varpi_1^{k+1}(\thom(C_{2r+1},G))\neq 0$. By

574: Theorem~\ref{thmmain1}(b), we have

575: $\varpi_1^{k+1}(\thom(C_{2r+1},K_{k+3}))=0$. Hence, applying

576: Theorem~\ref{thmmain} for $T=C_{2r+1}$ we get $\chi(G)\geq k+4$.

577: 

578: Assume now that $k$ is odd, and that $\chi(G)\leq k+3$. Let $\phi:G\ra

579: K_{k+3}$ be a~vertex-coloring map. Combining the

580: Remark~\ref{rem:conn}, the fact that $\thom(C_{2r+1},-)$ is

581: a~covariant functor from loopfree graphs to $\zz$-spaces, and the map

582: $\iota:K_2\ra C_{2r+1}$, we get the following diagram of $\zz$-spaces

583: and $\zz$-maps:

584: \begin{multline*}

585: S_a^{k+1}\stackrel{f}{\lra}\thom(C_{2r+1},G)

586: \stackrel{\phi^{C_{2r+1}}}{\lra}\thom(C_{2r+1},K_{k+3})

587: \stackrel{\iota_{K_{k+3}}}{\lra}\\

588: \stackrel{\iota_{K_{k+3}}}{\lra}\thom(K_2,K_{k+3})\cong S_a^{k+1}.

589: \end{multline*}

590: This gives a~homomorphism on the corresponding cohomology groups in

591: dimension $k+1$, $h^*=f^*\circ(\phi^{C_{2r+1}})^*

592: \circ(\iota_{K_{k+3}})^*:\dz\ra\dz$. It is well-known, see, e.g.,

593: \cite[Proposition 2B.6, p.\ 174]{Hat}, that a $\zz$-map $S_a^n\ra

594: S_a^n$ cannot induce a~$0$-map on the $n$th cohomology groups (in fact

595: it must be of odd degree). Hence, we have a~contradiction, and so

596: $\chi(G)\geq k+4$.  \qed \vspace{5pt}

597: 

598: Let us make a~couple of remarks.

599: 

600: \begin{rem}

601: As is apparent from our argument, we are actually proving a~sharper

602: statement than the original Lov\'asz Conjecture. First of all, the

603: condition ``$\thom(C_{2r+1},G)$ is $k$-connected'' can be replaced by

604: a~weaker condition ``the coindex of $\thom(C_{2r+1},G)$ is at least

605: $k+1$''. Furthermore, for even~$k$, that condition can be weakened

606: even further to ``$\varpi_1^{k+1}(\thom(C_{2r+1},G))\neq

607: 0$''. Conjecture~\ref{conj:sw} would imply that this weakening can be

608: done for odd $k$ as well.

609: \end{rem}

610: 

611: \begin{rem}

612: It follows from \cite[Proposition 5.1]{BK2} that Lov\'asz Conjecture

613: is true if $C_{2r+1}$ is replaced by any graph $T$, such that $T$ can

614: be reduced to $C_{2r+1}$, by a~sequence of folds.

615: \end{rem}

616: 

617: 

618: \section{$\thomp$ and filtrations}

619: 

620: \subsection{The $+$ construction.} $\,$ \vspace{5pt}

621: 

622: For a finite graph $H$, let $H_+$ be the graph obtained from $H$ by

623: adding an extra vertex $b$, called the base vertex, and connecting it

624: by edges to all the vertices of $H_+$ including itself, i.e.,

625: $V(H_+)=V(H)\cup\{b\}$, and $E(H_+)=E(H)\cup\{(v,b),(b,v)\,|\,v\in

626: V(H_+)\}$.

627: 

628: \begin{df}

629:   Let $G$ and $H$ be two graphs. The simplicial complex $\thomp(G,H)$

630:   is defined to be the link in $\thom(G,H_+)$ of the homomorphism

631:   mapping every vertex of $G$ to the base vertex in $H_+$.

632: \end{df}

633: 

634: So the cells in $\thomp(G,H)$ are indexed by all $\eta:V(G)\ra

635: 2^{V(H)}$ satisfying the same condition as in the

636: Definition~\ref{dfhom}. The closure of $\eta$ is also defined

637: identical to how it was defined for $\thom$. Note, that $\thomp(G,H)$

638: is simplicial, and that $\thomp(G,-)$ is a~covariant functor from {\bf

639:   Graphs} to {\bf Top}. One can think of $\thomp(G,H)$ as a~cell

640: structure imposed on the set of all partial homomorphisms from $G$

641: to~$H$.

642: 

643: 

644: 

645: %\[\pstexInput{ex+}\]

646: %\[\text{Figure 1.}\]

647: 

648: 

649: \begin{figure}[hbt]

650: \begin{center}

651: \pstexInput{ex+}

652: \end{center}

653: \caption{The hom plus construction.}

654: \label{fig:ex+}

655: \end{figure}

656: 

657: 

658: For an arbitrary graph $G$, let $\ind(G)$ denote the independence

659: complex of $G$, i.e., the vertices of $\ind(G)$ are all vertices

660: of~$G$, and simplices are all the independent sets of~$G$.

661: The~dimension of $\thomp(G,H)$, unlike that of $\thom(G,H)$ is easy to

662: find:

663: \[

664: \dim(\thomp(G,H))=|V(H)|\cdot(\dim \ind(G)+1)-1.

665: \]

666: 

667: 

668: Recall that for any graph $G$, the strong complement $\com G$ is

669: defined by $V(\com G)=V(G)$, $E(\com G)=V(G)\times V(G)\sm

670: E(G)$. Also, for any two graphs $G$ and $H$, the direct product

671: $G\times H$ is defined by $V(G\times H)=V(G)\times V(H)$, $E(G\times

672: H)=\{((x,y),(x',y'))\,|\,(x,x')\in E(G),(y,y')\in E(H)\}$.

673: 

674: Sometimes, it is convenient to view $\thomp(G,H)$ as an~independence

675: complex of a~certain graph.

676: 

677: \begin{prop} \label{pr_homp}

678: The complex $\thomp(G,H)$ is isomorphic to $\ind(G\times\com H)$. In

679: particular, $\thomp(G,K_n)$ is isomorphic to $\ind(G)^{*n}$, where $*$

680: denotes the simplicial join.

681: \end{prop}

682: 

683: \pr By the definition, $V(G\times\com H)=V(G)\times V(H)$. Let

684: $S\subseteq V(G)\times V(H)$, $S=\{(x_i,y_i)\,|\,i\in I, x_i\in V(G),

685: y_i\in V(H)\}$. Then $S\in\ind(G\times\com H)$ if and only if, for any

686: $i,j\in I$, we have either $(x_i,x_j)\notin E(G)$ or $(y_i,y_j)\in

687: E(H)$, since the forbidden constellation is when $(x_i,x_j)\in E(G)$

688: and $(y_i,y_j)\notin E(H)$.

689: 

690: Identify $S$ with $\eta_S:V(G)\ra 2^{V(H)}$ defined by: for $v\in

691: V(G)$, set $\eta_S(v):=\{w\in V(H)\,|\,(v,w)\in S\}$. The condition

692: for $\eta_S\in\thomp(G,H)$ is that, if $(v_1,v_2)\in E(G)$, and

693: $w_1\in\eta_S(v_1)$, $w_2\in\eta_S(v_2)$, then $(w_1,w_2)\in E(H)$,

694: which is visibly identical to the condition for $S\in\ind(G\times\com

695: H)$. Hence $\thomp(G,H)=\ind(G\times\com H)$.

696: 

697: To see the second statement note first that $\com K_n$ is the

698: disjoint union of $n$ looped vertices. Since taking direct

699: products is distributive with respect to disjoint unions, and a

700: direct product of $G$ with a loop is again $G$, we see that

701: $G\times\com K_n$ is a disjoint union of $n$ copies of $G$.

702: Clearly, its independence complex is precisely the $n$-fold join

703: of $\ind(G)$. \qed

704: 

705: \subsection{Cochain complexes for $\thom(G,H)$ and $\thomp(G,H)$.} 

706: $\,$ \vspace{5pt} \label{ss3.2}

707: 

708: For any CW complex $K$, let $K^{(i)}$ denote the $i$-th skeleton

709: of~$K$. Let $R$ be a~commutative ring with a~unit. In this paper we

710: will have two cases: $R=\dz$ and $R=\zz$. For any $\eta\in K^{(i)}$,

711: we fix an~orientation on $\eta$, and let

712: $C_i(K;R):=R[\eta\,|\,\eta\in K^{(i)}]$, where

713: $R[\alpha\,|\,\alpha\in I]$ denotes the free $R$-module generated

714: by $\alpha\in I$. Furthermore, let $C^i(K;R)$ be the dual

715: $R$-module to $C_i(K;R)$. For arbitrary $\alpha\in C_i(K;R)$ let

716: $\alpha^*$ denote the element of $C^i(K;R)$ which is dual

717: to~$\alpha$. Clearly, $C^i(K;R)=R[\eta^*\,|\,\eta\in K^{(i)}]$,

718: and the cochain complex of $K$ is

719: \[\dots\stackrel{\bo^{i-1}}\lra C^i(K;R)\stackrel{\bo^i}\lra

720: C^{i+1}(K;R)\stackrel{\bo^{i+1}}\lra\dots.\] 

721: 

722: For $\eta\in K^{(i)}$, $\ti\eta\in K^{(i+1)}$, we have the incidence

723: number $[\eta:\ti\eta]$, which is $0$ if $\eta\notin\ti\eta$. In these

724: notations $\bo^i(\eta^*)=\sum_{\ti\eta\in K^{(i+1)}}$

725: $[\eta:\ti\eta]\,\ti\eta^*$. For arbitrary $\alpha\in C_i(K;R)$,

726: resp.\ $\alpha^*\in C^i(K;R)$, we let $[\alpha]$, resp.\ 

727: $[\alpha^*]$, denote the corresponding element of $H_i(K;R)$, resp.\ 

728: $H^i(K;R)$.

729: 

730: When coming after the name of a~cochain complex, the brackets $[-]$

731: will denote the index shifting (to the left), that is for the cochain

732: complex $C^*$, the cochain complex $C^*[s]$ is defined by

733: $C^i[s]:=C^{i+s}$, and the differential is the same (we choose not to

734: change the sign of the differential).

735: 

736: We now return to our context. Let $G$ and $H$ be two graphs, and let

737: us choose some orders on $V(G)=\{v_1,\dots,v_{|V(G)|}\}$ and on

738: $V(H)=\{w_1,\dots,w_{|V(H)|}\}$. Through the end of this subsection we

739: assume the coefficient ring to be $\dz$; the situation over $\zz$ is

740: simpler and can be described by tensoring with $\zz$.

741: 

742: Vertices of $\thomp(G,H)$ are indexed with pairs $(x,y)$, where $x\in

743: V(G)$, $y\in V(H)$, such that if $x$ is looped, then so is~$y$. We

744: order these pairs lexicographically:

745: $(v_{i_1},w_{j_1})\prec(v_{i_2},w_{j_2})$ if either $i_1<i_2$, or

746: $i_1=i_2$ and $j_1<j_2$. Orient each simplex of $\thomp(G,H)$

747: according to this order on the vertices. We call this orientation {\it

748: standard}, and call the oriented simplex~$\eta_+$. If

749: $\ti\eta_+\in\thomp^{(i+1)}(G,H)$ is obtained from

750: $\eta_+\in\thomp^{(i)}(G,H)$ by adding a~vertex~$v$, then

751: $[\eta_+:\ti\eta_+]$ is $(-1)^{k-1}$, where $k$ is the position of $v$

752: in the order of the vertices of~$\ti\eta_+$.

753: 

754: Let us now turn to $C^*(\thom(G,H))$. We can fix an orientation, which

755: we also call standard, on each cell $\eta\in\thom(G,H)$ as follows:

756: orient each simplex $\eta(i)$ according to the chosen order on the

757: vertices of $H$, then, order these simplices in the direct product

758: according to the chosen order on the vertices of~$G$. To simplify our

759: notations, we still call this oriented cell $\eta$, even though

760: a~choice of orders on the vertex sets of $G$ and $H$ is implicitly

761: present.

762: 

763: We remark for later use, that permuting the vertices of the simplex

764: $\eta(i)$ by some $\sigma\in\cs_{|\eta(i)|}$ changes the orientation

765: of the cell $\eta$ by $\sgn(\sigma)$, whereas swapping the simplices

766: with vertex sets $\eta(i)$ and $\eta(i+1)$ in the direct product

767: changes the orientation by

768: $(-1)^{(|\eta(i)|-1)(|\eta(i+1)|-1)}=(-1)^{\dim\eta(i)\cdot\dim\eta(i+1)}$.

769: 

770: If $\ti\eta\in\thom^{(i+1)}(G,H)$ is obtained from

771: $\eta\in\thom^{(i)}(G,H)$ by adding a~vertex~$v$ to the list

772: $\eta(t)$, then $[\eta:\ti\eta]$ is $(-1)^{k+d-1}$, where $k$ is the

773: position of $v$ in $\ti\eta(t)$, and $d$ is the dimension of the

774: product of the simplices with the vertex sets

775: $\eta(1),\dots,\eta(t-1)$, that is $d=1-t+\sum_{j=1}^{t-1}|\eta(j)|$.

776: To see this, note that $[\eta:\ti\eta]=1$ if the first vertex in the

777: first simplex is inserted. The general case follows from the

778: previously described rules for changing the sign of the orientation

779: under permuting simplices in the product and permuting vertices within

780: simplices.

781: 

782: \subsection{The support map and the relation between $\thom(G,H)$ 

783: and $\thomp(G,H)$.} \label{ss3.3a} $\,$ \vspace{5pt}

784: 

785: For each simplex of $\thomp(G,H)$, $\eta:V(G)\ra 2^{V(H)}$, define the

786: support of $\eta$ to be

787: $\supp\eta:=V(G)\setminus\eta^{-1}(\emptyset)$. A~concise way to

788: phrase the definition of $\supp$ differently is to consider the map

789: $t^G:\thomp(G,H)\ra\thomp(G,\com K_1)\simeq\Delta_{|V(G)|-1}$ induced

790: by the homomorphism $t:H\ra\com K_1$. Then, for each

791: $\eta\in\thomp(G,H)$ we have $\supp\eta=t^G(\eta)$, where the

792: simplices in $\Delta_{|V(G)|-1}$ are identified with the subsets of

793: $V(G)$.

794: 

795: Let $\wti C^*$ be the subcomplex of $C^*(\thomp(G,H))$ generated by

796: all $\eta^*_+$, for $\eta:V(G)\ra 2^{V(H)}$, such that

797: $\supp\eta=V(G)$ (cf.\ filtration in subsection \ref{ss3.3}). Set

798: \begin{equation} \label{eq:x*}

799: X^*(G,H):=\wti C^*[|V(G)|-1].

800: \end{equation}

801: Note that both $C^i(\thom(G,H))$ and $X^i(G,H)$ are free $\dz$-modules

802: with the bases $\{\eta^*\}_\eta$ and $\{\eta^*_+\}_\eta$ indexed by

803: $\eta:V(G)\ra 2^{V(H)}\sm\{\emptyset\}$, such that

804: $\sum_{j=1}^{|V(G)|}|\eta(j)|=|V(G)|+i$.

805: 

806: At this point we introduce the following notations: for $\eta:V(G)\ra

807: 2^{V(H)}\sm\{\emptyset\}$, set 

808: $$c(\eta):=\sum_{\begin{subarray}{c}i\text{ is even}\\

809: 1\leq i\leq|V(G)|\end{subarray}}

810: |\eta(i)|.$$

811: 

812: For any $\eta:V(G)\ra 2^{V(H)}\sm\{\emptyset\}$, set

813: $\rho(\eta_+):=(-1)^{c(\eta)}\eta$. Obviously, the induced map

814: $\rho^*:X^i(G,H)\ra C^i(\thom(G,H))$ is a~$\dz$-module isomorphism for

815: any~$i$.

816: 

817: \begin{prop}

818:  The map $\rho^*:X^*(G,H)\ra C^*(\thom(G,H))$ is an~isomorphism of the

819:   cochain complexes.

820: \end{prop}

821: 

822: \pr Indeed, let $\ti\eta:V(G)\ra 2^{V(H)}\sm\{\emptyset\}$ be obtained

823: from $\eta$ by adding a~vertex $v$ to the list $\eta(t)$, and let $k$

824: be the position of $v$ in $\ti\eta(t)$. By our previous computation:

825: $[\eta_{+}:\ti\eta_+]=(-1)^{k+d+t}$, whereas

826: $[\eta:\ti\eta]=(-1)^{k+d-1}$, where

827: $d=1-t+\sum_{j=1}^{t-1}|\eta(j)|$. This shows that

828: $$

829: [\rho(\eta_+):\rho(\ti\eta_+)]=(-1)^{c(\eta)+c(\ti\eta)}

830: [\eta:\ti\eta]= (-1)^{c(\eta)+c(\ti\eta)+t+1}[\eta_+:\ti\eta_+],

831: $$

832: but

833: $$c(\eta)+c(\ti\eta)+t+1=\sum_{i\text{ is even}}|\eta(i)|+

834: \sum_{i\text{ is even}}|\ti\eta(i)|+t+1\equiv 0\,\,(\text{mod } 2)$$

835: for~any~$t$, hence $[\rho(\eta_+):\rho(\ti\eta_+)]=

836: [\eta_+:\ti\eta_+]$. \qed

837: 

838: 

839: \subsection{Relating $\zz$-actions on $\thom(G,H)$ and $\thomp(G,H)$.}  

840: $\,$ \vspace{5pt}

841: 

842: Assume that we have $\gamma\in\chom(G,G)$, and $0\leq r\leq |V(G)|/2$,

843: such that

844: \[\gamma(i)=\begin{cases}

845: 2r+1-i,& \text{ if }1\leq i\leq 2r; \\

846: i,&\text{ if }2r+1\leq i\leq |V(G)|,

847: \end{cases}\]

848: where we identified $V(G)$ with the numbering $[1,V(G)]$. In

849: particular, we have $\gamma^2=1$.

850: 

851: The homomorphism $\gamma$ induces $\zz$-action both on $\thom(G,H)$,

852: and on $\thomp(G,H)$. We shall see how $\rho^*$ behaves with respect

853: to this $\zz$-action. For any $\eta:V(G)\ra 2^{V(H)}\sm\{\emptyset\}$,

854: $\gamma$ takes $\eta$ to $\eta\circ\gamma$. By a~slight abuse of

855: notations we let $\gamma$ denote the induced actions both on

856: $C^*(\thom(G,H))$ and on $X^*(G,H)$.

857: 

858: Let $(u_1,\dots,u_q)$ be the vertices of the simplex $\eta_+$ listed

859: in the increasing order. By definition,

860: $\gamma(\eta_+)=(\gamma(u_1),\dots,\gamma(u_q))$, where

861: $\gamma((v,w)):=(\gamma(v),w)$, for $v\in V(G)$, $w\in V(H)$. Clearly,

862: $\gamma(\eta_+)$ has the same set of vertices as

863: $(\eta\circ\gamma)_+$, so we just need to see how their orientations

864: relate. To order the vertices of $\gamma(\eta_+)$ we need to invert

865: the order of the blocks with cardinalities

866: $|\eta(1)|,\dots,|\eta(2r)|$ without changing the vertex orders within

867: the blocks. The sign of this permutation is $(-1)^c$, where

868: $c=\sum_{1\leq i<j\leq 2r}|\eta(i)|\cdot|\eta(j)|$, so we conclude

869: that 

870: \begin{equation}

871:   \label{eq:geta+}\gamma(\eta^*_+)=(-1)^c(\eta\circ\gamma)^*_+.

872: \end{equation}

873: 

874: Consider now the oriented cell $\eta$. It is a~direct product of

875: simplices $\da^1,\dots,\da^{|V(G)|}$ of dimensions

876: $|\eta(1)|-1,\dots,|\eta(|V(G)|)|-1$, with the standard orientation as

877: defined above. The cell $\gamma(\eta)$ is the direct product of

878: $\gamma(\da^1)=\da^{2r},\gamma(\da^2)=\da^{2r-1},\dots,

879: \gamma(\da^{2r})=\da^1,\gamma(\da^{2r+1})=\da^{2r+1},$ $\dots,

880: \gamma(\da^{|V(G)|})=\da^{|V(G)|}$, with the order of the vertices

881: (hence the orientation) within each simplex being the same as

882: in~$\eta$.

883: 

884: We see that $\gamma(\eta)$ is, up to the orientation, the same cell as

885: $\eta\circ\gamma$. To relate their orientations, we need to permute

886: the simplices $\da^{2r},\dots,\da^1$ back in order, which, by the

887: previous observations, changes the orientation by $(-1)^{\ti d}$,

888: where

889: \begin{multline*}

890: \ti d=\sum_{1\leq i<j\leq 2r}\dim\da^i\cdot\dim\da^j=

891: \sum_{1\leq i<j\leq 2r}(|\eta(i)|-1)(|\eta(j)|-1)\\

892: =c-(2r-1)\sum_{i=1}^{2r}|\eta(i)|+{\binom{2r}{2}}.

893: \end{multline*}

894: Reducing modulo 2, we conclude that 

895: \begin{equation}

896:   \label{eq:geta}

897:   \gamma(\eta^*)=(-1)^d(\eta\circ\gamma)^*,

898: \end{equation}

899: where $d=c+\sum_{i=1}^{2r}|\eta(i)|+r$.

900: 

901: Let us now see how $\rho^*$ interacts with $\gamma$. We have

902: \begin{equation}

903:   \label{eq:3.3}

904:   \rho^*(\gamma(\eta^*_+))=(-1)^c\rho^*((\eta\circ\gamma)^*_+)=

905: (-1)^{c+c(\eta\circ\gamma)}(\eta\circ\gamma)^*,

906: \end{equation}

907: where the first equality is by \eqref{eq:geta+} and the second one is

908: by definition of $\rho$, and

909: \begin{equation}

910:   \label{eq:3.4}

911:   \gamma(\rho^*(\eta^*_+))=(-1)^{c(\eta)}\gamma(\eta^*)=

912: (-1)^{d+c(\eta)}(\eta\circ\gamma)^*,

913: \end{equation}

914: where the first equality is by definition of $\rho$ and second one is

915: by~\eqref{eq:geta}. Comparing \eqref{eq:3.3} with \eqref{eq:3.4}, and

916: using the computation

917: \begin{multline*}

918: c(\eta)+c(\eta\circ\gamma)=\sum_{\begin{subarray}{c}{i\text{ is even}}\\

919: {1\leq i\leq|V(G)|}\end{subarray}}|\eta(i)|+

920: \sum_{\begin{subarray}{c}{i\text{ is even}}\\{1\leq i\leq|V(G)|}

921: \end{subarray}}|\eta\circ\gamma(i)|\\

922: =\sum_{i=1}^{2r}|\eta(i)|+2\cdot\sum_{\begin{subarray}{c}{i\text{ is even}}\\

923: {2r+1\leq i\leq|V(G)|}\end{subarray}}|\eta(i)|,

924: \end{multline*}

925: we see that, for any $\eta$

926: \begin{equation}

927:   \label{eq:act_disc}

928:    \rho^*(\gamma(\eta^*_+))=(-1)^r \gamma(\rho^*(\eta^*_+)).

929: \end{equation}

930: 

931: \subsection{The filtration of $C^*(\thomp(G,H);\dz)$ and the 

932:   $E_0^{*,*}$-tableau.} \label{ss3.3} $\,$ \vspace{5pt}

933: 

934: We shall now filter $C^*(\thomp(G,H);\dz)$. Define the subcomplexes of

935: $C^*(\thomp(G,H);\dz)$, $F^p=F^pC^*(\thomp(G,H);\dz)$, as follows:

936: \[ F^p: \dots\stackrel{\bo^{q-1}}\lra F^{p,q}\stackrel{\bo^{q}}\lra

937: F^{p,q+1}\stackrel{\bo^{q+1}}\lra\dots,\]

938: where

939: \[ F^{p,q}=F^p C^q(\thomp(G,H);\dz)=\dz\left[\eta^*_+\,\left|\,

940: \eta_+\in\thomp^{(q)}(G,H),|\supp\eta|\geq p+1\right]\right.,

941: \]

942: and $\bo^*$ is the restriction of the differential in

943: $C^*(\thomp(G,H);\dz)$. Then,

944: \[

945: C^q(\thomp(G,H);\dz)=F^{0,q}\supseteq F^{1,q}\supseteq\dots\supseteq

946: F^{|V(G)|-1,q}\supseteq F^{|V(G)|,q}=0.

947: \]

948: 

949: 

950: \begin{prop}

951: For any $p$,

952: \begin{equation}\label{eq:e0tab}

953: F^p/F^{p+1}=\bigoplus_{\begin{subarray}{c}{S\subseteq V(G)}\\{|S|=p+1}

954: \end{subarray}}

955: C^*(\thom(G[S],H);\dz)[-p].

956: \end{equation}

957: Hence, the 0th tableau of the spectral sequence associated to

958: the cochain complex filtration $F^*$ is given by

959: \begin{equation}

960:   \label{eq:E0}

961: E_0^{p,q}=C^{p+q}(F^p,F^{p+1})=\bigoplus_{\begin{subarray}{c}

962: {S\subseteq V(G)}\\{|S|=p+1}\end{subarray}}

963: C^q(\thom(G[S],H);\dz).

964: \end{equation}

965: \end{prop}

966: \pr By construction

967: \begin{multline*}

968:   F^{p,q}/F^{p+1,q}=\dz\left[\eta^*_+\,\left|\,\eta_+\in\thomp^{(q)}(G,H),

969:   |\supp\eta|=p+1\right]\right.=\\

970:   \bigoplus_{\begin{subarray}{c}{S\subseteq V(G)}\\{|S|=p+1}\end{subarray}}

971: X^{q-p}(G[S],H;\dz)

972:   \stackrel{\rho^*}{=} 

973: \bigoplus_{\begin{subarray}{c}{S\subseteq V(G)}\\{|S|=p+1}\end{subarray}}

974: C^{q-p}(\thom(G[S],H);\dz),

975: \end{multline*} 

976: where $X^*$ is defined in \eqref{eq:x*}, and $\rho^*$ is the map

977: defined in subsection~\ref{ss3.3a}.  \qed \vspace{5pt}

978: 

979: Note, that in particular we have

980: $F^{|V(G)|-1,q}/F^{|V(G)|,q}=F^{|V(G)|-1,q}=C^q(\thom(G,H);\dz)[1-|V(G)|]$.

981: 

982: \section{$\zz$-action on $H^*(\thomp(C_{2r+1},K_n);\dz)$}

983: \label{sect4}

984: 

985: In this section we shall derive some information about the

986: $\dz[\zz]$-modules $H^*(\thom(C_{2r+1},K_n);\dz)$, for $r\geq 2$,

987: $n\geq 4$. For $r=2$ our computation will be complete.

988: 

989: We adopt the following convention: we think of $C_{2r+1}$ as a~unit

990: circle on the plane with $2r+1$ marked points with numbers

991: $1,\dots,2r+1$ following each other in the clockwise increasing order.

992: The directions {\it left}, resp.\ {\it right} on this circle will

993: denote counterclockwise, resp.\ clockwise. 

994: 

995: Furthermore, before we start our computation, we introduce the

996: following terminology. For $S\subset V(C_{2r+1})$, we call those

997: connected components of $C_{2r+1}[S]$ which have at least 2 vertices

998: the {\it arcs}.  For $x,y\in\dz$, we let $[x,y]_{2r+1}$ denote the arc

999: starting from $x$ and going clockwise to $y$, that is

1000: $[x,y]_{2r+1}=\{[x]_{2r+1},[x+1]_{2r+1},\dots,[y-1]_{2r+1},[y]_{2r+1}\}$.

1001: 

1002: \subsection{The simplicial complex of partial homomorphisms from a~cycle

1003:   to a~complete graph.} $\,$ \vspace{5pt}

1004: 

1005: Here and in the next subsection we summarize some previously published

1006: results which are necessary for our present computations. To start

1007: with, recall that the homotopy type of the independence complexes of

1008: cycles was computed in~\cite{Koz1}.

1009: 

1010: \begin{prop}\label{pr_indcomp}{\rm (\cite[Proposition 5.2]{Koz1})}. 

1011: 

1012: \nin For any integer $m\geq 2$, we have

1013: \[\ind(C_m)\simeq\begin{cases}

1014: S^{k-1}\vee S^{k-1},&\text{ if } m=3k;\\

1015: S^{k-1},&\text{ if } m=3k\pm 1.

1016: \end{cases}\]

1017: \end{prop}

1018: 

1019: Combining Propositions~\ref{pr_homp} and \ref{pr_indcomp} we get the

1020: following formula.

1021: 

1022: \begin{crl}\label{crl_indcomp} 

1023:  For any integers $m\geq 2$, $n\geq 1$, we have

1024: \begin{equation}\label{eq_indcomp}

1025: \thomp(C_m,K_n)\simeq\begin{cases}

1026: \bigvee_{2^n\text{ copies}} S^{nk-1},&\text{ if } m=3k;\\

1027: S^{nk-1},&\text{ if } m=3k\pm 1.

1028: \end{cases}

1029: \end{equation}

1030: \end{crl}

1031:  

1032: The following estimates will be needed later for our spectral sequence

1033: computations.

1034: 

1035: \begin{crl}\label{crl:h+}

1036: We have $\wti H^i(\thomp(C_{2r+1},K_n))=0$ for $r\geq 2$, $n\geq 4$,

1037: and $i\leq n+2r-2$; except for the two cases $(n,r)=(4,3)$ and

1038: $(5,3)$.

1039: \end{crl}

1040: \pr Note, that if $2r+1=3k+\epsilon$, with $\epsilon\in\{-1,0,1\}$,

1041: then we have $\wti H^i(\thomp(C_{2r+1},K_n))=0$, for $i\leq nk-2$.

1042: 

1043: Assume first $2r+1=3k$. The inequality $nk-2\geq n+2r-2$ is equivalent

1044: to $n\geq 3+2/(k-1)$, and the latter is always true since $k\geq 3$

1045: and $n\geq 4$.

1046: 

1047:  Assume now $2r+1=3k+1$. This time, $nk-2\geq n+2r-2$ is equivalent to

1048: $n\geq 3+3/(k-1)$. If $k\geq 4$, this is always true, since $n\geq 4$.

1049: If $k=2$, this reduces to saying that $n\geq 6$. This yields the two

1050: exceptional cases: $r=3$ and $n=4,5$.

1051: 

1052: Finally, assume $2r+1=3k-1$. Here, $nk-2\geq n+2r-2$ is equivalent to

1053: $n\geq 3+1/(k-1)$, which is always true, since $k\geq 2$, $n\geq 4$.

1054: \qed \vspace{5pt}

1055: 

1056: The Corollary \ref{crl_indcomp} can be strengthened to include the

1057: information on the $\zz$-action.

1058: 

1059: \begin{prop}\label{pr:h+/z2}

1060: For any positive integers $r$ and $n$ we have

1061: \begin{equation}\label{eq:h+/z2}

1062: \thomp(C_{2r+1},K_n)/\zz\simeq

1063: \begin{cases}

1064: \bigvee_{2^{n-1}\text{ copies}} S^{nk-1},&\text{ if } 2r+1=3k;\\

1065: S^{kn/2-1}*\rp^{kn/2-1},&\text{ if } 2r+1=3k\pm 1.

1066: \end{cases}

1067: \end{equation}

1068: \end{prop}

1069: \pr

1070: By Proposition~\ref{pr_homp} we know that $\thomp(C_{2r+1},K_n)$

1071: is isomorphic to $\ind(C_{2r+1})^{*n}$. We analyze $\zz$-action on

1072: $\ind(C_{2r+1})$ in more detail.

1073: 

1074: Assume first $2r+1=3k-1$, in particular, $k$ is even. It was shown

1075: in~\cite[Proposition~5.2]{Koz1} that

1076: $X=\ind(C_{2r+1})\sm\{1,4,\dots,2r-3,2r\}$ is contractible (here

1077: ``$\sm$'' just means the removal of an open maximal simplex).  It

1078: follows from the standard fact in the theory of transformation groups,

1079: see e.g., \cite[Theorem 5.16, p.\ 222]{TtD}, that $X/\zz$ is

1080: contractible as well. Hence $\ind(C_{2r+1})$ is $\zz$-homotopy

1081: equivalent to the unit sphere $S^{k-1}\subset \dr^k$ with the $\zz$

1082: acting by fixing $k/2$ coordinates and multiplying the other $k/2$

1083: coordinates by $-1$.

1084: 

1085: Assume $2r+1=3k+1$. The link of the vertex $2r+1$ is $\zz$-homotopy

1086: equivalent to a~point. Hence, deleting the open star of the vertex

1087: $2r+1$ produces a~complex $X$, which is $\zz$-homotopy equivalent to

1088: $\ind(C_{2r+1})$.  It was shown in \cite[Proposition~5.2]{Koz1}, that

1089: $X\sm\{2,5,\dots,2r-4,2r-1\}$ is contractible.  By an~argument,

1090: similar to the previous case, we conclude that $\ind(C_{2r+1})$ is

1091: $\zz$-homotopy equivalent to the unit sphere $S^{k-1}\subset \dr^k$

1092: with the $\zz$ acting by fixing $k/2$ coordinates and multiplying the

1093: other $k/2$ coordinates by $-1$.

1094: 

1095: In both cases we see that $\thomp(C_{2r+1},K_n)$ is $\zz$-homotopy

1096: equivalent to $\susp^{kn/2}S^{kn/2-1}$, with the $\zz$-action and the

1097: latter space being induced by the antipodal action on $S^{kn/2-1}$. It

1098: follows that $\thomp(C_{2r+1},K_n)/\zz$ is homotopy equivalent to

1099: $\susp^{kn/2}\rp^{kn/2-1}$.

1100: 

1101: Consider the remaining case $2r+1=3k$. It was shown in

1102: \cite[Proposition~5.2]{Koz1} that $\ind(C_{2r+1})$ becomes

1103: contractible if one removes the simplices $\{1,4,\dots,2r-1\}$ and

1104: $\{2,5,\dots,2r\}$. It follows that $\ind(C_{2r+1})$ is $\zz$-homotopy

1105: equivalent to the wedge of two unit spheres $S^{k-1}$ with the $\zz$

1106: acting by swapping the spheres. Thus $\thomp(C_{2r+1},K_n)$ is

1107: $\zz$-homotopy equivalent to a~wedge of $2^n$ $(nk-1)$-dimensional

1108: spheres, with the $\zz$-action swapping them in pairs, and so

1109: $\thomp(C_{2r+1},K_n)/\zz$ is homotopy equivalent to a~wedge of

1110: $2^{n-1}$ $(nk-1)$-dimensional spheres.

1111: \qed \vspace{5pt}

1112: 

1113: We summarize the estimates which we will need later.

1114: 

1115: \begin{crl}

1116: We have $\wti H^i(\thomp(C_{2r+1},K_n)/\zz)=0$ for $r\geq 2$, $n\geq

1117: 5$, and $i\leq n+r-2$. Except for the case $r=3$.

1118: \end{crl}

1119: 

1120: \pr If $2r+1=3k$, the inequality $nk-2\geq n+r-2$ is equivalent to

1121: $n\geq 3r/(2r-2)$, which is true for $n\geq 3$, $r\geq 2$. If

1122: $2r+1=3k-1$, then $nk/2\geq n+r-2$ is equivalent to $(n-3)(k-2)\geq

1123: 0$, again true for our parameters.

1124: 

1125: If $2r+1=3k+1$, then $nk/2\geq n+r-2$ is equivalent to $(n-3)(k-2)\geq

1126: 2$. This is true for all parameters $n\geq 5$, $k\geq 2$, except for

1127: $k=2$.  

1128: \qed

1129: 

1130: \subsection{The cell complex of homomorphisms from a~tree

1131:   to a~complete graph.} $\,$ %\vspace{5pt}

1132: \label{ss4.2}

1133: 

1134: In the next proposition we summarize several results proved

1135: in~\cite{BK2,K4}.

1136: \begin{prop}\label{pr:BK2}\cite[Propositions 4.3, 5.4, and 5.5]{BK2}, 

1137: \cite{K4}.

1138: 

1139: \nin Let $T$ be a~tree with at least one edge.

1140: 

1141: \nin {\rm (i)} The map $i_{K_n}:\thom(T,K_n)\ra\thom(K_2,K_n)$ induced

1142: by any inclusion $i:K_2\hookrightarrow T$ is a~homotopy equivalence.

1143: 

1144: \nin {\rm (ii)} $\thom(K_2,K_n)$ as a boundary complex of a polytope

1145: of dimension $n-2$, in particular $\thom(T,K_n)\simeq S^{n-2}$.

1146: 

1147: \nin {\rm (iii)} Given a~$\zz$-action determined by an~invertible

1148: graph homomorphism $\gamma:T\ra T$, if $\gamma$ flips an~edge in $T$,

1149: then $\thom(T,K_n)\simeq_\zz S^{n-2}_a$, otherwise

1150: $\thom(T,K_n)\simeq_\zz S^{n-2}_t$.

1151: \end{prop}

1152: 

1153: \nin Here $S^{m}_a$ denotes the $m$-sphere equipped with an antipodal

1154: $\zz$-action, whereas $S^{m}_t$ is the $m$-sphere equipped with the

1155: trivial one.

1156: 

1157: Let $F$ be any graph, with $F_1,\dots,F_t$ being the list of all those

1158: connected components of $F$ which have at least 2 vertices. For any

1159: $\emptyset\neq S\subseteq[1,t]$, and $V=\{v_i\}_{i\in S}$, such that

1160: $v_i\in V(F_i)$, for any $i\in S$, set

1161: $$\alpha_+(F,V):=\sum_{\eta}\eta^*_+,\quad\alpha(F,V):=\sum_{\eta}\eta^*,$$

1162: where both sums are taken over all $\eta:V(F)\ra

1163: 2^{[1,n]}\sm\{\emptyset\}$, such that

1164: \begin{itemize}

1165: \item $\eta(v_i)=[1,n-1]$, for all $i\in S$;

1166: \item $|\eta(w)|=1$, for all $w\in V(F)\sm V$.

1167: \end{itemize}

1168: Note that, for fixed $S$ and $V$, $(-1)^{c(\eta)}$ does not depend on

1169: the choice of $\eta$ as long as $\eta$ satisfies these two

1170: conditions. In our previous notations we have $\alpha_+(F,V)\in

1171: X^{|S|(n-2)}(F,K_n)$, and $\alpha(F,V)\in

1172: C^{|S|(n-2)}(\thom(F,K_n))$. When $|S|=1$, $V=\{v\}$, we shall simply

1173: write $\alpha_+(F,v)$ and $\alpha(F,v)$.

1174: 

1175: Assume now that $F$ is a forest. For $w\in V(F_i)$, such that

1176: $(v_i,w)\in E(F)$, set

1177: $W:=\{v_1,\dots,v_{i-1},w,v_{i+1},\dots,v_t\}=(V\cup\{w\})\sm\{v_i\}$.

1178: We have a~graph homomorphism $K_2\ra(v,w)$, which induces

1179: a~$\zz$-equivariant map $\varphi^*:H^*(\thom(K_2,K_n))\ra

1180: H^*(\thom(F,K_n))$.  We know that $\thom(K_2,K_n)\cong_\zz S^{n-2}_a$,

1181: and that the dual of any $(n-2)$-dimensional cell of $\thom(K_2,K_n)$

1182: is a~generator of $H^{n-2}(\thom(K_2,K_n);\dz)$. Comparing

1183: orientations of the cells of $\thom(K_2,K_n)$ we see that

1184: $[\alpha(K_2,1)]=(-1)^{n-1}[\alpha(K_2,2)]$, where $1$ and $2$ denote

1185: the vertices of $K_2$.  Applying $\varphi^*$ we conclude that

1186: \[ [\alpha(F,V)]=(-1)^{n-1}[\alpha(F,W)].\]

1187: Since $\rho^*$

1188: is a~cochain isomorphism and

1189: $\rho^*(\alpha_+(F,V))=(-1)^{c(\eta)}\alpha(F,V)$, we have

1190: \begin{equation} \label{eq:my7}

1191: [\alpha_+(F,V)]=

1192:   \begin{cases}

1193:     -[\alpha_+(F,W)],& \text{if $v$ and $w$ have different} \\ 

1194:                      & \text{parity in the order on $V(F)$}; \\

1195:     (-1)^{n-1}[\alpha_+(F,W)],& \text{if they have the same parity.}

1196:   \end{cases}

1197: \end{equation}

1198: 

1199: 

1200: \subsection{The $E_1^{*,*}$-tableau for $E_1^{p,q}\Rightarrow 

1201: H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}\footnote{The calculations performed in the subsections 

1202: \ref{ssect:4.3}--\ref{ssect:4.8} have been verified and generalized 

1203: in~\cite{ccc}.}  $\,$ \vspace{5pt} \label{ssect:4.3}

1204: 

1205: Let us fix integers $r\geq 2$ and $n\geq 4$. Let

1206: $(F^p)_{p=0,\dots,|V(G)|-1}$ be the filtration on

1207: $C^*(\thomp(C_{2r+1},K_n);\dz)$ defined in subsection~\ref{ss3.3}, and

1208: consider the corresponding spectral sequence. The entries of the

1209: $E_1$-tableau are given by $E_1^{p,q}=H^{p+q}(F^p,F^{p+1})$.  Since

1210: all proper subgraphs of $C_{2r+1}$ are forests, we can now use the

1211: formula~\eqref{eq:E0} to obtain almost complete information about the

1212: $E_1$-tableau. See Figure~\ref{fig:1e1}, where all the entries outside

1213: of the shaded area are equal to~0

1214: 

1215: Let $\emptyset\neq S\subset V(C_{2r+1})$, and let

1216: $S_1,\dots,S_{l(S)}$ be the connected components of $C_{2r+1}[S]$,

1217: with $|S_{1}|\geq|S_{2}|\geq\dots\geq |S_{d(S)}|>|S_{d(S)+1}|=

1218: \dots=|S_{l(S)}|=1$, where $l(S)\geq 1$, but we may have $d(S)=0$ or

1219: $d(S)=l(S)$. By Proposition~\ref{pr:BK2} together with property (3)

1220: from \cite[subsection 2.4]{BK2} we see that

1221: \begin{equation}\label{eq:4.3a}

1222: \thom(C_{2r+1}[S],K_n)\simeq\prod_{i=1}^{d(S)}S^{n-2}.

1223: \end{equation}

1224: Combining with the formula~\eqref{eq:E0} we conclude

1225: \begin{equation}

1226:   \label{eq:4.3b}

1227:   H^{p+q}(F^p,F^{p+1})=\bigoplus_{\begin{subarray}{c}

1228: {S\subset V(C_{2r+1})}\\{|S|=p+1}\end{subarray}}

1229: H^q\left(\prod_{i=1}^{d(S)}S^{n-2};\dz\right),

1230: \end{equation}

1231: for $p\leq 2r-1$.

1232: 

1233: %\[\pstexInput{1e1}\]

1234: %\[\text{ Figure 2. The $E_1^{*,*}$-tableau, for 

1235: %$E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}\]

1236: 

1237: 

1238: \begin{figure}[hbt]

1239: \begin{center}

1240: \pstexInput{1e1}

1241: \end{center}

1242: \caption{The $E_1^{*,*}$-tableau, for 

1243: $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1244: \label{fig:1e1}

1245: \end{figure}

1246: 

1247: 

1248: 

1249: Since the spectral sequence converges to $\wti H^*

1250: (\thomp(C_{2r+1},K_n) ;\dz)$, and, since by Corollary~\ref{crl:h+},

1251: $\wti H^{i} (\thomp(C_{2r+1},K_n);\dz)=0$ for $i\leq n+2r-2$, we know

1252: that the entries on the diagonals $p+q=n+2r-2$, and $p+q=n+2r-3$,

1253: should eventually all become~$0$.

1254: 

1255: \subsection{The cochain complex $(D_0^*,d_1)=(E_1^{*,0},d_1)$, for

1256: $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1257: $\,$ \vspace{5pt}

1258: 

1259: Let $(D_i^*,d_1)$ denote the cochain complex in the $i(n-2)$-th row of

1260: $E_1^{*,*}$, for any

1261: $i=0,\dots,\left\lfloor\dfrac{2r+1}{3}\right\rfloor$. Next we show

1262: that $(D_0^*,d_1)$ is isomorphic to the cochain complex of a~simplex.

1263: 

1264: \begin{lm}

1265: We have $E_2^{0,0}=\dz$, and

1266: $E_2^{1,0}=E_2^{2,0}=\dots=E_2^{2r,0}=0$.

1267: \end{lm}

1268: 

1269: \pr Let $\Delta_{2r}$ denote and abstract simplex with $2r+1$ vertices

1270: indexed by $[1,2r+1]$, and identify simplices of $\Delta_{2r}$ with

1271: the subsets of $[1,2r+1]$. Let $(C^*(\Delta_{2r};\dz),d^*)$ be the

1272: cochain complex of $\Delta_{2r}$ corresponding to the order on the

1273: vertices given by this indexing. By (\ref{eq:4.3b}), each $S\subseteq

1274: V(C_{2r+1})$, $|S|=p+1$, contributes one independent generator

1275: (over~$\dz$) to $E_1^{p,0}$. Identifying it with the generator in

1276: $C^*(\Delta_{2r};\dz)$ of the corresponding $p$-simplex in $\da_{2r}$,

1277: we see that $(D_0^*,d_1)$ and $(C^*(\Delta_{2r};\dz),d^*)$ are

1278: isomorphic as cochain complexes.

1279: 

1280: Indeed, for such $S$,

1281: $\tau_S:=\sum_{\varphi\in\chom(C_{2r+1}[S],K_n)}\varphi^*_+$ is

1282: a~representative of the corresponding generator in $E_1^{p,0}$.  This

1283: is true even for $S=V(C_{2r+1})$, since $\thom(C_{2r+1},K_n)$ is

1284: connected for $n\geq 4$, as was shown in~\cite[Proposition 2.1]{BK2}.

1285: Clearly, 

1286: \begin{multline*}  %\label{eq:4.3c}

1287:  d_1(\tau_S)=

1288: \sum_{\varphi\in\chom(C_{2r+1}[S],K_n)}\sum_{v\notin S}

1289: \sum_{\psi|S=\varphi}[\varphi_+:\psi_+]\psi^*_+=\\

1290: \sum_{v\notin S}[S:S\cup\{v\}]

1291: \sum_{\psi\in\chom(C_{2r+1}[S\cup\{v\}],K_n)}\psi^*_+=

1292: \sum_{v\notin S}[S:S\cup\{v\}]\tau_{S\cup\{v\}},

1293: \end{multline*}

1294: where the second equality is true since $[\varphi_+:\psi_+]$ only

1295: depends on $S$ and $v$, not on the specific choice of $\varphi$

1296: and~$\psi$.  This shows that the following diagram commutes

1297: \[\begin{CD}

1298: C^p(\Delta_{2r};\dz)@>d^p>>C^{p+1}(\Delta_{2r};\dz) \\

1299:         @V\tau_\cdot VV           @VV\tau_\cdot V\\

1300: E_1^{p,0} @>d_1>>E_1^{p+1,0}, \\

1301: \end{CD}\]

1302: where $\tau_\cdot:C^*(\Delta_{2r};\dz)\ra E_1^{*,0}$ is the linear

1303: extension of the map taking $S$ to $\tau_S$, for $S\subseteq

1304: V(C_{2r+1})$.  It follows that $(D_0^*,d_1)$ is isomorphic to

1305: $(C^*(\Delta_{2r};\dz),d^*)$, therefore $E_2^{0,0}=\dz$, and

1306: $E_2^{1,0}=E_2^{2,0}=\dots=E_2^{2r,0}=0$.

1307: \qed

1308: 

1309: \subsection{The cochain complexes $(D_t^*,d_1)=(E_1^{*,(n-2)t},d_1)$, 

1310: for $\left\lfloor(2r+1)/3\right\rfloor\geq t\geq 2$, and

1311: $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.} 

1312: $\,$ \vspace{5pt}

1313: \label{ss4.5}

1314: 

1315: We shall perform only a~partial analysis of the cohomology groups

1316: of $(D_t^*,d_1)$, which will however be sufficient for our purpose.

1317: 

1318: For $S\subset V(C_{2r+1})$ and $v\in S$, let $a(S,v)$ denote the arc

1319: of $S$ to which $v$ belongs (assuming this arc exists). Furthermore,

1320: for an~arbitrary arc $a$ of $S$, let $a=[a_\bu,a^\bu]_{2r+1}$. Let

1321: $|a|$ denote the number of vertices on $a$, and set $\widehat a:=

1322: [a_\bu-1,a^\bu+1]_{2r+1}$ (so $|\widehat a|=|a|+2$, if $|a|\leq 2r-1$).

1323: 

1324: For any $V\subseteq S\subseteq V(C_{2r+1})$, as in

1325: subsection~\ref{ss4.2}, set

1326: $\sigma_{S,V}:=\alpha_+(C_{2r+1}[S],V)$. By our previous observations,

1327: $E_1^{0,t(n-2)}=0$. Furthermore, for any $1\leq i\leq 2r-1$,

1328: $E_1^{i,t(n-2)}$ is a~free $\dz$-module with the basis

1329: $\{[\sigma_{S,V}]\}$, where $S\subset V(C_{2r+1})$, $|S|=i+1$,

1330: $|V|=t$, and $v=a_\bu(S,v)$ (i.e., $[v-1]_{2r+1}\notin S$) for all

1331: $v\in V$. Since $\sigma_{S,v}$ is a~cocycle in

1332: $X^{n-2}(C_{2r+1}[S],K_n;\dz)$, we have

1333: \begin{equation} \label{eq:my6}

1334:   d_1([\sigma_{S,v}])=\sum_{w\notin S}(-1)^{z(w)}[\sigma_{S\cup\{w\},v}],

1335: \end{equation}

1336: where

1337: $$z(w)=

1338: \begin{cases}

1339:   \left|S\cap[1,w-1]\right|,& \text{if } v\notin[1,w-1];\\

1340: n+\left|S\cap[1,w-1]\right|,& \text{if } v\in[1,w-1].

1341: \end{cases}

1342: $$ Note, that if $i\leq 2r-2$ and $[w]_{2r+1}=[v-1]_{2r+1}$, then

1343: $v\neq a_\bu(S\cup\{w\},w)$, so $[\sigma_{S\cup\{w\},v}]$ may differ

1344: by a~sign from one of the elements in our chosen basis.  We shall not

1345: need the analog of the equation~\eqref{eq:my6} for the case $|V|\geq

1346: 2$.

1347: 

1348: Let $A^*_1$ be the subcomplex of $D_1^*$ defined by:

1349: $$A^*_1:0\lra\wti E_1^{2r-2,n-2}{\stackrel{d_1}\lra} E_1^{2r-1,n-2}

1350: {\stackrel{d_1}\lra} E_1^{2r,n-2}\lra 0,$$

1351: where the $\dz$-modules

1352: indexed with $0,\dots,2r-3$ are equal to~0, and $\wti E_1^{2r-2,n-2}$

1353: is generated by $\{[\sigma_{S,v}]\}$, such that $S$ and $v$ satisfy

1354: all the previously required conditions and, in addition, $C_{2r+1}[S]$

1355: is connected. 

1356: 

1357: In general, let $A^*_t$ be the subcomplex of $D^*_t$ generated by all

1358: $\{[\sigma_{S,V}]\}$, such that

1359: \begin{equation}\label{eq:hat}

1360: \bigcup_{v\in V}\widehat{a(S,v)}=V(C_{2r+1}).

1361: \end{equation}

1362: In words: the gaps between those arcs of $S$ which have points in $V$

1363: are of length at most~$2$. For future reference, we note, that

1364: \eqref{eq:hat} implies that $|S|+2|V|\geq 2r+1$, i.e., $|S|-1\geq

1365: 2r-2t$, hence $A^j_t=0$ for $j<2r-2t$.

1366: 

1367: \begin{lm}

1368: We have $H^*(D_t^*)=H^*(A_t^*)$.

1369: \end{lm}

1370: 

1371: \pr

1372: Let us set up another spectral sequence for computing the cohomology

1373: of the relative complex $(D_t^*,A^*_t)$. We filter by $\sum_{v\in V}

1374: |a(S,v)|$. More precisely,

1375: $F^p(D_t^*,A^*_t)=\dz[[\sigma_{S,V}]\,\big{|}\,\sum_{v\in V}

1376: |a(S,v)|\geq p]$.  We see that $F^p(D_t^*,A^*_t)/F^{p+1}(D_t^*,A^*_t)=

1377: \dz[[\sigma_{S,v}]\,\big{|}\,\sum_{v\in V}|a(S,v)|=p]$, hence

1378: $$E_1^{p,q}(D_t^*,A^*_t)=H^{p+q}(F^p(D_t^*,A^*_t)/F^{p+1}(D_t^*,A^*_t))=

1379: \bigoplus_{a_1,\dots,a_t} H^{p+q}(M_{a_1,\dots,a_t}^*),$$

1380: where the sum is taken over all possible $t$-tuples of

1381: arcs $a_1,\dots,a_t$ such that

1382: \begin{enumerate}

1383: \item[(1)] $a_i\cap\widehat{a_j}=\emptyset$, for any $i\neq j$,

1384: $i,j\in[1,t]$;

1385: \item[(2)] $|a_1|+\dots+|a_t|=p$;

1386: \item[(3)] $\bigcup_{v\in V}\widehat{a(S,v)}\neq V(C_{2r+1})$,

1387: \end{enumerate}

1388: and $M_{a_1,\dots,a_t}^*$ is the~cochain subcomplex generated by all

1389: $\{[\sigma_{S,v}]\}$, such that the arcs with vertices in $V$ are

1390: precisely $a_1,\dots,a_t$, i.e., $\{a(S,v)\,|\,v\in

1391: V\}=\{a_1,\dots,a_t\}$.

1392: 

1393: Restricting the formula~\eqref{eq:my6} to $M_a^*$, we see that $M_a^*$

1394: is isomorphic to the cochain complex $C^*(\da_{2r-p-2};\dz)$.  More

1395: generally, we see that $M_{a_1,\dots,a_t}^*$ is isomorphic to

1396: $C^*(\da_{2r-\ti p};\dz)$, where $\ti p=\left|\bigcup_{v\in

1397: V}\widehat{a(S,v)}\right|$.

1398: 

1399: As mentioned, $\ti p\leq 2r$, hence $M_{a_1,\dots,a_t}^*$ is acyclic

1400: for any $a_1,\dots,a_t$ satisfying the above conditions. We conclude

1401: that $(D_t^*,A^*_t)$ is acyclic. The long exact sequence for the

1402: relative cohomology implies that $H^*(A^*_t)=H^*(D^*_t)$. \qed \vspace{5pt}

1403:  

1404: Now, we can show that $E^{i,t(n-2)}_2=0$ for $t\geq 2$,

1405: $i<2r-(t-1)(n-2)$, that is $E^{*,*}_2$ is $0$ in the region strictly

1406: above row $n-2$ and strictly below the diagonal $x+y=2r+n-2$, see

1407: Figure~\ref{fig:3e1}. Indeed, this is immediate when $2r-2t\geq

1408: 2r-(t-1)(n-2)$, which after cancellations reduces to $(n-4)(t-1)\geq

1409: 2$. The only cases when this inequality is false are $(t,n)=(2,5)$,

1410: and $n=4$.

1411: 

1412: %\[\pstexInput{3e1}\]

1413: %\begin{center}

1414: % Figure 3.  The possibly nonzero entries in $E_2^{*,*}$-tableau,

1415: % for $E_2^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.

1416: %\end{center}

1417: 

1418: 

1419: \begin{figure}[hbt]

1420: \begin{center}

1421: \pstexInput{3e1}

1422: \end{center}

1423: \caption{The possibly nonzero entries in $E_2^{*,*}$-tableau,

1424:  for $E_2^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1425: \label{fig:3e1}

1426: \end{figure}

1427: 

1428: 

1429: \subsection{The case $n=5$, for $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1430: $\,$ \vspace{5pt}

1431: 

1432: Assume now that $n=5$, $t=2$.

1433: 

1434: \begin{lm} 

1435: We have $E^{2r-4,6}_2=0$. 

1436: \end{lm}

1437: 

1438: \pr By dimensional argument, this is true if $2r+1<8$, so we can

1439: assume that $r\geq 4$.

1440: 

1441: By our previous arguments we need to see that $d_1:A^{2r-4,6}_1\ra

1442: A^{2r-3,6}_1$ is an~injective map. The generators of $A^{2r-4,6}_1$

1443: can be indexed with unordered pairs $\{v,w\}$, $v,w\in V(C_{2r+1})$, such

1444: that 

1445: \[[v-1,v+2]_{2r+1}\cap[w-1,w+2]_{2r+1}=\emptyset,\] 

1446: whereas the generators of $A^{2r-3,6}_1$ can be indexed with ordered

1447: pairs $(v,w)$, $v,w\in V(C_{2r+1})$, such that

1448: \[[v-1,v+2]_{2r+1}\cap[w-1,w+1]_{2r+1}=\emptyset.\] 

1449: 

1450: In these notations we have

1451: \begin{equation}

1452: \label{eq:n5diff}

1453: d_1(\{v,w\})=\epsilon_1 (v,w)+\epsilon_2 (v,[w+1]_{2r+1})+

1454: \epsilon_3 (w,v)+\epsilon_4 (w,[v+1]_{2r+1}),

1455: \end{equation}

1456: where $\epsilon_1,\epsilon_2,\epsilon_3,\epsilon_4\in\{-1,1\}$.

1457: 

1458: Take $0\neq\sum_{v,w}\alpha_{v,w}\{v,w\}\in\ker d_1$. Choose $v,w$

1459: such that $\alpha_{v,w}\neq 0$, and the minimum of the two distances

1460: between the arcs $\{v,[v+1]_{2r+1}\}$ and $\{w,[w+1]_{2r+1}\}$ is

1461: minimized. By symmetry we may assume $[w-v-1]_{2r+1}\leq

1462: [v-w-1]_{2r+1}$. Then, it follows from \eqref{eq:n5diff} that

1463: $\alpha_{[v+1]_{2r+1},w}\neq 0$ as well. 

1464: 

1465: Either $\{[v+1]_{2r+1},w\}$ is not a~well-defined pair or the minimal

1466: distance between the two arcs is smaller for this pair, than for

1467: $\{v,w\}$: $[w-v-1]_{2r+1}\geq [w-v-2]_{2r+1}$. Both ways we get

1468: a~contradiction to the assumption that there exists $\{v,w\}$, such

1469: that $\alpha_{v,w}\neq 0$. We conclude that $d_1:A^{2r-4,6}_1\ra

1470: A^{2r-3,6}_1$ is injective, hence $E^{2r-4,6}_2=0$. \qed \vspace{5pt}

1471: 

1472: This shows, that when $n\geq 5$, there are no higher differentials

1473: $d_i$, $i\geq 2$, in our spectral sequence, originating in the region

1474: above row $n-2$ and below diagonal $x+y=2r+n-2$. Hence, to figure out

1475: what happens to the entries $E_{\infty}^{2r,n-2}$ and

1476: $E_{\infty}^{2r,n-3}$, it is sufficient to consider rows $n-2$ and

1477: $n-3$.

1478: 

1479: 

1480: %\[\pstexInput{e1n5}\]

1481: %\begin{center} 

1482: %Figure 4. A part of the $E_1^{*,*}$-tableau, for $n=5$, and 

1483: %$E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.

1484: %\end{center}

1485: 

1486: 

1487: 

1488: \begin{figure}[hbt]

1489: \begin{center}

1490: \pstexInput{e1n5}

1491: \end{center}

1492: \caption{A part of the $E_1^{*,*}$-tableau, for $n=5$, and 

1493: $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1494: \label{fig:e1n5}

1495: \end{figure}

1496: 

1497: 

1498: \subsection{The case $n=4$, for $E_1^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n);\dz)$.}

1499: $\,$ \vspace{5pt}

1500: 

1501: For $n=4$ the nonzero rows of $E_1^{*,*}$ are too close to each other,

1502: so we are to do the computation by hand in a~somewhat detailed

1503: way. Since the reduction from $(D_t^*,d_1)$ to $(A_t^*,d_1)$ described

1504: in subsection~\ref{ss4.5} was valid when $n=4$, we may concentrate on

1505: the study of the latter complex. Let us first deal with $(A_2^*,d_1)$. 

1506: 

1507: \begin{lm}

1508: We have $H^{2r-2}(A_2^*)=H^{2r-3}(A_2^*)=\dz$, and

1509: $H^{i}(A_2^*)=0$, for $i\neq 2r-2,2r-3$.

1510: \end{lm}

1511: 

1512: \pr We filter $A_2^*$ by

1513: \[F^p A_2^*=\dz\left[\left. [\sigma_{S,\{v_1,v_2\}}]\,\right|\, 

1514: \min(|a(S,v_1)|,|a(S,v_2)|)\geq p\right].\]

1515: Clearly, $\dots\subseteq F^p\subseteq F^{p+1}\subseteq \dots$.

1516: Inspecting the case $p\leq r-2$, we see that in this situation

1517: \[C^*(F^p A_2^*/F^{p-1} A_2^*)=\bigoplus_{i=1}^{2r+1}B_i,\]

1518: where each $B_i$ is isomorphic to $C^*(\Delta_1)$, hence is acyclic.

1519: 

1520: It follows that $H^*(A_2^*)=H^*(F^{r-1} A_2^*/F^{r-2} A_2^*)$.  Let

1521: $\sigma_i=\sigma_{S_i,V_i}$, where

1522: $S_i=[i+1,i+r-1]_{2r+1}\cup[i+r+2,i-1]_{2r+1}$, $V_i=\{i+1,i+r+2\}$,

1523: and let $\tau_i=\sigma_{\wti S_i,V_i}$, where $\wti

1524: S_i=S_i\cup\{[i+r]_{2r+1}\}$. Clearly,

1525: $d_1(\sigma_i)=\pm\tau_i\pm\tau_{i+r}$, and to verify the statement of

1526: the lemma we need to show that the number of those $\sigma_i$, for

1527: which $d_1(\sigma_i)=\pm(\tau_i+\tau_{i+r})$ is even. By

1528: \eqref{eq:my7} and~\eqref{eq:my6} we see that 

1529: $d_1(\sigma_i)=\pm(\tau_i-\tau_{i+r})$ if $i+r\neq 2r,2r+1$, i.e., the

1530: only cases we need to consider are $i=r$ and $i=r+1$. 

1531: 

1532: If $i=2r+1$, the different sign comes from \eqref{eq:my6}, and the

1533: sign contribution is $2r+2$. This is an even number, hence again

1534: $d_1(\sigma_i)=\pm(\tau_i-\tau_{i+r})$. If $i=2r$, the different sign

1535: comes from \eqref{eq:my7}, but since $n=4$ is even, the sign remains

1536: the same. \qed \vspace{5pt}

1537: 

1538: Next, we consider $(A_t^*,d_1)$, for $t\geq 3$.  First, we introduce

1539: some additional notations. Since the sign will not matter in our

1540: argument, we write $\sigma_S$ instead of $\sigma_{S,V}$, it is then

1541: defined only up to a~sign. For $S\subset V(C_{2r+1})$, $\bar

1542: S=V(C_{2r+1})\sm S$; the connected components of $C_{2r+1}[\bar S]$

1543: are called {\it gaps}. Each gap consists of either one or two

1544: elements, we call the first ones {\it singletons}, and the second ones

1545: {\it double gaps}.  Let $m(S)$ be the leftmost element of the gap

1546: which contains $\min(\bar S \cap[2,2r]_{2r+1})$. For $s\in\bar S$, let

1547: $\overleftarrow{s}$ be the leftmost element of the first gap to the

1548: left of the gap containing $s$, and let $\overrightarrow{s}$ be the

1549: leftmost element of the first gap to the right of the gap containing

1550: $s$. For $x,y\in V(C_{2r+1})$, let $d(x,y)$ denote

1551: $\left|[x,y]_{2r+1}\right|-1$.

1552: 

1553: \begin{lm} \label{lm:4.11}

1554: We have $E_2^{2r-2t,2t}=E_2^{2r-2t+1,2t}=0$.

1555: \end{lm}

1556: 

1557: \nin{\bf Proof.}\footnote{It is possible to rephrase this argument 

1558: in terms of matchings on chain complexes, see~\cite{dmt}.}  Clearly

1559: $A_t^i=0$, unless $2r-2t\leq i\leq 2r-t$. Note that if $\sigma_S$ is

1560: a~generator of $A_t^{2r-2t+1}$, then $S$ has exactly one singleton. We

1561: decompose $A_t^{2r-2t+1}=B_1\oplus B_2\oplus B_3\oplus B_4$, where

1562: each $B_i$ is spanned by the generators $\sigma_S$, for which certain

1563: conditions are satisfied, see Figure~\ref{fig:bcases}:

1564: \begin{enumerate}

1565: \item[($B_1$)] $\overleftarrow{m(S)}$ is the singleton and

1566: $d(\overleftarrow{m(S)},m(S))=3$, or $m(S)$ is the singleton and

1567: $d(\overleftarrow{m(S)},m(S))=4$;

1568: \item[($B_2$)] $\overleftarrow{m(S)}$ is

1569: the singleton, and $d(\overleftarrow{m(S)},m(S))\geq 4$;

1570: \item[($B_3$)] $m(S)$ and $\overleftarrow{m(S)}$ are in double gaps;

1571: \item[($B_4$)] $m(S)$ is the singleton, and

1572: $d(\overleftarrow{m(S)},m(S))\geq 5$.

1573: \end{enumerate}

1574: 

1575: \begin{figure}[hbt]

1576: \begin{center}

1577: \pstexInput{bcases}

1578: \end{center}

1579: \caption{The 4 cases in the proof of Lemma~\ref{lm:4.11}.}

1580: \label{fig:bcases}

1581: \end{figure}

1582: 

1583: Let $(\wti A_t^*,d_1)$ be the complex spanned by $B_1$, $B_2$, $B_3$,

1584: and $A^i_t$, for $2r-2t+2\leq i\leq 2r-t$. The relative complex

1585: $(A_t^*/\wti A_t^*,d_1)$ has only cochains in dimensions $2r-2t$ and

1586: $2r-2t+1$. It is easy to see that the projection $d_1:A_t^{2r-2t}\ra

1587: A_t^{2r-2t+1}/\wti A_t^{2r-2t+1}=B_4$ is an isomorphism, with the

1588: inverse given by $\sigma_S\mapsto \sigma_{S\sm\{m(S)-1\}}$, where

1589: $\sigma_S$ is a~generator from $B_4$. Hence $(A_t^*/\wti A_t^*,d_1)$

1590: is acyclic, and we are led to study the complex $(\wti A_t^*,d_1)$.

1591: 

1592: Next, we show that $H^{2r-2t+1}(\wti A_t^*)=0$, which is the same as

1593: to say that $d_1$ is injective on $B=B_1\oplus B_2\oplus B_3$.  Let

1594: $\sigma\neq 0$ be in $\ker d_1(B)$. We think of $\sigma$ as a~linear

1595: combination of the generators from the descriptions of $B_1$, $B_2$,

1596: and $B_3$, and let $M$ be the set of generators which have a~nonzero

1597: coefficient in~$\sigma$. 

1598: 

1599: Assume $M$ contains a generator $\sigma_S$ from $B_3$, and choose

1600: $\sigma_S$ so that $d(m(S),\overrightarrow{m(S)})$ is minimized.  The

1601: coboundary $d_1(\sigma_S)$ contains a~copy of

1602: $\sigma_{S\cup\{m(S)\}}$.  At most~3 other generators will contain

1603: $\sigma_{S\cup\{m(S)\}}$ in the coboundary, depending on which element

1604: we remove from $S\cup\{m(S)\}$ instead of $m(S)$. Since we have chosen

1605: $d(m(S),\overrightarrow{m(S)})$ to be minimal, we cannot remove

1606: $m(S)+2$. Hence, we must remove an~element extending the singleton gap

1607: which is not $m(S)+1$. This gives a~generator of $B_4$, yielding

1608: a~contradiction. 

1609: 

1610: From now on we may presume that $M$ contains no generators from~$B_4$

1611: or~$B_3$.  Assume now $M$ contains a generator $\sigma_S$ from

1612: $B_2$. Again, choose $\sigma_S$ so that

1613: $d(m(S),\overrightarrow{m(S)})$ is minimized, and note that

1614: $d_1(\sigma_S)$ contains a~copy of $\sigma_{S\cup\{m(S)\}}$.

1615: Examining the generators which contain $\sigma_{S\cup\{m(S)\}}$ in the

1616: coboundary, we see again that, since removing $m(S)+2$ from

1617: $S\cup\{m(S)\}$ would contradict the minimality, we must remove

1618: an~element extending the singleton gap $\overleftarrow{m(S)}$. This

1619: way we will produce a~generator of $B_4$, except for one case: when

1620: $\overleftarrow{m(S)}=1$, and we remove vertex~2. In this case we

1621: produce a~generator from~$B_3$, hence again a~contradiction.

1622: 

1623: Finally, assume $M$ consists only of generators from $B_1$. Choose

1624: $\sigma_S$ so that $m(S)$ is maximized. Assume first that $m(S)$ is in

1625: a~double gap, and consider the copy of $\sigma_{S\cup\{m(S)\}}$ in

1626: $d_1(\sigma_S)$. There are 3 possibilities. Removing $m(S)+2$ gives

1627: a~generator of $B_2$, whereas removing $m(S)-4$ gives a generator

1628: of~$B_4$. Removing $m(S)-2$ gives either a~generator of $B_3$, if

1629: $m(S)=4$, or a~generator $\sigma_T$ of $B_1$, such that $m(T)=m(S)+1$.

1630: In either case we get a~contradiction.

1631: 

1632: Assume now that $m(S)$ is a~singleton, and examine the copy of

1633: $\sigma_{S\cup\{\overleftarrow{m(S)}\}}$ in $d_1(\sigma_S)$.  There is

1634: only one possibility for deletion: remove $m(S)+1$.  This will produce

1635: a generator $\sigma_T$ of $B_1$, with $m(T)=m(S)$, but such that

1636: $m(T)$ is in a~double gap, the case which we have already dealt with.

1637: 

1638: This finishes the proof that $H^{2r-2t+1}(\wti A_t^*)=0$, which,

1639: combined with the acyclicity of $(A_t^*/\wti A_t^*,d_1)$, and the fact

1640: that $H^*(A^*_t)=H^*(D^*_t)$, yields

1641: $E_2^{2r-2t,2t}=E_2^{2r-2t+1,2t}=0$. \qed

1642: 

1643: 

1644: 

1645: \subsection{Finishing the computation of $H^{n-2}(\thom(C_{2r+1},K_n);R)$

1646: and of $H^{n-3}(\thom(C_{2r+1},K_n);R)$,  for

1647: $R=\zz$ or $\dz$.} $\,$ \vspace{5pt} \label{ssect:4.8}

1648: 

1649: Let us now turn our attention to the cochain complex $A^*_1$. The

1650: generators of $\wti E_1^{2r-2,n-2}$ correspond to arcs of length

1651: $2r-1$ and can be indexed with the elements of $V(C_{2r+1})$: we set

1652: $\tau_{v,2}:=\sigma_{V(C_{2r+1})\sm\{v-2,v-1\},v}$, for any $v\in

1653: V(C_{2r+1})$.  The same way, the generators of $E_1^{2r-1,n-2}$

1654: correspond to arcs of length $2r$, we denote them by setting

1655: $\tau_{v,1}:=\sigma_{V(C_{2r+1})\sm\{v-1\},v}$, for any $v\in

1656: V(C_{2r+1})$.  It follows from~\eqref{eq:my6} that

1657: $$d_1([\tau_{v,2}])=(-1)^{v+1}[\tau_{v,1}]+(-1)^v[\tau_{v-1,1}],$$ for

1658: $v=3,\dots,2r+1$, where the second sign follows from~\eqref{eq:my7};

1659: $$d_1([\tau_{2,2}])=(-1)^{n+1}[\tau_{2,1}]-[\tau_{1,1}],$$

1660: where

1661: the~first sign is determined by the fact that there are $n+2r-3$

1662: vertices before the one inserted at position $2r+1$;

1663: $$d_1([\tau_{1,2}])=(-1)^{n+1}[\tau_{1,1}]+(-1)^n[\tau_{2r+1,1}],$$

1664: where we use again that there are $n+2r-3$ vertices before the

1665: inserted one, and, for determining the second sign, we use the fact

1666: that positions $1$ and $2r+1$ have different parity in

1667: $[1,2r+1]\sm\{2r\}$.

1668: 

1669: Summarizing, we have the following matrix for the first differential

1670: in $A^*_1$:

1671: $$M=

1672: \begin{bmatrix}

1673: (-1)^{n+1} &          0 & 0 &  0 & \dots & 0 & (-1)^n \\

1674:         -1 & (-1)^{n+1} & 0 &  0 & \dots & 0 & 0 \\

1675:          0 &         -1 & 1 &  0 & \dots & 0 & 0 \\

1676:          0 &          0 & 1 & -1 & \dots & 0 & 0 \\

1677:        \dots & \dots & \dots & \dots & \dots & \dots & \dots \\

1678:          0 &          0 & 0 &  0& \dots & -1 & 1

1679: \end{bmatrix}$$

1680: 

1681: Assume first that $n\geq 5$, and $(n,r)\neq(5,3)$.

1682: 

1683: \nin{\bf Case 1:} {\it $n$ is odd.}  It is easy to see that the 

1684: kernel of the differential $d_1:\wti E_1^{2r-2,n-2}\ra E_1^{2r-1,n-2}$

1685: is one-dimensional and is spanned by

1686: \[ [\tau_{1,2}]+[\tau_{2,2}]+[\tau_{3,2}]-[\tau_{4,2}]+[\tau_{5,2}]-

1687: [\tau_{6,2}]+\dots+[\tau_{2r+1,2}],\]

1688: while the image is

1689: \[\left\{\left[\sum_{i=1}^{2r+1}c_i\tau_{i,1}\right]\,\left|\,

1690: \sum_{i=1}^{2r+1}c_i=0\right.\right\}.\]

1691: 

1692: It follows that $E_2^{2r-2,n-2}=\dz$. Recall that, by

1693: Corollary~\ref{crl:h+}, the cohomology groups of

1694: $\thomp(C_{2r+1},K_n)$ vanish in dimension $n+2r-2$ and less. Hence,

1695: since $d_2:E_2^{2r-2,n-2}\ra E_2^{2r,n-3}$ must be an~isomorphism, we

1696: have $E_1^{2r,n-3}=E_2^{2r,n-3}=\dz$. On the other hand, the map

1697: $d_1:E_1^{2r-1,n-2}\ra E_1^{2r,n-2}$ is surjective, and

1698: $E_2^{2r-1,n-2}=E_2^{2r,n-2}=0$, so $E_1^{2r,n-2}=\dz$.

1699: 

1700: \nin{\bf Case 2:} {\it $n$ is even.} In this case the map

1701: $d_1:\wti E_1^{2r-2,n-2}\ra E_1^{2r-1,n-2}$ is injective. It follows

1702: that $E_2^{2r-2,n-2}=0$, and hence $E_1^{2r,n-3}=E_2^{2r,n-3}=0$. The

1703: image on the other hand is not the whole $E_1^{2r-1,n-2}$, but only

1704: \[\left\{\left[\sum_{i=1}^{2r+1}c_i\tau_{i,1}\right]\,\left|\,

1705: \sum_{i=1}^{2r+1}c_i\equiv 0\,(\text{mod }2)\right\}.\right.\]

1706: The fact that $E_2^{2r-1,n-2}=E_2^{2r,n-2}=0$ and the surjectivity of

1707: the map $d_1:E_1^{2r-1,n-2}\ra E_1^{2r,n-2}$ imply that

1708: $E_1^{2r,n-2}=\zz$. Again, we used that ${\wti H}^i

1709: (\thomp(C_{2r+1},K_n))$ vanish in dimension $n+2r-2$ and less.

1710: 

1711: 

1712: If $(n,r)=(5,3)$, then the argument above essentially holds, with the

1713: exception that $d_1:E_1^{2r-1,n-2}\lra E_1^{2r,n-2}$ does not have to

1714: be surjective. Instead, $\im d_1=\dz$ and $E_1^{2r,n-2}/\im

1715: d_1=\dz$. Thus $H^{n-2}(\thom(C_7,K_5);\dz)=E_1^{2r,n-2}=\dz^2$.

1716: 

1717: \begin{table}[hbt]

1718: \[\begin{array}{l|c|c|c|} 

1719: (n,r)&\,\,\,\,\,\,R\,\,\,\,\,\,& \,\,\,\,\,\,H^{n-2}\,\,\,\,\,\, &

1720: \,\,\, \,\,\,H^{n-3}\,\,\, \,\,\,\\ 

1721: \hline &&&\\[-0.35cm]

1722: 2\not{|}\,\,n, n\geq 5, (n,r)\neq (5,3)\,\,\, \,\,\, & \dz &\dz&\dz\\ 

1723: \hline &&&\\[-0.3cm]

1724: (n,r)=(5,3)& \dz  &\dz^2&\dz\\ \hline

1725: 2\,\,|\,\,n, n\geq 6, \text{ or } &&&\\

1726: n=4, r\leq 3 & \dz&\zz&0\\ \hline &&&\\[-0.35cm]

1727: n=4, r\geq 4& \dz  &\dz\oplus\zz&0\\ \hline

1728: n\geq 5, (n,r)\neq (5,3),\text{ or } &&&\\

1729: n=4, r\leq 3& \zz &\zz&\zz\\ \hline 

1730: (n,r)=(5,3), \text{ or } &&&\\

1731: n=4, r\geq 4 & \zz  &{\mathbb Z}_2^2&\zz\\ \hline

1732: \end{array}\]

1733: \caption{$\,$}

1734: \label{tab:answer}

1735: \end{table}

1736: 

1737: Assume, finally, that $n=4$. If $2r+1=5$, then the computations above

1738: hold. If $2r+1\geq 7$, the argument above still shows that the image

1739: of the map $d_1:E_1^{2r-1,2}\ra E_1^{2r,2}$ is $\dz[d_1(\tau_{1,r})]$,

1740: and $2 d_1(\tau_{1,r})=d_1(2\tau_{1,r})=0$. If $2r+1\geq 9$, we can

1741: compute $E_1^{2r,2}$ and $E_1^{2r,1}$ completely, since

1742: $H^i(\thomp(C_{2r+1},K_n))$ vanish in dimension $n+2r-2$ and less. In

1743: this case, the image of the map $d_1:E_1^{2r-1,2}\ra E_1^{2r,2}$ is

1744: $\zz$, and the map $d_2:E_2^{2r-3,4}\ra E_2^{2r,2}$ is

1745: an~isomorphism. Since, as we have shown earlier, $E_2^{2r-3,4}=\dz$,

1746: we conclude that $E_1^{2r,2}=\dz\oplus\zz$, and $E_1^{2r,1}=0$.

1747: 

1748: It follows that, for all $(n,r)$, $H^i(\thom(C_{2r+1},K_n);R)=0$, if

1749: $i\in [1,n-4]$, and $R=\dz$ or $\zz$\footnote{This has been

1750: strengthened to yield connectivity in \cite{CK2}, later a shorter

1751: proof appeared in~\cite{En05}.}. We summarize our computations of the

1752: next two cohomology groups in Table~1, where

1753: $H^i=H^i(\thom(C_{2r+1},K_n);R)$, and the case $(n,r)=(4,3)$ is

1754: conjectural.

1755: 

1756: \noindent

1757: {\bf Proof of Theorem~\ref{thm:even_n}.}

1758: 

1759: \noindent For $n\geq 6$, and $n=4$, $r\leq 3$, this follows from 

1760: the fact that the target group of the map is~$\zz$. For $n=4$, $r\geq

1761: 4$, we have shown above, that $2d_1(\tau_{1,r})=0$. By the

1762: construction, $\tau_{1,r}=\sigma_{V(C_{2r+1})\sm\{r-1\},r}$, so

1763: $d_1(\tau_{1,r})=\pm\sigma_{V(C_{2r+1}),r}$. Let $V(K_2)=\{1,2\}$, and

1764: pick a~nontrivial element $\alpha\in H^{n-2}(\thom(K_2,K_n);\dz)$ by

1765: setting $\alpha:=\eta^*$, $\eta(1):=[1,n-1]$,

1766: $\eta(2):=\{n\}$. Clearly,

1767: $\iota^*_{K_n}(\alpha)=\pm\sigma_{V(C_{2r+1}),r}$, where $\iota(1)=r$,

1768: $\iota(2)=r+1$. Thus, we see that $2\cdot\iota^*_{K_n}(\alpha)=\pm 2

1769: d_1(\tau_{1,r})=0$. \qed

1770: 

1771: 

1772: \subsection{The $\zz$-action on the cohomology groups of 

1773:   $\thom(C_{2r+1},K_n)$ for odd~$n$.} $\,$ %\vspace{5pt}

1774: 

1775: Throughout this subsection we assume that $n$ is odd, and that

1776: $(n,r)\neq (5,3)$. We tensor all our groups with $\mathbb C$ to

1777: simplify the representations. We denote by~$\chi_i$ the

1778: one-dimensional representation of $\zz$ given by the multiplication

1779: by~$(-1)^i$.

1780: 

1781: \begin{lm}\label{charlm1}

1782: We have $E_1^{2r,n-2}=\chi_r$, as a~$\zz$-module.

1783: \end{lm}

1784: \pr Recall that $\sigma_{V(C_{2r+1}),2r+1}:=\sum_{\eta}\eta^*_+$,

1785: where the sum is taken over all $\eta$, such that

1786: $\eta(2r+1)=[1,n-1]$, and $|\eta(i)|=1$, for all

1787: $i=1,\dots,2r$. $\sigma_{V(C_{2r+1}),2r+1}$ is a~representative of the

1788: generator of $E_1^{2r,n-2}$. Clearly, $\{\eta\circ\gamma\}=\{\eta\}$

1789: as a~collection of cells. To orient the cells in the standard way we

1790: need to reverse $\gamma$ as the permutation of $V(C_{2r+1})$. The sign

1791: of this is $(-1)^r$, hence

1792: $\gamma([\sigma_{V(C_{2r+1}),2r+1}])=(-1)^r[\sigma_{V(C_{2r+1}),2r+1}]$. \qed

1793: 

1794: \begin{lm}\label{charlm2}

1795: We have $E_1^{2r-1,n-2}=r\chi_0+r\chi_1+\chi_{n+r+1}$, as a~$\zz$-module.

1796: \end{lm}

1797: \pr $\tau_{1,1},\dots,\tau_{2r+1,1}$ can be taken as the

1798: representatives of the generators of $E_1^{2r-1,n-2}$. We see first

1799: that

1800: \begin{equation}

1801:   \label{eq:tau1,1}

1802:   \gamma([\tau_{1,1}])=(-1)^{n+r+1}[\tau_{1,1}].

1803: \end{equation}

1804: Indeed, $\gamma([\tau_{1,1}])=\sgn\pi\cdot[\sigma_{[1,2r],2r}]$,

1805: where $\pi$ is the permutation induced by $\gamma$ on the vertices

1806: of each support simplex of $\tau_{1,1}$, i.e.,

1807: $$\pi=(n+2r-2,n+2r-3,\dots,n+1,n,1,\dots,n-1).$$

1808: Since $\pi$ consists of inverting the sequence $(1,\dots,n+2r-2)$,

1809: and then inverting the subsequence $(1,\dots,n-1)$, we see that

1810: $$\sgn\pi=(-1)^{\left\lfloor\frac{n+2r-2}{2}\right\rfloor+

1811: \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r-1+\left\lfloor\frac{n}{2}\right\rfloor+

1812: \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r+n},$$

1813: where we used the fact that the sign of inverting a~sequence

1814: $[1,\dots,m]$ is $(-1)^{\left\lfloor\frac{m}{2}\right\rfloor}$, and that, for any

1815: natural number $m$, we have $\left\lfloor\frac{m}{2}\right\rfloor+

1816: \left\lfloor\frac{m-1}{2}\right\rfloor=m-1$. Additionally,

1817: $[\sigma_{[1,2r],2r}]=-[\sigma_{[1,2r],1}]$ by~(\ref{eq:my7}),

1818: hence~\ref{eq:tau1,1} follows.

1819: 

1820: Next, we shall see that

1821: \begin{equation}

1822:   \label{eq:taui,1}

1823:   \gamma([\tau_{2r+2-i,1}])=(-1)^{r}[\tau_{i+1,1}],

1824: \end{equation}

1825: for $i=1,\dots,2r$. Again,

1826: $\gamma([\tau_{2r+2-i,1}])=\sgn\pi\cdot[\sigma_{V(C_{2r+1})\sm\{i\},i-1}]$,

1827: where $\pi$ consists of inverting the sequence $(1,\dots,n+2r-3)$, and

1828: then inverting some subsequence of length $n-1$ back. It follows that

1829: \[\sgn\pi=(-1)^{\left\lfloor\frac{n+2r-3}{2}\right\rfloor+

1830: \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r-1+\left\lfloor\frac{n-1}{2}\right\rfloor+

1831: \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r+1}.\]

1832: On the other hand, by~(\ref{eq:my7})

1833: $[\sigma_{V(C_{2r+1})\sm\{i\},i-1}]=-[\sigma_{V(C_{2r+1})\sm\{i\},i+1}]$,

1834: hence we get~\eqref{eq:taui,1}. The actual sign has no bearing on our

1835: final conclusion.

1836: 

1837: Since the permutation action of $\zz$ on a 2-dimensional space

1838: decomposes as $\chi_0+\chi_1$, the formulae~(\ref{eq:tau1,1})

1839: and~(\ref{eq:taui,1}) together yield the claim of the lemma.  \qed

1840: 

1841: \begin{lm}\label{charlm3}

1842: We have $E_1^{2r-2,n-2}=r\chi_0+r\chi_1+\chi_{r+1}$, as a~$\zz$-module.

1843: \end{lm}

1844: \pr $\tau_{1,2},\dots,\tau_{2r+1,2}$ can be taken as the

1845: representatives of the generators of $E_1^{2r-2,n-2}$. We see first

1846: that

1847: \begin{equation}

1848:   \label{eq:taur,2}

1849:   \gamma([\tau_{r+2,2}])=(-1)^{r+1}[\tau_{r+2,2}].

1850: \end{equation}

1851: We have

1852: $\gamma([\tau_{r+2,2}])=\sgn\pi\cdot[\sigma_{V(C_{2r+1})\sm\{r,r+1\},r-1}]$,

1853: where $\pi$ consists of inverting the sequence of length $n+2r-4$, and

1854: then inverting some subsequence of length $n-1$ back. It follows that

1855: $$\sgn\pi=(-1)^{\left\lfloor\frac{n+2r-4}{2}\right\rfloor+

1856:   \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r-2+\left\lfloor\frac{n}{2}\right\rfloor+

1857:   \left\lfloor\frac{n-1}{2}\right\rfloor}=(-1)^{r+n+1}.$$

1858: Furthermore,

1859: by~(\ref{eq:my7})

1860: $[\sigma_{V(C_{2r+1})\sm\{r,r+1\},r-1}]=(-1)^n[\sigma_{V(C_{2r+1})\sm\{r,r+1\},r+2}]$,

1861: where $(-1)^n$ is composed of $2r-3$ steps changing the sign, and one step

1862: changing the sign by $(-1)^{n+1}$, since $1$ and $2r+1$ have the same

1863: parity in $V(C_{2r+1})\sm\{r,r+1\}$. Summarizing we get~(\ref{eq:taur,2}).

1864: 

1865: Second we note that

1866: \begin{equation}

1867:   \label{eq:taui,2}

1868:   \gamma([\tau_{2r+2-i,1}])=\pm[\tau_{i+2,1}],

1869: \end{equation}

1870: for $i\in V(C_{2r+1})\sm\{r+2\}$. Indeed, as before we see that

1871: $\gamma([\tau_{2r+2-i,1}])=\pm[\sigma_{V(C_{2r+1})\sm\{i,i+1\},i-1}]=

1872: \pm[\sigma_{V(C_{2r+1})\sm\{i,i+1\},i+2}]$.

1873: 

1874: Equations~\eqref{eq:taur,2} and~\eqref{eq:taui,2} show that the

1875: $\zz$-representation splits into $\chi_{r+1}$ and the $r$-fold

1876: permutation action, yielding the claim of the lemma. 

1877: \qed

1878: 

1879: \begin{crl}

1880: The group $\zz$ acts trivially on

1881: $H^{n-2}(\thom(C_{2r+1},K_n);\dz)=\dz$, and as a~multiplication

1882: by~$-1$ on $H^{n-3}(\thom(C_{2r+1},K_n);\dz)=\dz$.

1883: \end{crl}

1884: \pr It follows from Lemmata~\ref{charlm1}, \ref{charlm2}, and

1885: \ref{charlm3} that $E_1^{2r,n-3}=\chi_{r+1}$, as a~$\zz$-module. The

1886: result follows now from the equation~(\ref{eq:act_disc}). \qed

1887: 

1888: 

1889: 

1890: \section{Cohomology groups of $\zz$-quotients of products of spheres.}

1891: 

1892: \label{ss5.4}

1893: 

1894: From now on, unless explicitely stated otherwise, we shall only work

1895: with $\zz$-coefficients.

1896: 

1897: We begin by introducing another piece of terminology: for a~positive

1898: integer~$d$, let {\it $d$-symbols} be elements of the set

1899: $\{*,\infty\}$, where $*$ will denote a~open $d$-cell, and $\infty$

1900: denote a~$0$-cell. We assume throughout this section that $d\geq

1901: 2$. For example, $S^d$ is decomposed into $*$ and $\infty$, whereas

1902: a~direct product of $t$ $d$-dimensional spheres decomposes into cells,

1903: indexed by all possible $t$-tuples of $d$-symbols. We let $\dim *=d$,

1904: $\dim\infty=0$, and we set the dimension of a~tuple of $d$-symbols be

1905: the sum of the dimensions of the constituting symbols.

1906: 

1907: \subsection{Cohomology groups of $\zz$-quotients of products 

1908: of odd number of spheres.} \label{sssodd} $\,$

1909: 

1910: Let $X$ be a direct product of $2t+1$ $d$-dimensional spheres, and let

1911: $\zz$ act on $X$ be swapping spheres numbered $2i+1$ and $2i$, for

1912: $i\in[1,t]$, and acting on the first sphere by an antipodal map.  We

1913: shall decompose $X/\zz$ into cells, and describe its cohomology

1914: groups.

1915: 

1916: Clearly, $X/\zz$ is a total space of a fiber bundle over $\rp^d$ with

1917: fiber homeomorphic to a~direct product of $2t$ $d$-dimensional

1918: spheres.  Consider the standard cell decomposition of $\rp^d$ with one

1919: cell in each dimension $i\in [0,d]$.

1920: 

1921: \begin{prop}\label{pr:odd}

1922: The space $X/\zz$ can be decomposed into cells indexed with $(i,x,y)$,

1923: where $x$ and $y$ are $t$-tuples of $d$-symbols, $0\leq i\leq d$. The

1924: dimension of this cell is $\dim(i,x,y)=i+\dim x+\dim y$.

1925: 

1926: The coboundary is given by the equation

1927: \begin{equation}\label{eq:oddcb}

1928: d^{i+\dim x+\dim y}

1929: ((i,x,y)^*)=(i+1,x,y)^*+(i+1,y,x)^*,

1930: \end{equation}

1931: where the cochains are considered with $\zz$ coefficients.

1932: \end{prop}

1933: \pr Divide $X$ into cells, by taking the product cell structure, where

1934: spheres 2 to $2t+1$ have one $0$-cell and one $d$-cell, whereas the

1935: first sphere is subdivided as a join of $d+1$ $0$-spheres, with $\zz$

1936: acting antipodally on each of these $0$-spheres. The cells can then be

1937: indexed with triples $(i,x,y)_+$ and $(i,x,y)_-$. The coboundary is

1938: given by

1939: \begin{equation}\label{eq:prcb}

1940: d((i,x,y)_+^*)=(i+1,x,y)_+^*+(i+1,x,y)_-^*.

1941: \end{equation}

1942: This cell structure is $\zz$-equivariant, and no cells are preserved

1943: by the involution. This means that it induces a cell structure on

1944: $X/\zz$. Let $(i,x,y)$ denote the orbit $\{(i,x,y)_+,(i,y,x)_-\}$.

1945: After taking the quotient, \eqref{eq:prcb} becomes~\eqref{eq:oddcb}.

1946: \qed \vspace{5pt}

1947: 

1948: It follows from the Proposition \ref{pr:odd} that the generators 

1949: of $H^*(X/\zz;\zz)$ are indexed with 

1950: \begin{itemize}

1951: \item $(i,x,x)$, for any $0\leq i\leq d$, and a $t$-tuple of

1952: $d$-symbols $x$, here $(i,x,x)^*$ is the cocycle;

1953: \item $(0,x,y)$, for any $t$-tuples of $d$-symbols $x\neq y$, here

1954: $(0,x,y)^*+(0,y,x)^*$ is the cocycle; $(0,x,y)$ and $(0,y,x)$ index

1955: the same generator;

1956: \item $(d,x,y)$, for any $t$-tuples of $d$-symbols $x\neq y$, here

1957: $(d,x,y)^*$ is the cocycle; $(d,x,y)$ and $(d,y,x)$ index the same

1958: generator.

1959: \end{itemize}

1960: 

1961:  In other words, the cohomology generators are indexed by pairs

1962: $(\langle A\rangle,i)$, where A is a~$2\times t$ array of $d$-symbols,

1963: and $i\in[0,d]$, if $A$ is fixed by $\zz$, while $i\in\{0,d\}$, if $A$

1964: is not fixed by $\zz$. Here $\zz$ acts on the set of all $2\times t$

1965: arrays of $d$-symbols by swapping the two rows, and $\langle -\rangle$

1966: denotes an~orbit of this action.

1967: 

1968: For future reference, we remark the following property: these

1969: generators behave functorially, under the maps which insert additional

1970: pairs of spheres. More specifically, assume $q\geq t$, and let

1971: $f:[1,t]\hookrightarrow [1,q]$ be an injection. Let $\ti f:

1972: \underbrace{S^d\times\dots\times S^d}_{2q+1}\ra

1973: \underbrace{S^d\times\dots\times S^d}_{2t+1}$ be the following map:

1974: $\ti f$ is identity on the first sphere, it maps isomorphically the

1975: spheres indexed $2i$ and $2i+1$, for $i\in\im f$, to the spheres

1976: indexed by $2f^{-1}(i)$ and $2f^{-1}(i)+1$, and it maps the remaining

1977: spheres to the base point. Then, the induced map on the cohomology

1978: $\ti f^*$ maps the generator $(\langle A\rangle,i)$ to the generator

1979: $(\langle \wti A\rangle,i)$, where $\wti A$ is the $2\times q$ array

1980: obtained from $A$ as follows: the column $f(i)$ in $\wti A$ is equal

1981: to the column~$i$ in $A$, and, for $j\notin\im f$, the column $j$ in

1982: $\wti A$ consists of two~$\infty$'s.

1983: 

1984: 

1985: \subsection{Cohomology groups of $\zz$-quotients of products 

1986: of even number of spheres.}\label{ssseven} $\,$

1987: 

1988: Let $X$ be a direct product of $2t$ $d$-dimensional spheres, and let

1989: $\zz$ act on $X$ be swapping spheres $2i-1$ and $2i$, for $i\in[1,t]$.

1990: A~customary notation for $X/\zz$ is

1991: $SP^2(\underbrace{S^d\times\dots\times S^d}_{t})$.  Again, we shall

1992: decompose $X/\zz$ into cells, and describe its cohomology groups.

1993: 

1994: \begin{prop}\label{pr:even}

1995: The space $X/\zz$ can be decomposed into cells indexed with two types

1996: of labels:

1997: \begin{enumerate}

1998: \item [Type 1.] the unordered pairs $\{x,y\}$, where $x$ and $y$ are

1999: $t$-tuples of $d$-symbols, $x\neq y$; the dimension is $\dim x+\dim y$;

2000: \item [Type 2.] $(x,x,k)$, where $x$ is a~$t$-tuple of $d$-symbols,

2001: and $0\leq k\leq\dim x$; the dimension is $\dim  x+k$.

2002: \end{enumerate}

2003: 

2004: With $\zz$ coefficients, the coboundary is equal to 0 for all

2005: generators, except for $(x,x,0)^*$, when $\dim x\geq 1$, in which case

2006: $d^{\dim x}(x,x,0)^*=(x,x,1)^*$. In particular, the generators of

2007: $\wti H^i(X/\zz;\zz)$ are indexed with the same symbols as the cells

2008: in our decomposition, except for $(x,x,0)$ and $(x,x,1)$.

2009: \end{prop}

2010: \pr Start with a usual subdivision of a direct product of $2t$

2011: $d$-spheres, with the cells indexed by pairs $(x,y)$ of $t$-tuples of

2012: $d$-symbols. For $x\neq y$, the set $(x,y)\cup(y,x)/\zz$ is a~cell in

2013: $X/\zz$, which we label with $\{x,y\}$.

2014: 

2015: To do the same for $x=y$, we need to take a finer subdivision of

2016: $(x,x)$. Let $(x,x,k)^+$, resp.\ $(x,x,k)^-$, be the set of all points

2017: $\bar\alpha\in{\mathbb R}^{2\dim x}$, $\bar\alpha=(\alpha_i)_{i\in

2018: [2\dim x]}$, such that $\alpha_j=\alpha_{j+\dim x}$, for $k+1\leq

2019: j\leq \dim x$, and $\alpha_k>\alpha_{k+\dim x}$, resp.\

2020: $\alpha_k<\alpha_{k+\dim x}$. Obviously, $(x,x,k)^+$ and $(x,x,k)^-$

2021: are cells, which are mapped to each other by the $\zz$-action. These

2022: cells are different for $k\geq 1$, whereas $(x,x,0)^+=(x,x,0)^-$ is

2023: fixed pointwise.

2024: 

2025: Set $(x,x,0):=(x,x,0)^+$, and $(x,x,k):=(x,x,k)^+\cup(x,x,k)^-/\zz$,

2026: for $k\geq 1$. The statements about the coboundary map and the

2027: indexing of the cohomology generators follow immediately from our

2028: construction. \qed \vspace{5pt}

2029: 

2030: Rephrasing Proposition~\ref{pr:even} in the language of arrays, the

2031: generators of $H^*(X/\zz;\zz)$ are indexed with $\zz$-orbits $\langle

2032: A\rangle$ of $2\times t$ arrays of $d$-symbols, with an additional

2033: index $2\leq i\leq \dim A/2$, if $A$ is fixed by the $\zz$-action.

2034: Here $\dim A$ is the sum of the dimensions of all entries of~$A$.

2035: 

2036: Again, we have functoriality in the following sense: if $q\geq t$, and

2037: $f:[1,t]\hookrightarrow [1,q]$ is an~injection, define $\ti f:

2038: \underbrace{S^d\times\dots\times S^d}_{2q}\ra

2039: \underbrace{S^d\times\dots\times S^d}_{2t}$ analogously to the one in

2040: subsection~\ref{sssodd}.  Then $\ti f^*$ maps $\langle A\rangle$,

2041: resp.\ $(\langle A\rangle,i)$, to $\langle \wti A\rangle$, resp.\

2042: $(\langle \wti A\rangle,i)$, where $\wti A$ is a~$2\times q$ array of

2043: $d$-symbols obtained from $A$ by inserting the columns consisting

2044: entirely of $\infty$'s in the places indexed by $[1,q]\sm\im f$.

2045: 

2046: 

2047: 

2048: 

2049: \section{Spectral sequence for $H^*(\thomp(C_{2r+1},K_n)/\zz;\zz)$}

2050: \label{sect5}

2051: 

2052: Next, we would like to show Theorem~\ref{thmmain1}(b). We assume

2053: that $\varpi_1^{n-2}(\thom(C_{2r+1},K_n))\neq 0$, and arrive to

2054: a~contradiction by doing computations in a~spectral sequence, which we

2055: now proceed to set~up.

2056:  

2057: \subsection{$\zz$-equivariant cell decomposition of $\thomp(C_{2r+1},K_n)$.}

2058: $\,$ \vspace{5pt}

2059: 

2060: For convenience, we give following names to the vertices of

2061: $C_{2r+1}$: $c:=[0]_{2r+1}$, $a_i:=[r+i]_{2r+1}$, and

2062: $b_i:=[r+1-i]_{2r+1}$, for $i\in[1,r]$.  That is $\gamma:C_{2r+1}\ra

2063: C_{2r+1}$ fixes $c$, and $\gamma(a_i)=b_i$, for any $i\in[1,r]$.

2064: Identify $V(C_{2r+1})$ with the vertices of an abstract simplex

2065: $\Delta_{2r}$ of dimension~$2r$. It is also convenient to have

2066: multiple notations for $c$, namely $a_{r+1},b_{r+1}:=c$, see

2067: Figure~\ref{fig:cycle}.

2068: 

2069: We subdivide the simplex $\Delta_{2r}$ by adding $r$ more vertices,

2070: which we denote $c_1,c_2,\dots,c_r$, and defining a~new abstract

2071: simplicial complex $\tilde\Delta_{2r}$ on the set

2072: $\{c,a_1,\dots,a_r,b_1,\dots,b_r,c_1,\dots,c_r\}=V(\tilde\Delta_{2r})$.

2073: The simplices of $\tilde\Delta_{2r}$ are all the subsets of

2074: $V(\tilde\Delta_{2r})$ which do not contain the subset $\{a_i,b_i\}$,

2075: for any $i\in[1,r]$. We set $\cc=\{c,c_1,\dots,c_r\}$. The complex

2076: $\tilde\Delta_{2r}$ comes equipped with a~simplicial $\zz$-action,

2077: which fixes $\cc$ and swaps $a_i$ and $b_i$, for all $i\in[1,r]$. For

2078: $S\subseteq V(\tilde\Delta_{2r})$ we let $\langle S\rangle$ denote the

2079: $\zz$-orbit of~$S$.

2080: 

2081: %\[\pstexInput{cycle}\]

2082: %\[\text{ Figure 5. Summary of notations.}\]

2083: 

2084: 

2085: \begin{figure}[hbt]

2086: \begin{center}

2087: \pstexInput{cycle}

2088: \end{center}

2089: \caption{Summary of notations.}

2090: \label{fig:cycle}

2091: \end{figure}

2092: 

2093: One can think of this new complex $\tilde\Delta_{2r}$ as the one

2094: obtained from $\Delta_{2r}$ by representing it as a~topological join

2095: $\{c\}*[a_1,b_1]*\dots*[a_r,b_r]$, with the additional simplicial

2096: structure defined by inserting an~extra vertex $c_i$ into the middle

2097: of each $[a_i,b_i]$, and then taking the join of $\{c\}$ and the

2098: subdivided intervals.  For $\tilde\sigma\in\tilde\Delta_{2r}$ we

2099: obtain $\vartheta(\tilde\sigma)\in\Delta_{2r}$ by replacing every

2100: $c_i$ in $\ti\sigma$ by $\{a_i,b_i\}$, i.e., $\vartheta(\tilde\sigma)=

2101: (\tilde\sigma\sm\{c_1,\dots,c_r\})\cup\bigcup_{c_i\in\ti\sigma}\{a_i,b_i\}$.

2102: 

2103: The simplicial complex $\tilde\Delta_{2r}$ has an additional property:

2104: if a~simplex of $\tilde\Delta_{2r}$ is $\gamma$-invariant, then it is

2105: fixed pointwise.  This allows us to introduce a~simplicial structure

2106: (strictly speaking - a~structure of triangulated space) on

2107: $\tilde\Delta_{2r}/\zz$ by taking the orbits of the simplices of

2108: $\tilde\Delta_{2r}$ as the simplices of $\tilde\Delta_{2r}/\zz$.

2109: 

2110: \vspace{5pt}

2111: 

2112: \subsection{The chain complex of the subdivision of $\thomp(C_{2r+1},K_n)$.}

2113: $\,$ \vspace{5pt}

2114:  

2115: Since we are working over $\zz$, from now on we shall drop the $+$

2116: notation for the simplices of $\thomp(C_{2r+1},K_n)$, e.g., we shall

2117: write $\eta^*$ instead of $\eta^*_+$ (here we refer to notations

2118: introduced in subsection~\ref{ss3.2}).

2119: 

2120: Let us now describe a~cochain complex $\wti C^*

2121: (\thomp(C_{2r+1},K_n);\zz)$, which comes from a~triangulation of the

2122: simplicial complex $\thomp(C_{2r+1},K_n)$. The cochain complex

2123: consists of vector spaces over~$\zz$, whose generators are pairs

2124: $(\eta,\sigma)^*$, where $\eta\in\thomp(C_{2r+1},K_n)$, and

2125: $\sigma\in\tilde\Delta_{2r}$, such that

2126: $\vartheta(\sigma)=\supp\eta$. Such a~pair indexes the cochain which

2127: is dual to the cell $\eta\cap\supp^{-1}(\sigma)$. The coboundary of

2128: $(\eta,\sigma)^*$ is the sum of the following generators:

2129: \begin{enumerate}

2130: \item[(1)] $(\ti\eta,\sigma)^*$, if $\supp\ti\eta=\supp\eta$,

2131: $\eta\in\bo\ti\eta$, and $\dim\ti\eta=\dim\eta+1$;

2132: \item[(2)] $(\eta,\sigma\cup\{x\})^*$, if $x\in

2133: V(\ti\Delta_{2r})\sm\sigma$, and $\vt(\sigma)=\vt(\sigma\cup\{x\})$;

2134: \item[(3)] $(\ti\eta,\sigma\cup\{x\})^*$, if $x\in

2135: V(\ti\Delta_{2r})\sm\sigma$, $\ti\eta|_{\vt(\sigma)}=\eta$, and all

2136: the values of $\ti\eta$ on $\vt(\sigma\cup\{x\})\sm\vt(\sigma)$ have

2137: cardinality~1.

2138: \end{enumerate}

2139:  The degree of $(\eta,\sigma)^*$ in $\wti C^*(\thomp(C_{2r+1},K_n))$

2140: is given by

2141: $$\deg(\eta,\sigma)^*=|\sigma|-1+\sum_{v\in\supp\eta}

2142: (|\eta(v)|-1)=\deg\eta+|\sigma|-|\vt(\sigma)|.$$

2143: 

2144: $\zz$ acts on $\wti C^*(\thomp(C_{2r+1},K_n))$ and we let $\wti

2145: C^*_\zz(\thomp(C_{2r+1},K_n))$ denote its subcomplex consisting of the

2146: invariant cochains. By construction of the subdivision, $\wti

2147: C^*_\zz(\thomp(C_{2r+1},K_n))$ is a~cochain complex for

2148: a~triangulation of the space $\thomp(C_{2r+1},K_n)/\zz$.

2149: 

2150: \subsection{The filtration of $\wti C^*_\zz(\thomp(C_{2r+1},K_n);\zz)$.}

2151: $\,$ \vspace{5pt}

2152: 

2153: This time, we consider the natural filtration $(\wti F^{0}

2154: \supseteq\wti F^{1}\supseteq\dots)$ on the cochain complex $\wti

2155: C^*_\zz(\thomp(C_{2r+1},K_n);\zz)$ by the cardinality of~$\sigma$.

2156: Namely, $\wti F^p=\wti F^p C^*_\zz(\thomp(C_{2r+1},K_n);\zz)$, is

2157: a~cochain subcomplex of $C^*_\zz(\thomp(C_{2r+1},K_n);\zz)$ defined

2158: by:

2159: \[\wti F^p: \dots\stackrel{\bo^{q-1}}\lra \wti F^{p,q}\stackrel{\bo^{q}}

2160: \lra\wti F^{p,q+1}\stackrel{\bo^{q+1}}\lra\dots,\] 

2161: where

2162: \[ \wti F^{p,q}=\zz\left[(\eta,\sigma)^*\,\left|\,

2163: (\eta,\sigma)\in C^q_\zz(\thomp(C_{2r+1},K_n);\zz),|\sigma|\geq p+1\right]\right.,

2164: \]

2165: and $\bo^*$ is the restriction of the differential in

2166: $C^*_\zz(\thomp(C_{2r+1},K_n);\zz)$. 

2167: 

2168: The following formula is the analog of \eqref{eq:e0tab}.

2169: 

2170: \begin{prop}

2171: For any $p$,

2172: \begin{equation}\label{eq:qe0tab}

2173: \begin{aligned}

2174: \wti F^p/\wti F^{p+1}=&

2175: \bigoplus_{\sigma}C^*(\thom(C_{2r+1}[\vt(\sigma)],K_n)/\zz;\zz)[-p]\\

2176: &\bigoplus_{\langle\tau\rangle}C^*(\thom(C_{2r+1}[\vt(\tau)],K_n);\zz)[-p],

2177: \end{aligned}

2178: \end{equation}

2179: where the first sum is taken over all $\sigma\subseteq\cc$,

2180: $|\sigma|=p+1$, and the second sum is taken over all orbits

2181: $\langle\tau\rangle$, such that $\tau\subseteq V(\ti\Delta_{2r})$,

2182: $|\tau|=p+1$, $\tau\sm\cc\neq\emptyset$.

2183: 

2184: Hence, the 0th tableau of the spectral sequence associated to

2185: the cochain complex filtration $\wti F^*$ is given by

2186: \begin{equation}

2187:   \label{eq:qE0}

2188: \begin{aligned}

2189: E_0^{p,q}=&

2190: \bigoplus_{\sigma}C^{q}(\thom(C_{2r+1}[\vt(\sigma)],K_n)/\zz;\zz)\\

2191: &\bigoplus_{\langle\tau\rangle}C^{q}(\thom(C_{2r+1}[\vt(\tau)],K_n);\zz),

2192: \end{aligned}

2193: \end{equation}

2194: with the summations over the same sets as in \eqref{eq:qe0tab}.

2195: \end{prop}

2196: 

2197: 

2198: \subsection{The analysis of the spectral sequence converging to 

2199: $H^*(\thomp(C_{2r+1},K_n)/\zz;\zz)$.}

2200: $\,$ \vspace{5pt}

2201: 

2202: The $E_1^{*,*}$-tableau of this spectral sequence is given by

2203: $E_1^{p,q}=H^{p+q}(\wti F^p,\wti F^{p+1})$.  It follows immediately

2204: from the formula \eqref{eq:qE0} that each $E_1^{p,q}$ splits as

2205: a~vector space over $\zz$ into direct sums of

2206: $H^q(\thom(C_{2r+1}[S],K_n);\zz)$, and of

2207: $H^q(\thom(C_{2r+1}[S],K_n)/\zz;\zz)$. More precisely,

2208: \begin{equation}

2209:   \label{eq:qE1}

2210: \begin{aligned}

2211: E_1^{p,q}=&

2212: \bigoplus_{\sigma\subseteq\cc}

2213: H^{q}(\thom(C_{2r+1}[\vt(\sigma)],K_n)/\zz;\zz)\\

2214: &\bigoplus_{\langle\tau\rangle,\,\tau\not\subseteq\cc}

2215: H^{q}(\thom(C_{2r+1}[\vt(\tau)],K_n);\zz).

2216: \end{aligned}

2217: \end{equation}

2218: 

2219: The generators of $E_1^{p,q}$ stemming from $\sigma\subseteq\cc$ will

2220: be called {\it symmetric}, whereas the generators stemming from

2221: $\langle \tau\rangle$ for $\tau\not\subseteq\cc$ will be called {\it

2222: asymmetric}.

2223: 

2224: For $i\in[1,r]$, we shall denote the arc

2225: $\{a_i,a_{i-1},\dots,a_1,b_1,b_2,\dots,b_i\}$ by $\smile_i$. For

2226: $2\leq i\leq r$, we denote the arc

2227: $\{a_i,a_{i+1},\dots,a_r,c,b_r,b_{r-1},\dots,b_i\}$ by $\frown_i$.

2228: For $2\leq i<j\leq r$, let $\big{(}_{i,j}$ denote the arc

2229: $\{a_i,a_{i+1},\dots,a_j\}$, let $\big{)}_{i,j}$ denote the arc

2230: $\{b_j,b_{j-1},\dots,b_i\}$, and let $\big{(}\big{)}_{i,j}$ denote

2231: the symmetric pair of arcs $\big{(}_{i,j}$ and~$\big{)}_{i,j}$.

2232: 

2233: \begin{prop} The map

2234: \begin{equation}\label{eqquot}

2235: q^{n-3}:H^{n-3}(\thom(C_{2r+1},K_n)/\zz;\zz)\ra

2236: H^{n-3}(\thom(C_{2r+1},K_n);\zz),

2237: \end{equation}

2238: is a 0-map.

2239: \end{prop}

2240: 

2241: \pr First of all, since we are working over the field $\zz$,

2242: the map $q^{n-3}$ is dual to the map on homology

2243: 	\[q_{n-3}:H_{n-3}(\thom(C_{2r+1},K_n);\zz)

2244: 	\longrightarrow H_{n-3}(\thom(C_{2r+1},K_n)/\zz;\zz),

2245: \]

2246: hence it is enough to prove that $q_{n-3}$ is a $0$-map.

2247: 

2248: We start by proving that $q_{n-3}=0$ over integers. The map $q_{n-3}$

2249: commutes with the $\zz$-action. Recall that we have proven that

2250: \[H^{n-3}(\thom(C_{2r+1},K_n);\dz)=H^{n-2}(\thom(C_{2r+1},K_n);\dz)=\dz,\] 

2251: so it follows that $H_{n-3}(\thom(C_{2r+1},K_n);\dz)=\dz$. Let $\xi$

2252: be a~generator of the group $H_{n-3}(\thom(C_{2r+1},K_n);\dz)$. By our

2253: previous computations $\gamma^{K_n}(\xi)=-\xi$, since

2254: $H_{n-3}(\thom(C_{2r+1},K_n);\dc)=\chi_1$ as a $\zz$-module (it is

2255: a~dual $\zz$-module to $H^{n-3}(\thom(C_{2r+1},K_n);\dc))$, and since

2256: $H_{n-3}(\thom(C_{2r+1},K_n);\dz)$ is torsion-free.  On the other

2257: hand, the $\zz$-action on $H_{n-3}(\thom(C_{2r+1},K_n)/\zz;\dz)$ is

2258: trivial, hence

2259: \[-q_{n-3}(\xi)=q_{n-3}(-\xi)=q_{n-3}(\gamma^{K_n}(\xi))=

2260: \gamma^{K_n}(q_{n-3}(\xi))=q_{n-3}(\xi).

2261: \]

2262: We conclude that $q_{n-3}(\xi)=0$.

2263: 

2264: Second, by the universal coefficient theorem the map

2265: \[

2266: \tau:H_{n-3}(\thom(C_{2r+1},K_n);\dz)\otimes\zz\longrightarrow

2267: H_{n-3}(\thom(C_{2r+1},K_n);\zz)

2268: \]

2269: is injective and functorial. In our concrete situation, this map is

2270: also surjective, hence the claim follows from the following

2271: commutative diagram:

2272: $$\begin{CD}

2273: H^{n-3}(\thom(C_{2r+1},K_n);\dz)\otimes\zz   @>\text{0-map}>>

2274: H^{n-3}(\thom(C_{2r+1},K_n)/\zz;\dz)\otimes\zz  \\

2275:         @V\tau V\text{iso}V           @VVV\\

2276: H^{n-3}(\thom(C_{2r+1},K_n);\zz) @>q^{n-3}>>

2277: H^{n-3}(\thom(C_{2r+1},K_n)/\zz;\zz) \\

2278: \end{CD}

2279: $$

2280: $\qquad$\hfill  \qed

2281: 

2282: 

2283: \begin{lm}\label{lm:e2r+1n-3}

2284: We have $E_2^{r+1,n-3}=\zz$.

2285: \end{lm}

2286: \pr To start with, the only contribution to $E_1^{r,n-3}$ comes from

2287: $\sigma=\cc$, so the fact that $q^{n-3}$ in \eqref{eqquot} is a~0-map

2288: implies that the differential $d_1:E^{r,n-3}_1\ra E^{r+1,n-3}_1$ is

2289: a~0-map as well.

2290: 

2291: %\[\pstexInput{2e1}\]

2292: %\[\text{ Figure 6. The $E_2^{*,*}$-tableau, 

2293: %$E_2^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n)/\zz;\zz)$.}\]

2294: 

2295: \begin{figure}[hbt]

2296: \begin{center}

2297: \pstexInput{2e1}

2298: \end{center}

2299: \caption{The $E_2^{*,*}$-tableau, 

2300: $E_2^{p,q}\Rightarrow H^{p+q}(\thomp(C_{2r+1},K_n)/\zz;\zz)$.}

2301: \label{fig:2e1}

2302: \end{figure}

2303: 

2304: Consider the cochain complex 

2305: \[

2306: A^*:E^{r+1,n-3}_1\stackrel{d_1}{\lra} E^{r+2,n-3}_1\stackrel{d_1}{\lra}

2307: \dots\stackrel{d_1}{\lra} E^{2r,n-3}_1.

2308: \]

2309: The generators of $E_1^{r+i,n-3}$ come from $\tau=\cc\cup I$, for

2310: $I\subseteq\{a_1,\dots,a_r,b_1,\dots,b_r\}$, $|I|=i$. We can identify

2311: the generator indexed by $\langle\tau\rangle$ with the simplex of

2312: ${\mathbb R}{\mathbb P}^{r-1}\cong\{a_1,b_1\}*\dots*\{a_r,b_r\}/\zz$,

2313: indexed by $\langle I\rangle$, where the $\zz$-action swaps $a_i$ and

2314: $b_i$, for $i\in[1,r]$.

2315: 

2316: By inspecting the description of the differential $d_1$ we see that

2317: $A^*$ is isomorphic to the~chain complex $C^*({\mathbb R}{\mathbb

2318: P}^{r-1};\zz)$.  It follows that $E_2^{r+1,n-3}=H^0({\mathbb

2319: R}{\mathbb P}^{r-1};\zz)=\zz$.  \qed \vspace{5pt}

2320: 

2321: In the proof of the next lemma we shall often use the chain homotopy

2322: between $0$ and the identity.

2323: 

2324: \noindent

2325: {\it Let $(C^*,d)$ be a cochain complex, and assume there exist

2326: linear maps $\phi^n:C^n\ra C^{n-1}$, $\forall n$, such that

2327: \begin{equation}\label{eq:phid}

2328: \phi^{n+1}(d(\alpha))+d(\phi^n(\alpha))=\alpha, \text{ for all }

2329: \alpha\in C^n.

2330: \end{equation}

2331: Then $C^*$ is acyclic.}

2332: 

2333: The proof is immediate, since modulo the coboundaries, every

2334: $\alpha\in C^n$ is equal to $\phi^{n+1}(d(\alpha))$, hence

2335: $d(\alpha)=0$ implies $\alpha=0$ modulo the coboundaries.

2336: 

2337: 

2338: \begin{lm}\label{lm:e2r-1n-2}

2339: We have $E_2^{r-1,n-2}=0$.

2340: \end{lm}

2341: \pr Set

2342: \[

2343: A^*:E^{0,n-2}_1\stackrel{d_1}{\lra} E^{1,n-2}_1\stackrel{d_1}{\lra}

2344: \dots\stackrel{d_1}{\lra} E^{2r,n-2}_1.

2345: \]

2346: Clearly, to show $E_2^{r-1,n-2}=0$ is the same as to show that

2347: $H^{r-1}(A^*)=0$.

2348: 

2349: For dimensional reasons, every generator in $A^*$ is indexed either by

2350: $\sigma\subset V(\ti\Delta_{2r})$ with an~arc selected in

2351: $\vt(\sigma)$ (which we call the {\it indexing arc}), or the whole set

2352: $V(C_{2r+1})$ (namely, those coming from

2353: $H^{n-2}(\thom(C_{2r+1},K_n)/\zz)$ and from

2354: $H^{n-2}(\thom(C_{2r+1},K_n))$). To simplify the terminology, we shall

2355: call the set $V(C_{2r+1})$ an~arc as well. Filter the cochain complex

2356: $A^*=G^{2r+1}\supseteq G^{2r}\supseteq\dots\supseteq G^2\supseteq

2357: G^1=0$, where $G^l$ is spanned by the generators whose indexing arc

2358: has the length at least~$l$. We shall compute $H^{r-1}(A^*)$ by

2359: considering the corresponding spectral sequence $\wti E^{p,q}_0:=

2360: C^{p+q}(G^p/G^{p-1})$.

2361: 

2362: In the same pattern as we have already encountered, the cochain

2363: complex $(G^p/G^{p-1},d_0)$ splits into a~direct sum of subcomplexes

2364: which are indexed by different arcs. For an~arc~$a$, let $B^*_a$

2365: denote the corresponding summand. Hence $\wti E^{p,q}_1= \bigoplus_a

2366: H^{p+q}(B^*_a)$, where the sum is taken over all arcs $a$ of

2367: length~$p$.

2368: 

2369: Next, by considering all possible arcs case-by-case, we compute the

2370: entries $\wti E^{p,r-1-p}_1$, for $p=2,\dots,2r+1$. To start with,

2371: since $\vt(\sigma)=V(C_{2r+1})$ implies $|\sigma|\geq r+1$, $\wti

2372: E^{2r+1,-r-2}_0=C^{r-1}(G^{2r+1}/G^{2r})=0$ for dimensional reasons,

2373: hence $\wti E^{2r+1,-r-2}_1=0$.

2374: 

2375: We shall only consider the cases where we cannot use dimensional

2376: reasons to immediately conclude that $B_a^{r-1}=0$.

2377: 

2378: {\bf Case 1.} Let $a=\smile_r$. Then, $B_a^{r-2}=0$ for dimensional

2379: reasons, and $B_a^{r-1}=\zz$ coming from $\sigma=\{c_1,\dots,c_r\}$.

2380: The differential $d:B_a^{r-1}\ra B_a^r$ is a~0-map since

2381: $f^{n-2}:H^{n-2}(\rp^{n-2};\zz)\ra H^{n-2}(S^{n-2};\zz)$ is a~0-map,

2382: where $f:S^{n-2}\ra S^{n-2}/\zz=\rp^{n-2}$ denotes the covering map.

2383: Hence, in this case, $H^{r-1}(B_a^*)=\zz$.

2384: 

2385: {\bf Case 2.} Let $a=\frown_2$. Again, $B_a^{r-2}=0$ for dimensional

2386: reasons, and $B_a^{r-1}=\zz$ coming from $\sigma=\{c_2,\dots,c_r,c\}$.

2387: However, this time $d(B_a^{r-1})\neq 0$, since it is induced by the

2388: map $f^{n-2}:H^{n-2}(\thom(G,K_n)/\zz)\ra H^{n-2}(\thom(G,K_n))$,

2389: which, as we have seen, is not a~0-map; here $G$ is the tree on 3

2390: vertices and $\zz$ action is swapping the leaves. Hence

2391: $H^{r-1}(B_a^*)=0$.

2392: 

2393: {\bf Case 3.} Let $a=\smile_k$, for $1\leq k\leq r-1$.  Let $\alpha\in

2394: B_a^m$ be a~generator indexed by $\sigma\subset V(\ti\Delta_{2r})$. If

2395: $\sigma\subset\cc$, and $x\in\{a_1,\dots,a_r,b_1,\dots,b_r\}\sm

2396: \{a_{k+1},b_{k+1}\}$, then the differential maps $\alpha$ to the

2397: generator indexed by $\langle\sigma\cup\{x\}\rangle$ (again $\smile_k$

2398: is selected) as a~0-map, for the reason described in Case~1. This

2399: means that the complex $B_a^*$ splits into two direct summands, one

2400: containing all generators indexed by $\sigma\subset\cc$, and the other

2401: those indexed by $\langle\sigma\rangle$, such that

2402: $\sigma\sm\cc\neq\emptyset$.

2403: 

2404: In both summands, define $\phi^m(\alpha)$ to be the generator indexed

2405: by $\langle\sigma\sm\{c\}\rangle$, if $c\in\sigma$, and

2406: $\phi^m(\alpha)=0$ otherwise. The equation \eqref{eq:phid} is

2407: satisfied, so both summands are acyclic, hence so is $B^*_a$, in

2408: particular $H^{r-1}(B_a^*)=0$.

2409: 

2410: {\bf Case 4.} Let $a=\frown_k$, for $3\leq k\leq r$. We do the same as

2411: in the case~3 with $c$, replaced with $c_{k-2}$. However, in this

2412: complex, if $\sigma\subset\cc$, and $x\in

2413: \{a_1,\dots,a_r,b_1,\dots,b_r\}\sm\{a_{k-1},b_{k-1}\}$, then the

2414: differential maps $\alpha$ to the generator indexed with

2415: $\langle\sigma\cup\{x\}\rangle$ (again $\frown_k$ is selected) as an

2416: identity, hence the complex does not split and the equation

2417: \eqref{eq:phid} can be applied to the whole complex, yielding

2418: $H^{r-1}(B_a^*)=0$.

2419: 

2420: {\bf Case 5.} Let $a=\big{(}_{2,r}$. For each indexing orbit

2421: $\langle\sigma\rangle$ choose the representative $\sigma$ such that

2422: $a_1\notin\sigma$.  Define $\phi^*$ as in the Case~3, taking $b_1$

2423: instead of $c$. The equation~\eqref{eq:phid} is rather

2424: straightforward. We just need to pay attention to what the

2425: differential does to the generator indexed by

2426: $\sigma=\cc\sm\{c_1,c\}$.

2427: 

2428: It follows from the description of the cell decomposition and the

2429: cohomology of $SP^2(S^{n-2})$, given as a~special case in

2430: subsection~\ref{ssseven}, that the map

2431: $f^{n-2}:H^{n-2}(SP^2(S^{n-2});\zz)\ra H^{n-2}(S^{n-2}\times

2432: S^{n-2};\zz)$, induced by the quotient map, takes the nonzero element

2433: to the sum of the two generators of $H^{n-2}(S^{n-2}\times

2434: S^{n-2};\zz)$ corresponding to each of the two spheres. In $B_a^*$

2435: this means that the differential of $\sigma$ will contain the

2436: generator of $\langle\sigma\cup\{b_1\}\rangle$, with the indexing arc

2437: $\big{(}_{2,r}$, but not the generator of

2438: $\langle\sigma\cup\{b_1\}\rangle$, with the indexing arc

2439: $\big{)}_{1,r}$. Thus we conclude again that $B^*_a$ is acyclic, and

2440: $H^{r-1}(B_a^*)=0$.

2441: 

2442: {\bf Case 6.} Let $a=\big{(}_{1,r}$. $B_a^{r-2}=0$ for dimensional

2443: reasons. The space $B_a^{r-1}$ is spanned by the $2^{r-1}$ generators

2444: which we can index with sets $\{a_1,\xi_2,\dots,\xi_r\}$, where

2445: $\xi_i\in\{a_i,c_i\}$, for $2\leq i\leq r$. Denote by $\bar a$ the

2446: generator indexed by the set $\{a_1,a_2,\dots,a_r\}$.  The coboundary

2447: of every generator $\alpha\neq\bar a$ contains some generator $\beta$

2448: indexed by $\{b_i,a_1,\xi_2,\dots,\xi_r\}$. Since $\alpha$ is uniquely

2449: reconstructible from $\beta$, and the coboundary of $\bar a$ does not

2450: contain such generators as $\beta$, we see that an element in

2451: $\ker(d:B_a^{r-1}\ra B_a^r)$ cannot contain $\alpha$ with a~nonzero

2452: coefficient. Thus, the only chance for this kernel to be nontrivial

2453: would be that $\bar a$ lies in it, but, obviously, $d(\bar a)\neq 0$.

2454: Hence, once again, $B^*_a$ is acyclic, and $H^{r-1}(B_a^*)=0$.

2455: 

2456: {\bf Case 7.} Let $a$ be an~assymetric arc, such that

2457: $a\cap\{a_r,b_r,c\}=\emptyset$. The complex $B^*_a$ is isomorphic to

2458: a~simplicial complex of a cone with an apex in the vertex~$c$. Hence

2459: $B^*_a$ is acyclic, and $H^{r-1}(B_a^*)=0$.

2460: 

2461: {\bf Case 8.} Let $a$ be such that

2462: $a\cap\{a_1,a_2,b_1,b_2\}=\emptyset$.  The complex $B^*_a$ is

2463: isomorphic to a~simplicial complex of a cone with an apex in the

2464: vertex~$c_1$. Hence $B^*_a$ is acyclic, and $H^{r-1}(B_a^*)=0$.

2465: 

2466: {\bf Case 9.} Let

2467: $a=\{c,a_i,a_{i+1},\dots,a_r,b_r,b_{r-1},\dots,b_j\}$, for $2\leq

2468: i<j$, where possibly $j=r+1$, which means $a$ does not contain any

2469: $b_i$'s. The complex $B^*_a$ is isomorphic to a~simplicial complex of

2470: a cone with an apex in the vertex~$b_1$. Hence $B^*_a$ is acyclic, and

2471: $H^{r-1}(B_a^*)=0$.

2472: 

2473: {\bf Case 10.} Let $a=\{a_i,a_{i-1},\dots,a_1,b_1,b_2,\dots,b_j\}$,

2474: for $r\geq i>j\geq 1$. The complex $B^*_a$ is isomorphic to

2475: a~simplicial complex of a cone with an apex in the vertex $a_1$. Hence

2476: $B^*_a$ is acyclic, and $H^{r-1}(B_a^*)=0$.

2477: 

2478: We can now summarize our computations as follows: $\wti E^{p,r-1-p}_1

2479: =0$, for $p=2,\dots,2r-1$, whereas $\wti E^{2r,-r-1}_1=\zz$.  The

2480: generator of $\wti E^{2r,-r-1}_1$ comes from $a=\smile_r$, which in

2481: turn comes from $\varpi_1^{n-2}(\thom(K_2,K_n))$. The map $d_1:\wti

2482: E^{2r,-r-1}_1\ra \wti E^{2r+1,-r-1}_1$ is the same as 

2483: \[(\iota_{K_n})^{n-2}:H^{n-2}(\thom(K_2,K_n)/\zz;\zz)\ra

2484: H^{n-2}(\thom(C_{2r+1},K_n)/\zz;\zz),\] 

2485: where $\iota:K_2\hookrightarrow C_{2r+1}$ is either of the two

2486: $\zz$-equivariant inclusion maps which take the vertices of $K_2$ to

2487: $\{a_1,b_1\}$.

2488: 

2489: Since we assumed that $\varpi_1^{n-2}(\thom(C_{2r+1},K_n))\neq 0$, and

2490: the Stiefel-Whitney characteristic classes are functorial, we see that

2491: $d_1:\wti E^{2r,-r-1}_1\ra \wti E^{2r+1,-r-1}_1$ has rank~$1$, hence

2492: $\wti E^{2r,-r-1}_2=0$.  Thus $\wti E^{p,r-1-p}_2 =0$, for

2493: $p=2,\dots,2r+1$, and we conclude that $E_2^{r-1,n-2}=0$.  \qed

2494: 

2495: 

2496: 

2497: 

2498: \begin{lm}\label{lm:r-i}

2499: We have $E_2^{r-i,n-3+i}=0$, for all $i=2,3,\dots,r$.

2500: \end{lm}

2501: \pr First of all, we note, that the generators in the columns indexed

2502: by $r-1$ and less come from $H^*(\thom(C_{2r+1}[S],K_n)/\zz;\zz)$ and

2503: from $H^*(\thom(C_{2r+1}[S],K_n);\zz)$, with $S\neq V(C_{2r+1})$ in

2504: both cases.

2505: 

2506:  For each row $q$, $q>n-2$, we shall show that the subcomplex

2507: $A_q^*=(E_1^{*,q},d_1)$ is acyclic in the entry $n+r-q-3$.

2508: 

2509: 

2510: We begin by dealing with the case $q=n-1$ separately, that is we

2511: analyze the entry $E_2^{r-2,n-1}$. It follows from

2512: Propositions~\ref{pr:odd}, \ref{pr:even}, and for dimensional reasons,

2513: that the entries $E_1^{0,n-1},E_1^{1,n-1},\dots,E_1^{r-1,n-1}$ are

2514: generated by the contributions whose indexing collections of arcs are

2515: $(\smile_i,\frown_j)$, for $1\leq i<j-1\leq r-1$.

2516: 

2517: The contributing spaces are homotopy equivalent to $S^{n-2}\times

2518: X/\zz$, where $X$ is a~direct product of $2t+1$ $(n-2)$-dimensional

2519: spheres, and $\zz$-action is as in Section~\ref{ss5.4}. The generators

2520: appearing in the first $r$ entries of the $(n-1)$th row are coming

2521: from the $(n-2)$-cocycle of $S^{n-2}$ and the 1-cocycle of

2522: $\rp^{n-2}$. The analysis of the differentials shows that the complex

2523: $E_1^{0,n-1}\stackrel{d_1}{\ra}E_1^{1,n-1}\stackrel{d_1}{\ra}\dots

2524: \stackrel{d_1}{\ra}E_1^{r-1,n-1}$ computes the nonreduced homology of

2525: a~simplex with $r-2$ vertices (which could be identified with the set

2526: $\{c_2,\dots,c_{r-1}\}$). It follows that the entry $E_2^{r-2,n-1}$,

2527: which computes the first homology group is equal to~$0$.

2528: 

2529: We assume from now on that $q\geq n$.  Similar to

2530: subsection~\ref{ss4.5} we filter the complexes $A_q^*$.  To describe

2531: the filtration, we sort all generators into 5 groups. The first

2532: group~(Gr1) contains all asymmetric generators, i.e., those coming

2533: from $\langle\sigma\rangle$, for $\sigma\not\subseteq\cc$. The

2534: symmetric generators, coming from $\sigma\subseteq\cc$, are divided

2535: into 4~groups, depending on whether the indexing collection of arcs

2536: \begin{enumerate}

2537: \item[(Gr2)] contains both an $\frown$-arc, and an $\smile$-arc,

2538: \item[(Gr3)] contains an $\smile$-arc, but not an $\frown$-arc,

2539: \item[(Gr4)] contains an $\frown$-arc, but not an $\smile$-arc, 

2540: \item[(Gr5)] contains no $\frown$-arc, and no $\smile$-arc.

2541: \end{enumerate}

2542: The groups are ordered as above. We filter the complex $A_q^*$ by

2543: first sorting the generators by the groups, and then, within each

2544: group we filter additionally by the total length of the indexing arcs.

2545: 

2546: Let $\wti E^{*,*}_*$ denote the tableaux of the spectral sequence

2547: computing the cohomology of $A_q^*$.  In complete analogy to the

2548: situation in subsection~\ref{ss4.5}, $\wti E^{*,*}_0$ splits into

2549: pieces indexed by various collections of arcs, which we shall call

2550: {\it layers}.

2551: 

2552: We start by analyzing the contributions of the asymmetric generators.

2553: Consider the subcomplex $B^*$ in the splitting indexed by a~collection

2554: $A$ of $t$ arcs of total length~$l$.  Since the asymmetric generators

2555: come from the direct products of $(n-2)$-spheres, the only nontrivial

2556: cases are $q=t(n-2)$, for $t\geq 2$.

2557: 

2558: Assume first there is a~gap between some pair of arcs of length at

2559: least~3, and let $x\in V(C_{2r+1})$ be one of the internal points of

2560: a~gap.  If $x=c$, then $B^*$ is isomorphic to the chain complex of

2561: a~cone with apex in~$c$.  Without loss of generality, we can assume

2562: that $x=b_i$, for some $i$.  By the previous assumption,

2563: $b_{i-1},b_{i+1}\notin A$. If $a_i \notin A$, but either $a_{i-1}$, or

2564: $a_{i+1}$ (or both) is in $A$, then $B^*$ is isomorphic to a~chain

2565: complex of a~cone with apex $b_i$. If $a_{i-1},a_i,a_{i+1}\notin A$,

2566: then $B^*$ is isomorphic to a~chain complex of a~cone with apex

2567: $c_i$. Finally, assume $a_i\in A$.  Define $\phi^k:B^k\ra B^{k-1}$ as

2568: follows: for a~generator $\sigma\in B^k$,

2569: \[\phi^k(\sigma)=\begin{cases}

2570: \sigma\setminus\{b_i\}, & \text{ if } b_i\in\sigma,\text{ i.e., if } 

2571: \sigma\cap\{a_i,b_i,c_i\}\text{ is } \{b_i\}, \text{ or }\{b_i,c_i\};\\

2572: \sigma\setminus\{c_i\}, & \text{ if }\sigma\cap\{a_i,b_i,c_i\}=\{a_i,c_i\};\\

2573: 0, & \text{ if } \sigma\cap\{a_i,b_i,c_i\}\text{ is } \emptyset, 

2574: \text{ or }\{c_i\},\text{ or }\{a_i\}.

2575: \end{cases}\]

2576: Let $\wti B^*$ be the subcomplex of $B^*$ generated by all $\sigma$,

2577: such that $a_i\in\sigma$. Clearly, the \eqref{eq:phid} is fulfilled

2578: both for $\wti B^*$ and for $B^*/\wti B^*$. It implies that they are

2579: both acyclic, hence so is $B^*$.

2580: 

2581: If all gaps are of length at most 2, then $l+2t\geq 2r+1$. On the

2582: other hand, $B^p=0$ for $p\leq l/2-1$, since $|\vt(\sigma)|\leq

2583: 2|\sigma|-1$, for $\sigma\not\subseteq\cc$. Recall that $q=t(n-2)$, it

2584: follows that the entry $n+r-q-3$ is~0, since

2585: \begin{multline*}

2586: l/2-1-(n+r-t(n-2)-3)>r-t-1-n-r+tn-2t+3=  \\

2587: tn-3t-n+2=(t-1)(n-3)-1\geq 1.

2588: \end{multline*}

2589: Hence $B^*$ is acyclic in the required entry, for all $B^*$ in the

2590: group~(Gr1).

2591: 

2592: Next, we move on to the symmetric generators. For $\sigma\subseteq\cc$

2593: we call $|\cc\sm\sigma|$ the {\it total length of gaps}. Let $B^*$ be

2594: a~subcomplex in the splitting corresponding to a~layer from the

2595: group~(Gr2). The contributing space here is $S^{n-2}\times X/\zz$,

2596: where $X$ is a~direct product of $2t+1$ $(n-2)$-spheres and

2597: $\zz$-action is as in Section~\ref{ss5.4}.

2598: 

2599: If $t=0$, since the column number is at most $r-3$, the gap between

2600: the $\smile$-arc and the $\frown$-arc is at least 3. This means that

2601: $B^*$ is isomorphic to a cochain complex of the simplex, hence is

2602: acyclic.

2603: 

2604: Assume now $t\geq 1$. By examining the cohomology groups of

2605: $S^{n-2}\times X/\zz$, and taking into account that each of the $t$

2606: pairs of spheres must contribute nontrivially, we see that the

2607: dimension of the contributing cocycle of $S^{n-2}\times X/\zz$ is at

2608: least $n-2+t(n-2)=(t+1)(n-2)$, hence the total length of gaps is at

2609: least $t(n-2)+1$. Assume the total length of gaps is at most $2(t+1)$,

2610: as otherwise $B^*$ is isomorphic to a cochain complex of the simplex.

2611: By assumptions, $t\geq 1$ and $n\geq 5$, so unless $(t,n)=(1,5)$, we

2612: have

2613: \[t(n-2)+1-(2t+2)=t(n-4)-1>0,\]

2614: yielding a~contradiction.

2615: 

2616: Consider the remaining case $(t,n)=(1,5)$. This is the first situation

2617: in which we need to analyze the particular entries of $\wti

2618: E^{*,*}_1$. Since we must have a precise equality, the total length of

2619: gaps is 4, and the only nontrivial case is provided by generators

2620: indexed with $\smile_i$, $\frown_j$, and $\big{(}_{i+3,j-3}$.  The

2621: contributing cohomology generator must be indexed $(0,*,\infty)$, so

2622: just the set of arcs determines everything.

2623: 

2624: Let $\alpha_{i,j}$ denote such a~generator, and let $\beta_{i,j}$

2625: denote the generator whose indexing set of arcs is $\smile_i$,

2626: $\frown_j$, and $\big{(}_{i+2,j-3}$, and which is also indexed by

2627: $(0,*,\infty)$. Both $\alpha_{i,j}$'s, and $\beta_{i,j}$'s are

2628: generators in $\wti E^{*,*}_1$.  Consider a linear combination

2629: $\sum_{i,j}p_{i,j}\alpha_{i,j}$ lying in the kernel of $d_1$. Since

2630: $d_1(\alpha_{i,j})$ contains $\beta_{i,j}$, $\beta_{i+1,j}$, and no

2631: other $\beta_{i',j'}$'s we see that $p_{i,j}\neq 0$ implies

2632: $p_{i-1,j}\neq 0$.  This leads obviously to $p_{i,j}=0$ for all $i,j$,

2633: hence $B^*$ is acyclic in the required entry.

2634: 

2635:  Now consider $B^*$ corresponding to a layer from group~(Gr3).  The

2636: contributing space here is $X/\zz$, where $X$ is a~direct product of

2637: $2t+1$ $(n-2)$-spheres and the $\zz$-action is as above. Since we are

2638: in the row $n$ or higher, we must have $t\geq 1$.  The total length of

2639: gaps cannot be larger than $2t+1$, since otherwise $B^*$ is isomorphic

2640: to a~cochain complex of the simplex. On the other hand, since the

2641: dimension of the contributing cohomology generator is at least

2642: $t(n-2)$, the total length of gaps must be at least $(t-1)(n-2)+1$.

2643: Comparing these two we see that

2644: \[(t-1)(n-2)+1-(2t+1)=(t-1)(n-4)-2>0,\]

2645: with exceptions: $t=1$, $n$ is any, $t=2$, $n=5,6$, and $(t,n)=(3,5)$.

2646: 

2647: Consider first $t=1$. Since we can have at most 3 gaps, we must have

2648: precisely 3 gaps, so the contributing cohomology generators of

2649: $S^{n-2}\times S^{n-2}\times S^{n-2}/\zz$ must have

2650: dimension~$n$. Inspecting the cohomology description of this space

2651: from Section~\ref{ss5.4} we see that there are no generators in

2652: dimensions between $n-2$ and $2n-4$. Since $2n-4>n$ we verify this

2653: case.

2654: 

2655: Assume now $(t,n)=(2,5)$. The only nontrivial case is when the total

2656: length of gaps is~4 or~5, and $c$ is in the gaps. Let $\alpha_{i,j}$

2657: denote the generator where the gaps are $\{c,i,i+1,j\}$, $r-2\geq

2658: j\geq i+4$, $i\geq 2$, and $\beta_{i,j}$ denote the generator where

2659: the gaps are $\{c,i,j,j+1\}$, $r-3\geq j\geq i+3$, $i\geq 2$. Let

2660: $\gamma_{i,j}$ denote the generator where the gaps are $\{c,i,j\}$,

2661: $r-2\geq j\geq i+3$, $i\geq 2$. Clearly

2662: $d_1(\alpha_{i,j})=\gamma_{i,j}+\gamma_{i+1,j}$, and

2663: $d_1(\beta_{i,j})=\gamma_{i,j}+\gamma_{i,j+1}$. We see that,

2664: restricted to the generators $\alpha_{i,j}$, $\beta_{i,j}$, and

2665: $\gamma_{i,j}$, we have a~chain complex of the graph on Figure~\ref{fig:gr1}.

2666: 

2667: 

2668: %\[\pstexInput{gr1}\]

2669: %\[\text{ Figure 7.}\]

2670: 

2671: \begin{figure}[hbt]

2672: \begin{center}

2673: \pstexInput{gr1}

2674: \end{center}

2675: \caption{$\,$}

2676: \label{fig:gr1}

2677: \end{figure}

2678: 

2679: 

2680: The kernel is generated by the elementary squares, so it is enough so

2681: see that each square is a coboundary. Indeed, the elementary square

2682: with the lower left corner $(i,j)$ is a~coboundary of the generator

2683: with gaps $\{c,i,i+1,j,j+1\}$.

2684: 

2685: Finally, assume $(t,n)=(2,6)$ or $(3,5)$. These are the tight cases,

2686: in the sense that the lengths of all gaps are predetermined: the top

2687: gap consists of just $c$, and the other 2, resp.~3, gaps are of

2688: length~2. Assume that the kernel of $d_1$ is not zero, and let

2689: $\alpha$ be an~element in $\ker d_1$. Let $g$ be a~generator, which is

2690: contained in $\alpha$ with a~nonzero coefficient, such that this $g$

2691: maximizes the height of the top gap over all generators appearing with

2692: a~nonzero coefficient in~$\alpha$. Removing the lower element of the

2693: top gap of $g$ gives a~generator which cannot be cancelled out by the

2694: coboundaries of other elements in $\alpha$, due to the assumed

2695: maximality property. This yields a~contradiction, and hence $\ker

2696: d_1=0$.

2697: 

2698: We move on to group (Gr4), and let $B^*$ correspond to a~generator

2699: indexed by $\frown_j$, and $t$ side arcs.  We can have at most $2t+2$

2700: gaps. The dimension of the contributing cohomology generator is at

2701: least $t(n-2)+n-2$, thus total length of the gaps is at least

2702: $t(n-2)+1$. Comparing these inequalities we get

2703: \[t(n-2)+1-(2t+2)=t(n-4)-1>0,\]

2704: with the only exception $t=1,n=5$.

2705: 

2706: Let $(t,n)=(1,5)$. The interesting dimension here is 6, thus the total

2707: length of gaps must be exactly~4. The generators $\alpha_i$ indexed

2708: with the collection of arcs $\{\big{(}_{3,i},\frown_{i+3}\}$, for

2709: $4\leq i\leq r-3$.  Since $d_1(\alpha_i)$ contains the generator

2710: indexed with $\{\big{(}_{2,i},\frown_{i+3}\}$, and this generator is

2711: different for different $\alpha_i$, we see that the only linear

2712: combination of $\alpha_i$'s in the kernel of $d_1$ is the trivial one.

2713: Hence, we conclude that the contribution to $\wti E_2^{*,*}$ is~0.

2714: 

2715: Finally, we consider the case of generators indexed with collections

2716: of arcs avoiding all $\smile$- and $\frown$-arcs. Let us assume there

2717: are $t$ such arcs. To avoid a~cochain complex of a~simplex, the total

2718: length of the gaps must be at most $2t+1$. On the other hand, since

2719: the dimension of the generator is at least $t(n-2)$, the total length

2720: of the gaps must be at least $(t-1)(n-2)+1$. Comparing we see that

2721: \[(t-1)(n-2)+1-(2t+1)=(t-1)(n-4)-2>0,\]

2722: with the exceptions $t=1$, any $n$, $n=5$, $t\leq 3$, and $n=6$,

2723: $t=2$.

2724: 

2725: If $t=1$, the only nontrivial case occurs in the row $n$. Then, in the

2726: entry of interest we have only one generator: the one indexed by the

2727: arcs $\big{(}\big{)}_{3,r}$. Its coboundary will contain the generator

2728: $\big{(}\big{)}_{2,r}$, hence it is different from~0.

2729: 

2730: Let $(t,n)=(2,5)$. Since we are in the row 6, for dimensional reasons,

2731: the total length of gaps in the contributing generator is~4. Thus, we

2732: have two types of generators: $\alpha_i^1$ indexed with arc

2733: collections $\{\bbr_{2,i},\bbr_{i+3,r}\}$, $3\leq i\leq r-4$, and

2734: $\alpha_i^2$ indexed with arc collections

2735: $\{\bbr_{3,i},\bbr_{i+2,r}\}$, $4\leq i\leq r-3$. Considering the

2736: value of $d_1$ on the generator indexed with

2737: $\{\bbr_{3,i},\bbr_{i+3,r}\}$, we see that for $i>3$, modulo

2738: coboundaries, any generator $\alpha_i^1$ is a~linear combination of

2739: the generators $\alpha_j^2$. The coboundary of $\alpha_3^1$ contains

2740: the generator indexed with $\{\bbr_{2,3},\bbr_{5,r}\}$, hence no

2741: element in the kernel of $d_1$ can contain $\alpha_3^1$ with a~nonzero

2742: coefficient.  Finally, a~nonzero linear combination of $\alpha_j^2$'s

2743: cannot lie in the kernel of $d_1$, since $d_1(\alpha_j^2)$ contains

2744: the generator indexed with $\{\bbr_{2,j},\bbr_{j+2,r}\}$, which is

2745: different for different~$j$.  Again, we conclude that the contribution

2746: to $\wti E_2^{*,*}$ is~0.

2747: 

2748: Let $(t,n)=(3,5)$. For dimensional reasons, the total length of the

2749: gaps is precisely~7, thus we have the generators $\alpha_{i,j}$

2750: indexed with arc collections

2751: $\{\bbr_{3,i},\bbr_{i+3,j},\bbr_{j+3,r}\}$, for $4\leq i$, $i+4\leq

2752: j\leq r-4$. Since $d_1(\alpha_{i,j})$ contains the generator indexed

2753: with $\{\bbr_{2,i},\bbr_{i+3,j},\bbr_{j+3,r}\}$, and these generators

2754: are different for different $\alpha_{i,j}$'s, we see that $d_1$ is

2755: injective on the space spanned by $\alpha_{i,j}$'s.  Therefore, in

2756: this case the contribution to $\wti E_2^{*,*}$ is~0. The case

2757: $(t,n)=(2,6)$ is completely analogous.

2758: \qed

2759: 

2760: \begin{lm} \label{lm6.6}

2761: We have $E_2^{r+i,n-i-1}=0$, for all $i=3,\dots,n-1$.

2762: \end{lm}

2763: \pr Since $|\cc|=r+1$, the entries $E_1^{r+i,n-i-1}$, for

2764: $i=3,\dots,n-1$, come from

2765: $H^{n-i-1}(\thom(C_{2r+1}[\vt(\tau)],K_n))$, for

2766: $\tau\not\subseteq\cc$. We have shown before that these cohomology

2767: groups vanish in dimension $n-4$ and less, which implies

2768: $E_1^{r+i,n-i-1}=0$, hence $E_2^{r+i,n-i-1}=0$.

2769: \qed \vspace{5pt}

2770: 

2771: We conclude that $E_\infty^{r+1,n-3}=\zz$, contradicting the fact that

2772: $H^{r+n-2}(\thomp(C_{2r+1},K_n)/\zz;\zz)=0$. Therefore, our original

2773: assumption that $\varpi_1^{n-2}(\thom(C_{2r+1},K_n))\neq 0$ is wrong,

2774: and Theorem~\ref{thmmain1}(b) is proved.

2775: 

2776: 

2777: 

2778: 

2779: 

2780: \begin{thebibliography}{00}

2781: 

2782: \bibitem{BK1} E.\ Babson, D.N.\ Kozlov, {\it Topological obstructions

2783: to graph colorings}, Electron.\ Res.\ Announc.\ Amer.\ Math.\ Soc.\

2784: {\bf 9} (2003), 61--68.

2785: 

2786: \bibitem{BK2} E.\ Babson, D.N.\ Kozlov, {\it Complexes of graph

2787: homomorphisms}, Israel J.\ Math., in press. 

2788: \newline

2789: {\tt arXiv:math.CO/0310056}.

2790: 

2791: \bibitem{CF} P.E.\ Conner, E.E.\ Floyd, {\it Fixed point free involutions

2792: and equivariant maps}, Bull.\ Amer.\ Math.\ Soc.\ {\bf 66} (1960),

2793: 416--441.

2794: 

2795: \bibitem{CK2} S.Lj.\ \v{C}uki\'c, D.N.\ Kozlov, {\it Higher connectivity 

2796: of graph coloring complexes}, Int.\ Math.\ Res.\ Not.\ {\bf 25}

2797: (2005), 1543--1562.

2798: 

2799: \bibitem{TtD} T.\ tom Dieck, {\it Transformation groups}, Studies in

2800: Math.\ 8, de Gruyter, Berlin, New York, 1987.

2801: 

2802: \bibitem{En05} A.\ Engstr\"om, {\it A short proof of a conjecture 

2803: on the higher connectivity of graph coloring complexes}, preprint, 3

2804: pages, 2005.

2805: \newline

2806: {\tt arXiv:math.CO/0505460}

2807: 

2808: \bibitem{Hat} A.\ Hatcher, {\it Algebraic topology}, Cambridge

2809: University Press, Cambridge, 2002.

2810: 

2811: \bibitem{Knes} M.\ Kneser, Aufgabe 360, {\it Jber.\ Deutsch.\

2812: Math.-Verein.\ } {\bf 58} (1955).

2813: 

2814: \bibitem{Koz1} D.N.\ Kozlov, {\it Complexes of directed trees}, J.\

2815: Combin.\ Theory Ser.\ A {\bf 88} (1999), no.~1, 112--122.

2816: 

2817: \bibitem{dmt} D.N.\ Kozlov, {\em Discrete Morse theory for chain complexes}, 

2818: C.\ R.\ Acad.\ Sci.\ Paris, Ser.\ I {\bf 340} (2005), pp.\ 867--872.

2819: 

2820: \bibitem{K4} D.N.\ Kozlov, {\em A simple proof for folds on both 

2821: sides in complexes of graph homomorphisms}, Proc.\ Amer.\ Math.\ Soc., in press.      

2822: \newline

2823: {\tt arXiv:math.CO/0408262}

2824: 

2825: \bibitem{IAS} D.N.\ Kozlov, 

2826: {\em Chromatic numbers, morphism complexes, and Stiefel-Whitney

2827: characteristic classes}, in: {\em Geometric Combinatorics} (eds.\ E.\

2828: Miller, V.\ Reiner, B.\ Sturmfels), IAS/Park City Mathematics Series

2829: {\bf 14}, American Mathematical Society, Providence, RI; Institute for

2830: Advanced Study (IAS), Princeton, NJ; in press.

2831: \newline

2832: {\tt arXiv:math.AT/0505563}

2833: 

2834: \bibitem{K5} D.N.\ Kozlov, {\em Simple homotopy type of some

2835: combinatorially defined complexes}, preprint, 12 pages, 2005.

2836: \newline

2837: {\tt arXiv:math.AT/0503613}

2838:  

2839: \bibitem{ccc} D.N.\ Kozlov, {\em Cohomology of colorings of cycles}, 

2840: preprint, 24 pages, 2005.

2841: \newline

2842: {\tt arXiv:math.AT/0507117}

2843:  

2844: 

2845: \bibitem{Lo} L.\ Lov\'asz, {\it Kneser's conjecture, chromatic number,

2846:     and homotopy}, J.\ Combin.\ Theory Ser.~A {\bf 25}, (1978), no.~3,

2847:   pp.~319--324.

2848: 

2849: \bibitem{Qu} D.\ Quillen, {\em Higher algebraic K-theory} I, Lecture

2850:   Notes in Mathematics {\bf 341} (1973), pp.~77--139,

2851:   Springer-Verlag.

2852: 

2853: 

2854: \bibitem{Zi02} G.M.\ Ziegler, {\it Generalized Kneser coloring

2855:     theorems with combinatorial proofs}, Invent.~Math. {\bf 147}

2856:   (2002), pp.~671-691.

2857: 

2858: \bibitem{Z1} R.T.\ \v{Z}ivaljevi\'{c}, {\em Parallel transport of 

2859: $\thom$-complexes and the Lov\'asz conjecture}, preprint, 17 pages, 2005.

2860:   \newline {\tt arXiv:math.CO/0506075}

2861: 

2862: \end{thebibliography}

2863: 

2864: \end{document}

2865: