1:
2: \documentclass{article}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \usepackage{amssymb}
5: \usepackage{amsfonts}
6: \usepackage{graphicx}
7: \usepackage{amsmath}
8:
9: \setcounter{MaxMatrixCols}{10}
10: %TCIDATA{OutputFilter=LATEX.DLL}
11: %TCIDATA{Version=4.10.0.2363}
12: %TCIDATA{Created=Monday, June 07, 2004 17:50:34}
13: %TCIDATA{LastRevised=Tuesday, July 20, 2004 16:42:45}
14: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
15: %TCIDATA{<META NAME="DocumentShell" CONTENT="Standard LaTeX\Standard LaTeX Article">}
16: %TCIDATA{Language=American English}
17: %TCIDATA{CSTFile=LaTeX article (bright).cst}
18:
19: \textwidth 14cm
20: \newtheorem{theorem}{Theorem}
21: \newtheorem{acknowledgement}[theorem]{Acknowledgement}
22: \newtheorem{algorithm}[theorem]{Algorithm}
23: \newtheorem{axiom}[theorem]{Axiom}
24: \newtheorem{case}[theorem]{Case}
25: \newtheorem{claim}[theorem]{Claim}
26: \newtheorem{conclusion}[theorem]{Conclusion}
27: \newtheorem{condition}[theorem]{Condition}
28: \newtheorem{conjecture}[theorem]{Conjecture}
29: \newtheorem{corollary}[theorem]{Corollary}
30: \newtheorem{criterion}[theorem]{Criterion}
31: \newtheorem{definition}[theorem]{Definition}
32: \newtheorem{example}[theorem]{Example}
33: \newtheorem{exercise}[theorem]{Exercise}
34: \newtheorem{lemma}[theorem]{Lemma}
35: \newtheorem{notation}[theorem]{Notation}
36: \newtheorem{problem}[theorem]{Problem}
37: \newtheorem{proposition}[theorem]{Proposition}
38: \newtheorem{remark}[theorem]{Remark}
39: \newtheorem{solution}[theorem]{Solution}
40: \newtheorem{summary}[theorem]{Summary}
41: \newenvironment{proof}[1][Proof]{\noindent\textbf{#1.} }{\ \rule{0.5em}{0.5em}}
42: %\input{tcilatex}
43:
44: \begin{document}
45:
46: \title{\textsf{Topology of partition of measures by fans and the second
47: obstruction}}
48: \author{Pavle V.M. Blagojevi\'{c}\thanks{{\small Supported by St. John the
49: Baptist Serbian Orthodox Church in San Francisco, the grant 1854 of the
50: Serbian Ministry of Science, Technology and Development, by Town goverments
51: of Leskovac and Sremska Mitrovica and Water supply company of Leskovac.}} \\
52: %EndAName
53: Mathematical Institute SANU, Belgrade}
54: \maketitle
55:
56: \begin{abstract}
57: \noindent The simultaneous partition problems are classical problems of the
58: combinatorial geometry which have the natural flavor of the equivariant
59: topology. The $k$-fan partition problems have attracted a lot of attention
60: \cite{Aki2000}, \cite{BaMa2001}, \cite{BaMa2002} and forced some hard
61: concrete combinatorial calculations in the equivariant cohomology \cite%
62: {Bl-Vr-Ziv}. These problems can be reduced, by a beautiful scheme of \cite%
63: {BaMa2001}, to a \textquotedblright typical\textquotedblright\ question of
64: the existence of a $\mathbb{D}_{2n}$ equivariant map $f:V_{2}(\mathbb{R}%
65: ^{3})\rightarrow W_{n}-\cup \mathcal{A}(\alpha )$, \ where $V_{2}(\mathbb{R}%
66: ^{3})$ is the space of all orthonormal $2$-frames in $\mathbb{R}^{3}$ and $%
67: W_{n}-\cup \mathcal{A}(\alpha )$ is the complement of the appropriate
68: arrangement. We introduce the \textit{target extension scheme} which allow
69: us to use the equivariant obstruction theory as a tool for proving that: for
70: every two proper measures on the sphere $S^{2}$, and any $\alpha
71: =(a,a+b,b)\in \mathbb{R}_{>0}^{3}$, there exists an $\alpha $-partition of
72: theses measures by a $3$-fan.
73:
74: \noindent The significance of these results, among other, is that, beside
75: negative results \cite{Bl-Vr-Ziv}, the equivariant obstruction theory can
76: pull off some positive results, which were not attained by other means.
77: \end{abstract}
78:
79: \section{\textsf{Introduction}}
80:
81: \subsection{Problem}
82:
83: \noindent A $k$-fan $(x;l_{1},l_{2},\ldots ,l_{k})$ on the sphere $S^{2}$ is
84: formed of a point $x$, called the center of the fan, and $k$ great
85: semicircles $l_{1},\ldots ,l_{k}$ emanating from $x$. We always assume
86: counter clockwise enumeration on great semicircles $l_{1},\ldots ,l_{k}$ of
87: a $k$-fan. Sometimes instead of lines we use open angular sectors $\sigma
88: _{i}$ between $l_{i}$ and $l_{i+1},\,i=1,\ldots ,k$. In that case we denote
89: a $k$-fan with $(x;\sigma _{1},\sigma _{2},\ldots ,\sigma _{k})$.
90:
91: \noindent Let $\mu _{1},\mu _{2},\ldots ,\mu _{m}$ be \textit{proper} Borel
92: probability measures on $S^{2}$. Measure $\mu $ is \textit{proper} if $\mu
93: ([a,b])=0$ for any circular arc $[a,b]\subset S^{2}$ and $\mu (U)>0$ for
94: each nonempty open set $U\subset S^{2}$. All results can be extended to more
95: general measures, including the counting measures of finite sets, see \cite%
96: {BaMa2001} for related examples.
97:
98: \noindent Let $(\alpha _{1},\alpha _{2},\ldots ,\alpha _{k})\in \mathbb{R}%
99: _{>0}^{k}$ be a vector where $\alpha _{1}+\alpha _{2}+\ldots +\alpha _{k}=1$%
100: . The general problem stated in \cite{BaMa2001} is:
101:
102: \begin{problem}
103: Find all triples $(m,k,\alpha )\in \mathbb{N\times }\mathbb{N\times R}^{k}$
104: such that for any collection of $m$ measures $\{\mu _{1},\mu _{2},\ldots
105: ,\mu _{m}\}$, there exists a $k$-fan $(x;l_{1},\ldots ,l_{k})$ with the
106: property%
107: \begin{equation*}
108: (\forall i=1,\ldots ,k)\,(\forall j=1,\ldots ,m)\,\mu _{j}(\sigma
109: _{i})=\alpha _{i}
110: \end{equation*}%
111: That kind of a $k$-fan $(x;l_{1},\ldots ,l_{k})$ is called an $\alpha $%
112: \textit{-partition} for the collection of measures $\{\mu _{j}\}_{j=1}^{m}$.
113: \end{problem}
114:
115: \noindent The analysis given in \cite{BaMa2001} shows that the most
116: interesting triples are $(3,2,\alpha ),(2,3,\alpha ),(2,4,\alpha )$. Known
117: results can be summed in the following table:%
118: \begin{equation*}
119: \begin{tabular}{|l|l|l|l|}
120: \hline
121: $m/k$ & $2$ & $3$ & $4$ \\ \hline
122: $2$ & $R_{\geq 0}^{2},\text{\cite{BaMa2001}}$ & $%
123: \begin{array}{c}
124: (\tfrac{1}{3},\tfrac{1}{3},\frac{1}{3}) \\
125: (\frac{1}{2},\frac{1}{4},\frac{1}{4}) \\
126: (\frac{\ast }{5},\frac{\ast }{5},\frac{\ast }{5})%
127: \end{array}%
128: \text{\cite{BaMa2001}}$ & $%
129: \begin{array}{c}
130: (\frac{2}{5},\frac{1}{5},\frac{1}{5},\frac{1}{5})\text{\cite{BaMa2001}} \\
131: (\frac{1}{4},\frac{1}{4},\frac{1}{4},\frac{1}{4})\text{\cite{BaMa2002}} \\
132: \text{{\small Topology can't help any more \cite{Bl-Vr-Ziv}.}}%
133: \end{array}%
134: $ \\ \hline
135: $3$ & $%
136: \begin{array}{c}
137: (\frac{1}{2},\frac{1}{2}) \\
138: (\frac{1}{3},\frac{2}{3}) \\
139: (\frac{2}{3},\frac{1}{3})%
140: \end{array}%
141: \text{\cite{BaMa2001}}$ & $\emptyset $ & $\emptyset $ \\ \hline
142: \end{tabular}%
143: \
144: \end{equation*}%
145: In this paper we try to fill this table a little bit more.
146:
147: \subsection{The statement of results}
148:
149: \noindent We are interested in the problem of $3$-fan partitions of two
150: measures on $S^{2}$, respectively. We prove the following result:
151:
152: \begin{theorem}
153: \label{th:main1}Lets choose $\alpha =(a,a+b,b)\in \mathbb{R}_{>0}^{3}$ such
154: that $2a+2b=1$. Then any two \textit{proper} measures $\mu $ and $\nu $ on
155: the sphere $S^{2}$ admit an $\alpha $-partition by a $3$-fan $\mathfrak{p}%
156: =(x;l_{1},l_{2},l_{3})$, i. e.
157: \begin{equation*}
158: \mathcal{A}_{2,3}\supseteq \{(a,a+b,b)\in \mathbb{R}^{3}~|~a,b>0,~2a+b=1\}.
159: \end{equation*}
160: \end{theorem}
161:
162: \begin{remark}
163: The case $a=b=1$ was already considered in \cite{BaMa2001}
164: \end{remark}
165:
166: \subsection{The solution scheme}
167:
168: \noindent The solution of the problem has two natural parts. The first part
169: is the reduction of the problem to the question of the existence of the
170: appropriate equivariant map. The second part is and topological effort to
171: disprove the existence of a such map. There is also a third part of the
172: proof, the limit argument. It extends the result from rational triples to
173: real triples, but we omited it because it is the standard part of every
174: similiar proof.
175:
176: \medskip
177:
178: \noindent \textit{Reduction to the equivariant problem:}
179:
180: \begin{itemize}
181: \item The configuration space / test map procedure of Imre B\'{a}r\'{a}ny
182: and Ji\v{r}i Matou\v{s}ek from \cite{BaMa2001} reduces the problem to the
183: question: Is there an $\alpha =(\frac{a_{1}}{n}$,$\frac{a_{2}}{n}$,$\frac{%
184: a_{3}}{n})$, such that there is no $\mathbb{D}_{2n}$-map $V_{2}(\mathbb{R}%
185: ^{3})\rightarrow W_{n}\setminus \cup \mathcal{A}(\alpha )$?
186:
187: \item The extension of scalars\ equivalence from homological algebra, \cite%
188: {Brown}, allows us to change the initial equivariant question to: Is there
189: an $\alpha =(\frac{a_{1}}{n},\frac{a_{2}}{n},\frac{a_{3}}{n})$, such that
190: there is no $\mathbb{Q}_{4n}$-map $S^{3}\rightarrow W_{n}\setminus \cup
191: \mathcal{A}(\alpha )$?
192:
193: \item The elementary obstruction theory can convince us that the actual free
194: $\mathbb{Q}_{4n}$ action on $S^{3}$ is not of the essential importance. Thus
195: we can always change it the way it pleases us.
196: \end{itemize}
197:
198: \medskip
199:
200: \noindent \textit{Obstruction theory approach:}
201:
202: \begin{itemize}
203: \item The dimensional reasons of the problem, $\pi _{1}(W_{n}\setminus \cup
204: \mathcal{A}(\alpha ))\neq \varnothing $, implies that there are two
205: essential obstructions. This forces us to introduce the target extension
206: scheme, and instead of dealing with two obstructions we have only one to
207: compute. The target extension scheme extends the $\mathbb{Q}_{4n}$ space $%
208: W_{n}\setminus \cup \mathcal{A}(\alpha )$ in such a way that connectivity
209: increases by one. Thus, the problem of the existence of the $\mathbb{Q}_{4n}
210: $-map from $S^{3}$ to this extension has only one obstruction.
211:
212: \item Once we extend the complement $W_{n}\setminus \cup \mathcal{A}(\alpha )
213: $ we use the classical obstruction theory and the "map in general position"
214: method to prove that appropriate obstruction is not zero. This part of the
215: proof goes in a small number of steps, like we described in the section \ref%
216: {sec:HowToComputeTheOstructionCocycle}.
217:
218: \item The main step is the identification of the obstruction cocycle. This
219: is the point where many previous papers have been broken (\cite{VreZiv2002},
220: proof of theorem 4.2, equality (25) and \cite{limun2002}, proof of theorem
221: 6.1). We use the clear geometrical picture and simple testing methods (in
222: sections \ref{Sec:Step7} and \ref{sec:Step8}) to interpret the obstruction
223: element as a non-zero element of the appropriate coinvariant group.
224: \end{itemize}
225:
226: \noindent This scheme of the solution, with different extensions, can be
227: applied to other cases of the $3$-fan / $2$-measures problem, as well on the
228: $2$-fan / $3$-measures problem. The main idea of extending the target space,
229: of course with some variations, can be applied on every problem of the
230: existace of the equivaiant map to a complement of an arrangement, or to any
231: space that has natural extension candidates.
232:
233: \section{\textsf{From a partition problem to an equivariant problem}}
234:
235: \subsection{Reduction to the equivariant problem}
236:
237: \noindent The reduction of the fan partition problem to the equivariant
238: problem is done by the configurations space / test map scheme. The main idea
239: is to look at the space of all possible solutions and to rephrase the
240: question in terms of coincidences of the associated test map. Imre B\'{a}r%
241: \'{a}ny and Ji\v{r}i Matou\v{s}ek demonstrated in \cite{BaMa2001} that the
242: test map scheme can be applied on the problem of $\alpha $-partitioning of $%
243: m $-measures on $S^{2}$ by spherical $k$-fans. In a very elegant way this
244: problem was reduced to the problem of the existence of the appropriate
245: equivariant map. We briefly review this reduction for the $(3,2)$ case of
246: this problem.
247:
248: \noindent \textbf{The configuration space.} Let $\mu $ and $\nu $ be two
249: proper Borel probability measures on $S^{2}$, and $F_{k}$ the space of all $%
250: k $-fans on the sphere $S^{2}$. The space $X_{\mu }$ of all possible
251: solutions associated to the measure $\mu $ is defined by
252: \begin{equation*}
253: X_{\mu }=\{(x;l_{1},\ldots ,l_{n})\in F_{n}\mid (\forall i=1,\ldots ,n)\,\mu
254: (\sigma _{i})=\tfrac{{\small 1}}{{\small n}}\}.
255: \end{equation*}%
256: Observe that every $n$-fan $(x;l_{1},\ldots ,l_{n})\in X_{\mu }$ is
257: completely determined by the pair $(x,l_{1})$ or equivalently, the pair $%
258: (x,y)$, where $y$ is the unit tangent vector to $l_{1}$ at $x$. Thus, the
259: space $X_{\mu }$ is a Stiefel manifold $V_{2}(\mathbb{R}^{3})$ of all
260: orthonormal $2$-frames in $\mathbb{R}^{3}$. Keep in mind that $V_{2}(\mathbb{%
261: R}^{3})\cong SO(3)\cong \mathbb{R}P^{3}$.
262:
263: \noindent \textbf{Test map.} Let $\mathbb{R}^{n}$ be an Euclidean
264: space with the standard orthonormal basis $e_{1},e_{2},\ldots
265: ,e_{n}$ and the associated coordinate functions
266: $x_{1},x_{2},\ldots ,x_{n}$. Let $W_{n}$ be
267: the hyperplane $\{x\in \mathbb{R}^{n}\mid x_{1}+x_{2}+\ldots +x_{n}=0\}$ in $%
268: \mathbb{R}^{n}$, and suppose that $\alpha $-vectors have the following form
269: \begin{equation*}
270: \alpha =(\tfrac{a_{1}}{n},\tfrac{a_{2}}{n},\tfrac{a_{3}}{n})\in \frac{1}{n}\,%
271: \mathbb{N}^{3}\subseteq \mathbb{Q}^{3},
272: \end{equation*}%
273: where $a_{1}+a_{2}+a_{3}=n$. Then test maps for the $(3,2)$ fan problem are
274: defined by%
275: \begin{equation*}
276: \begin{array}{lll}
277: F_{\nu }:X_{\mu }\rightarrow W_{n} & & F_{\nu }(\mathfrak{p})=(\nu (\sigma
278: _{1})-\tfrac{{\small 1}}{{\small n}},\ldots ,\nu (\sigma _{n})-\tfrac{%
279: {\small 1}}{{\small n}})%
280: \end{array}%
281: \end{equation*}%
282:
283: \noindent \textbf{The action.} The dihedral group
284: $\mathbb{D}_{2n}=\langle j,\varepsilon \,|\,\varepsilon
285: ^{n}=j^{2}=1,\,\varepsilon j=j\varepsilon ^{n-1}\,\rangle $ acts
286: both on the possible solution space $X_{\mu }$ and the linear
287: subspace $W_{n}\subseteq \mathbb{R}^{n}$ by
288: \begin{equation*}
289: X_{\mu }:\left\{
290: \begin{array}{l}
291: \varepsilon (x;l_{1},\ldots ,l_{n})=(x;l_{n},l_{1},\ldots ,l_{n-1}) \\
292: j(x;l_{1},\ldots ,l_{n})=(-x;l_{1},l_{n},l_{n-1,}\ldots ,l_{2})%
293: \end{array}%
294: ,\right. W_{n}:\left\{
295: \begin{array}{c}
296: \varepsilon (x_{1},\ldots ,x_{n})=(x_{2},\ldots ,x_{n},x_{1}) \\
297: j(x_{1},\ldots ,x_{n})=(x_{n},\ldots ,x_{2},x_{1})%
298: \end{array}%
299: ,\right.
300: \end{equation*}%
301: for $(x;l_{1},\ldots ,l_{n})\in X_{\mu }$ and $(x_{1},\ldots
302: ,x_{n})\in W_{n} $. The action of $\mathbb{D}_{2n}$ on $X_{\mu }$
303: is \textbf{free}.
304:
305: \noindent Observe that the space of possible solutions $X_{\mu }$ is $%
306: \mathbb{D}_{2n}$-homeomorphic to the manifold
307: $V_{2}(\mathbb{R}^{3})$, where $V_{2}(\mathbb{R}^{3})$ is a
308: $\mathbb{D}_{2n}$-space given by
309: \begin{equation*}
310: \varepsilon (x,y)=(x,R_{x}({\frac{2\pi }{n}})(y))\text{, }j(x,y)=(-x,y)\text{%
311: ,}
312: \end{equation*}%
313: and $R_{x}(\theta ):\mathbb{R}^{3}\rightarrow \mathbb{R}^{3}$ is
314: the rotation round the axes determined by $x$ through the angle
315: $\theta $.
316:
317: \noindent \textbf{The test space. }The test space in this problem
318: is the union $\cup \mathcal{A}(\alpha )\subset W_{n}$ of a
319: smallest $\mathbb{D}_{2n} $-invariant linear subspace arrangement
320: $\mathcal{A}(\alpha )$, which contains linear subspace $L(\alpha
321: )\subset W_{n}$. The subspace $L(\alpha )$ is defined by
322: \begin{equation*}
323: L(\alpha )=\{x\in \mathbb{R}^{n}\mid \xi _{1}(x)=\xi _{2}(x)=\xi
324: _{3}(x)=0\}\subseteq W_{n},
325: \end{equation*}%
326: where
327: \begin{equation*}
328: \begin{array}{lll}
329: \xi _{1}(x)=x_{1}+\ldots +x_{a_{1}}, & \xi
330: _{2}(x)=x_{a_{1}+1}+\ldots
331: +x_{a_{1}+a_{2}}, & \xi _{3}(x)=x_{a_{1}+a_{2}+1}+\ldots +x_{n}\text{.}%
332: \end{array}%
333: \end{equation*}
334:
335:
336: \noindent Since the test map $F_{\nu }$ is obviously $\mathbb{D}_{2n}$%
337: -equivariant, the following standard proposition is proved.
338:
339: \begin{proposition}
340: \label{prop:VezaProblem-Ekvivarijantan}Let $\alpha =(\frac{a_{1}}{n},\frac{%
341: a_{2}}{n},\frac{a_{3}}{n})\in \frac{1}{n}\,\mathbb{N}^{3}\subseteq \mathbb{Q}%
342: ^{3}$ be a vector such that $a_{1}+a_{2}+a_{3}=n$. If there is no $\mathbb{D}%
343: _{2n}$-equivariant map%
344: \begin{equation*}
345: F:V_{2}(\mathbb{R}^{3})\rightarrow W_{n}\setminus \cup \mathcal{A}(\alpha )
346: \end{equation*}%
347: then for any two measures $\mu $ and $\nu $ on $S^{2}$, there exists an $%
348: \alpha $-partition $(x;l_{1},l_{2},l_{3})$ of measures $\mu $ and $\nu $.
349: \end{proposition}
350:
351: \subsection{Modifying Problem}
352:
353: \noindent Knowing the fact that for an odd $n$, there always exists a $%
354: \mathbb{Z}_{n}$-map $f:S^{3}\rightarrow V_{2}(\mathbb{R}^{3})$, B\'{a}r\'{a}%
355: ny and Matou\v{s}ek in \cite{BaMa2001} questioned if there is a $\mathbb{Z}%
356: _{n}$-map from the sphere $S^{3}$ to the complements of appropriate
357: arrangements. To do something similar we extend the group, like in \cite%
358: {limun2002} and \cite{Bl-Vr-Ziv}. We use well known \textquotedblleft
359: extension of scalars\textquotedblright\ equivalence from homological
360: algebra, \cite{Brown} Section III.3.
361:
362: \noindent \textbf{The generalized quaternion group.} Let $S^{3}=S(\mathbb{H}%
363: )=Sp(1)$ be the group of all unit quaternions and let $\epsilon =\epsilon
364: _{2n}=\cos \frac{\pi }{n}+i\sin \frac{\pi }{n}\in S(\mathbb{H})$ be a root
365: of unity. Group $\langle \epsilon \rangle $ is a subgroup of $S(\mathbb{H})$
366: of the order $2n$. Then, the generalized quaternion group\textit{, }\cite%
367: {CaEi}\textit{\ }p. 253\textit{, }is the subgroup
368: \begin{equation*}
369: \mathbb{Q}_{4n}=\{1,\epsilon ,\ldots ,\epsilon ^{2n-1},j,\epsilon j,\ldots
370: ,\epsilon ^{2n-1}j\}
371: \end{equation*}%
372: of $S^{3}$, of the order $4n$. Let $H=\{1,\epsilon ^{n}\}=\{1,-1\}\subset
373: \mathbb{Q}_{4n}$. Then, it is not hard to see that the quotient group $%
374: \mathbb{Q}_{4n}/H$ is isomorphic to the dihedral group $\mathbb{D}_{2n}$ of
375: the order $2n$.\textbf{\ }
376:
377: \begin{proposition}
378: \label{prop:PrelazakNaNovuGrupu}Let the generalized quaternion group\textit{%
379: \ }$\mathbb{Q}_{4n}$ act on $S^{3}$ as a subgroup, and on $W_{n}$ via
380: already defined $\mathbb{D}_{2n}$ action by the quotient homomorphism $%
381: \mathbb{Q}_{4n}\rightarrow \mathbb{Q}_{4n}/\{1,-1\}\cong \mathbb{D}_{2n}$.
382: Then the following maps coexist:%
383: \begin{equation*}
384: \mathbb{D}_{2n}\text{-map }V_{2}(\mathbb{R}^{3})\rightarrow W_{n}\setminus
385: \cup \mathcal{A}(\alpha )\text{ \ and \ }\mathbb{Q}_{4n}\text{-map }%
386: S^{3}\rightarrow W_{n}\setminus \cup \mathcal{A}(\alpha ).
387: \end{equation*}
388:
389: \noindent By the coexistence we mean that the one map exists if and only if
390: the other map exists, i.e. the one can't exist without the other.
391: \end{proposition}
392:
393: \begin{proof}
394: Let us denote the target space $W_{n}\setminus \cup \mathcal{A}(\alpha )$
395: with $T$. Also, observe that $S^{3}/\{1,-1\}\cong \mathbb{R}P^{3}\cong
396: SO(3)\cong V_{2}(\mathbb{R}^{3})$
397:
398: \noindent $\Rightarrow :$ Let $F:V_{2}(\mathbb{R}^{3})\rightarrow T$ be a $%
399: \mathbb{D}_{2n}$-map. The quotient map $p:S^{3}\rightarrow
400: S^{3}/\{1,-1\}\cong V_{2}(\mathbb{R}^{3})$ is a $\mathbb{Q}_{4n}$-map where
401: the $\mathbb{Q}_{4n}$-acts on $V_{2}(\mathbb{R}^{3})$ by the quotient
402: homomorphism $\mathbb{Q}_{4n}\rightarrow \mathbb{Q}_{4n}/\{1,-1\}\cong
403: \mathbb{D}_{2n}$. Since, $\mathbb{Q}_{4n}$ acts on both $V_{2}(\mathbb{R}%
404: ^{3})$ and $T$ via the quotient homomorphism, the given $\mathbb{D}_{2n}$%
405: -map $F:V_{2}(\mathbb{R}^{3})\rightarrow T$ can also be seen as the $\mathbb{%
406: Q}_{4n}$-map. Thus, the composition
407: \begin{equation*}
408: F\circ p:S^{3}\longrightarrow S^{3}/\{1,-1\}\cong V_{2}(\mathbb{R}%
409: ^{3})\longrightarrow T
410: \end{equation*}%
411: is the required $\mathbb{Q}_{4n}$-map $S^{3}\rightarrow T$.
412:
413: \noindent $\Rightarrow :$ Let $G:S^{3}\rightarrow T$ be a $\mathbb{Q}_{4n}$%
414: -map. Observe that $S^{3}/\{1,-1\}\cong V_{2}(\mathbb{R}^{3})$ can be seen
415: as a $\mathbb{D}_{2n}\cong \mathbb{Q}_{4n}/\{1,-1\}$ space by $\{1,-1\}x%
416: \overset{g\{1,-1\}}{\longmapsto }\{1,-1\}(gx)$, where $x\in S^{3}$ and $g\in
417: \mathbb{Q}_{4n}$. Since the subgroup $\{1,-1\}$ acts trivially on $T$, there
418: is a factorization of the map $G$ through the quotient $h:S^{3}/\{1,-1\}%
419: \rightarrow T$ such that $G=h\circ p$. The map $h$ is the required $\mathbb{D%
420: }_{2n}$-map%
421: \begin{equation*}
422: h(g\{1,-1\}\cdot \{1,-1\}x)=G(g\cdot x)=g\cdot G(x)=g\cdot
423: h(\{1,-1\}x)=(g\{1,-1\})\cdot h(\{1,-1\}x).
424: \end{equation*}
425: \end{proof}
426:
427: \begin{remark}
428: The $\mathbb{Q}_{4n}$ action on $S^{3}$ is \textbf{free}. Also, the $\mathbb{%
429: Q}_{4n}$ action on $W_{n}$ is the restriction of the following $\mathbb{Q}%
430: _{4n}$ action on $\mathbb{R}^{n}$. Let $e_{1},..,e_{n}$ be the standard
431: orthonormal basis in $\mathbb{R}^{n}$. The action is defined by
432: \begin{equation*}
433: \epsilon \cdot e_{i}=e_{i \ \textrm{mod}\ n+1} \text{ and }j\cdot e_{i}=e_{n-i+1}%
434: \text{.}
435: \end{equation*}
436: \end{remark}
437:
438: \noindent \textbf{The free action on }$S^{3}$\textbf{.} Since the sphere $%
439: S^{3}$ is $2$-connected it turns out that the particular $\mathbb{Q}_{4n}$%
440: -action on $S^{3}$ is not something we have to live with. The elementary
441: equivariant obstruction theory allows us to prove the following useful fact.
442:
443: \begin{proposition}
444: \label{prop:promenaDejstva}If $\gamma _{1}$ and $\gamma _{2}$ are $G$%
445: -actions on $S^{3}$ and $\gamma _{1}$ is free, then there exists a $G$-map $%
446: f:S^{3}\rightarrow S^{3}$ such that
447: \begin{equation*}
448: (\forall g\in G)\,(\forall x\in S^{3})\,\,f(g\cdot _{1}x)=g\cdot _{2}f(x)%
449: \text{.}
450: \end{equation*}
451: \end{proposition}
452:
453: \begin{proof}
454: The statement is true because $S^{3}$ is $2$-connected, $\gamma _{1}$ is
455: free and so there are no obstructions to extend a $G$-map from $0$-skeleton
456: to $S^{3}$.
457: \end{proof}
458:
459: \noindent Thus, when the time comes we will be able to use any free $\mathbb{%
460: Q}_{4n}$ action on $S^{3}$ and we will have the favorite one.
461:
462: \section{\textsf{Obstruction theory approach}}
463:
464: \noindent Once again, denote the target space $W_{n}\setminus \cup \mathcal{A%
465: }(\alpha )$ with $T$. To answer a question of the existence of $\mathbb{Q}%
466: _{4n}$-map $S^{3}\rightarrow T$ we will try to employ the classical
467: obstruction theory. Since the maximal elements of the arrangement $\mathcal{A%
468: }(\alpha )$ are of the codimension $2=(n-1)-(n-3)$ in $W_{n}$, the
469: complement is connected and the first obstruction lives in $H_{\mathbb{Q}%
470: _{4n}}^{2}(S^{3},\pi _{1}(T))$. In this case it is very hard even to
471: identify this group, not to mention to identify the particular element in
472: it. Even if we do manage to identify and calculate the first obstruction,
473: there is a good chance that it is zero, so the second obstructions should be
474: calculated. To omit this difficulties we use the nature of the target
475: spaces, introduce the target extension scheme and then use the equivariant
476: obstruction theory.
477:
478: \subsection{The Target extension scheme}
479:
480: \noindent According to the proposition \ref{prop:VezaProblem-Ekvivarijantan}
481: we would prefer to prove that there are no equivariant $\mathbb{Q}_{4n}$%
482: -maps $S^{3}\rightarrow T$. Thus the following scheme can be of some help.
483:
484: \noindent \textbf{The basic idea} of the target extension scheme is to
485: \textit{find a }$\mathbb{Q}_{4n}$\textit{\ space }$E$\textit{\ which
486: contains the target space }$T$\textit{\ and to prove that there is no }$%
487: \mathbb{Q}_{4n}$\textit{-map }$S^{3}\rightarrow E$. Thus, this would imply
488: that there is no $\mathbb{Q}_{4n}$\textit{-map }$S^{3}\rightarrow T$.
489:
490: \noindent Since the target space $T$ is the complement of the arrangement,
491: the basic idea can be refined as follows:
492:
493: (A) \textit{Increase the codimension of the arrangement:} Take an arbitrary
494: hyper arrangement $\mathcal{J}$ in $W_{n}$. By the \textit{hyper}
495: arrangement we mean the arrangement of hyperplanes and / or closed
496: hyperplane halfspaces. Then form the minimal $\mathbb{Q}_{4n}$-invariant
497: arrangement $\mathcal{A}(\mathcal{J},\alpha )$ containing the family $%
498: \mathcal{J}\cap L(\alpha )=\{J\cap L(\alpha )~|~J\in \mathcal{J}\}$. Then
499: inclusion $\cup \mathcal{A}(\mathcal{J},\alpha )\subseteq \cup \mathcal{A}%
500: (\alpha )$ implies that $W_{n}-\cup \mathcal{A}(\mathcal{J},\alpha
501: )\supseteq W_{n}-\cup \mathcal{A}(\alpha )$. Observe that the dimension of
502: maximal elements of the arrangement $\mathcal{A}(\mathcal{J},\alpha )$ is $%
503: n-4$. Let us denote (when it suits us) the new union $\cup \mathcal{A}(%
504: \mathcal{J},\alpha )$ by $U^{\ast }$, and the new complement $W_{n}-\cup
505: \mathcal{A}(\mathcal{J},\alpha )$ by $T^{\ast }$.
506:
507: (B) \textit{Apply the obstruction theory to the new question:} Is there a $%
508: \mathbb{Q}_{4n}$-map $S^{3}\rightarrow T^{\ast }$. Since the codimension of
509: maximal elements of the new defined arrangement in $W_{n}$ is $3$, the
510: target space $T^{\ast }$ is $1$-connected and consequently $2$-simple in the
511: sense that $\pi _{1}(T^{\ast })$ acts trivially on $\pi _{2}(T^{\ast })$.
512: Thus by Hurewicz theorem $\pi _{2}(T^{\ast })\cong \lbrack S^{2},T^{\ast
513: }]\cong H_{2}(T^{\ast };\mathbb{Z})$. Then the part of the obstruction exact
514: sequence (\cite{Dieck87}, \cite{guide2}) we are interested in is
515:
516: \begin{equation*}
517: \begin{array}{ccccc}
518: \lbrack S^{3},T^{\ast }]_{\mathbb{Q}_{4n}} & \overset{\theta }{%
519: \longrightarrow } & \mathrm{Im}\left\{ [S_{(2)}^{3},T^{\ast }]_{\mathbb{Q}%
520: _{4n}}\longrightarrow \lbrack S_{(1)}^{3},T^{\ast }]_{\mathbb{Q}%
521: _{4n}}\right\} & \overset{\tau _{\mathbb{Q}_{4n}}}{\longrightarrow } & H_{%
522: \mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))%
523: \end{array}%
524: \end{equation*}%
525: where $S_{(1)}^{3}$ and $S_{(2)}^{3}$ are respectively the $1$- and $2$%
526: -skeleton of $S^{3}=P_{2n}\ast P_{2n}$ and $H_{2}(T^{\ast };\mathbb{Z})$ is
527: viewed as a $\mathbb{Q}_{4n}$-module. Since $[S_{(1)}^{3},T^{\ast }]_{%
528: \mathbb{Q}_{4n}}=\{\ast \}$ is a one-element set and $[S_{(2)}^{3},T^{\ast
529: }]_{\mathbb{Q}_{4n}}\neq \varnothing $, the sequence becomes
530: \begin{equation*}
531: \begin{array}{ccccc}
532: \lbrack S^{3},T^{\ast }]_{\mathbb{Q}_{4n}} & \longrightarrow & \{\ast \} &
533: \overset{\tau _{\mathbb{Q}_{4n}}}{\longrightarrow } & H_{\mathbb{Q}%
534: _{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z})).%
535: \end{array}%
536: \end{equation*}%
537: The exactness means that the set $[S^{3},T^{\ast }]_{\mathbb{Q}_{4n}}\neq
538: \emptyset $ if and only if $\tau _{\mathbb{Q}_{4n}}(\ast )\in H_{\mathbb{Q}%
539: _{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$ is equal to zero. The element
540: $\tau _{\mathbb{Q}_{4n}}(\ast )$ depends only on $T^{\ast }$. Thus, the main
541: question transforms in
542:
543:
544:
545: \subsection{A map in the general position}
546:
547: \noindent We evaluate the class $\tau _{\mathbb{Q}_{4n}}(\ast )$ by the so
548: called \textquotedblright map in the general position\textquotedblright\
549: standard procedure, \cite{guide2}. Since the set $[S_{(2)}^{3},T^{\ast }]_{%
550: \mathbb{Q}_{4n}}$ is a point, it suffices to calculate an obstruction
551: cocycle for the particular map $h$.
552:
553: \noindent Let $h:S^{3}\rightarrow W_{n}$ be an arbitrary $\mathbb{Q}_{4n}$%
554: -simplicial map which is in the general position. What we mean by the
555: general position is that for any simplex $\sigma $ in $S^{3}$
556: \begin{equation*}
557: h(\sigma )\cap U^{\ast }\neq \emptyset \text{ }\Rightarrow \mathrm{dim}%
558: (h(\sigma ))=3\text{ and }h(\sigma )\cap U^{\ast }=\{p_{1},..,p_{k}\}\subset
559: \mathrm{int}h(\sigma ).
560: \end{equation*}%
561: Now let $h:S^{3}\rightarrow W_{n}$ be a $\mathbb{Q}_{4n}$-map in the general
562: position. Then the associated cohomology class of the obstruction cocycle $%
563: c_{\mathbb{Q}_{4n}}(h)\in C_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };%
564: \mathbb{Z}))=\mathrm{Hom}_{\mathbb{Q}_{4n}}(C_{3}(S^{3}),H_{2}(T^{\ast };%
565: \mathbb{Z}))$ for the map $h$ is equal to $\tau (\ast )$, i.e. $[c_{\mathbb{Q%
566: }_{4n}}(h)]=\tau (\ast )$. In order to calculate it let us describe the
567: cocycle $c_{\mathbb{Q}_{4n}}(h)$ a little bit closer. Let $\sigma $ be an
568: oriented $3$-simplex in $S^{3}$. Then $c_{\mathbb{Q}_{4n}}(h)(\sigma )\in
569: H_{2}(T^{\ast };\mathbb{Z})$ is the $h_{\ast }$ image of the fundamental
570: class of $\partial (\sigma )\cong S^{2}$ by the map $h_{\ast
571: }:H_{2}(\partial (\sigma );\mathbb{Z})\rightarrow H_{2}(T^{\ast };\mathbb{Z}%
572: ) $, i.e.
573: \begin{equation*}
574: c_{\mathbb{Q}_{4n}}(h)(\sigma )=h_{\ast }[\partial (\sigma )].
575: \end{equation*}
576:
577: \subsection{The nature of the obstruction cocycle}
578:
579: \noindent The following proposition will narrow our attention to
580: the torsion part of the group \\
581: $H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$ and
582: will be the perfect control factor in our calculations. It can be
583: also found in \cite{Bl-Vr-Ziv}.
584:
585: \begin{proposition}
586: \label{prop:ObCoCycleTorzioni} The cohomology class of the obstruction
587: cocycle $c_{\mathbb{Q}_{4n}}(h)$ is a torsion element of the group $H_{%
588: \mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$.
589: \end{proposition}
590:
591: \begin{proof}
592: Let $H$ be a subgroup of $\mathbb{Q}_{4n}$. The restriction map
593:
594: \begin{equation*}
595: r:H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))\rightarrow
596: H_{H}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))
597: \end{equation*}%
598: on the cochain level sends $c\in C_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast
599: };\mathbb{Z}))$ a $\mathbb{Q}_{4n}$-cochain to now $H$-cochain $c\in
600: C_{H}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$. The definition of the
601: obstruction cocycle implies that $r(c_{\mathbb{Q}_{4n}}(h))$ is the
602: obstruction cocycle for the extension of the $H$-map in the general position
603: $h$. It is a known fact (\cite{Brown} Section III.9. Proposition 9.5.(ii))
604: that the composition of the restriction with the transfer $\tau
605: :H_{H}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))\rightarrow H_{\mathbb{Q}%
606: _{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$ is just a multiplication by
607: the index $[\mathbb{Q}_{4n}:H]$,
608: \begin{equation*}
609: \begin{array}{ccccc}
610: H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z})) & \rightarrow &
611: H_{H}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z})) & \rightarrow & H_{\mathbb{Q}%
612: _{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z})) \\
613: \lbrack c_{\mathbb{Q}_{4n}}(h)] & \longmapsto & [c_{H}(h)] & \longmapsto & [%
614: \mathbb{Q}_{4n}:H]\cdot \lbrack c_{\mathbb{Q}_{4n}}(h)]\text{.}%
615: \end{array}%
616: \end{equation*}%
617: Particularly, let $H$ be the trivial subgroup of $\mathbb{Q}_{4n}$. The
618: sphere $S^{3}$ is $2$-connected and $H$ is\ a trivial group, so the map $h$
619: can be extended to a $H$-map $S^{3}\rightarrow M$. Thus, $[c_{H}(h)]=0$ and
620: consequently $[\mathbb{Q}_{4n}:H]\cdot \lbrack c_{\mathbb{Q}_{4n}}(h)]=0$ in
621: $H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$.
622: \end{proof}
623:
624: \subsection{The $\mathbb{Q}_{4n}$ cellular structures on $S^{3}$}
625:
626: \noindent In order to start efficient computations of the obstruction
627: cocycle we need to describe the concrete $\mathbb{Q}_{4n}$ $CW$-structures
628: of the sphere $S^{3}$. The proposition \ref{prop:promenaDejstva} allows us
629: to be very picky in selecting the adequate cellular structures. We describe
630: two $\mathbb{Q}_{4n}$ cellular structures the natural one and the most
631: economical one, and the cellular map joining them. The direct consequence of
632: these discussions is the isomorphism
633: \begin{equation*}
634: H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))\cong
635: H_{2}(T^{\ast };\mathbb{Z})_{\mathbb{Q}_{4n}}.
636: \end{equation*}
637:
638: \noindent \textbf{The natural }$\mathbb{Q}_{4n}$ \textbf{cellular-simplicial
639: structure.} This structure comes from the join decomposition $%
640: S^{3}=S^{1}\ast S^{1}$ of the $3$-sphere. Let the sphere $S^{1}$ be
641: represented by the simplicial complex of the regular $2n$-gon $P_{2n}$. Then
642: the sphere $S^{3}$, as the simplicial complex, is the join $P_{2n}^{(1)}\ast
643: P_{2n}^{(2)}$ of two copies of $P_{2n}$. Let the vertex of $P_{2n}^{(1)}$
644: and $P_{2n}^{(2)}$ be denoted by $a_{1},..,a_{2n}$ and $b_{1},..,b_{2n}$,
645: respectively. The action of the group $\mathbb{Q}_{4n}$ on $S^{3}$ is
646: defined on vertices by%
647: \begin{equation*}
648: \epsilon \cdot a_{i}=a_{i\ \textrm{mod}\ 2n+1}\text{, }\epsilon \cdot b_{i}=b_{i%
649: \ \textrm{mod}\ 2n+1}\text{, }j\cdot a_{1}=b_{1}
650: \end{equation*}%
651: and it extends equivariantly to upper skeletons. Then, for example%
652: \begin{equation*}
653: \begin{array}{c}
654: \text{ }j\cdot a_{i}=j\epsilon ^{i-1}\cdot a_{1}=\epsilon
655: ^{(2n-1)(i-1)}j\cdot a_{1}=\epsilon ^{(2n-1)(i-1)}\cdot b_{1}=b_{(2n-i+1)%
656: \ \textrm{mod}\ 2n+1} \\
657: j\cdot \lbrack a_{1},a_{2};b_{1},b_{2}]=[b_{1},b_{2n};a_{n+1},a_{n}].%
658: \end{array}%
659: \end{equation*}%
660: The associated chain complex $\mathfrak{C}=\{\emph{C}_{i}\}$ has the form%
661: \begin{equation*}
662: 0\longrightarrow \mathbb{Z}^{4n^{2}}\longrightarrow \mathbb{Z}%
663: ^{8n^{2}}\longrightarrow \mathbb{Z}^{4n^{2}+4n}\longrightarrow \mathbb{Z}%
664: ^{4n}\longrightarrow 0.
665: \end{equation*}%
666: This $\mathbb{Q}_{4n}$ cellular structure makes the beautiful
667: rectangular middle section of the join representation
668: $S^{3}=S^{1}\ast S^{1}=[0,1]\ast \lbrack 0,1]/\approx $, (with
669: some identifications). The action is indicated in the figure
670: \ref{fig:Fig1}.
671:
672: %\begin{center}
673: %\FRAME{ftbhFU}{3.0519in}{2.5936in}{0pt}{\Qcb{{\protect\small The
674: %model for
675: %two }$\mathbb{Q}_{4n}${\protect\small \ cellular structures of }$S^{3}$%
676: %{\protect\small .}}}{\Qlb{Fig1}}{picture1.wmf}{\special{language "Scientific
677: %Word";type "GRAPHIC";maintain-aspect-ratio TRUE;display "USEDEF";valid_file
678: %"F";width 3.0519in;height 2.5936in;depth 0pt;original-width
679: %6.0473in;original-height 5.1307in;cropleft "0";croptop "1";cropright
680: %"1";cropbottom "0";filename 'Picture1.wmf';file-properties "XNPEU";}}
681: %\end{center}
682:
683: \begin{figure}[htb]
684: \centering
685: \includegraphics[scale=0.70]{picture1.eps}
686: \caption{The model for two $\mathbf{Q}_{4n}$ cellular structures
687: of $S^3$.} \label{fig:Fig1}
688: \end{figure}
689:
690: \noindent \textbf{The economic }$\mathbb{Q}_{4n}$ \textbf{cellular structure.%
691: } The second one comes from the minimal resolution of $\mathbb{Z}$ by free $%
692: \mathbb{Q}_{4n}$-modules described in \cite{CaEi}, p.~253. The associated
693: cellular complex has one $\mathbb{Q}_{4n}$ $0$-cell $a$, two $\mathbb{Q}_{4n}
694: $ $1$-cells $b$ and $b^{\prime }$, two $\mathbb{Q}_{4n}$ $2$-cells $c$ and $%
695: c^{\prime }$, and finally then one $\mathbb{Q}_{4n}$ $3$-cell $e$. The
696: associated chain complex $\mathfrak{D}=\{\emph{D}_{i}\}$ has the form
697: \begin{equation*}
698: 0\rightarrow \mathbb{Z}[\mathbb{Q}_{4n}]e\overset{\partial }{\rightarrow }%
699: \mathbb{Z}[\mathbb{Q}_{4n}]c\oplus \mathbb{Z}[\mathbb{Q}_{4n}]c^{\prime }%
700: \overset{\partial }{\rightarrow }\mathbb{Z}[\mathbb{Q}_{4n}]b\oplus \mathbb{Z%
701: }[\mathbb{Q}_{4n}]b^{\prime }\overset{\partial }{\rightarrow }\mathbb{Z}[%
702: \mathbb{Q}_{4n}]e\rightarrow 0
703: \end{equation*}%
704: where
705: \begin{equation*}
706: \begin{array}{lll}
707: \partial e=(\epsilon -1)c-(\epsilon j-1)c^{\prime } & \partial
708: c=(1+..+\epsilon ^{n-1})b-(j+1)b^{\prime } & \partial c^{\prime }=(\epsilon
709: j+1)b+(\epsilon -1)b^{\prime } \\
710: & \partial b=(\epsilon -1)a & \partial b^{\prime }=(j-1)a%
711: \end{array}%
712: .
713: \end{equation*}%
714: Thus, it will be enough to look at the obstruction cocycle $c_{\mathbb{Q}%
715: _{4n}}(h)$ on the maximal cell $e$, and to prove that its image is or is not
716: zero, when we pass to cohomology.
717:
718: \noindent \textbf{Computing equivariant cohomology.} The explicit formulas
719: for the chain complex $\{\emph{D}_{i}\}$ is lurking us to apply the $\mathrm{%
720: Hom}_{\mathbb{Q}_{4n}}(\cdot ,H_{2}(T^{\ast };\mathbb{Z}))$ functor. The
721: result is the equivariant cochain complex
722: \begin{equation*}
723: 0\longleftarrow H_{2}(T^{\ast };\mathbb{Z})\overset{\Gamma }{\longleftarrow }%
724: H_{2}(T^{\ast };\mathbb{Z})\oplus H_{2}(T^{\ast };\mathbb{Z})\leftarrow
725: H_{2}(T^{\ast };\mathbb{Z})\oplus H_{2}(T^{\ast };\mathbb{Z})\longleftarrow
726: H_{2}(T^{\ast };\mathbb{Z})\longleftarrow 0
727: \end{equation*}%
728: where $\Gamma (p,q)=(\epsilon -1)p-(\epsilon j-1)q$ for $p,q\in
729: H_{2}(T^{\ast };\mathbb{Z})$. The definition of the equivariant cohomology
730: and a standard calculation imply that
731: \begin{equation*}
732: H_{\mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))=H_{2}(T^{\ast };%
733: \mathbb{Z})/\mathrm{Im}\Gamma \cong H_{2}(T^{\ast };\mathbb{Z})_{\mathbb{Q}%
734: _{4n}}
735: \end{equation*}%
736: where $H_{2}(T^{\ast };\mathbb{Z})_{\mathbb{Q}_{4n}}$ denote the group of
737: coinvariants of the $\mathbb{Q}_{4n}$-module $H_{2}(T^{\ast };\mathbb{Z})$,
738: \cite{Brown}.
739:
740: \noindent \textbf{The chain map.} There exist a cellular map $\mathfrak{f}:%
741: \mathfrak{D}\rightarrow \mathfrak{C}$, which can be visualized on the figure %
742: \ref{fig:Fig1} by identifying top dimensional cell $e$ of the
743: economic cell structure with the transparent-shaded fundamental
744: domain.
745:
746: \noindent Since we are interested in $3$-cochains, i.e. elements of $C_{%
747: \mathbb{Q}_{4n}}^{3}(S^{3},H_{2}(T^{\ast };\mathbb{Z}))$, we only need to
748: know the concrete expression for the cellular map on the top dimensional
749: cell $e$,%
750: \begin{equation*}
751: \mathfrak{f}(e)=[a_{1},\epsilon a_{1}]\ast \lbrack b_{1},\epsilon
752: b_{1}]+[\epsilon a_{1},\epsilon ^{2}a_{1}]\ast \lbrack b_{1},\epsilon
753: b_{1}]+...+[\epsilon ^{n-1}a_{1},\epsilon ^{n}a_{1}]\ast \lbrack
754: b_{1},\epsilon b_{1}]\text{,}
755: \end{equation*}%
756: where the simplexes on the right hand side are appropriately oriented.
757:
758: \subsection{\label{sec:pointClasses}Point classes}
759:
760: \noindent It is of utmost importance for computation of the obstruction
761: cocycle $c(h)_{\mathbb{Q}_{4n}}$ to pinpoint some elements from the
762: coefficient module $H_{2}(T^{\ast };\mathbb{Z})$. The similar discussion can
763: be found in \cite{Bl-Vr-Ziv}.
764:
765: \noindent Let $\{W_{1},W_{2},\ldots ,W_{k}\}$ be the family on maximal
766: elements of the arrangement $\mathcal{A}$ of linear (closed half-)subspaces
767: in an $(n+m)$-dimensional, Euclidean space $E$. Let also maximal elements
768: have the constant dimension $\dim W_{i}=n$. Let $\hat{D}(\mathcal{A})=\cup
769: \mathcal{A}\cup \{+\infty \}\subset E\cup \{+\infty \}\cong S^{n+m}$ be the
770: compactified union and $M(\mathcal{A})=E\setminus \cup \mathcal{A}$ the
771: complement of the arrangement.
772:
773: \noindent Let the point $x\in \mathrm{int}(W_{i}\setminus \cup _{j\neq
774: i}W_{j})$ and let $D_{\varepsilon }(x)=x+D_{\varepsilon }$ be a disc around $%
775: x$, where $D_{\varepsilon }$ is the $\varepsilon $-disc in the orthogonal
776: complement $W_{i}^{\perp }$. For a sufficiently small $\varepsilon $ the
777: intersection $D_{\epsilon }(x)\cap (\cup _{j\neq i}W_{j})$ vanishes and we
778: assume fix such an $\varepsilon $. We assume that $D_{\epsilon }(x)$ is
779: oriented by the orientation inherited from the ambient orientation and the
780: orientation prescribed of $W_{i}$.
781:
782: \noindent \textbf{The point class} $[x]\in H_{m}(E,M(\mathcal{A});\mathbb{Z}%
783: ) $ of $x$ is inclusion image of the fundamental class of the pair $%
784: (D_{\epsilon }(x),\partial D_{\epsilon }(x))$. By the Excision axiom
785: described $\varepsilon $ has no effect on the class $[x]$. In addition, by
786: the Homotopy axiom, $[x]$ is uniquely determined by the connected component
787: of $W_{i}\setminus \cup _{j\neq i}W_{j}$. The image $\left\Vert x\right\Vert
788: :=\partial \lbrack x]$ of the point class $[x]$ by the isomorphism $%
789: H_{m}(E,M(\mathcal{A});\mathbb{Z})\rightarrow H_{m-1}(M(\mathcal{A});\mathbb{%
790: Z})$ is also called the \textit{point class }of\emph{\ }$x$ and has all the
791: properties of the original one.
792:
793: %\FRAME{fhFU}{5.5247in}{1.829in}{0pt}{\Qcb{{\protect\small The point classes
794: %and the broken point classes.}}}{\Qlb{Fig2}}{picture2.wmf}{\special{language
795: %"Scientific Word";type "GRAPHIC";maintain-aspect-ratio TRUE;display
796: %"USEDEF";valid_file "F";width 5.5247in;height 1.829in;depth
797: %0pt;original-width 6.1084in;original-height 2.0026in;cropleft "0";croptop
798: %"1";cropright "1";cropbottom "0";filename 'Picture2.wmf';file-properties
799: %"XNPEU";}}
800:
801: \begin{figure}[htb]
802: \centering
803: \includegraphics[scale=1]{picture2.eps}
804: \caption{The point classes and the broken point classes}
805: \label{fig:Fig2}
806: \end{figure}
807:
808:
809: \begin{proposition}
810: \label{prop:PointClass}Let $\mathcal{A}$ be the arrangement of linear
811: subspaces as above. Let $x\in \mathrm{int}(W_{i}\setminus \cup _{j\neq
812: i}W_{j})=W_{i}\setminus \cup _{j\neq i}W_{j}$.
813:
814: (A) The class $[x]\in H_{m}(E,M(\mathcal{A});\mathbb{Z})$ does not vanish.
815:
816: (B) If we assume that $(\forall j)~i\neq j~\Rightarrow ~\mathrm{codim}%
817: _{W_{i}}(W_{i}\cap W_{j})>1$, then the class $[x]$ does not depend on
818: particular $x$.
819:
820: (C) If there is a subspace $W_{j}$ such that $\mathrm{codim}%
821: _{W_{i}}(W_{i}\cap W_{j})=1$ and $x_{1}$, $x_{2}$ belong to different
822: connected components of $W_{i}\setminus \cup _{j\neq i}W_{j}$, then $%
823: [x_{1}]\neq \lbrack x_{2}]$.
824: \end{proposition}
825:
826: \begin{proof}
827: (A) By the Ziegler-\v{Z}ivaljevi\'{c} formula \cite{ZZ}, has he wedge
828: decomposition%
829: \begin{equation*}
830: \hat{D}(\mathcal{A})\simeq \hat{W}_{1}\vee \hat{W}_{2}\vee \ldots \vee \hat{W%
831: }_{k}\vee ...
832: \end{equation*}%
833: where the displayed factors correspond to maximal elements. Let $\Delta (%
834: \hat{W}_{i})\in H^{m}(E,M(\mathcal{A});\mathbb{Z})$ be the Poincar\'{e}%
835: -Alexander dual of the fundamental homology class $[\hat{W}_{i}]\in H_{n}(%
836: \hat{D}(\mathcal{A});\mathbb{Z})$ associated with the sphere $\hat{W}_{i}$
837: in $\hat{E}$. Then $\Delta (\hat{W}_{i})([N])\in \mathbb{Z}$ is the
838: intersection number $[\hat{W}_{i}]\cap \lbrack N]$, whenever this number is
839: correctly defined, for example if $N$ is a manifold and the intersection $%
840: \hat{W}_{i}\cap N$ is transversal. From here we see that $\Delta (\hat{W}%
841: _{i})([x])=1$ and so $[x]$ does not vanish.
842:
843: (B) Since the assumption implies the connectness of the complement $%
844: W_{i}\setminus \cup _{j\neq i}W_{j}$, the statement follows by some homotopy.
845:
846: (C) When $\mathrm{codim}_{W_{i}}(W_{i}\cap W_{j})=1$ the subspaces $W_{i}$
847: and $W_{j}$ are decomposed by the hyperplane $W_{i}\cap W_{j}$ into unions
848: of closed half-spaces, $W_{i}=W_{i}^{1}\cup W_{i}^{2}$ and $%
849: W_{j}=W_{j}^{1}\cup W_{j}^{2}$, respectively. Let us assume that $x_{1}\in
850: W_{i}^{1}$ and $x_{2}\in W_{i}^{2}$. The half-space $W_{i}^{1}$ and $%
851: W_{j}^{1}$ glued along the common boundary and compactified determine a
852: sphere $S\subset \hat{E}$. Since the Ziegler-\v{Z}ivaljevi\'{c}
853: decomposition \cite{ZZ} involves a choice of generic points, the points can
854: be chosen in such a way that the sphere $S$ appears in the decomposition.
855: Precisely, the sphere $S$ is the factor $\widehat{W_{i}\cap W_{j}}\ast
856: \Delta (P_{<W_{i}\cap W_{j}})\cong S^{n-1}\ast S^{0}\cong S^{n}$, where $P$
857: denote the intersection poset of the arrangement $\mathcal{A}$. thus, the
858: fundamental class $[S]$ is nontrivial in $H_{n}(\hat{D}(\mathcal{A});\mathbb{%
859: Z})$. Like in the previous case, let $\Delta (S)\in H^{m}(E,M(\mathcal{A});%
860: \mathbb{Z})$ be the Poincar\'{e}-Alexander dual to $[S]$. Then%
861: \begin{equation*}
862: \Delta (S)([x_{1}])=\pm 1\text{ and }\Delta (S)([x_{2}])=0
863: \end{equation*}%
864: and consequently $[x_{1}]\neq \lbrack x_{2}]$.
865: \end{proof}
866:
867: \noindent \textbf{The broken point class.} In order to simplify the
868: exposition let $\mathcal{A}$ be a very special arrangement consisting of two
869: vector spaces $W_{1}$ and $W_{2}$ of the same dimension $m$ inside the
870: vector space $E=\mathbb{R}^{n+m}$. Let $U=W_{1}\cap W_{2}$ be the
871: intersection and $k=\dim (W_{1}\cap W_{2})$. Let us also fix $V$, a vector
872: space of dimension $n$ such that $E=W_{1}\oplus V=W_{2}\oplus V$, where $%
873: \oplus $ denotes the direct sum of vector spaces. For $x\in W_{1}\cap W_{2}$
874: let $D_{\varepsilon }(x)$ be a disc $x+D_{\varepsilon }$, where $%
875: D_{\varepsilon }$ is the $\varepsilon $-disc in $V$. Then the \textit{broken
876: point class} of $x$, denoted by $[x]\in H_{n}(E,E\backslash W_{1}\cup W_{2};%
877: \mathbb{Z})$, is the fundamental class of the pair $(D_{\epsilon
878: }(x),\partial D_{\epsilon }(x))$. In addition the image $\left\Vert
879: x\right\Vert :=\partial \lbrack x]$ of the broken point class $[x]$ by the
880: isomorphism $H_{n}(E,E\backslash W_{1}\cup W_{2};\mathbb{Z})\rightarrow
881: H_{n-1}(E\backslash W_{1}\cup W_{2};\mathbb{Z})$ is also called the broken
882: point class\textit{\ }of\emph{\ }$x$ and has all the properties of the
883: original one.
884:
885: \noindent An illustration for the case $m=1$ and $n=2$ can be seen
886: in the figure \ref{fig:Fig2}. The same definition stands even if
887: $W_{1}$ and $W_{2}$ are closed half spaces intersecting over
888: linear space $U$. The broken point classes can be naturally
889: expressed as linear combinations of the ordinary point classes.
890: The third picture in the figure \ref{fig:Fig2} illustrates this
891: situation.
892:
893: %\FRAME{fhFU}{4.9142in}{2.5569in}{0pt}{\Qcb{Two posible point classes in the
894: %case $\dim (W_{1}\cap W_{2})=m-1$}}{\Qlb{Fig5}}{picture5.wmf}{\special%
895: %{language "Scientific Word";type "GRAPHIC";maintain-aspect-ratio
896: %TRUE;display "USEDEF";valid_file "F";width 4.9142in;height 2.5569in;depth
897: %0pt;original-width 6.1084in;original-height 3.0282in;cropleft "0";croptop
898: %"1";cropright "1";cropbottom "0";filename 'Picture5.wmf';file-properties
899: %"XNPEU";}}
900:
901: \begin{figure}[htb]
902: \centering
903: \includegraphics[scale=1.3]{picture5.eps}
904: \caption{Two posible point classes in the case $dim(W1 \cap
905: W2)=m-1$} \label{fig:Fig5}
906: \end{figure}
907:
908: \noindent Let us now suppose that $k=\dim (W_{1}\cap W_{2})=\dim
909: (W_{1})-1=m-1$. Then there are two different broken point classes
910: associated to every point $x\in W_{1}\cap W_{2}$, i.e. the broken
911: point class $[x]$ depends on the choice of the vector space $V$.
912: Let us omit the format proof of this fact and illustrate this
913: situation by the figure \ref{fig:Fig5}. The
914: formal proof would go along the lines of the proposition \ref%
915: {prop:PointClass}. The main property that differs these two classes,
916: brusquely explained, is that if we substitute each sphere with the disc and
917: move these discs a little bit:
918:
919: (A) the first disc will intersect simultaneously $W_{1}^{+}$ and $W_{2}^{+}$%
920: , or $W_{1}^{-}$ and $W_{2}^{-}$; and
921:
922: (B) the second disc will intersect simultaneously $W_{1}^{+}$ and $W_{2}^{-}$%
923: , or $W_{1}^{-}$ and $W_{2}^{+}$.
924:
925: \subsection{Calculating $H_{2}(T^{\ast };\mathbb{Z})$ and $H_{2}(T^{\ast };%
926: \mathbb{Z})_{\mathbb{Q}_{4n}}$}
927:
928: \noindent \textbf{The homology. }In order to analyze the second homology of
929: the complement $T^{\ast }$ we use the following equality%
930: \begin{equation*}
931: W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha )=S^{n-1}\backslash \cup
932: \widehat{\mathcal{A}}(\mathcal{J},\alpha )
933: \end{equation*}%
934: where $\widehat{\mathcal{A}}(\mathcal{J},\alpha )$ denotes the one point
935: compactification of the arrangement $\mathcal{A}(\mathcal{J},\alpha )$. This
936: allow us to use the Poincar\'{e}-Alexander duality and to work with the
937: arrangement $\widehat{\mathcal{A}}(\mathcal{J},\alpha )$ instead of its
938: complement. The Poincar\'{e}-Alexander duality and the Universal Coefficient
939: isomorphisms give us the sequence of isomorphisms (assuming $\mathbb{Z}$
940: coefficients)
941: \begin{eqnarray}
942: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha ))
943: &=&H_{2}(S^{n-1}\backslash \cup \widehat{\mathcal{A}}(\mathcal{J},\alpha
944: ))\cong H^{(n-1)-2-1}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha ))
945: \label{Poincare-Alexander 1} \\
946: &\cong &\mathrm{Hom}(H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha
947: )),\mathbb{Z})\oplus \mathrm{Ext}(H_{n-3}(\cup \widehat{\mathcal{A}}(%
948: \mathcal{J},\alpha )),\mathbb{Z}). \notag
949: \end{eqnarray}%
950: Since the maximal elements of the arrangement $\mathcal{A}(\mathcal{J}%
951: ,\alpha )$ are $(n-4)$-dimensional linear (half-) subspaces, respectively,
952: the $\mathrm{Ext}$ factor in the above equality vanish. Precisely, $%
953: H_{n-3}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})=0$.
954: Therefore,%
955: \begin{equation}
956: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})\cong
957: \mathrm{Hom}(H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{%
958: Z}),\mathbb{Z}). \label{Isomorphisms Complement - Arrangement}
959: \end{equation}%
960: The Ziegler-\v{Z}ivaljevi\'{c} formula implies the following homology
961: decomposition (assuming $\mathbb{Z}$ coefficients)
962: \begin{equation}
963: H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})\cong
964: \underset{d=0}{\overset{n-4}{\bigoplus }}\underset{V\in P(\alpha ):\dim V=d}%
965: {\bigoplus }H_{n-4}(\Delta (P_{<V})\ast \hat{V}) \label{Z-Z
966: homology decomposition - 1}
967: \end{equation}%
968: where $P(\alpha )$ is the intersection poset of the arrangement $\mathcal{A}%
969: (J,\alpha )$ and $\hat{V}$ one-point compactification of the element $V$.
970: Thus, the property of the hom functor imply%
971: \begin{equation}
972: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha
973: );\mathbb{Z})\cong \underset{d=0}{\overset{n-4}{\bigoplus
974: }}\mathrm{Hom}\left( \underset{V\in
975: P(\alpha ):\dim V=d}{\bigoplus }H_{n-4}(\Delta (P_{<V})\ast \hat{V});%
976: \mathbb{Z}\right) \label{hom decomposition - 1}
977: \end{equation}%
978: \noindent \textbf{Poincar\'{e} dual of the (broken) point class.} Let us
979: illustrate the isomorphism (\ref{Isomorphisms Complement - Arrangement}) on
980: the point classes for the general arrangement of linear (half-) spaces $%
981: \mathcal{A}$ in $\mathbb{R}^{n+m}$. Let maximal elements $%
982: \{W_{1},W_{2},\ldots ,W_{k}\}$ have the constant dimension $\dim W_{i}=n$.
983: Let $x\in W_{1}\backslash \cup _{i\neq 1}W_{i}$ and $S=\partial
984: D_{\varepsilon }(x)$, where $D_{\varepsilon }(x)=x+D_{\varepsilon }$ and $%
985: D_{\varepsilon }$ is a small disk in $W_{1}{}^{\bot }$. Then the isomorphism
986: , $\vartheta :H_{m-1}(\mathbb{R}^{n+m}\backslash \cup \mathcal{A};\mathbb{Z}%
987: )\rightarrow \mathrm{Hom}(H_{n}(\cup \widehat{\mathcal{A}};\mathbb{Z}),%
988: \mathbb{Z})$ can be expressed for $t\in H_{n}(\cup \widehat{\mathcal{A}};%
989: \mathbb{Z});\mathbb{Z})$ by%
990: \begin{equation}
991: \vartheta (\left\Vert x\right\Vert ):H_{n}(\cup \widehat{\mathcal{A}};%
992: \mathbb{Z})\rightarrow \mathbb{Z}\text{, }\vartheta (\left\Vert x\right\Vert
993: )(t)=\mathrm{link}(S,T) \label{link-isomorphism}
994: \end{equation}%
995: if $T$ is a submanifold in $\cup \widehat{\mathcal{A}}$ representing the
996: homology class $t$ and linking number $\mathrm{link}(S,T)$ is correctly
997: defined.
998:
999: \begin{example}
1000: Let the arrangement $\mathcal{A}$ be given by the figure
1001: \ref{fig:Fig4},(A). Then the Hasse diagram of the arrangement
1002: $\mathcal{A}$ is like the in the figure \ref{fig:Fig4},(B).
1003: %\FRAME{ftbhFU}{4.2615in}{1.966in}{0pt}{\Qcb{%
1004: %{\protect\small The arrangement} $\mathcal{A}$ {\protect\small with
1005: %indicated generators in homology.}}}{\Qlb{Fig4}}{picture4.wmf}{\special%
1006: %{language "Scientific Word";type "GRAPHIC";maintain-aspect-ratio
1007: %TRUE;display "USEDEF";valid_file "F";width 4.2615in;height 1.966in;depth
1008: %0pt;original-width 6.0473in;original-height 2.7679in;cropleft "0";croptop
1009: %"1";cropright "1";cropbottom "0";filename 'Picture4.wmf';file-properties
1010: %"XNPEU";}}
1011:
1012: \begin{figure}[htb]
1013: \centering
1014: \includegraphics[scale=1]{picture4.eps}
1015: \caption{The arrangement $\mathcal{A}$ with indicated generators
1016: in homology.} \label{fig:Fig4}
1017: \end{figure}
1018:
1019: Then $H_{1}(\cup \widehat{\mathcal{A}};\mathbb{Z})\cong \mathbb{%
1020: Z\oplus Z\oplus Z}$, where the generators / spheres $T$, $T_{1}$ and $T_{2}$
1021: indicated in the figure. Then we can read of isomorphism $\vartheta :H_{1}(%
1022: \mathbb{R}^{3}\backslash \cup \mathcal{A};\mathbb{Z})\rightarrow \mathrm{Hom}%
1023: (H_{1}(\cup \widehat{\mathcal{A}};\mathbb{Z}),\mathbb{Z})$ for point classes
1024: $\left\Vert x_{1}\right\Vert $ and $\left\Vert x_{2}\right\Vert $ from the
1025: picture,%
1026: \begin{eqnarray*}
1027: \vartheta (\left\Vert x_{1}\right\Vert )(t_{1}) &=&1,\vartheta (\left\Vert
1028: x_{1}\right\Vert )(t_{2})=0,\vartheta (\left\Vert x_{1}\right\Vert )(t)=-1%
1029: \text{,} \\
1030: \vartheta (\left\Vert x_{2}\right\Vert )(t_{1}) &=&0,\vartheta (\left\Vert
1031: x_{2}\right\Vert )(t_{2})=1,\vartheta (\left\Vert x_{2}\right\Vert )(t)=0%
1032: \text{.}
1033: \end{eqnarray*}
1034: \end{example}
1035:
1036: \noindent \textbf{The coinvariants. }When time comes to compute coinvariants
1037: the best news would be that the Poincar\'{e}-Alexander duality map and the
1038: Universal coefficient isomorphism are equivariant maps. Then the
1039: isomorphisms (\ref{Isomorphisms Complement - Arrangement}) would be
1040: isomorphisms of $\mathbb{Q}_{4n}$-modules, and there would be no difference
1041: in what module we work. (Un)fortunately, the Poincar\'{e}-Alexander duality
1042: map is not a $\mathbb{Q}_{4n}$-map, but a $\mathbb{Q}_{4n}$-map up to a
1043: orientation character, while Universal coefficient isomorphism is an $%
1044: \mathbb{Q}_{4n}$-map. If $o$ is an orientation of the $\mathbb{Q}_{4n}$%
1045: -sphere $S^{n-1}$ or $S^{2(n-1)}$, then $o$ determines Poincar\'{e}%
1046: -Alexander duality map $\gamma _{o}$ \cite{Mu}. In particular, $g\cdot
1047: o=\det (g)\cdot o$, where $g\in \mathbb{Q}_{4n}\subseteq GL_{n}(\mathbb{R})$%
1048: , or $g\in \mathbb{Q}_{4n}\subseteq GL_{2n}(\mathbb{R})$.
1049:
1050: \noindent The natural $\mathbb{Q}_{4n}$-action on the union $\cup \widehat{%
1051: \mathcal{A}}(\mathcal{J},\alpha )$, inherited from the ambient $\mathbb{Q}%
1052: _{4n}$-action, respects the dimensional decomposition (\ref{Z-Z homology
1053: decomposition - 1}) of the $(n-4)$-homology. When these homologies are free,
1054: then the decomposition (\ref{hom decomposition - 1}) becomes true even
1055: without $\mathrm{Hom}$. Let us assume that $(n-4)$-homology of the
1056: arrangement $\widehat{\mathcal{A}}(\mathcal{J},\alpha )$ is free. In some
1057: calculation we will see that we actually do not need the whole group to be
1058: free, but just some factors in the decomposition (\ref{hom decomposition - 1}%
1059: ). Therefore, if we would rather work with $\mathbb{Q}_{4n}$-module $%
1060: H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})$ instead
1061: of $\mathbb{Q}_{4n}$-module $H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{%
1062: J},\alpha );\mathbb{Z})$ we have to modify the $\mathbb{Q}_{4n}$-action.
1063: Specifically, let $l\in H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J}%
1064: ,\alpha );\mathbb{Z})$ and $g\in \mathbb{Q}_{4n}$, then
1065: \begin{equation*}
1066: g\ast l=\det (g)~g\cdot l
1067: \end{equation*}%
1068: where $\ast $ is the new modified action, and $\cdot $ the old one. Let $%
1069: \sim $ denote the congruence relation on $H_{n-4}(\cup \widehat{\mathcal{A}}(%
1070: \mathcal{J},\alpha ),\mathbb{Z})$ which class of zero is a subgroup
1071: generated by the elements of the form $g\ast x-x$, $g\in \mathbb{Q}_{4n}$, $%
1072: x\in H_{n-4}(\widehat{\mathcal{A}}(\mathcal{J},\alpha ),\mathbb{Z})$. Then
1073: there is an isomorphism%
1074: \begin{equation}
1075: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})_{%
1076: \mathbb{Q}_{4n}}\cong H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha
1077: ),\mathbb{Z})/\sim \label{isomorphism of the modified action}
1078: \end{equation}
1079:
1080: \noindent \textbf{The torsion of the coinvariants.} The proposition \ref%
1081: {prop:ObCoCycleTorzioni} directs us to search for the obstruction cohomology
1082: class in the torsion part of the coinvariants. Therefore, if the
1083: coinvariants are free, then the obstruction is zero and the map exists. Thus
1084: everythig we have done is in vain. It is of utmost importance to get a
1085: feeling, when some $\mathbb{G}$ arrangement can produce nontrivial torsion
1086: group in appropriate coinvariant group.
1087:
1088: \noindent Let us discuss a few simple examples, which met all the above
1089: assumptions. These examples come from the fan partition problems, and we
1090: work with cyclic groups in order to simplify computations.
1091:
1092: \begin{example}
1093: \label{Ex:Coinvariants1}Let $\mathcal{A}$ be the minimal $\mathbb{Z}_{8}$
1094: arrangement in $\mathbb{R}^{8}$ containing subspace
1095: \begin{equation*}
1096: L=\{\mathbf{x}\in \mathbb{R}%
1097: ^{4}~|~x_{1}+x_{2}=x_{3}+x_{4}=x_{5}+x_{6}=x_{7}+x_{8}=0\}\subset W_{8}.
1098: \end{equation*}%
1099: The group $\mathbb{Z}_{8}=\langle \varepsilon \rangle $ acts by a cyclic
1100: permutation, i.e. $\varepsilon \cdot (x_{1},..,x_{8})=(x_{8},x_{1},..,x_{7})$%
1101: . Then $\det (\varepsilon )=(-1)^{8+1}$.
1102:
1103: \noindent It is not hard to see that $\mathcal{A}=\{L,\varepsilon L,L\cap
1104: \varepsilon L\}$, $L\cap \varepsilon L=\{0\}$, and consequently the by Z-Z
1105: decomposition (\ref{Z-Z homology decomposition - 1}),
1106: \begin{equation*}
1107: H_{4}(\cup \widehat{\mathcal{A}};\mathbb{Z})\cong \underset{d=0}{\overset{4}{%
1108: \bigoplus }}\underset{V\in P:\dim V=d}{\bigoplus }H_{4}(\Delta
1109: (P_{<V})\ast \hat{V};\mathbb{Z})\cong 0\oplus 0\oplus 0\oplus (\mathbb{Z}%
1110: \oplus \mathbb{Z})
1111: \end{equation*}%
1112: where $P$ is the intersection poset of the arrangement $\mathcal{A}$. Since,
1113: homology $H_{4}(\cup \widehat{\mathcal{A}};\mathbb{Z})$ has no torsion we
1114: have that $H_{2}(W_{8}\backslash \cup \mathcal{A};\mathbb{Z})\cong \mathbb{Z}%
1115: \oplus \mathbb{Z}$.
1116:
1117: \noindent Let $l$ and $\varepsilon \cdot l$ be the generators of $H_{4}(\cup
1118: \widehat{\mathcal{A}};\mathbb{Z})$ corresponding to spheres $\widehat{L}$
1119: and $\widehat{\varepsilon L}$. The the equality $L=\varepsilon ^{2}L$ in $%
1120: \mathbb{R}^{8}$ imply the equality $l=\epsilon (\varepsilon ^{2}\cdot l)$ in
1121: $H_{4}(\cup \widehat{\mathcal{A}};\mathbb{Z})$, where $\epsilon \in \{1,-1\}$%
1122: . The sign $\epsilon $ depends on the nature of $\varepsilon ^{2}$. If $%
1123: \varepsilon ^{2}$ changes the orientation of $L$, then $\epsilon =-1$,
1124: otherwise $\epsilon =1$. To calculate the sign $\epsilon $ we use a $%
1125: \varepsilon ^{2}$-invariant decomposition $\mathbb{R}^{8}=L\oplus L^{\bot }$%
1126: , where $L^{\bot }$ denote the orthogonal complement of $L$. Then,
1127: \begin{equation*}
1128: \det_{\mathbb{R}^{8}}\varepsilon ^{2}=\det_{L}\varepsilon ^{2}\cdot
1129: \det_{L^{\bot }}\varepsilon ^{2}~\Rightarrow ~\det_{L}\varepsilon ^{2}=\det_{%
1130: \mathbb{R}^{8}}\varepsilon ^{2}\det_{L^{\bot }}\varepsilon ^{2}=(-1)^{2\cdot
1131: (8+1)}\det_{L^{\bot }}\varepsilon ^{2}\text{.}
1132: \end{equation*}%
1133: Let $e_{1},..,e_{8}$ be the standard base of $\mathbb{R}^{8}$. Then one base
1134: for $L^{\bot }$ is%
1135: \begin{equation*}
1136: f_{1}=e_{1}+e_{2}\text{, }f_{2}=e_{3}+e_{4}\text{, }f_{3}=e_{5}+e_{6}\text{,
1137: }f_{4}=e_{1}+e_{2}+..+e_{7}+e_{8}\text{, }
1138: \end{equation*}%
1139: and $\varepsilon ^{2}$ acts on it by%
1140: \begin{equation*}
1141: \varepsilon ^{2}\cdot f_{1}=f_{2}\text{, }\varepsilon ^{2}\cdot f_{2}=f_{3}%
1142: \text{, }\varepsilon ^{2}\cdot f_{3}=f_{4}-f_{1}-f_{2}-f_{3}\text{, }%
1143: \varepsilon ^{2}\cdot f_{4}=f_{4}\text{.}
1144: \end{equation*}%
1145: Therefore,%
1146: \begin{equation*}
1147: \det_{L^{\bot }}\varepsilon ^{2}=\det \left[
1148: \begin{array}{cccc}
1149: {\small 0} & {\small 1} & {\small 0} & {\small 0} \\
1150: {\small 0} & {\small 0} & {\small 1} & {\small 0} \\
1151: {\small -1} & {\small -1} & {\small -1} & {\small 1} \\
1152: 0 & 0 & 0 & {\small 1}%
1153: \end{array}%
1154: \right] =-1~\Rightarrow \epsilon =\det_{L}\varepsilon ^{2}=-\det_{\mathbb{R}%
1155: ^{8}}\varepsilon ^{2}=-1~\Rightarrow l=-(\varepsilon ^{2}\cdot l)
1156: \end{equation*}%
1157: Now, we calculate coinvariants $H_{2}(W_{n}\backslash \cup \mathcal{A};%
1158: \mathbb{Z})_{\mathbb{Z}_{8}}$ by using the modified action $g\ast l=\det
1159: (g)~g\cdot l$, $g\in \mathbb{Z}_{8}$ on $\mathbb{Z}_{8}$-module $H_{4}(\cup
1160: \widehat{\mathcal{A}};\mathbb{Z})$. Let us observe that
1161: \begin{equation*}
1162: l\sim \varepsilon \ast l=\det (\varepsilon )\varepsilon \cdot l=-\varepsilon
1163: \cdot l\text{ and }l\sim \varepsilon ^{2}\ast l=\det (\varepsilon
1164: ^{2})\varepsilon ^{2}\cdot l=-\det (\varepsilon ^{2})\det (\varepsilon
1165: ^{2})l=-l\text{.}
1166: \end{equation*}%
1167: Thus, from this relations we can conclude that $H_{2}(W_{n}\backslash \cup
1168: \mathcal{A};\mathbb{Z})_{\mathbb{Z}_{8}}\cong \mathbb{Z}_{2}$.
1169: \end{example}
1170:
1171: \begin{example}
1172: \label{Ex:Coinvariants2}Let $\mathcal{B}$ be the minimal $\mathbb{Z}_{4}$
1173: arrangement in $\mathbb{R}^{8}=\mathbb{R}^{4}\oplus \mathbb{R}^{4}$
1174: containing subspace
1175: \begin{equation*}
1176: L=\{\mathbf{x}\in \mathbb{R}%
1177: ^{8}~|~x_{1}=x_{5}=x_{1}+..+x_{4}=x_{5}+..+x_{8}=x_{3}+x_{7}=0\}\subset
1178: W_{4}\oplus W_{4}.
1179: \end{equation*}%
1180: Let the action of $\mathbb{Z}_{4}=\langle \varepsilon \rangle $ be defined
1181: by $\varepsilon \cdot
1182: (x_{1},..,x_{8})=(x_{4},x_{1},x_{2},x_{3};x_{8},x_{5},x_{6},x_{7})$. Then $%
1183: \det (\varepsilon )=(-1)^{4+1}(-1)^{4+1}=1$. The arrangement $\mathcal{A}$
1184: has $4$ maximal elements $L$, $\varepsilon L$, $\varepsilon ^{2}L$, $%
1185: \varepsilon ^{3}L$ and the Hasse diagram of the intersection poset
1186: $P$ is like in the figure \ref{fig:Fig3}.
1187:
1188: %\FRAME{ftbhFU}{1.8423in}{1.3599in}{0pt}{\Qcb{{\protect\small The Hasse
1189: %diagram of }$\mathcal{B}$.}}{\Qlb{Fig3}}{picture3.wmf}{\special{language
1190: %"Scientific Word";type "GRAPHIC";maintain-aspect-ratio TRUE;display
1191: %"USEDEF";valid_file "F";width 1.8423in;height 1.3599in;depth
1192: %0pt;original-width 6.0473in;original-height 4.4417in;cropleft "0";croptop
1193: %"1";cropright "1";cropbottom "0";filename 'Picture3.wmf';file-properties
1194: %"XNPEU";}}From the Hasse diagram and Z-Z decomposition (\ref{Z-Z homology
1195: %decomposition - 1}) follows that%
1196:
1197: \begin{figure}[htb]
1198: \centering
1199: \includegraphics[scale=0.6]{picture3.eps}
1200: \caption{The Hasse diagram of $\mathcal{B}$.} \label{fig:Fig3}
1201: \end{figure}
1202:
1203:
1204: \begin{equation*}
1205: H_{3}(\cup \widehat{\mathcal{B}};\mathbb{Z})\cong \underset{d=0}{\overset{3}{%
1206: \bigoplus }}\underset{V\in Q(\alpha ):\dim V=d}{\bigoplus
1207: }H_{2n-5}(\Delta
1208: (P_{<V})\ast \hat{V})\cong 0\oplus (\mathbb{Z}\oplus \mathbb{Z})\oplus (%
1209: \mathbb{Z}\oplus \mathbb{Z\oplus Z}\oplus \mathbb{Z}).
1210: \end{equation*}%
1211: Like in the previous example the homology $H_{3}(\cup \widehat{\mathcal{B}};%
1212: \mathbb{Z})$ has no torsion and so $H_{2}(W_{4}\oplus W_{4}\backslash \cup
1213: \mathcal{B};\mathbb{Z})\cong H_{3}(\cup \widehat{\mathcal{B}};\mathbb{Z})$.
1214:
1215: \noindent Let $l$, $\varepsilon \cdot l$, $\varepsilon ^{2}\cdot l$, $%
1216: \varepsilon ^{3}\cdot l$ and $k$, $\varepsilon \cdot k$ be the generators of
1217: $H_{3}(\cup \widehat{\mathcal{B}};\mathbb{Z})$ corresponding to spheres $%
1218: \widehat{L}$, $\widehat{\varepsilon L}$, $\widehat{\varepsilon ^{2}L}$, $%
1219: \widehat{\varepsilon ^{3}L}$ and $\widehat{L\cap \varepsilon ^{2}L}\ast S^{0}
1220: $, $\widehat{\varepsilon L\cap \varepsilon ^{3}L}\ast S^{0}$ respectively.
1221: The equality $L\cap \varepsilon ^{2}L=\varepsilon ^{2}(L\cap \varepsilon
1222: ^{2}L)$ in $\mathbb{R}^{8}$ implie the equality $k=\epsilon (\varepsilon
1223: ^{2}k)$, where $\epsilon \in \{1,-1\}$. The sign $\epsilon $ depends on the
1224: nature of $\varepsilon ^{2}$. If $\varepsilon ^{2}$ changes the orientation
1225: of the sphere $\widehat{L\cap \varepsilon ^{2}L}\ast S^{0}$, then $\epsilon
1226: =-1$, otherwise $\epsilon =1$. Again, we use the same decomposition $\mathbb{%
1227: R}^{8}=(L\cap \varepsilon ^{2}L)\oplus (L\cap \varepsilon ^{2}L)^{\bot }$
1228: and calculate%
1229: \begin{equation*}
1230: \det_{\mathbb{R}^{8}}\varepsilon ^{2}=\det_{(L\cap \varepsilon
1231: ^{2}L)}\varepsilon ^{2}\cdot \det_{(L\cap \varepsilon ^{2}L)^{\bot
1232: }}\varepsilon ^{2}~\Rightarrow ~\det_{(L\cap \varepsilon ^{2}L)}\varepsilon
1233: ^{2}=\det_{\mathbb{R}^{8}}\varepsilon ^{2}\det_{(L\cap \varepsilon
1234: ^{2}L)^{\bot }}\varepsilon ^{2}=\det_{(L\cap \varepsilon ^{2}L)^{\bot
1235: }}\varepsilon ^{2}\text{.}
1236: \end{equation*}%
1237: If $e_{1},..,e_{8}$ is the standard base of $\mathbb{R}^{8}$, then one base
1238: for $(L\cap \varepsilon ^{2}L)^{\bot }$ is%
1239: \begin{eqnarray*}
1240: f_{1} &=&e_{1}\text{, }f_{2}=e_{1}+e_{2}+e_{3}+e_{4}\text{, }f_{3}=e_{5}%
1241: \text{, } \\
1242: f_{4} &=&e_{5}+e_{6}+e_{7}+e_{8}\text{, }f_{5}=e_{3}\text{, }f_{6}=e_{7}%
1243: \text{, }
1244: \end{eqnarray*}%
1245: and $\varepsilon ^{2}$ acts on it by%
1246: \begin{equation*}
1247: \varepsilon ^{2}\cdot f_{1}=f_{5}\text{, }\varepsilon ^{2}\cdot f_{2}=f_{2}%
1248: \text{, }\varepsilon ^{2}\cdot f_{3}=f_{6}\text{, }\varepsilon ^{2}\cdot
1249: f_{4}=f_{4}\text{, }\varepsilon ^{2}\cdot f_{5}=f_{1}\text{, }\varepsilon
1250: ^{2}\cdot f_{6}=f_{3}.
1251: \end{equation*}%
1252: Since the sign of the permutation $\left(
1253: \begin{array}{llllll}
1254: {\small 1} & {\small 2} & {\small 3} & {\small 4} & {\small 5} & {\small 6}
1255: \\
1256: {\small 5} & {\small 2} & {\small 6} & {\small 4} & {\small 1} & {\small 3}%
1257: \end{array}%
1258: \right) $ is one then $\det_{(L\cap \varepsilon ^{2}L)^{\bot }}\varepsilon
1259: ^{2}=1$.
1260:
1261: \noindent In order to verify that the element $\varepsilon ^{2}$ does not
1262: change the orientation of the sphere $\widehat{L\cap \varepsilon ^{2}L}\ast
1263: S^{0}$, we analyze the action of $\varepsilon ^{2}$ on the homology $H_{3}(%
1264: \widehat{L\cap \varepsilon ^{2}L}\ast S^{0};\mathbb{Z})$. Since $\widehat{%
1265: L\cap \varepsilon ^{2}L}$ is a $2$-sphere and, we saw, $\varepsilon ^{2}$
1266: acts trivially on it, the isomorphism%
1267: \begin{equation*}
1268: H_{3}(\widehat{L\cap \varepsilon ^{2}L}\ast S^{0};\mathbb{Z})\cong \tilde{H}%
1269: _{0}(S^{0};\mathbb{Z})\otimes H_{2}(\widehat{L\cap \varepsilon ^{2}L};%
1270: \mathbb{Z})
1271: \end{equation*}%
1272: instructs us that it remains to look at the action of the $\varepsilon ^{2}$
1273: on $\tilde{H}_{0}(S^{0};\mathbb{Z})$. Keeping in mind that $S^{0}$ is the
1274: order complex of the lower cone of the element $L\cap \varepsilon ^{2}L$, it
1275: is obvious that $\varepsilon ^{2}$ acts on $H_{0}(S^{0};\mathbb{Z})\cong
1276: \mathbb{Z\oplus Z}$ by permuting generators of the two copies of $\mathbb{Z}$
1277: and remembering the orientation, i. e.%
1278: \begin{equation*}
1279: (x,y)~\mapsto ~\det \varepsilon ^{2}(y,x)
1280: \end{equation*}%
1281: Therefore, the definition of the augmentation implies that $\varepsilon ^{2}$
1282: acts by multiplication on the reduced homology $\tilde{H}_{0}(S^{0};\mathbb{Z%
1283: })\cong \mathbb{Z}$. Thus, $k=(\det \varepsilon ^{2})~\varepsilon ^{2}\cdot k
1284: $.
1285:
1286: \noindent Again we compute the coinvariants $H_{2}(W_{4}\oplus
1287: W_{4}\backslash \cup \mathcal{B};\mathbb{Z})_{\mathbb{Z}_{4}}$ using the
1288: modified action. The following relations
1289: \begin{equation*}
1290: l\sim \varepsilon ^{i}\ast l=\det (\varepsilon ^{i})\varepsilon ^{i}\cdot
1291: l=\varepsilon ^{i}\cdot l\text{, }k\sim \varepsilon \ast k=\det (\varepsilon
1292: )\varepsilon \cdot k=\varepsilon k\text{, }k\sim \varepsilon ^{2}\ast
1293: k=(\det \varepsilon ^{2})\varepsilon ^{2}\cdot k=k
1294: \end{equation*}%
1295: imply that actually nothing "torsion-like" happens and that $%
1296: H_{2}(W_{4}\oplus W_{4}\backslash \cup \mathcal{B};\mathbb{Z})_{\mathbb{Z}%
1297: _{4}}\cong \mathbb{Z\oplus Z}$.
1298: \end{example}
1299:
1300: \subsection{\label{sec:HowToComputeTheOstructionCocycle}How to compute the
1301: obstruction cocycle}
1302:
1303: \noindent Finally, we are ready to give an algorithm for computing the
1304: cohomology class of the obstruction cocycle of the map in general position.
1305: Let us assume that additional hyper arrangement $\mathcal{J}$ is already
1306: chosen.
1307:
1308: \textit{Step 1:} Let $S^{3}$ be $\mathbb{Q}_{4n}$ the simplicial complex
1309: earlier defined. Define $\mathbb{Q}_{4n}$-map $h\,:S^{3}\rightarrow W_{n}$
1310: by defining them on $0$-skeleton. It is enough to define an image of a
1311: single vertex, because everything else is defined by the equivariant request.
1312:
1313: \textit{Step 2:}\textbf{\ }Find all simplexes $\sigma
1314: _{ij}=[a_{i},a_{i+1};b_{j},b_{j+1}]$ such that
1315: \begin{equation*}
1316: h([a_{i},a_{i+1};b_{j},b_{j+1}])\cap (L(\alpha )\cap \mathcal{J})=\{pt.\}%
1317: \text{.}
1318: \end{equation*}%
1319: The intersection can't have more than one point, because then at least the
1320: whole interval will be in the intersection. This would mean that the map $h$
1321: is not in the general position.
1322:
1323: \textit{Step 3:} With the help of the cellular map between two introduced $%
1324: \mathbb{Q}_{4n}$ cellular structures on $S^{3}$, and previous step, count
1325: the following sets%
1326: \begin{equation*}
1327: h(e)\cap (\cup \mathcal{A}(\mathcal{J},\alpha ))\text{ and }h^{-1}(h(e)\cap
1328: (\cup \mathcal{A}(\mathcal{J},\alpha )))\subset e\text{.}
1329: \end{equation*}
1330:
1331: \textit{Step 4:} Let us assume that every element $y\in h(e)\cap (\cup
1332: \mathcal{A}(\mathcal{J},\alpha ))$ is contained in just one and only one
1333: maximal element $L_{y}$ of the arrangement $\mathcal{A}$. Then
1334: \begin{equation}
1335: c_{\mathbb{Q}_{4n}}(h)(e)=\sum_{x\in h^{-1}(h(e)\cap (\cup \mathcal{A}(%
1336: \mathcal{J},\alpha )))}\mathrm{I}(e,L_{h(x)})\left\Vert h(x)\right\Vert
1337: \label{eq:ObstructionCocycle}
1338: \end{equation}%
1339: where $\mathrm{I}(h(\theta ),L_{h(x)})$ is the intersection number of the
1340: oriented cell $e$ and appropriate oriented element $L_{y}$ ($L_{y}\cap
1341: h(\theta )=\{y\}$) of the arrangement $\mathcal{A}(J,\alpha )$.
1342:
1343: \noindent If some $y\in h(e)\cap (\cup \mathcal{A}(\mathcal{J},\alpha ))$
1344: belongs to codimension one intersection of maximal elements $%
1345: L_{y}^{(1)},..,L_{y}^{(k)}$ the formulas is unchanged except the class $%
1346: \left\Vert h(x)\right\Vert $ is a broken point class. The real trouble with
1347: this situation is that we need an extra effort to identify this broken point
1348: class. As we mentioned in section \ref{sec:pointClasses} the broken point
1349: class depends on the embedding of the simplex (or its linear span) which
1350: intersects arrangement in $y\in h(e)\cap (\cup \mathcal{A}(\mathcal{J}%
1351: ,\alpha ))$.
1352:
1353: \textit{Step 5:} Compute $H_{n-4}(\cup \widehat{\mathcal{A}}(J,\alpha );%
1354: \mathbb{Z})$ using decomposition (\ref{Z-Z homology decomposition - 1}). If
1355: there are no torsion we actually calculated $H_{2}(W_{n}\backslash \cup
1356: \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})$.
1357:
1358: \textit{Step 6:} Compute the coinvariants $H_{2}(W_{n}\backslash \cup
1359: \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})_{\mathbb{Q}_{4n}}$ by working
1360: in $\mathbb{Q}_{4n}$-module $H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J}%
1361: ,\alpha );\mathbb{Z})$ using modified action - isomorphism (\ref{isomorphism
1362: of the modified action}).
1363:
1364: \textit{Step 7:} Express the obstruction element $c_{\mathbb{Q}%
1365: _{4n}}(h)(e)\in H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );%
1366: \mathbb{Z})$ as the element of the group $\mathrm{Hom}(H_{n-4}(\cup \widehat{%
1367: \mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z}),\mathbb{Z})$ via the
1368: isomorphism (\ref{Isomorphisms Complement - Arrangement}).
1369:
1370: \textit{Step 8:} Identify the class of the obstruction cocycle $c_{\mathbb{Q}%
1371: _{4n}}(h)(e)$ in the group
1372: \begin{equation*}
1373: H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha ),\mathbb{Z})/\sim
1374: \text{.}
1375: \end{equation*}%
1376: If it is not zero we proved that the $\mathbb{Q}_{4n}$-map in question can
1377: not exist. Thus, propositions \ref{prop:VezaProblem-Ekvivarijantan} and \ref%
1378: {prop:PrelazakNaNovuGrupu} imply that the appropriate fan partition exists.
1379: If it is zero, then the whole effort was in vane.
1380:
1381: \section{\textsf{Computations and proof of theorem \protect\ref{th:main1}}}
1382:
1383: \noindent The proof of the theorem \ref{th:main1} has two stages
1384:
1385: \begin{itemize}
1386: \item The proof for the wide class of special cases; in particular we prove
1387: theorem for all $\alpha =(\tfrac{a}{n},\tfrac{a+b}{n},\tfrac{b}{n})\in \frac{%
1388: 1}{n}\,\mathbb{N}^{3}\subseteq \mathbb{Q}^{3}$ such that $2a+2b=n$, $a,b\geq
1389: 1$. This proof goes along the lines described in section \ref%
1390: {sec:HowToComputeTheOstructionCocycle}.
1391:
1392: \item The limit argument extend the result from the class of special cases
1393: to the whole class of triples $\alpha =(a,a+b,b)\in \mathbb{R}_{>0}^{3}$, $%
1394: 2a+2b=1$.
1395: \end{itemize}
1396:
1397: \subsection{Finding hyper arrangement $\mathcal{J}$}
1398:
1399: \noindent Before we try to prove the main theorem in steps described in the
1400: section \ref{sec:HowToComputeTheOstructionCocycle}, let us make the
1401: fundamental step by defining the additional hyper arrangement $\mathcal{J}$
1402: in $\mathbb{R}^{n}$. Let us define hyperplanes $H_{1}$, $H_{2}$, $K$ and
1403: half-spaces $K^{+}$, $K^{-}$, by%
1404: \begin{equation*}
1405: \begin{array}{l}
1406: H_{1}=\{\mathbf{x}\in \mathbb{R}^{n}|\text{ }%
1407: (a+b)(x_{a}-x_{2a+b}+x_{1}-x_{a+b+1})+x_{a+1}-x_{2a+b+1}+x_{n}-x_{a+b}=0\},
1408: \\
1409: H_{2}=\{\mathbf{x}\in \mathbb{R}^{n}|\text{ }x_{a+1}+..+x_{a+b}=0\},K=\{%
1410: \mathbf{x}\in \mathbb{R}^{n}|\text{~}x_{1}+..+x_{a+b}=0\} \\
1411: K^{+}=\{\mathbf{x}\in \mathbb{R}^{n}|\text{ }x_{1}+..+x_{a+b}\geq 0\}\text{
1412: and }K^{-}=\{\mathbf{x}\in \mathbb{R}^{n}|\text{~}x_{1}+..+x_{a+b}\leq 0\}.%
1413: \end{array}%
1414: \end{equation*}%
1415: The hyper arrangement $\mathcal{J}$ we would like to consider is%
1416: \begin{equation*}
1417: \mathcal{J}=\{H_{1}\cap K_{1}^{+},~H_{2}\}.
1418: \end{equation*}%
1419: Let us consider some properties of this arrangement which will produce
1420: crucial arguments in the proof of the main theorem. The first property is
1421: that $\varepsilon ^{a+b}H_{1}=H_{1}$ and $\varepsilon ^{a+b}$ does change
1422: the orientation of the subspace $H_{1}^{\bot }$. Indeed, let $%
1423: e=(a+b)(e_{a}-e_{2a+b}+e_{1}-e_{a+b+1})+e_{a+1}-e_{2a+b+1}+e_{n}-e_{a+b}$ be
1424: a base vector of $H_{1}^{\bot }$, then $\varepsilon ^{a+b}\cdot
1425: e=(a+b)(e_{2a+b}-e_{a}+e_{a+b+1}-e_{1})+e_{2a+b+1}-e_{a+1}+e_{a+b}-e_{n}$.
1426: The other property is that $\varepsilon ^{a+b}(K^{+}\cap W_{n})=K^{-}\cap
1427: W_{n}$.
1428:
1429: \noindent Thus the arrangement $\mathcal{A}(\mathcal{J},\alpha )$ is the
1430: minimal $\mathbb{Q}_{4n}$-arrangement containing half-subspace $L_{1}^{\ast
1431: }=L(\alpha )\cap H_{1}\cap K^{+}$ and subspace $L_{2}^{\ast }=L(\alpha )\cap
1432: H_{2}$ defined by
1433: \begin{eqnarray*}
1434: L_{1}^{\ast } &:&\left\{
1435: \begin{tabular}{l}
1436: $x_{1}+..+x_{a}=x_{a+1}+..+x_{2a+b}=x_{2a+b+1}+..+x_{n}=0,~x_{1}+..+x_{a+b}%
1437: \geq 0,$ \\
1438: $(a+b)(x_{a}-x_{2a+b}+x_{1}-x_{a+b+1})+x_{a+1}-x_{2a+b+1}+x_{n}-x_{a+b}=0.$%
1439: \end{tabular}%
1440: \right. \\
1441: L_{2}^{\ast }
1442: &:&x_{1}+x_{2}+..+x_{a}=x_{a+1}+..+x_{a+b}=x_{a+b+1}+..+x_{2a+b}=x_{2a+b+1}+..+x_{n}=0.
1443: \end{eqnarray*}%
1444: The introduced setting provides us with the following facts:
1445:
1446: \textit{(i)} The intersection
1447: \begin{equation*}
1448: I=L_{1}^{\ast }\cap \varepsilon ^{a+b}L_{1}^{\ast }\cap \varepsilon
1449: ^{a}jL_{1}^{\ast }\cap \varepsilon ^{2a+b}jL_{1}^{\ast }\cap L_{2}^{\ast }
1450: \end{equation*}%
1451: is a linear subspace of codimension one in both $L_{1}^{\ast }=L(\alpha
1452: )\cap H_{1}\cap K^{+}$ and $L_{2}^{\ast }=L(\alpha )\cap H_{2}$;
1453:
1454: \textit{(ii)} The following set of equalities stands%
1455: \begin{equation*}
1456: \varepsilon ^{a+b}(L_{1}^{\ast }\cap \varepsilon ^{a+b}L_{1}^{\ast
1457: })=L_{1}^{\ast }\cap \varepsilon ^{a+b}L_{1}^{\ast },~\varepsilon
1458: ^{a+b}(\varepsilon ^{a}jL_{1}^{\ast }\cap \varepsilon ^{2a+b}jL_{1}^{\ast
1459: })=\varepsilon ^{a}jL_{1}^{\ast }\cap \varepsilon ^{2a+b}jL_{1}^{\ast
1460: },~\varepsilon ^{a+b}L_{2}^{\ast }=L_{2}^{\ast }.
1461: \end{equation*}
1462:
1463: \textit{(iii)} The element $\varepsilon ^{a+b}$ changes the orientation on $%
1464: I^{\bot }$.
1465:
1466: \noindent These properties will be the key arguments that the appropriate
1467: group of coinvariants $H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J}%
1468: ,\alpha ),\mathbb{Z})_{\mathbb{Q}_{4n}}$ have the nontrivial
1469: torsion part. The part of our arrangement $\cup
1470: \mathcal{A}(\mathcal{J},\alpha )$ can be pictured like in the
1471: figure \ref{fig:Fig6}.
1472:
1473: %\FRAME{ftbhFU}{1.803in}{2.4705in}{0pt}{\Qcb{{\protect\small The part of the
1474: %arrangement }$\mathcal{A}(\mathcal{J},\protect\alpha )${\protect\small .}}}{%
1475: %\Qlb{Fig6}}{picture6.wmf}{\special{language "Scientific Word";type
1476: %"GRAPHIC";maintain-aspect-ratio TRUE;display "USEDEF";valid_file "F";width
1477: %1.803in;height 2.4705in;depth 0pt;original-width 4.4413in;original-height
1478: %6.1084in;cropleft "0";croptop "1";cropright "1";cropbottom "0";filename
1479: %'Picture6.wmf';file-properties "XNPEU";}}
1480:
1481: \begin{figure}[htb]
1482: \centering
1483: \includegraphics[scale=1]{picture6.eps}
1484: \caption{The part of the arrangement
1485: $\mathcal{A}(\mathcal{J},\alpha )$} \label{fig:Fig6}
1486: \end{figure}
1487:
1488:
1489:
1490: \subsection{Step 1}
1491:
1492: \noindent\ Let us recall that we substituted the sphere $S^{3}$ with the
1493: simplicial complex $P_{2n}^{(1)}\ast P_{2n}^{(2)}$, where $P_{2n}$ is the
1494: regular $2n$-gon. Earlier we have denoted vertices of two copies of $P_{2n}$
1495: by $a_{1},..,a_{2n}$ and $b_{1},..,b_{2n}$, respectively. To simplify
1496: notation in calculation ahead let $t=a_{1}$. Let $e_{1},..,e_{n}$ be the
1497: standard base of $\mathbb{R}^{n}$ and $u_{i}=e_{i}-\tfrac{1}{n}%
1498: \sum\limits_{j=1}^{n}e_{j}$, $i\in \{1,..,n\}$.
1499:
1500: \noindent We define the $\mathbb{Q}_{4n}$-map $h:S^{3}\rightarrow
1501: W_{n}\subset \mathbb{R}^{n}$ which is in the general position by defining it
1502: on the vertex $t$ by $h(t)=u_{1}$. Then the request that $h$ is $\mathbb{Q}%
1503: _{4n}$-map imply
1504: \begin{equation*}
1505: \begin{array}{l}
1506: h(\varepsilon ^{i}t)=\varepsilon ^{i}h(t)=\varepsilon ^{i\
1507: \textrm{mod}\ %
1508: n}h(t)=u_{i\ \textrm{mod}\ n+1} \\
1509: h(jt)=jh(t)=u_{n} \\
1510: h(\varepsilon ^{i}jt)=\varepsilon ^{i}jh(t)=\varepsilon ^{i\
1511: \textrm{mod}\ %
1512: n}jh(t)=u_{i\ \textrm{mod}\ n}.%
1513: \end{array}%
1514: \end{equation*}%
1515: In the future we will omit the "$\textrm{mod}\ n$" part in the
1516: indexes on the
1517: right hand side, i.e. all the indexes in $W_{n}$ are calculated $\textrm{mod}\ n$%
1518: .
1519:
1520: \subsection{Step 2\textbf{\ }}
1521:
1522: \noindent It is not hard to see from the definition that the image
1523: of the map $h$ is
1524: $h(S^{3})=\bigcup\limits_{i,j=1}^{n}[u_{i},u_{i+1}]\ast \lbrack
1525: u_{j},u_{j+1}]\subset W_{n}$, remembering that $n+1$ is actually
1526: $1$. Let us list all the simplexes of the form
1527: $[u_{i},u_{i+1}]\ast \lbrack u_{j},u_{j+1}]\equiv \lbrack
1528: u_{i},u_{i+1};u_{j},u_{j+1}]$ which intersect
1529: the subspace%
1530: \begin{equation*}
1531: L(\alpha )=\{\mathbf{x}\in W_{n}|~x_{1}+\ldots +x_{a}=0,x_{a+1}+\ldots
1532: +x_{2a+b}=0,x_{2a+b+1}+\ldots +x_{n}=0\}\text{.}
1533: \end{equation*}%
1534: With a little linear algebra and combinatorics we see that the only
1535: simplexes from the image $h(S^{3})$ that intersects $L(\alpha )$ are:%
1536: \begin{equation*}
1537: \begin{tabular}{|l|l|l|}
1538: \hline
1539: \ \ {\small Simplex} & {\small Intersection with }${\small L(\alpha )}$ &
1540: {\small For }${\small r\in }$ \\ \hline
1541: {\small 1.}${\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{2a+b}{\small ,u}%
1542: _{2a+b+1}{\small ]}$ & ${\small \{}\tfrac{a}{n}{\small u}_{a}{\small +}%
1543: \tfrac{\beta }{n}{\small u}_{a+1}{\small +}\tfrac{\gamma }{n}{\small u}%
1544: _{2a+b}{\small +}\tfrac{b}{n}{\small u}_{2a+b+1}{\small \}}$ & \\ \hline
1545: {\small 2.}${\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{r}{\small ,u}_{r+1}%
1546: {\small ]}$ & $\{\tfrac{a}{n}{\small u}_{a}{\small +}\tfrac{a+b}{n}{\small u}%
1547: _{a+1}{\small +}\tfrac{\gamma }{n}{\small u}_{r}{\small +}\tfrac{\delta }{n}%
1548: {\small u}_{r+1}\}$ & ${\small [2a+b+1,n-1]}$ \\ \hline
1549: {\small 3.}${\small [u}_{r}{\small ,u}_{r+1}{\small ;u}_{2a+b}{\small ,u}%
1550: _{2a+b+1}{\small ]}$ & $\{\tfrac{\alpha }{n}{\small u}_{r}{\small +}\tfrac{%
1551: \beta }{n}{\small u}_{r+1}{\small +}\tfrac{a+b}{n}{\small u}_{2a+b}{\small +}%
1552: \tfrac{b}{n}{\small u}_{2a+b+1}{\small \}}$ & ${\small [1,a-1]}$ \\ \hline
1553: {\small 4.}${\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{n}{\small ,u}_{1}%
1554: {\small ]}$ & $\{\tfrac{\alpha }{n}{\small u}_{a}{\small +}\tfrac{a+b}{n}%
1555: {\small u}_{a+1}{\small +}\tfrac{b}{n}{\small u}_{n}{\small +}\tfrac{\delta
1556: }{n}{\small u}_{1}{\small \}}$ & \\ \hline
1557: {\small 5.}${\small [u}_{2a+b}{\small ,u}_{2a+b+1}{\small ;u}_{n}{\small ,u}%
1558: _{1}{\small ]}$ & $\{\tfrac{a+b}{n}{\small u}_{2a+b}{\small +}\tfrac{\beta }{%
1559: n}{\small u}_{2a+b+1}{\small +}\tfrac{\gamma }{n}{\small u}_{n}{\small +}%
1560: \tfrac{a}{n}{\small u}_{1}{\small \}}$ & \\ \hline
1561: {\small 6.}${\small [u}_{r}{\small ,u}_{r+1}{\small ;u}_{n}{\small ,u}_{1}%
1562: {\small ]}$ & $\{\tfrac{\alpha }{n}{\small u}_{r}{\small +}\tfrac{\beta }{n}%
1563: {\small u}_{r+1}{\small +}\tfrac{b}{n}{\small u}_{n}{\small +}\tfrac{a}{n}%
1564: {\small u}_{1}{\small \}}$ & ${\small [a+1,2a+b-1]}$ \\ \hline
1565: \end{tabular}%
1566: \end{equation*}%
1567: Whit the assumption that $a,b\geq 1$ we analize two cases.
1568:
1569: \textit{(1)} If we introduce the hyperplane $H_{1}$, then the only simplex
1570: that intersects $L(\alpha )\cap H_{1}$ are%
1571: \begin{equation*}
1572: \rho _{1}={\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{2a+b}{\small ,u}%
1573: _{2a+b+1}{\small ]}\text{, }\rho _{2}={\small [u}_{a+b}{\small ,u}_{a+b+1}%
1574: {\small ;u}_{n}{\small ,u}_{1}{\small ]}\text{, }\rho _{3}={\small [u}%
1575: _{a+b+1}{\small ,u}_{a+b+2}{\small ;u}_{n}{\small ,u}_{1}{\small ]}\text{.}
1576: \end{equation*}%
1577: Precisely,%
1578: \begin{eqnarray*}
1579: (L(\alpha )\cap H_{1})\cap \rho _{1} &=&\{\tfrac{a}{n}u_{a}+\tfrac{b}{n}%
1580: u_{a+1}+\tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}u_{2a+b+1}\} \\
1581: (L(\alpha )\cap H_{1})\cap \rho _{2} &=&\{\tfrac{b}{n}u_{a+b}+\tfrac{a}{n}%
1582: u_{a+b+1}+\tfrac{b}{n}u_{n}+\tfrac{a}{n}u_{1}\} \\
1583: (L(\alpha )\cap H_{1})\cap \rho _{3} &=&\{\tfrac{a+\frac{b}{a+b}}{n}{\small u%
1584: }_{a+b+1}{\small +}\tfrac{b-\frac{b}{a+b}}{n}{\small u}_{a+b+2}{\small +}%
1585: \tfrac{b}{n}{\small u}_{n}{\small +}\tfrac{a}{n}{\small u}_{1}\}.
1586: \end{eqnarray*}%
1587: But when we add the inequality condition $x_{1}+..+x_{a+b}\geq 0$, the
1588: intersection of $L_{1}^{\ast }=L(\alpha )\cap H_{1}\cap K^{+}$ with the
1589: simplex $\rho _{3}$ vanishes. Thus there are only two simplexes
1590: \begin{equation*}
1591: {\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{2a+b}{\small ,u}_{2a+b+1}%
1592: {\small ]}\text{ and }{\small [u}_{a+b}{\small ,u}_{a+b+1}{\small ;u}_{n}%
1593: {\small ,u}_{1}{\small ]}
1594: \end{equation*}%
1595: which intersects $L_{1}^{\ast }$, each of them in the single interior point,%
1596: \begin{equation*}
1597: v=\tfrac{a}{n}u_{a}+\tfrac{b}{n}u_{a+1}+\tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}%
1598: u_{2a+b+1}\in L_{1}^{\ast }\text{ and }w=\tfrac{b}{n}u_{a+b}+\tfrac{a}{n}%
1599: u_{a+b+1}+\tfrac{b}{n}u_{n}+\tfrac{a}{n}u_{1}\in L_{1}^{\ast }\text{.}
1600: \end{equation*}%
1601: Thus, $h(S^{3})\cap L_{1}^{\ast }=\{v,w\}$.
1602:
1603: \textit{(2)} If we introduce the hyperplane $H_{2}$ instead the hyperplane $%
1604: H_{1}$, there are only two simplexes from $h(S^{3})$ that intersects $%
1605: L_{2}^{\ast }=L(\alpha )\cap H_{2}$. Those are
1606: \begin{equation*}
1607: \rho _{1}={\small [u}_{a}{\small ,u}_{a+1}{\small ;u}_{2a+b}{\small ,u}%
1608: _{2a+b+1}{\small ]}\text{ and }\rho _{2}={\small [u}_{a+b}{\small ,u}_{a+b+1}%
1609: {\small ;u}_{n}{\small ,u}_{1}{\small ]}
1610: \end{equation*}%
1611: and by another miracle they intersect $L_{2}^{\ast }$ in the same points as
1612: in the previous case, $h(S^{3})\cap L_{2}^{\ast }=\{v,w\}$.
1613:
1614: \begin{conclusion}
1615: This means that in order to analize the set $h^{-1}(h(e)\cap (\cup \mathcal{A%
1616: }(\mathcal{J},\alpha )))\subset e$ we only have to track down the pre images
1617: of $v$ and $w$.
1618: \end{conclusion}
1619:
1620: \subsection{Step 3}
1621:
1622: \noindent There are 16 simplexes in the sphere $P_{2n}^{(1)}\ast P_{2n}^{(2)}
1623: $ which belong to the $h$ inverse image of the simplexes ${\small [u}_{a}%
1624: {\small ,u}_{a+1}{\small ;u}_{2a+b}{\small ,u}_{2a+b+1}{\small ]}$ and $%
1625: {\small [u}_{a+b}{\small ,u}_{a+b+1}{\small ;u}_{n}{\small ,u}_{1}{\small ]}$%
1626: . These simplexes are
1627: \begin{equation*}
1628: \begin{array}{ll}
1629: \sigma _{1}=[\epsilon ^{a-1}t,\epsilon ^{a}t;\epsilon ^{2a+b}jt,\epsilon
1630: ^{2a+b+1}jt], & \sigma _{2}=[\epsilon ^{n+a-1}t,\epsilon ^{n+a}t;\epsilon
1631: ^{2a+b}jt,\epsilon ^{2a+b+1}jt], \\
1632: \sigma _{3}=[\epsilon ^{a-1}t,\epsilon ^{a}t;\epsilon ^{n+2a+b}jt,\epsilon
1633: ^{n+2a+b+1}jt], & \sigma _{4}=[\epsilon ^{n+a-1}t,\epsilon ^{n+a}t;\epsilon
1634: ^{n+2a+b}jt,\epsilon ^{n+2a+b+1}jt], \\
1635: \sigma _{5}=[\epsilon ^{2a+b-1}t,\epsilon ^{2a+b}t;\epsilon ^{a}jt,\epsilon
1636: ^{a+1}jt], & \sigma _{6}=[\epsilon ^{n+2a+b-1}t,\epsilon ^{n+2a+b}t;\epsilon
1637: ^{a}jt,\epsilon ^{a+1}jt], \\
1638: \sigma _{7}=[\epsilon ^{2a+b-1}t,\epsilon ^{2a+b}t;\epsilon
1639: ^{n+a}jt,\epsilon ^{n+a+1}jt], & \sigma _{8}=[\epsilon ^{n+2a+b-1}t,\epsilon
1640: ^{n+2a+b}t;\epsilon ^{n+a}jt,\epsilon ^{n+a+1}jt], \\
1641: \theta _{1}=[\epsilon ^{a+b-1}t,\epsilon ^{a+b}t;jt,\epsilon jt], & \theta
1642: _{2}=[\epsilon ^{n+a+b-1}t,\epsilon ^{n+a+b}t;jt,\epsilon jt], \\
1643: \theta _{3}=[\epsilon ^{a+b-1}t,\epsilon ^{a+b}t;\epsilon ^{n}jt,\epsilon
1644: ^{n+1}jt], & \theta _{4}=[\epsilon ^{n+a+b-1}t,\epsilon ^{n+a+b}t;\epsilon
1645: ^{n}jt,\epsilon ^{n+1}jt], \\
1646: \theta _{5}=[\epsilon ^{2n-1}t,t;\epsilon ^{a+b}jt,\epsilon ^{a+b+1}jt], &
1647: \theta _{6}=[\epsilon ^{2n-1}t,t;\epsilon ^{n+a+b}jt,\epsilon ^{n+a+b+1}jt],
1648: \\
1649: \theta _{7}=[\epsilon ^{n-1}t,\epsilon ^{n}t;\epsilon ^{a+b}jt,\epsilon
1650: ^{a+b+1}jt], & \theta _{8}=[\epsilon ^{n-1}t,\epsilon ^{n}t;\epsilon
1651: ^{n+a+b}jt,\epsilon ^{n+a+b+1}jt]%
1652: \end{array}%
1653: \end{equation*}%
1654: By some miracle all these simplexes are in the orbit of the single simplex $%
1655: \sigma $ in the maximal cell $e$. Moreover, the pre-images of intersection
1656: points with $L^{\ast }$%
1657: \begin{equation*}
1658: v=\tfrac{a}{n}u_{a}+\tfrac{b}{n}u_{a+1}+\tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}%
1659: u_{2a+b+1}\text{ and }w=\tfrac{b}{n}u_{a+b}+\tfrac{a}{n}u_{a+b+1}+\tfrac{b}{n%
1660: }u_{n}+\tfrac{a}{n}u_{1}
1661: \end{equation*}%
1662: are all in the orbit of two points
1663: \begin{equation*}
1664: v^{\ast }=\tfrac{a}{n}\epsilon ^{a+b-1}t+\tfrac{b}{n}\epsilon ^{a+b}t+\tfrac{%
1665: a}{n}jt+\tfrac{b}{n}\epsilon jt\text{ and }w^{\ast }=\tfrac{b}{n}\epsilon
1666: ^{a+b-1}t+\tfrac{a}{n}\epsilon ^{a+b}t+\tfrac{b}{n}jt+\tfrac{a}{n}\epsilon jt
1667: \end{equation*}%
1668: in the simplex $\sigma $. In particular,%
1669: \begin{equation*}
1670: \begin{array}{ll}
1671: \sigma =[\epsilon ^{a+b-1}t,\epsilon ^{a+b}t;jt,\epsilon jt] & =\epsilon
1672: j\epsilon ^{-a+1}\sigma _{1}=\epsilon ^{-2a-b}\sigma _{2}=\epsilon
1673: ^{-n-2a-b}\sigma _{3}=\epsilon j\epsilon ^{-n-a+1}\sigma _{4} \\
1674: & =\epsilon ^{-a}\sigma _{5}=\epsilon j\epsilon ^{-n-2a-b+1}\sigma
1675: _{6}=\epsilon j\epsilon ^{-2a-b+1}\sigma _{7}=\epsilon ^{-a-n}\sigma _{8} \\
1676: & =\theta _{1}=\epsilon j\epsilon ^{-n-a-b+1}\theta _{2}=\epsilon j\epsilon
1677: ^{-a-b+1}\theta _{3}=\epsilon ^{-n}\theta _{4} \\
1678: & =\epsilon j\epsilon \theta _{5}=\epsilon ^{-n-a-b}\theta _{6}=\epsilon
1679: ^{-a-b}\theta _{7}=\epsilon j\epsilon ^{-n+1}\theta _{8}%
1680: \end{array}%
1681: \end{equation*}%
1682: To illustrate the fact that points $v^{\ast }$ and $w^{\ast }$ from the cell
1683: $e$ are mapped in the arrangement, we track the change of the barycentric
1684: coordinates of points $v$, $w$ and $v^{\ast }$, $w^{\ast }$ along the
1685: actions. For example,
1686:
1687: \textit{(A)} From $\sigma =[\epsilon jt,jt;\epsilon ^{a+b-1}t,\epsilon
1688: ^{a+b}t]=\epsilon j\epsilon ^{-a+1}\sigma _{1}=\epsilon ^{a}j\sigma _{1}$,
1689: we have
1690: \begin{eqnarray*}
1691: h(\sigma _{1})\cap L_{1}^{\ast } &=&\{\tfrac{a}{n}u_{a}+\tfrac{b}{n}u_{a+1}+%
1692: \tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}u_{2a+b+1}\} \\
1693: h(\epsilon ^{a}j\sigma _{1})\cap \epsilon ^{a}jL_{1}^{\ast } &=&\{\tfrac{a}{n%
1694: }u_{1}+\tfrac{b}{n}u_{n}+\tfrac{a}{n}u_{a+b+1}+\tfrac{b}{n}u_{a+b}\} \\
1695: h^{-1}\left( h(\epsilon ^{a}j\sigma _{1})\cap \epsilon ^{a}jL_{1}^{\ast
1696: }\right) &=&\{\tfrac{a}{n}\epsilon jt+\tfrac{b}{n}jt+\tfrac{a}{n}\epsilon
1697: ^{a+b}t+\tfrac{b}{n}\epsilon ^{a+b-1}t\}=\{w^{\ast }\}.
1698: \end{eqnarray*}
1699:
1700: \textit{(B)} From $\sigma =[\epsilon ^{a+b-1}t,\epsilon ^{a+b}t;jt,\epsilon
1701: jt]=\epsilon ^{-2a-b}\sigma _{2}$, we conclude that
1702: \begin{eqnarray*}
1703: h(\sigma _{2})\cap L_{1}^{\ast } &=&\{\tfrac{a}{n}u_{a}+\tfrac{b}{n}u_{a+1}+%
1704: \tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}u_{2a+b+1}\} \\
1705: h(\epsilon ^{-2a-b}\sigma _{2})\cap \epsilon ^{-2a-b}L_{1}^{\ast } &=&\{%
1706: \tfrac{a}{n}u_{a+b}+\tfrac{b}{n}u_{a+b+1}+\tfrac{a}{n}u_{n}+\tfrac{b}{n}%
1707: u_{a+b}\} \\
1708: h^{-1}\left( h(\epsilon ^{-2a-b}\sigma _{2})\cap \epsilon
1709: ^{-2a-b}L_{1}^{\ast }\right) &=&\{\tfrac{a}{n}\epsilon ^{a+b-1}t+\tfrac{b}{n}%
1710: \epsilon ^{a+b}t+\tfrac{a}{n}jt+\tfrac{b}{n}\epsilon jt\}=\{v^{\ast }\};
1711: \end{eqnarray*}
1712:
1713: \textit{(C)} From $\sigma =[\epsilon ^{a+b-1}t,\epsilon ^{a+b}t;jt,\epsilon
1714: jt]=\theta _{1}$, it easily follows
1715:
1716: \begin{eqnarray*}
1717: h(\theta _{1})\cap L_{1}^{\ast } &=&\{\tfrac{b}{n}u_{a+b}+\tfrac{a}{n}%
1718: u_{a+b+1}+\tfrac{b}{n}u_{n}+\tfrac{a}{n}u_{1}\} \\
1719: h^{-1}\left( h(\theta _{1})\cap L_{1}^{\ast }\right) &=&\{\tfrac{b}{n}%
1720: \epsilon ^{a+b-1}t+\tfrac{a}{n}\epsilon ^{a+b}t+\tfrac{b}{n}jt+\tfrac{a}{n}%
1721: \epsilon jt\}=\{w^{\ast }\}.
1722: \end{eqnarray*}
1723:
1724: \begin{conclusion}
1725: Therefore,
1726: \begin{equation*}
1727: \mathrm{card}(h(e)\cap (\cup \mathcal{A}(J,\alpha )))=\mathrm{card}%
1728: (h^{-1}(h(e)\cap (\cup \mathcal{A}(J,\alpha ))))=2
1729: \end{equation*}%
1730: and
1731: \begin{equation*}
1732: h^{-1}(h(e)\cap (\cup \mathcal{A}(J,\alpha )))=\{v^{\ast },w^{\ast }\}\text{.%
1733: }
1734: \end{equation*}
1735: \end{conclusion}
1736:
1737: \subsection{Step 4}
1738:
1739: \noindent Before analyzing the obstruction cocycle, let us observe that%
1740: \begin{equation*}
1741: \begin{array}{lll}
1742: h(v^{\ast })=\epsilon ^{b}\cdot v\in & \epsilon ^{-2a-b}(L_{1}^{\ast }\cap
1743: L_{2}^{\ast })\cap \epsilon ^{-a}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap
1744: \epsilon j\epsilon ^{-a-b+1}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon
1745: j\epsilon (L_{1}^{\ast }\cap L_{2}^{\ast }) & = \\
1746: & \epsilon ^{b}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon
1747: ^{2b+a}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon ^{a+b}j(L_{1}^{\ast
1748: }\cap L_{2}^{\ast })\cap j(L_{1}^{\ast }\cap L_{2}^{\ast }) & = \\
1749: & \epsilon ^{b}(L_{1}^{\ast }\cap \epsilon ^{b+a}L_{1}^{\ast }\cap \epsilon
1750: ^{a}jL_{1}^{\ast }\cap \epsilon ^{2a+b}jL_{1}^{\ast }\cap L_{2}^{\ast }) &
1751: \end{array}%
1752: \end{equation*}%
1753: and similarly%
1754: \begin{equation*}
1755: \begin{array}{lll}
1756: h(w^{\ast })=w\in & (L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon
1757: ^{-a-b}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon j\epsilon
1758: ^{-a+1}(L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon j\epsilon
1759: ^{-2a-b+1}(L_{1}^{\ast }\cap L_{2}^{\ast }) & = \\
1760: & (L_{1}^{\ast }\cap L_{2}^{\ast })\cap \epsilon ^{a+b}(L_{1}^{\ast }\cap
1761: L_{2}^{\ast })\cap \epsilon ^{a}j(L_{1}^{\ast }\cap L_{2}^{\ast })\cap
1762: \epsilon ^{2a+b}j(L_{1}^{\ast }\cap L_{2}^{\ast }) & = \\
1763: & L_{1}^{\ast }\cap \epsilon ^{b+a}L_{1}^{\ast }\cap \epsilon
1764: ^{a}jL_{1}^{\ast }\cap \epsilon ^{2a+b}jL_{1}^{\ast }\cap L_{2}^{\ast } &
1765: \end{array}%
1766: \end{equation*}%
1767: Then the equality \ref{eq:ObstructionCocycle} applied in this situation
1768: implies that the obstruction cocycle is
1769: \begin{equation*}
1770: c_{\mathbb{Q}_{4n}}(h)(e)=\left\Vert h(v^{\ast })\right\Vert +\left\Vert
1771: h(w^{\ast })\right\Vert
1772: \end{equation*}%
1773: where $\tau _{1},\tau _{2}\in \{+1,-1\}$. In this situation classes $%
1774: \left\Vert h(v^{\ast })\right\Vert $ and $\left\Vert h(w^{\ast })\right\Vert
1775: $ are broken point classes determined by the $h$ embedding of the boundary
1776: of the simplex $\sigma $, i.e. by the linear subspace $\mathrm{span}%
1777: \{u_{a+b},u_{a+b+1},u_{n},u_{1}\}$.
1778:
1779: \noindent Since, $h(v^{\ast })=\epsilon ^{b}\cdot v$ and $h(w^{\ast })=w$
1780: the obstruction cocycle is
1781: \begin{equation*}
1782: c_{\mathbb{Q}_{4n}}(h)(e)=\epsilon ^{b}\cdot \left\Vert v\right\Vert
1783: +\left\Vert w\right\Vert \text{.}
1784: \end{equation*}%
1785: where the first broken point class is determined by the linear span of the
1786: simplex
1787: \begin{equation*}
1788: \epsilon
1789: ^{-b}[u_{a+b},u_{a+b+1},u_{n},u_{1}]=[u_{a},u_{a+1},u_{2a+b},u_{2a+b+1}].
1790: \end{equation*}
1791: Also, observe that the equality%
1792: \begin{equation*}
1793: \epsilon ^{a}j\cdot v=\epsilon ^{a}j\cdot \left( \tfrac{a}{n}u_{a}+\tfrac{b}{%
1794: n}u_{a+1}+\tfrac{a}{n}u_{2a+b}+\tfrac{b}{n}u_{2a+b+1}\right) =\tfrac{a}{n}%
1795: u_{1}+\tfrac{b}{n}u_{n}+\tfrac{a}{n}u_{a+b+1}+\tfrac{b}{n}u_{a+b}=w
1796: \end{equation*}%
1797: and the fact that the permutation $(4321)$ is even implies%
1798: \begin{equation*}
1799: \left\Vert w\right\Vert =\epsilon ^{a}j\cdot \left\Vert v\right\Vert \text{.}
1800: \end{equation*}
1801:
1802: \begin{conclusion}
1803: The cohomology class of the obstruction cocycle, as we have seen, lives in
1804: the group of coinvariants $H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J}%
1805: ,\alpha );\mathbb{Z})_{\mathbb{Q}_{4n}}$. Thus, instead of the original
1806: obstruction cocycle $c_{\mathbb{Q}_{4n}}(h)(e)=\epsilon ^{b}\cdot \left\Vert
1807: v\right\Vert +\left\Vert w\right\Vert =(\epsilon ^{b}+\epsilon ^{a}j)\cdot
1808: \left\Vert v\right\Vert $ we can analize the cocycle%
1809: \begin{equation*}
1810: c^{\prime }=2\left\Vert v\right\Vert
1811: \end{equation*}%
1812: where the broken point class $\left\Vert v\right\Vert $ is determined by the
1813: linead span of the simplex%
1814: \begin{equation*}
1815: \lbrack u_{a},u_{a+1},u_{2a+b},u_{2a+b+1}].
1816: \end{equation*}
1817: \end{conclusion}
1818:
1819: \subsection{Step 5}
1820:
1821: \noindent Now we will find the complete second homology $H_{2}(W_{n}%
1822: \backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})$ via the
1823: isomorphism (\ref{hom decomposition - 1}). In order to do so we describe the
1824: first two lower levels of the Hasse diagram of the intersection poset $%
1825: P(\alpha )$ of the arangement $\mathcal{A}(\mathcal{J},\alpha )$. With a
1826: little linear algebra and mentioned properties of the added arrangement $%
1827: \mathcal{J}$, it can be seen that the $n-4$ and $n-5$ levels of the Hasse
1828: diagram of the intersection poset $P(\alpha )$ are like in the figure \ref%
1829: {fig:Fig7}.
1830:
1831: %\FRAME{ftbhFU}{5.5247in}{0.7321in}{0pt}{\Qcb{{\protect\small The }$n-4$%
1832: %{\protect\small \ and }$n-5${\protect\small \ level of the intersection
1833: %poset.}}}{\Qlb{Fig7}}{picture7.wmf}{\special{language "Scientific Word";type
1834: %"GRAPHIC";maintain-aspect-ratio TRUE;display "USEDEF";valid_file "F";width
1835: %5.5247in;height 0.7321in;depth 0pt;original-width 6.1084in;original-height
1836: %0.7841in;cropleft "0";croptop "1";cropright "1";cropbottom "0";filename
1837: %'Picture7.wmf';file-properties "XNPEU";}}\noindent Thus, from the
1838: %decomposition (\ref{hom decomposition - 1}) and the facts
1839:
1840: \begin{figure}[htb]
1841: \centering
1842: \includegraphics[scale=1]{picture7.eps}
1843: \caption{The $n-4$ and $n-5$ level of the intersection poset.}
1844: \label{fig:Fig7}
1845: \end{figure}
1846:
1847: \textit{(A)} $L_{1}^{\ast }\cap \epsilon ^{b+a}L_{1}^{\ast }\cap \epsilon
1848: ^{a}jL_{1}^{\ast }\cap \epsilon ^{2a+b}jL_{1}^{\ast }\cap L_{2}^{\ast }$ is
1849: a linear space of dimension $n-5$
1850:
1851: \textit{(B)} $L_{1}^{\ast }$ is a half-subspace of dimension $n-4$
1852:
1853: \textit{(C)} $L_{2}^{\ast }$ is a linear space of dimension $n-4$
1854:
1855: \noindent we conclude that%
1856: \begin{eqnarray*}
1857: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})
1858: &\cong &\mathbb{Z}^{a+b}\oplus \mathbb{Z}^{4(a+b)}\oplus \underset{d=0}{%
1859: \overset{n-6}{\bigoplus }}\mathrm{Hom}\left( \underset{\dim
1860: V=d}{\bigoplus
1861: }H_{n-4}(\Delta (P_{<V})\ast \hat{V});\mathbb{Z}\right) \\
1862: &\cong &\mathbb{Z}^{a+b}\oplus \mathbb{Z}^{4(a+b)}\oplus \underset{d=0}{%
1863: \overset{n-6}{\bigoplus }}\mathrm{Hom}\left( \underset{\dim
1864: V=d}{\bigoplus }\tilde{H}_{n-5-\dim V}(\Delta
1865: (P_{<V}));\mathbb{Z}\right) .
1866: \end{eqnarray*}
1867:
1868: \begin{lemma}
1869: $(\forall V\in P(\alpha ))\,\dim V\leq n-6\,\Longrightarrow \,\tilde{H}%
1870: _{n-5-\dim V}(\Delta (P(\alpha )_{<V}))=0.$
1871: \end{lemma}
1872:
1873: \begin{proof}
1874: For every element $W\in P(\alpha )_{<V}$ such that $\dim W=n-4$ there exists
1875: a unique element $U_{W}\in P(\alpha )_{<V}$ with the property $\dim
1876: U_{W}=n-5 $ and $W<U_{W}$. There is a monotone map $f:P(\alpha
1877: )_{<V}\rightarrow P(\alpha )_{<V}-\{U~|~\dim U=n-4\}$ defined by
1878: \begin{equation*}
1879: \begin{array}{ccc}
1880: U & \longmapsto & \left\{
1881: \begin{array}{cc}
1882: U, & \text{for }\dim U\leq n-5 \\
1883: U_{W} & \text{for }\dim U=n-4%
1884: \end{array}%
1885: \right.%
1886: \end{array}%
1887: \end{equation*}%
1888: which satisfies conditions of the Quillen fiber lemma. Thus, $f$ induces a
1889: homotopy equivalence and so
1890: \begin{equation*}
1891: \tilde{H}_{n-5-\dim V}(\Delta (P(\alpha )_{<V}))=\tilde{H}_{n-5-\dim
1892: V}(\Delta (P(\alpha )_{<V}-\{U~|~\dim U=n-4\}))=0.
1893: \end{equation*}%
1894: because $\dim \Delta (P(\alpha )_{<V}-\{U~|~\dim U=n-4\})<n-5-\dim V$.
1895: \end{proof}
1896:
1897: \begin{conclusion}
1898: The previous lemma and equality impies that
1899: \begin{equation}
1900: H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})\cong
1901: H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})\cong
1902: \mathbb{Z}^{a+b}\oplus \mathbb{Z}^{4(a+b)}. \label{H2(complement)}
1903: \end{equation}
1904: \end{conclusion}
1905:
1906: \subsection{Step 6}
1907:
1908: \noindent We compute $\mathbb{Q}_{4n}$ coinvariants by working in the $%
1909: \mathbb{Q}_{4n}$-module $H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J}%
1910: ,\alpha );\mathbb{Z})$ with the modified action. Since action respects
1911: dimensional decomposition (\ref{hom decomposition - 1}) we analyze $n-4$ and
1912: $n-5$ dimension cases separatelly.
1913:
1914: \textit{(A)} Let $l\in H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha
1915: );\mathbb{Z})$ be the element which corresponds to the compactification of
1916: the subspace $L_{2}^{\ast }$. Then $l,\epsilon l,..,\epsilon ^{a+b-1}l$ is a
1917: base of the first factor in the equality (\ref{H2(complement)}). There are
1918: to set identities which should be considered.
1919:
1920: \noindent The identity $L_{2}^{\ast }=\epsilon ^{a+b}L_{2}^{\ast }$ produces
1921: equality $l=\det (\epsilon ^{a+b})\epsilon ^{a+b}\cdot
1922: l=(-1)^{(n+1)(a+b)}\epsilon ^{a+b}\cdot l$ in homology. Indeed, like in
1923: examples \ref{Ex:Coinvariants1} \ and \ref{Ex:Coinvariants2}, we look at the
1924: orthogonal complement of $L_{2}^{\ast }$ and see that $\epsilon ^{a+b}$ acts
1925: on a basis as the even permutation $(3412)$. Then coinvariant calculation
1926: with modified action implies a trivial relation,%
1927: \begin{equation*}
1928: l\sim \epsilon ^{a+b}\ast l=\det (\epsilon ^{a+b})\epsilon ^{a+b}\cdot
1929: l=\det (\epsilon ^{a+b})^{2}l.
1930: \end{equation*}%
1931: The second identity $L_{2}^{\ast }=\epsilon ^{-b}jL_{2}^{\ast }$ produces
1932: equality $l=(-1)\det (\epsilon ^{-b}j)\epsilon ^{-b}j\cdot l$ in homology.
1933: Like we mentioned, this is the consequence of the fact that $\epsilon ^{-b}j$
1934: acts on a base of $\left( L_{2}^{\ast }\right) ^{\bot }$ as the odd
1935: permutation $(3214)$. Then,%
1936: \begin{equation*}
1937: l\sim \epsilon ^{-b}j\ast l=\det (\epsilon ^{-b}j)\epsilon ^{-b}j\cdot
1938: l=-\det (\epsilon ^{-b}j)^{2}\cdot l=-l.
1939: \end{equation*}
1940:
1941: \noindent Thus, the first factor in the equality (\ref{H2(complement)})
1942: reduces in the coinvariants to a single $\mathbb{Z}_{2}$.
1943:
1944: \textit{(B) }Since we want to find the coinvariants we can concentrate on
1945: the "genterating" part of the Hasse diagram and its inner symmetries (figure %
1946: \ref{fig:Fig9})
1947:
1948: %\FRAME{ftbhFU}{4.2791in}{1.6466in}{0pt}{\Qcb{{\protect\small Generating part
1949: %of the Hasse diagram for coinvariants.} }}{\Qlb{Fig9}}{picture9-1.wmf}{%
1950: %\special{language "Scientific Word";type "GRAPHIC";maintain-aspect-ratio
1951: %TRUE;display "USEDEF";valid_file "F";width 4.2791in;height 1.6466in;depth
1952: %0pt;original-width 4.2289in;original-height 1.6103in;cropleft "0";croptop
1953: %"1";cropright "1";cropbottom "0";filename 'Picture9-1.wmf';file-properties
1954: %"XNPEU";}}
1955:
1956: \begin{figure}[htb]
1957: \centering
1958: \includegraphics[scale=1]{picture9.eps}
1959: \caption{Generating part of the Hasse diagram for coinvariants. }
1960: \label{fig:Fig9}
1961: \end{figure}
1962:
1963: \noindent Let $k$ and $h$ be the elements of $H_{n-4}(\cup \widehat{\mathcal{%
1964: A}}(\mathcal{J},\alpha );\mathbb{Z})$ which correspond to the
1965: compactifications of the intersections $L_{1}^{\ast }\cap L_{2}^{\ast }$
1966: and. $\epsilon ^{a+b}L_{1}^{\ast }\cap L_{1}^{\ast }$. Then the orbits of
1967: these two elements form a base of the second factor in the equality (\ref%
1968: {H2(complement)}). Let $t\in H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J}%
1969: ,\alpha );\mathbb{Z})$ also denotes the element which corresponds to the
1970: compactification of the intersection $\epsilon ^{a+b}L_{1}^{\ast }\cap
1971: L_{2}^{\ast }$ with the same generic point in $L_{2}^{\ast }$ as in the
1972: representation of the $k$. Here we have to be very carefull, because $%
1973: \epsilon ^{a+b}$ interchanges tha halfspaces of $L_{2}^{\ast }$
1974: genterated by the hyperplane $H_{1}$. Moreover Thus, like the
1975: figure \ref{fig:Fig9}
1976: indicates%
1977: \begin{equation}
1978: \epsilon ^{a+b}\cdot k=\det (\epsilon ^{a+b})(l+t) \label{rel1}
1979: \end{equation}%
1980: where $l\in H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z%
1981: })$ (as in part (A)) is homology class corresponding to $\widehat{%
1982: L_{2}^{\ast }}$.
1983:
1984: \noindent On the other hand $k+t=h$. Since the sphere $\widehat{L_{1}^{\ast
1985: }\cap \epsilon ^{a+b}L_{1}^{\ast }}\ast S^{0}$ is invariant of the element $%
1986: \epsilon ^{a+b}$, there is a possible equality $h=\pm \epsilon ^{a+b}\cdot h$%
1987: . Let us fix the basis $%
1988: \{e_{1}+..+e_{a},~e_{a+1}+..+e_{a+b},~e_{a+b+1}+..+e_{2a+b},~e_{2a+b+1}+..+e_{n},~(a+b)(e_{a}-e_{2a+b}+e_{1}-e_{a+b+1})+e_{a+1}-e_{2a+b+1}+e_{n}-e_{a+b}\}
1989: $ of the orthogonal complement of $L_{1}^{\ast }\cap \epsilon
1990: ^{a+b}L_{1}^{\ast }$. Then the matrix of $\epsilon ^{a+b}$ on this basis is%
1991: \begin{equation*}
1992: M=\left[
1993: \begin{array}{lllll}
1994: 0 & 0 & 1 & 0 & 0 \\
1995: 0 & 0 & 0 & 1 & 0 \\
1996: 1 & 0 & 0 & 0 & 0 \\
1997: 0 & 1 & 0 & 0 & 0 \\
1998: 0 & 0 & 0 & 0 & -1%
1999: \end{array}%
2000: \right]
2001: \end{equation*}%
2002: and $\det M=-1$. This produces the relation $h=-\det (\epsilon
2003: ^{a+b})\epsilon ^{a+b}\cdot h$ in homology. Thus,%
2004: \begin{equation}
2005: h\sim \epsilon ^{a+b}\ast h=\det (\epsilon ^{a+b})\epsilon ^{a+b}\cdot
2006: h=-\det (\epsilon ^{a+b})^{2}h=-h\text{.} \label{rel2}
2007: \end{equation}%
2008: Since, $\epsilon ^{a+b}\cdot k-\det (\epsilon ^{a+b})t=\det (\epsilon
2009: ^{a+b})l$ and $h=k+t$, the relations \ref{rel1} and \ref{rel2} imply that%
2010: \begin{equation*}
2011: l\sim \det (\epsilon ^{a+b})l=\det (\epsilon ^{a+b})(\epsilon ^{a+b}\cdot
2012: k-\det (\epsilon ^{a+b})t)=\epsilon ^{a+b}\ast k-t\sim k-t\text{ and }%
2013: k+t\sim -k-t\text{.}
2014: \end{equation*}%
2015: Actually, we conclude that $4k=0$, or the second factor produces one $%
2016: \mathbb{Z}_{4}$ .
2017:
2018: \begin{conclusion}
2019: $H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})_{%
2020: \mathbb{Q}_{4n}}\cong H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J}%
2021: ,\alpha );\mathbb{Z})_{\mathbb{Q}_{4n}}\cong \mathbb{Z}_{2}\oplus \mathbb{Z}%
2022: _{4}$.
2023: \end{conclusion}
2024:
2025: \subsection{\label{Sec:Step7}Step 7}
2026:
2027: \noindent Now we are ready to find the image of the obstruction cocycle $%
2028: c^{\prime }=2\left\Vert v\right\Vert $ via the isomorphism $\vartheta
2029: :H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z}%
2030: )\rightarrow \mathrm{Hom}\left( H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J%
2031: },\alpha ),\mathbb{Z}),\mathbb{Z}\right) $. Actually we have to compute
2032: (from (\ref{link-isomorphism})) the linking numbers of the sphere / boundary
2033: of the simplex $S=\partial \lbrack u_{a},u_{a+1},u_{2a+b},u_{2a+b+1}]$
2034: (which represents the broken point class $\left\Vert v\right\Vert $) and
2035: spheres which represents generators of the $H_{n}(\cup \widehat{\mathcal{A}}(%
2036: \mathcal{J},\alpha );\mathbb{Z})$. Since simplex $%
2037: [u_{a},u_{a+1},u_{2a+b},u_{2a+b+1}]$ intersects only $I=L_{1}^{\ast }\cap
2038: \epsilon ^{b+a}L_{1}^{\ast }\cap \epsilon ^{a}jL_{1}^{\ast }\cap \epsilon
2039: ^{2a+b}jL_{1}^{\ast }\cap L_{2}^{\ast }$, the homology elements which may
2040: have non-zero values are those from the local picture $H_{n}(\widehat{I}\ast
2041: \lbrack 5];\mathbb{Z})$. Since the coinvariant class of $l$ lives in $%
2042: \mathbb{Z}_{2}$ and our obstruction cocycle is $2\cdot (something)$ there is
2043: no need to compute the image of $l$ either.
2044:
2045: \noindent By moving simplex $[u_{a},u_{a+1},u_{2a+b},u_{2a+b+1}]$ and
2046: analyzing its intersection with subspaces $L_{1}^{\ast }$, $\epsilon
2047: ^{b+a}L_{1}^{\ast }$, $\epsilon ^{a}jL_{1}^{\ast }$, $\epsilon
2048: ^{2a+b}jL_{1}^{\ast }$, $L_{2}^{\ast }$ we decompose the broken point class $%
2049: \left\Vert v\right\Vert $ into a sum of ordinary point classes.
2050:
2051: \begin{theorem}
2052: \label{Lemma:Link}Let $v_{1}\in L_{1}^{\ast }$, $v_{2}\in \epsilon
2053: ^{b+a}L_{1}^{\ast }$, $v_{3}\in \epsilon ^{a}jL_{1}^{\ast }$, $v_{4}\in
2054: \epsilon ^{2a+b}jL_{1}^{\ast }$ and $y_{1}$, $y_{2}\in L_{2}^{\ast }$ are
2055: arbitrary elements with the property that thay do not belong to any other
2056: element of the arrangement $\mathcal{A}(\mathcal{J},\alpha )$. The request
2057: for $y_{1}$ and $y_{2}$ is that thay are in different connecting components
2058: of $L_{2}^{\ast }\backslash I$. Then our broken point class $\left\Vert
2059: v\right\Vert $ can be presented as the sum of the point classes%
2060: \begin{equation}
2061: \left\Vert v\right\Vert =\tau _{1}\left\Vert v_{1}\right\Vert +\tau
2062: _{2}\left\Vert v_{3}\right\Vert +\tau _{3}\left\Vert y_{1}\right\Vert =\mu
2063: _{1}\left\Vert v_{2}\right\Vert +\mu _{2}\left\Vert v_{4}\right\Vert +\mu
2064: _{3}\left\Vert y_{2}\right\Vert \label{Broken-point}
2065: \end{equation}%
2066: where $\tau _{i}$ and $\mu _{i}$ are appropriate signs \{+1,-1\}.
2067: \end{theorem}
2068:
2069: \begin{proof}
2070: The idea of the proof is very simple. \textit{Move our filled sphere /
2071: simplex a little and look where it hits our arrangement and we are done.}
2072: The reason for this to work is in the definition of the linking -
2073: intersection number and the deffinition of (broken) point classes. Let us
2074: recall that the linking /intersection number is "very" invariant under small
2075: movement of the sphere - disc. Thus if you want to check whether you
2076: calculated the intersection number properly you move your sphere - disc and
2077: see what happens. Since we are in linear / convex situation, besides the
2078: linking - intersection numbers (if correctly defined) are +1,-1 or 0, it is
2079: enought to consider small translatorions of a sphere - disc.
2080:
2081: \noindent Thus, let us move our simplex $\sigma $ by the generic
2082: "small" vector $s=\sum_{i=1}^{n}\xi _{i}e_{i}$. Because of
2083: complementary dimension affine space
2084: $\mathrm{span}\{u_{a+b},u_{a+b+1},u_{n},u_{1}\}+s$ must hit all
2085: linear spans of $L_{1}^{\ast }$, $\epsilon ^{a+b}L_{1}^{\ast }$,
2086: $\epsilon ^{a}jL_{1}^{\ast }$, $\epsilon ^{2a+b}jL_{1}^{\ast }$
2087: and $L_{2}^{\ast }$. Let us denote these intersection points by
2088: \begin{eqnarray*}
2089: v_{1} &\in &\mathrm{span}\left( L_{1}^{\ast }\right) \cap (\sigma +s)\text{,
2090: }v_{2}\in \mathrm{span}\left( \epsilon ^{a+b}L_{1}^{\ast }\right) \cap
2091: (\sigma +s)\text{, }v_{3}\in \mathrm{span}\left( \epsilon ^{a}jL_{1}^{\ast
2092: }\right) \cap (\sigma +s)\text{, } \\
2093: v_{4} &\in &\mathrm{span}\left( \epsilon ^{2a+b}jL_{1}^{\ast }\right) \cap
2094: (\sigma +s)\text{, }w\in \mathrm{span}\left( L_{2}^{\ast }\right) \cap
2095: (\sigma +s).
2096: \end{eqnarray*}%
2097: If the translation is small enough these points will remain in the interion
2098: of the simplex $\sigma +s$. Now the tiresome, but necessary part. The points
2099: are given by%
2100: \begin{eqnarray*}
2101: {\small v}_{1} &{\small =}&\tfrac{\alpha _{1}}{n}{\small e}_{a}{\small +}%
2102: \tfrac{\beta _{1}}{n}{\small e}_{a+1}{\small +}\tfrac{\gamma _{1}}{n}{\small %
2103: e}_{2a+b}{\small +}\tfrac{\delta _{1}}{n}{\small e}_{2a+b+1}{\small -}\tfrac{%
2104: 1}{n}\sum_{i=1}^{n}{\small e}_{i}{\small +}\sum_{i=1}^{n}{\small \xi }_{i}%
2105: {\small e}_{i}, \\
2106: {\small \alpha }_{1} &{\small =}&{\small a-n}\sum_{1}^{a}{\small \xi }_{i}%
2107: {\small ,~\gamma _{1}=a+b-n\sum_{a+1}^{2a+b}\xi _{i}-\beta
2108: _{1},~\delta
2109: _{1}=b-n}\sum_{2a+b+1}^{n}{\small \xi }_{i}, \\
2110: {\small \beta }_{1} &{\small =}&\tfrac{{\small -n}}{a+b+1}\left(
2111: \sum_{2a+b+2}^{n}{\small \xi }_{i}+{\small \xi }_{n}-{\small \xi }_{a+b}+%
2112: {\small \xi }_{a+1}{\small -}\left( {\small a+b}\right) \left(
2113: \sum_{2}^{a-1}{\small \xi }_{i}{\small -}\sum_{a+1}^{2a+b-1}{\small \xi }%
2114: _{i}+{\small \xi }_{a+b+1}\right) \right) {\small +b}
2115: \end{eqnarray*}%
2116: \begin{eqnarray*}
2117: {\small v}_{2} &{\small =}&\tfrac{\alpha _{2}}{n}{\small e}_{a}{\small +}%
2118: \tfrac{\beta _{2}}{n}{\small e}_{a+1}{\small +}\tfrac{\gamma _{2}}{n}{\small %
2119: e}_{2a+b}{\small +}\tfrac{\delta _{2}}{n}{\small e}_{2a+b+1}{\small -}\tfrac{%
2120: 1}{n}\sum_{i=1}^{n}{\small e}_{i}{\small +}\sum_{i=1}^{n}{\small \xi }_{i}%
2121: {\small e}_{i}, \\
2122: {\small \alpha }_{2} &{\small =}&{\small
2123: a+b-n}\sum_{2a+b+1}^{n}{\small \xi
2124: }_{i}{\small -n}\sum_{i=1}^{a}{\small \xi }_{i}-{\small \delta }_{2},~%
2125: {\small \beta }_{2}{\small =b-n}\sum_{a+1}^{a+b}{\small \xi }_{i}{\small %
2126: ,~\gamma }_{2}{\small =a-n}\sum_{a+b+1}^{2a+b}{\small \xi }_{i}, \\
2127: {\small \delta }_{2} &{\small =}&\tfrac{{\small -n}}{a+b+1}\left(
2128: \sum_{a+2}^{a+b}{\small \xi }_{i}{\small -\xi }_{n}{\small +\xi }_{a+b}%
2129: {\small +\xi }_{2a+b+1}+\left( {\small a+b}\right) \left( \sum_{2a+b+1}^{n}%
2130: {\small \xi }_{i}{\small +}\sum_{2}^{a-1}{\small \xi }_{i}{\small -}%
2131: \sum_{a+b+2}^{2a+b-1}{\small \xi }_{i}\right) \right) {\small +b;}
2132: \end{eqnarray*}
2133:
2134: \begin{eqnarray*}
2135: {\small v}_{3} &{\small =}&\tfrac{\alpha _{3}}{n}{\small e}_{a}{\small +}%
2136: \tfrac{\beta _{3}}{n}{\small e}_{a+1}{\small +}\tfrac{\gamma _{3}}{n}{\small %
2137: e}_{2a+b}{\small +}\tfrac{\delta _{3}}{n}{\small e}_{2a+b+1}{\small -}\tfrac{%
2138: 1}{n}\sum_{i=1}^{n}{\small e}_{i}{\small +}\sum_{i=1}^{n}{\small \xi }_{i}%
2139: {\small e}_{i}, \\
2140: {\small \alpha }_{3} &{\small =}&{\small a-n}\sum_{1}^{a}{\small \xi }_{i},~%
2141: {\small \beta }_{3}{\small =b-n}\sum_{a+1}^{a+b}{\small \xi }_{i}{\small %
2142: ,~\gamma }_{3}{\small =a+b-n}\sum_{a+b+1}^{n}{\small \xi }_{i}{\small %
2143: -\delta }_{3} \\
2144: {\small \delta }_{3} &{\small =}&{\small \tfrac{-n}{a+b-1}\left(
2145: (a+b)\left( \sum_{a+b+2}^{n}\xi _{i}-\sum_{2}^{a-1}\xi _{i}-\xi
2146: _{2a+b}\right) -\sum_{a+2}^{a+b}\xi _{i}-\xi _{2a+b+1}-\xi
2147: _{a+b}+\xi _{n}\right) +b;}
2148: \end{eqnarray*}
2149:
2150: \begin{eqnarray*}
2151: {\small v}_{4} &{\small =}&\tfrac{\alpha _{4}}{n}{\small e}_{a}{\small +}%
2152: \tfrac{\beta _{4}}{n}{\small e}_{a+1}{\small +}\tfrac{\gamma _{4}}{n}{\small %
2153: e}_{2a+b}{\small +}\tfrac{\delta _{4}}{n}{\small e}_{2a+b+1}{\small -}\tfrac{%
2154: 1}{n}\sum_{i=1}^{n}{\small e}_{i}{\small +}\sum_{i=1}^{n}{\small \xi }_{i}%
2155: {\small e}_{i}, \\
2156: {\small \alpha }_{4} &{\small =}&{\small a+b-n}\sum_{1}^{a+b}{\small \xi }%
2157: _{i}{\small -\beta }_{4}{\small ,}~{\small \delta
2158: _{4}=b-n\sum_{2a+b+1}^{n}\xi _{i},}~{\small \gamma _{4}=a-n}%
2159: \sum_{a+b+1}^{2a+b}{\small \xi }_{i} \\
2160: {\small \beta }_{4} &{\small =}&\tfrac{n}{a+b-1}\left( {\small (a+b)}\left(
2161: \sum_{a+b+1}^{2a+b}{\small \xi }_{i}{\small -}\sum_{2}^{a+b}{\small \xi }%
2162: _{i}{\small +\xi }_{a}\right) -\sum_{2a+b+1}^{n-1}{\small \xi }_{i}{\small %
2163: -\xi }_{2a+b+1}{\small +\xi }_{a+1}{\small -\xi }_{a+b}\right) {\small +b;}
2164: \end{eqnarray*}
2165:
2166: \begin{eqnarray*}
2167: {\small w}_{4} &{\small =}&\tfrac{\alpha _{5}}{n}{\small e}_{a}{\small +}%
2168: \tfrac{\beta _{5}}{n}{\small e}_{a+1}{\small +}\tfrac{\gamma _{5}}{n}{\small %
2169: e}_{2a+b}{\small +}\tfrac{\delta _{5}}{n}{\small e}_{2a+b+1}{\small -}\tfrac{%
2170: 1}{n}\sum_{i=1}^{n}{\small e}_{i}{\small +}\sum_{i=1}^{n}{\small \xi }_{i}%
2171: {\small e}_{i}, \\
2172: {\small \alpha }_{5} &=&{\small a-n\sum_{1}^{a}\xi _{i},}~{\small \beta }%
2173: _{4}{\small =b-n\sum_{a+1}^{a+b}\varepsilon _{i},~\gamma _{5}=a-n}%
2174: \sum_{a+b+1}^{2a+b}{\small \xi }_{i},~{\small \delta
2175: _{5}=b-n\sum_{2a+b+1}^{n}\xi _{i}}\text{.}
2176: \end{eqnarray*}
2177:
2178: \noindent Now we will prove the following equivalences%
2179: \begin{eqnarray*}
2180: &&\left( v_{1}\in K^{+}\Leftrightarrow v_{2}\in K^{+}\right) ~\vee ~\left(
2181: v_{1}\in K^{-}\Leftrightarrow v_{2}\in K^{-}\right) \\
2182: &&\left( v_{3}\in \epsilon ^{a}jK^{+}\Leftrightarrow v_{4}\in \epsilon
2183: ^{a}jK^{+}\right) ~\vee ~\left( v_{3}\in \epsilon ^{a}jK^{-}\Leftrightarrow
2184: v_{4}\in \epsilon ^{a}jK^{-}\right)
2185: \end{eqnarray*}%
2186: which have the following consequence
2187: \begin{equation}
2188: \left( v_{1}\in L_{1}^{\ast }\Leftrightarrow v_{2}\notin \epsilon
2189: ^{a+b}L_{1}^{\ast }\right) \text{ and }\left( v_{3}\in \epsilon
2190: ^{a}jL_{1}^{\ast }\Leftrightarrow v_{4}\notin \epsilon ^{2a+b}jL_{1}^{\ast
2191: }\right) \label{equvalencije}
2192: \end{equation}%
2193: and therefore the equality (\ref{Broken-point}).stands.
2194:
2195: \noindent In order to prove equivalences (\ref{equvalencije}) we evaluate $%
2196: v_{1}$, $v_{2}$, $v_{3}$ and $v_{4}$\ on linear forms $x_{1}+...+x_{a+b}$, $%
2197: x_{a+b+1}+...+x_{n}$, $x_{2a+b+1}+..+x_{n}+x_{1}+..+x_{a}$ and $%
2198: x_{a+1}+..+x_{2a+b}$, respectively. It can be calculated that
2199: \begin{eqnarray*}
2200: &&%
2201: \begin{array}{ll}
2202: {\small x}_{1}{\small +..+x}_{a+b}{\small ~|}_{v_{1}} & {\small =x}_{a+1}%
2203: {\small +..+x}_{a+b}{\small ~|}_{v_{1}}{\small =}\tfrac{\beta _{1}}{n}%
2204: {\small -}\tfrac{b}{n}{\small +}\sum_{a+1}^{a+b}{\small \xi }_{i} \\
2205: & {\small =}\tfrac{a+b}{a+b+1}\sum_{2}^{a-1}{\small \xi }_{i}{\small +}%
2206: \tfrac{1}{a+b+1}\sum_{a+2}^{a+b-1}{\small \xi }_{i}{\small +}\tfrac{2}{a+b+1%
2207: }{\small \xi }_{a+b} \\
2208: & {\small -}\tfrac{a+b}{a+b+1}\sum_{a+b+2}^{2a+b-1}{\small \xi }_{i}{\small %
2209: -}\tfrac{1}{a+b+1}\sum_{2a+b+2}^{n-1}{\small \xi }_{i}{\small -}\tfrac{2}{%
2210: a+b+1}{\small \xi }_{n}%
2211: \end{array}
2212: \\
2213: &&%
2214: \begin{array}{ll}
2215: {\small x}_{a+b+1}{\small +..+x}_{n}{\small ~|}_{v_{2}} & {\small =x}%
2216: _{2a+b+1}{\small +...+x}_{n}{\small ~|}_{v_{2}}{\small =}\tfrac{\delta _{2}}{%
2217: n}{\small -}\tfrac{b}{n}{\small +}\sum_{2a+b+1}^{n}{\small \xi }_{i} \\
2218: & ={\small -}\tfrac{a+b}{a+b+1}\sum_{i=2}^{a-1}{\small \xi }_{i}{\small -}%
2219: \tfrac{1}{a+b+1}\sum_{a+2}^{a+b-1}{\small \xi }_{i}{\small -}\tfrac{2}{a+b+1%
2220: }{\small \xi }_{a+b} \\
2221: & {\small +}\tfrac{a+b}{a+b+1}\sum_{i=a+b+2}^{2a+b-1}{\small \xi }_{i}%
2222: {\small +}\tfrac{1}{a+b+1}\sum_{2a+b+2}^{n-1}{\small \xi }_{i}{\small +}%
2223: \tfrac{2}{a+b+1}{\small \xi }_{n}%
2224: \end{array}
2225: \\
2226: &&%
2227: \begin{array}{ll}
2228: {\small x}_{2a+b+1}{\small +..+x}_{n}{\small +x}_{1}{\small +..+x}_{a}%
2229: {\small ~|}_{v_{3}} & {\small =x}_{2a+b+1}{\small +...+x}_{n}{\small ~|}%
2230: _{v_{2}}{\small =}\tfrac{\delta _{3}}{n}{\small -}\tfrac{b}{n}{\small +}%
2231: \sum_{2a+b+1}^{n}{\small \xi }_{i} \\
2232: & =\tfrac{a+b}{a+b-1}\sum_{2}^{a-1}{\small \xi }_{i}{\small +}\tfrac{1}{%
2233: a+b-1}\sum_{a+2}^{a+b-1}{\small \xi }_{i}{\small +}\tfrac{2}{a+b-1}{\small %
2234: \xi }_{a+b} \\
2235: & {\small -}\text{{\small $\tfrac{a+b}{a+b-1}\sum_{a+b+2}^{2a+b-1}\xi _{i}-%
2236: \tfrac{1}{a+b-1}\sum_{2a+b+2}^{n-1}\xi _{i}-\tfrac{2}{a+b-1}\xi _{n}$}}%
2237: \end{array}
2238: \\
2239: &&%
2240: \begin{array}{ll}
2241: x_{a+1}+..+x_{2a+b}{\small ~|}_{v_{4}} & {\small =x}_{a+1}{\small +...+x}%
2242: _{a+b}{\small ~|}_{v_{4}}{\small =}\tfrac{\beta _{4}}{n}{\small -}\tfrac{b}{n%
2243: }{\small +\sum_{a+1}^{a+b}\xi _{i}} \\
2244: & =-\tfrac{a+b}{a+b-1}\sum_{2}^{a-1}{\small \xi }_{i}{\small -}\tfrac{1}{%
2245: a+b-1}\sum_{a+2}^{a+b-1}{\small \xi }_{i}{\small -}\tfrac{2}{a+b-1}{\small %
2246: \xi }_{a+b} \\
2247: & {\small +}\text{{\small $\tfrac{a+b}{a+b-1}\sum_{a+b+2}^{2a+b-1}\xi _{i}+%
2248: \tfrac{1}{a+b-1}\sum_{2a+b+2}^{n-1}\xi _{i}+\tfrac{2}{a+b-1}\xi _{n}$}}%
2249: \end{array}%
2250: \end{eqnarray*}
2251:
2252: \noindent Thus, we have proved
2253: \begin{eqnarray*}
2254: \left( {\small a+b+1}\right) \left( {\small x}_{1}{\small +..+x}_{a+b}%
2255: {\small ~|}_{v_{1}}\right) &=&-\left( {\small a+b+1}\right) \left( {\small x}%
2256: _{a+b+1}{\small +...+x}_{n}{\small ~|}_{v_{2}}\right) \\
2257: &=&\left( {\small a+b-1}\right) \left( {\small x}_{2a+b+1}{\small +..+x}_{n}%
2258: {\small +x}_{1}{\small +..+x}_{a}{\small ~|}_{v_{3}}\right) \\
2259: &=&-\left( {\small a+b+1}\right) \left( x_{a+1}+..+x_{2a+b}{\small ~|}%
2260: _{v_{4}}\right)
2261: \end{eqnarray*}%
2262: which implies the equivalence (\ref{equvalencije}).
2263:
2264: \noindent It looks like we forgot about the space $L_{2}^{\ast }$ and point $%
2265: w$. The reason is that when ever we move the simplex $\sigma $ it will hit
2266: the space $L_{2}^{\ast }$. In respect of the direction we move $\sigma $ the
2267: point $w$ will be in one or in another connecting component of $L_{2}^{\ast
2268: }-I$. Thus, the broken point class $\left\Vert w\right\Vert $ appears in
2269: both equalities. Also, observe that we do not worry about the signs $\tau
2270: _{i}$ and $\mu _{i}$, they will not play any role in our calculation.
2271:
2272: %\FRAME{ftphFU}{5.2494in}{3.2984in}{0pt}{\Qcb{{\protect\small The dilemma of
2273: %Theorem \protect\ref{Lemma:Link}.}}}{\Qlb{Fig8}}{picture8.wmf}{\special%
2274: %{language "Scientific Word";type "GRAPHIC";maintain-aspect-ratio
2275: %TRUE;display "USEDEF";valid_file "F";width 5.2494in;height 3.2984in;depth
2276: %0pt;original-width 5.2215in;original-height 3.2711in;cropleft "0";croptop
2277: %"1";cropright "1";cropbottom "0";filename 'Picture8.wmf';file-properties
2278: %"XNPEU";}}
2279:
2280: \begin{figure}[htb]
2281: \centering
2282: \includegraphics[scale=0.8]{picture8.eps}
2283: \caption{Generating part of the Hasse diagram for coinvariants. }
2284: \label{fig:Fig8}
2285: \end{figure}
2286:
2287:
2288: \end{proof}
2289:
2290: \begin{conclusion}
2291: Let us observe that $\left\Vert v_{1}\right\Vert =\pm \epsilon
2292: ^{a}j\left\Vert v_{3}\right\Vert $ and $\left\Vert v_{2}\right\Vert =\pm
2293: \epsilon ^{a}j\left\Vert v_{4}\right\Vert $. Thus, since the class of our
2294: obstruction cocycle%
2295: \begin{equation*}
2296: c^{\prime }=2\left\Vert v\right\Vert =2\left( \tau _{1}\left\Vert
2297: v_{1}\right\Vert +\tau _{2}\left\Vert v_{3}\right\Vert +\tau _{3}\left\Vert
2298: y_{1}\right\Vert \right) =2\left( \mu _{1}\left\Vert v_{2}\right\Vert +\mu
2299: _{2}\left\Vert v_{4}\right\Vert +\mu _{3}\left\Vert y_{2}\right\Vert \right)
2300: \end{equation*}%
2301: lives in coinvariants $H_{2}(W_{n}\backslash \cup \mathcal{A}(\mathcal{J}%
2302: ,\alpha );\mathbb{Z})_{\mathbb{Q}_{4n}}\cong \mathbb{Z}_{2}\oplus \mathbb{Z}%
2303: _{4}$ we can shift our attention to the cohomological cocycle%
2304: \begin{equation*}
2305: c^{\prime \prime }=2\tau _{3}\left\Vert y_{1}\right\Vert .
2306: \end{equation*}
2307: \end{conclusion}
2308:
2309: \subsection{\label{sec:Step8}Step 8}
2310:
2311: \noindent Before proving that the cohomology class of the obstruction
2312: cocycle $c^{\prime \prime }=2\tau _{3}\left\Vert y_{1}\right\Vert $ is not
2313: zero, let us recall the nature of the Poincar\'{e} duality isomorphism $%
2314: \vartheta :H_{m-1}(\mathbb{R}^{n+m}\backslash \cup \mathcal{A};\mathbb{Z}%
2315: )\rightarrow \mathrm{Hom}(H_{n}(\cup \widehat{\mathcal{A}};\mathbb{Z}),%
2316: \mathbb{Z})$. As we have seen, for $t\in H_{n}(\cup \widehat{\mathcal{A}};%
2317: \mathbb{Z});\mathbb{Z})$ represented by a submanifold $T$ in $\cup \widehat{%
2318: \mathcal{A}}$, the image $\vartheta (\left\Vert x\right\Vert )$ of $t$ is
2319: given by the linking number%
2320: \begin{equation*}
2321: \vartheta (\left\Vert x\right\Vert )(t)=\mathrm{link}(S,T)\text{..}
2322: \end{equation*}
2323:
2324: \begin{theorem}
2325: As we have already denoted, let $l,k,h\in H_{n-4}(\cup \widehat{\mathcal{A}}(%
2326: \mathcal{J},\alpha );\mathbb{Z})$ be elements corresponding to spheres $%
2327: \widehat{L_{2}^{\ast }}$, $\widehat{L_{1}^{\ast }\cap L_{2}^{\ast }}\ast
2328: S^{\ast }$, $\widehat{\epsilon ^{a+b}L_{1}^{\ast }\cap L_{1}^{\ast }}\ast
2329: S^{\ast }$. Then there exists a basis of $H_{n-4}(\cup \widehat{\mathcal{A}}(%
2330: \mathcal{J},\alpha );\mathbb{Z})$ which contains $l$, $k$, and $h$ such that
2331: \begin{equation}
2332: \vartheta (\left\Vert y_{1}\right\Vert )(l)=\pm 1,\vartheta (\left\Vert
2333: y_{1}\right\Vert )(k)=\pm 1,\vartheta (\left\Vert y_{1}\right\Vert )(h)=0
2334: \label{linkovanje}
2335: \end{equation}%
2336: and for every other base element $d\in $ $H_{n-4}(\cup \widehat{\mathcal{A}}(%
2337: \mathcal{J},\alpha );\mathbb{Z})$, $\vartheta (\left\Vert y_{1}\right\Vert
2338: )(d)=0$. Thus,
2339:
2340: (1) $\vartheta (\left\Vert y_{1}\right\Vert )=\mu _{1}l+\mu _{2}k\in
2341: H_{n-4}(\cup \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})$ where $%
2342: \mu _{i}\in \{+1,-1\}$;
2343:
2344: (2) $c^{\prime \prime }=2k\neq 0$ in coinvariants $H_{2}(W_{n}\backslash
2345: \cup \mathcal{A}(\mathcal{J},\alpha );\mathbb{Z})_{\mathbb{Q}_{4n}}\cong
2346: H_{n}(\cup \widehat{\mathcal{A}};\mathbb{Z})_{\mathbb{Q}_{4n}}$.
2347:
2348: \noindent We abuse notation in (2) and (3) using the fact that $H_{n-4}(\cup
2349: \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z})$ is free ($\mathrm{Hom%
2350: }$ can be deleted) and omitting the notation of the cohomology class to
2351: simplify notation.
2352: \end{theorem}
2353:
2354: \begin{proof}
2355: The proof can be read from the figure \ref{fig:Fig8}. For example,
2356: the required basis of $H_{n-4}(\cup
2357: \widehat{\mathcal{A}}(\mathcal{J},\alpha );\mathbb{Z}) $ is
2358: composed of
2359:
2360: (A) the orbit of $l$,
2361:
2362: (B) $k$, $h$, $\epsilon ^{a+b}h$ and element $f$ corresponding to the sphere
2363: $\widehat{\epsilon ^{a+b}L_{1}^{\ast }\cap \epsilon ^{a}jL_{1}^{\ast }},$
2364:
2365: (C) anything you want outside the $I$ "neighbourhood", or precisely $%
2366: \epsilon k$, $\epsilon h$, $\epsilon ^{a+b+1}h$, $\epsilon f$, ..., $%
2367: \epsilon ^{a+b-1}k$, $\epsilon ^{a+b-1}h$, $\epsilon ^{n-1}h$, $\epsilon
2368: ^{a+b-1}f$.
2369:
2370: \noindent This basis satisfies the property (\ref{linkovanje}). Thus (1) and
2371: (2) follow directly with the knowing that $4k=0$ in coinvariants.
2372: \end{proof}
2373:
2374: \begin{conclusion}
2375: Since we proved that the cohomology class of the obstruction cocycle is not
2376: zero we are ready to sum up (remember Propositions \ref%
2377: {prop:PrelazakNaNovuGrupu} and \ref{prop:VezaProblem-Ekvivarijantan}):
2378: \begin{equation*}
2379: \begin{tabular}{lll}
2380: $\text{There is no }\mathbb{Q}_{4n}\text{-map }S^{3}\rightarrow \cup
2381: \mathcal{A}(\mathcal{J}\text{,}\alpha )\text{ }$ & $\Longrightarrow $ &
2382: There is no $\mathbb{Q}_{4n}$-map $S^{3}\rightarrow \cup \mathcal{A}$ \\
2383: & $\Longleftrightarrow $ & There is no $\mathbb{D}_{2n}$-map $V_{2}(\mathbb{R%
2384: }^{3})\rightarrow \cup \mathcal{A}$ \\
2385: & $\Longrightarrow $ & There is an $(\frac{a}{n},\frac{a+b}{n},\frac{b}{n})$%
2386: -partition.%
2387: \end{tabular}%
2388: \text{\ }
2389: \end{equation*}%
2390: Once more, we have just proved that for every $a,b>1$ and every two measures
2391: $\mu $ and $\nu $ on $S^{2}$, there exists an $(\frac{a}{n},\frac{a+b}{n},%
2392: \frac{b}{n})$-partition $(x;l_{1},l_{2},l_{3})$ of measures $\mu $ and $\nu $%
2393: .
2394: \end{conclusion}
2395:
2396: \subsection{The limit argument}
2397:
2398: \noindent The reasons for assuming nice properties for our measures will
2399: finally pop up. Let $\emph{S}\subseteq \mathbb{R}_{>0}^{3}$ be the space of
2400: all $3$-fan partitions, i.e. the space of all triples $\alpha =(a,b,c)$, $%
2401: a+b+c=1$ such that there exists an $\alpha $-partition of measures $\mu $
2402: and $\nu $ by a $3$-fan. Since our measures $\mu $ and $\nu $ are proper
2403: Borel probability measures, the space $\emph{S}$ is a closed subset of $%
2404: \mathbb{R}_{>0}^{3}$. All we did so far is proving the following inclusion
2405: \begin{equation*}
2406: \{(\tfrac{a}{n},\tfrac{a+b}{n},\tfrac{b}{n})\in \mathbb{Q}%
2407: _{>0}^{3}~|~2a+2b=n,~a,b\in \mathbb{Z}\}\subseteq \emph{S.}
2408: \end{equation*}%
2409: The fact that $\emph{S}$ is closed implies the inclusion%
2410: \begin{equation*}
2411: \{(a,b,c)\in \mathbb{R}_{>0}^{3}~|~a+b+c=1\}=\mathrm{cl}\{(\tfrac{a}{n},%
2412: \tfrac{a+b}{n},\tfrac{b}{n})~|~2a+2b=n,~a,b\in \mathbb{Z}\}\subseteq \emph{S}%
2413: \text{.}
2414: \end{equation*}%
2415: and therefore we are finally done.
2416:
2417: \begin{thebibliography}{99}
2418: \bibitem{Aki2000} J. Akiyama, A. Kaneko, M. Kano, G. Nakamura, E.
2419: Rivera-Campo, S. Tokunaga, and J.Urrutia, \textit{Radial perfect partitions
2420: of convex sets in the plane}. In Discrete and Computational Geometry (J.
2421: Akiyama et al. eds.), Lect. Notes Comput. Sci. 1763, pp. 1--13. Springer,
2422: Berlin 2000.
2423:
2424: \bibitem{BaMa2001} I.B\'{a}r\'{a}ny, J.Matou\v{s}ek, \textit{Simultaneous
2425: partitions of measures by }$k$\textit{-fans}, Discrete Comp. Geometry,
2426: 25\thinspace\ (2001), 317--334.
2427:
2428: \bibitem{BaMa2002} I.B\'{a}r\'{a}ny, J. Matou\v{s}ek, \textit{Equipartitions
2429: of two measures by a }$4$\textit{-fan}, Discrete Comp. Geometry 27 (2001),
2430: 317-334
2431:
2432: \bibitem{Bl-Vr-Ziv} P. Blagojevi\'{c}, S. Vre\'{c}ica, R. \v{Z}ivaljevi\'{c}%
2433: : \textit{Computational Topology of Equivariant Maps from Spheres to
2434: Complement of Arrangements}, arXiv:math.AT/0403161
2435:
2436: \bibitem{Brown} K.S. Brown, \textit{Cohomology of groups}, Springer-Verlag,
2437: New York, Berlin, 1982.
2438:
2439: \bibitem{CaEi} H. Cartan and S. Eilenberg, \textit{Homological Algebra}
2440: Princeton University Press, 1956.
2441:
2442: \bibitem{CoFl} P.E. Conner and E.E. Floyd, \textit{Differentiable periodic
2443: maps}, Springer-Verla, Berlin 1964.
2444:
2445: \bibitem{Dieck87} T. tom Dieck, \textit{Transformation groups}, de Gruyter
2446: Studies in Math. 8, Berlin, 1987.
2447:
2448: \bibitem{Mat94} J. Matou\v{s}ek, \textit{Topological methods in
2449: Combinatorics and Geometry}, Lecture notes, Prague 1994. (updated version,
2450: February 2002),
2451:
2452: \bibitem{Mu} J. R. Munkres, \textit{Elements of Algebraic Topology},
2453: Addison-Wesley, 1984
2454:
2455: \bibitem{VZ92} S. Vre\'{c}ica, R. \v{Z}ivaljevi\'{c}, \textit{The ham
2456: sandwich theorem revisited,} Israel J. Math. 78, 1992, pp. 21--32.
2457:
2458: \bibitem{VreZiv2002} S. Vre\'{c}ica, R. \v{Z}ivaljevi\'{c}, \textit{%
2459: Arrangements, equivariant maps and partitions of measures by }$4$\textit{%
2460: -fans}, (preprint).
2461:
2462: \bibitem{Wall} C.T.C. Wall, \textit{Surgery on Compact Manifolds}, Academic
2463: Press, 1970.
2464:
2465: \bibitem{elisat} R. \v{Z}ivaljevi\'{c}, \textit{Topological methods}, in CRC
2466: Handbook of Discrete and Computational Geometry, J.E. Goodman, J.O'Rourke,
2467: eds. CRC Press, Boca Raton 1997.
2468:
2469: \bibitem{T-V} R. \v{Z}ivaljevi\'{c}, \textit{The Tverberg--Vre\'{c}ica
2470: problem and the combinatorial geometry on vector bundles,} Israel J. Math.
2471: 111 (1999), 53--76.
2472:
2473: \bibitem{guide1} R. \v{Z}ivaljevi\'{c},\textit{\ User's guide to equivariant
2474: methods in combinatorics}, Publ. Inst. Math.Belgrade, 59(73), 1996, 114--130.
2475:
2476: \bibitem{guide2} R. \v{Z}ivaljevi\'{c}, \textit{User's guide to equivariant
2477: methods in combinatorics II}, Publ. Inst. Math. Belgrade, 64(78), 1998,
2478: 107--132.
2479:
2480: \bibitem{limun2002} R.T. \v{Z}ivaljevi\'{c}, \textit{Combinatorics and
2481: topology of partitions of spherical measures by }$2$\textit{\ and }$3$%
2482: \textit{\ fans, }preprint, (arXiv:math.CO/0203028 v2 4 Mar 2002).
2483:
2484: \bibitem{ZZ} G. M. Ziegler, R. T. \v{Z}ivaljevi\'{c}, Homotopy types of
2485: subspace arrangements via diagrams of spaces, \textit{Math. Ann.} \textbf{295%
2486: }:527-548, 1993
2487: \end{thebibliography}
2488:
2489: \end{document}
2490: