1: \documentclass[a4paper]{article}

2: \usepackage{amssymb,amsbsy,amsmath,amsfonts,amssymb,amscd}

3: \usepackage{latexsym}

4: \usepackage{geometry}

5: %\usepackage{showkeys}

6: % \usepackage{pdfsync}

7: \input{macros.tex}

8: 

9: \input xy

10: \xyoption{all}

11: 

12: 

13: \newcommand{\bysame}{\hspace{1.5cm}}

14: \newcommand{\ra}{\rightarrow}

15: \renewcommand{\H}{{\mathrm H}}

16: \newcommand{\cD}{{\cal D}}

17: \newcommand{\cE}{{\cal E}}

18: \newcommand{\cF}{{\cal  F}}

19: \newcommand{\cG}{{\cal G}}

20: \newcommand{\cK}{{\cal K}}

21: \newcommand{\cS}{{\cal S}}

22: \newcommand{\cT}{{\cal T}}

23: \newcommand{\cN}{{\cal N}}

24: \newcommand{\cH}{{\cal H}}

25: \newcommand{\cR}{{\cal R}}

26: 

27: \newcommand{\tGam}{{\tilde{\Gamma}}} 

28: \newcommand{\tA}{{\tilde{A}}}

29: \newcommand{\tF}{{\tilde{F}}}

30: 

31: \newcommand{\Ga}{{\bb G}_a}

32: \newcommand{\Gm}{{\bb G}_m}

33: 

34: 

35: % Fritz compatibility macros 

36: 

37: \def\normalbaselines{\baselineskip20pt\lineskip3pt\lineskiplimit3pt }

38: \def\mapright#1{\smash{\mathop{\longrightarrow}\limits^{#1}}}

39: \def\mapdown#1{\Big\downarrow\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}}

40: 

41: 

42: \def\g{{\Gamma}}

43: \def\a{{\Aut}}

44: \hsize= 15cm

45: \vsize= 20.5cm

46: \font\Bbb=msbm10

47: \font\Bbbb=msbm7

48: \def\I{{\bf I}}

49: \def\BZ{{\hbox{\Bbb Z}}}

50: \def\BL{{\hbox{\Bbb L}}}

51: \def\BH{{\hbox{\Bbb H}}}

52: \def\BK{{\hbox{\Bbb K}}}

53: \def\BP{{\hbox{\Bbb P}}}

54: \def\BF{{\hbox{\Bbb F}}}

55: \def\BG{{\hbox{\Bbb G}}}

56: \def\BN{\hbox{\Bbb N}}

57: \def\BR{{\hbox{\Bbb R}}}

58: \def\BC{{\hbox{\Bbb C}}}

59: \def\BQ{{\hbox{\Bbb Q}}}

60: \def\BT{\hbox{\Bbb T}}

61: \def\A{{\hbox{\rm Aut}}}

62: \def\Ro{{\bar R_{\rm old}}}

63: \def\A{{\rm Aut}}     

64: 

65: \def\op{{\rm op}}

66: \def\R{{\bf R}}

67: \def\Ad{{\rm Ad}}

68: \def\g{{\Gamma}}

69: \def\u{{{\mathsf R}_{\rm u}(G)}}

70: \def\s{{{\mathsf R}_{\rm u}(G_1)}}

71: \def\t{{{\mathsf R}_{\rm u}(G_2)}}

72: 

73: \baselineskip= 13pt

74: \parindent= 0pt

75: \parskip= 3pt plus 1pt minus 0.5pt

76: 

77: \def \Box{\lower .1 em

78:           \vbox{\hrule \hbox{\vrule \hskip .6 em \vrule height .6 em} \hrule}}

79: \def \Mid{\quad \vrule \quad}

80: \font \msbm= msbm10

81: \def \rfish{\mathbin {\hbox {\msbm \char'157}}}

82: 

83: % End of Fritz compatibility 

84: 

85: 

86: 

87: \title{ Linear Representations of the Automorphism Group of a 

88: Free Group}

89: 

90: \author{Fritz Grunewald\\

91: Mathematisches Institut \\ 

92: Heinrich-Heine-Universit\"at\\ 

93: D-40225 D\"usseldorf\\

94: e-mail: fritz@math.uni-duesseldorf.de 

95: \and 

96: Alexander Lubotzky\\

97: Einstein Institute of Mathematics\\

98: The Hebrew University of Jerusalem\\

99: Jerusalem, 91904, Israel\\

100: e-mail: alexlub@math.huji.ac.il}

101: 

102: 

103: 

104: \begin{document}

105: \maketitle 

106: 

107: 

108: \begin{abstract} 

109: Let $F_n$ be the free group on $n\ge 2$ elements and $\A(F_n)$ its group of

110: automorphisms. In this paper we present a rich collection of 

111: linear representations of

112: $\A(F_n)$ arising through the action of finite index subgroups of it on

113: relation modules of finite quotient groups of $F_n$. We show 

114: (under certain conditions) that the images

115: of our representations are arithmetic groups. 

116: \newline 

117: 

118: \noindent 

119: 2000 Mathematics Subject Classification: Primary 20F28, 20E05; \\

120: Secondary 20E36, 20F34, 20G05  

121: \end{abstract}

122: 

123: \medskip

124: 

125: 

126: \bigskip

127: \bigskip

128: 

129: 

130: \tableofcontents 

131: \medskip

132: 

133: \bigskip

134: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

135: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

136:  

137: 

138: 

139: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

140: \section{Introduction} 

141: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

142: 

143: Let $F_n$ be the free group on $n\ge 2$ elements and $\A(F_n)$ its group of

144: automorphisms. The latter is a much studied group, but very little seems to be

145: known about its (finite dimensional) complex representation theory

146: (see for example \cite{FP}, \cite{DJ}, \cite{PORA}, \cite{RA}, \cite{LP}). 

147: In fact,

148: as far as we know the only representations studied (at least when $n\ge 3$) 

149: are:

150: \begin{itemize}

151: \item The representation

152: \begin{equation}\label{furep}

153: \rho_1 : \A(F_n)\to \A(F_n/F'_n)\cong \GL(n,\BZ) 

154: \end{equation}

155: where $F_n'$ is the commutator subgroup of $F_n$.

156: \item The representations factoring through the homomorphisms

157: $$\rho_i : \A(F_n)\to \A(F_n/F^{(i+1)}_n)$$

158: where $F^{(i+1)}_n$ ($i=0,1,\ldots$) stands for the lower 

159: central series of $F_n$.  

160: \item Representations with finite image. 

161: \end{itemize}

162: What is common to all of the above mentioned reprentations $\rho$ is that if we

163: denote by ${\cal H}:=\overline{\rho(\A(F_n))}$ the Zariski closure of 

164: $\rho(\A(F_n))$, then the semi-simple part of the connected component 

165: ${\cal H}^\circ$ of ${\cal H}$ is either trivial or $\SL(n,\BC)$. Moreover,

166: $\rho({\rm IA}(F_n))$ is always virtually solvable for these representations,

167: where ${\rm IA}(F_n)$ is the kernel of $\rho_1$.

168: 

169: In this paper we will show that the representation theory of $\A(F_n)$ is in

170: fact much richer. Our main Theorem (Theorem \ref{teo} below) 

171: implies for example 

172: \begin{theorem}\label{teo4}

173: Let $n\ge 2$, $k\ge 1$, $h_1<\ldots <h_k$, $m_1,\ldots ,m_k$ be 

174: natural numbers.  Let $\BQ(\zeta_{m_i})$ be the field of 

175: $m_i$-th roots of unity and

176: $\BZ(\zeta_{m_i})$ its ring of integers. There is a 

177: subgroup $\Gamma\le \A(F_n)$ of finite index and a representation 

178: $$

179: \rho : \Gamma \to \prod_{i=1}^k\SL((n-1)h_i,\BQ(\zeta_{m_i}))^{m_i} 

180: $$

181: such that $\rho(\Gamma)$ is commensurable with 

182: $\prod_{i=1}^k\SL((n-1)h_i,\BZ(\zeta_{m_i}))^{m_i}$. 

183: \end{theorem}

184: Theorem \ref{teo4} shows that $\A(F_n)$ has a representation $\rho$ such that

185: the connected component of $\overline{\rho(\A(F_n))}$ is isomorphic to 

186: $$\prod_{i=1}^k\SL((n-1)h_i,\BC)^{m_i}$$

187: as above. In fact the proof shows that this is even true for 

188: $\overline{\rho({\rm IA}(F_n))}$ and in particular the latter 

189: is very far from being virtually solvable (see Section \ref{Concl1} for more). 

190: 

191: Specifying $m_1=\ldots =m_k=1$ in Theorem \ref{teo4}, we deduce

192: \begin{corollary}\label{coro4}

193: Let $n\ge 2$, $k\ge 1$ be natural numbers.  There is a 

194: subgroup $\Gamma\le \A(F_n)$ of finite index and a representation 

195: $$

196: \rho : \Gamma \to \prod_{i=1}^k\SL((n-1)i,\BZ) 

197: $$

198: such that $\rho(\Gamma)$ is of finite index in $\prod_{i=1}^k\SL((n-1)i,\BZ)$. 

199: \end{corollary}

200: Specialising Theorem \ref{teo4} even further to 

201: $n=3$, $k=1$, $h_1=m_1=1$ we get that

202: $\A(F_3)$ has a subgroup of finite index which can be mapped onto a subgroup

203: of finite index in $\SL(2,\BZ)$. This implies  

204: \begin{corollary}\label{kaz}

205: The automorphism group $\A(F_3)$ is large, that is it has a subgroup of finite

206: index which can be mapped onto a free nonabelian group. In particular

207: $\A(F_3)$ does not have Kazdhan's property (T).

208: \end{corollary}

209: The corollary is easy for $\A(F_2)$, see the discussion in Section 

210: \ref{Proofs}, but it is a

211: well known open problem for larger $n$. Our solution for 

212: $n=3$ does not indicate what should be the answer for $n>3$. 

213: 

214: We shall describe now how the representations to be considered here arise.  

215: Let $G$ be a finite group and $\pi : F_n\to G$ a surjective homomorphism

216: of the free group $F_n$ onto $G$. Let $R$ be the kernel of 

217: $\pi$. We define  

218: \begin{equation}

219: \Gamma(R):=\{\, \varphi\in \A(F_n)\Mid \varphi(R)=R \, \}

220: \end{equation}

221: and 

222: \begin{equation}\label{I1}

223: \Gamma(G,\pi):=\{\, \varphi\in \Gamma(R) \Mid  

224: \varphi\ {\rm induces\ the\ identity\ on}\ F_n/R \, \}.

225: \end{equation}

226: These are subgroups of finite index in $\A(F_n)$. We let

227: \begin{equation}

228: \bar R:=R/R'

229: \end{equation}

230: be the relation module of the presentation $\pi : F_n\to G$. 

231: The action of $F_n$ on $R$ by conjugation leads 

232: to an action of $G$ on $\bar R$.

233: 

234: The structure of $\bar R$ as a $G$-module is described by 

235: Gasch\"utz' theory (see Section \ref{suga} and the references therein). 

236: It says in particular that 

237: \begin{equation}\label{G1}

238: \BQ\otimes_{\bb Z} \bar R\cong \BQ\oplus \BQ[G]^{n-1}

239: \end{equation}

240: as a module over the rational group ring $\BQ[G]$ (see also (\ref{gasch})).

241: Our work can be described as an equivariant Gasch\"utz' theory where we try to

242: describe the action of $\Gamma(G,\pi)$ on $\bar R$: Every $\varphi\in

243: \Gamma(R)$ induces a linear automorphism $\bar \varphi$ 

244: of $\bar R$ and (as an application of the theorem of Gasch\"utz) 

245: $\Gamma(G,\pi)$ consists exactly of those elements  $\varphi\in\Gamma(R)$ 

246: for which $\bar\varphi$ is $G$-equivariant.

247: Already now we can see from (\ref{G1}) why the number $n-1$ plays such an

248: important role in Theorem \ref{teo4}.

249: 

250: To be more precise: 

251: The relation module $\bar R$ is a finitely generated 

252: free abelian group, let $t$ be its $\BZ$-rank. Then

253: \begin{equation}\label{I4}

254: {\cal G}_{G,\pi}:=\Aut_G(\BC\otimes_{\bb Z} \bar R)\le \GL(t,\BC)

255: \end{equation}

256: is a $\BQ$-defined algebraic subgroup of $\GL(t,\BC)$, in fact 

257: ${\cal G}_{G,\pi}$ is the centraliser of the group $G$ acting on 

258: $\BC\otimes_\BZ \bar R$ through matrices with rational entries. Let

259: \begin{equation}

260: {\cal G}_{G,\pi}^1 \le \SL(t,\BC)

261: \end{equation}

262: be the kernel of all $\BQ$-defined homomorphisms from the complex algebraic

263: group $\BC\otimes{\cal G}_{G,\pi}$ to the multiplicative group. This is a

264: $\BQ$-defined subgroup of ${\cal G}_{G,\pi}$. We shall describe it in more

265: detail in Section \ref{suga}.

266: We define 

267: \begin{equation}

268: {\cal G}_{G,\pi}(\BZ):=\{\, \phi\in {\cal G}_{G,\pi}\Mid \phi(\bar

269: R)=\bar R\,\}.

270: \end{equation}

271: This is an arithmetic subgroup of the $\BQ$-defined algebraic group

272: ${\cal G}_{G,\pi}$ which contains the arithmetic subgroup

273: \begin{equation}

274: {\cal G}^1_{G,\pi}(\BZ):=\{\, \phi\in {\cal G}^1_{G,\pi}\Mid \phi(\bar

275: R)=\bar R\,\}

276: \end{equation}

277: of ${\cal G}^1_{G,\pi}$.

278: Following the definitions we obtain an 

279: integral linear representation

280: \begin{equation}\label{I5}  

281: \rho_{G,\pi} : \Gamma(G,\pi)\to {\cal G}_{G,\pi}(\BZ),\qquad

282: \rho_{G,\pi}(\varphi)=\bar\varphi\quad (\varphi\in \Gamma(G,\pi)).

283: \end{equation}

284: Examples show (see Section \ref{Choose}) that in general $\rho_{G,\pi}$ is not

285: onto, neither is its image of finite index in ${\cal G}_{R,\pi}(\BZ)$. 

286: However, our main result shows that, at least for redundant presentations, the

287: image of $\rho_{G,\pi}$ captures the whole semi-simple part of 

288: ${\cal G}_{R,\pi}(\BZ)$. More precisely:

289: \begin{theorem}\label{teo} Assume $n$ is a natural number with $n\ge 4$.

290: Let $\pi : F_n\to G$ be a redundant presentation of the finite group $G$. Then

291: $\rho_{G,\pi}(\Gamma(G,\pi))\cap {\cal G}^1_{G,\pi}$ is 

292: of finite index in the arithmetic group ${\cal G}_{G,\pi}^1(\BZ)$.

293: \end{theorem}

294: A presentation $\pi : F_n\to G$ is called {\it redundant} if there is a basis

295: $x_1,\ldots ,x_n$ of $F_n$ such that $\pi(x_n)=1$. This is equivalent to the

296: kernel of $\pi$ containing at least one primitive element 

297: (that is a member of a basis) of $F_n$.

298: 

299: In general we do not know whether 

300: $\rho_{G,\pi}(\Gamma(G,\pi))\cap {\cal G}_{G,\pi}^1(\BZ)$ is of finite index

301: in $\rho_{G,\pi}(\Gamma(G,\pi))$. Section \ref{SL} contains criteria under

302: which this is going to happen.

303: 

304: Choosing the finite group in various ways we obtain a rich variety of 

305: linear representations of subgroups of finite index in $\A(F_n)$ with

306: arithmetic image groups. Let us start with some simple examples.

307: 

308: {\it Example 1:} Let $G=C_2$ be the group of order $2$ and $\pi :F_n\to G$

309: be any surjective homomorphism. Since we generally have assumed $n\ge 2$, the

310: representation is redundant. As an abelian group we have 

311: $\bar R\cong \BZ^{2(n-1)+1}$. Also $\BQ\otimes_{\bb Z} \bar R$ 

312: can be decomposed into

313: the $+1$-eigenspace of a generator of $G$ and the corresponding

314: $-1$-eigenspace. We obtain a decomposition 

315: $\BQ\otimes_{\bb Z} \bar R=\BQ^n\oplus \BQ^{n-1}$ 

316: which has to be respected by   

317: $\rho_{G,\pi}(\Gamma(G,\pi))$. Thus we obtain a representation

318: $$

319: \rho_{G,\pi} : \Gamma(G,\pi) \to \GL(n,\BQ)\times \GL(n-1,\BQ)

320: $$

321: which we show to have an image commensurable with $\SL(n,\BZ)\times

322: \SL(n-1,\BZ)$.

323: 

324: The first factor is the familiar one arising from the representation

325: (\ref{furep}). But the second is somewhat less expected. It already gives rise

326: to a representation which is suitable for proving Corollary \ref{kaz}.

327: 

328: {\it Example 2:} Let $p$ be a prime number, $\zeta_p\in\BC$ a non trivial

329: $p$-th root of unity. Let $K:=\BQ(\zeta_p)$ be the $p$-th cyclotomic field

330: and $\BZ(\zeta_p)$ its ring of integers. 

331: Taking $G=C_p$ the cyclic group of order 

332: $p$ we have $\bar R\cong \BZ^{p(n-1)+1}$. Now $p(n-1)+1=(p-1)(n-1)+n$ and we  

333: get a representation

334: $$

335: \rho_{G,\pi} : \Gamma(G,\pi) \to \GL(n,\BQ)\times \GL(n-1,K)

336: $$

337: which we show to have an image commensurable with  $SL(n,\BZ)\times

338: SL(n-1,\BZ(\zeta_p))$. This example is treated in Section \ref{Choose} along

339: with the case of a general cyclic group.

340: 

341: Let $G$ be now an arbitrary finite group and $\pi: F_n\to G$ a redundant

342: epimorphism. Decompose the group algebra $\BQ[G]$ 

343: as

344: $$\BQ[G]=\BQ\oplus \prod_{i=2}^\ell M(h_i,D_i)$$

345: where the $D_i$ are division algebras. Theorem \ref{teo} implies that there is 

346: a homomorphism $\rho$ from a finite index subgroup $\Gamma$ of $\A(F_n)$ into

347: a subgroup of $\prod_{i=2}^\ell\GL((n-1)h_i,D_i)$ where the image contains an

348: arithmetic subgroup of 

349: $$\Lambda:=\prod_{i=2}^\ell\SL((n-1)h_i,{\cal R}_i)$$ where 

350: ${\cal R}_i$ is an order in $D_i$. In general we do not know whether

351: $\rho(\Gamma)\cap\Lambda$ is of finite index in $\rho(\Gamma)$. It will be

352: especially interesting if it is not! In this case  

353: $\rho(\Gamma)/(\rho(\Gamma)\cap\Lambda)$ would be an infinite abelian group and

354: this would imply that $\A(F_n)$ does not have Kazdhan property (T). We have

355: some results in the opposite direction showing, for example, that if $G$ is

356: metabelian then $\rho(\Gamma)$ is in fact commensurable to $\Lambda$. 

357: 

358: In any event, whenever $\rho(\Gamma)\cap\Lambda$ is of finite index or not, it

359: becomes of finite index after we divide by the center of 

360: $\prod_{i=2}^\ell\GL((n-1)h_i,D_i)$. This shows that in all cases an

361: arithmetic group commensurable to $\Lambda$ is a quotient of some finite index

362: subgroup of $\A(F_n)$. Thus Theorem \ref{teo} provides a very rich class of

363: arithmetic subgroups which are quotients of (finite index subgroups) of

364: $\A(F_n)$. Theorem \ref{teo4} is obtained from the main result by some special

365: choices of the finite group $G$ and by an application of Margulis

366: super-rigidity (see Section \ref{Proofs}).

367: 

368: The present paper opens up a large number of interesting problems. Among

369: others it shows that the classical Torelli subgroup (i.e. ${\rm IA}(F_n)$) 

370: is just a first example in a series of infinitely many. 

371: The kernel of $\rho_{G,\pi}$, for any (redundant) homomorphism 

372: $\pi:F_n\to G$ of $F_n$ to a finite group $G$, 

373: can be considered as a natural generalisation of

374: ${\rm IA}(F_n)$ which corresponds to the case where $G$ is the trivial group.

375: Just as ${\rm IA}(F_n)$, the kernel of $\rho_{G,\pi}$ is residually

376: torsion-free nilpotent and, at least when $\pi$ is redundant,

377: the image of $\rho_{G,\pi}$ is an arithmetic group hence

378: is finitely presented, which implies that   

379: ${\rm ker}(\rho_{G,\pi})$ is a finitely generated normal subgroup.  

380: Is ${\rm ker}(\rho_{G,\pi})$ a finitely generated group? In case of the

381: classical Torelli group this is a result of Magnus (see \cite{MKS}).

382: 

383: Let us now describe the method of the proof of the main result and the layout

384: of the paper: In Section \ref{Two}, we describe in detail the rational

385: relation module $\BQ\otimes_{\bb Z}\bar R$ together 

386: with ${\cal G}_{G,\pi}$ which

387: is the centraliser of $G$ when acting on the relation module. In Section 

388: \ref{Drei}, we describe the homomorphism from $\Gamma (G,\pi)$ to 

389: ${\cal G}_{G,\pi}$, which is given by the action of $\Gamma (G,\pi)$

390: on $\BQ\otimes_\BZ\bar R$. We compute the action of various elements of    

391: $\Gamma (G,\pi)$ on $\BQ\otimes_{\bb Z}\bar R$ in Section \ref{Vier} 

392: using results

393: on Fox calculus from Section \ref{Drei}. A careful choice of elements from 

394: $\Gamma (G,\pi)$ allows us to produce sufficiently many unipotent elements

395: of ${\cal G}_{G,\pi}(\BZ)$ to appeal to results of Vaserstein \cite{Vas} 

396: (see also Venkataramana \cite{Venk} for generalisations) to deduce 

397: in Section \ref{Fuenf} that these

398: elements generate a finite index subgroup. Section \ref{Choose} is devoted to

399: a detailed study of the case when $G$ is a cyclic group. In this case we give

400: fairly complete results. Some of them are used in Section \ref{Proofs} for the

401: proof of Theorem \ref{teo4}. 

402: In Section \ref{SL} we discuss the question

403: whether the image of $\Gamma (G,\pi)$ is commensurable with 

404: ${\cal G}_{G,\pi}^1(\BZ)$ or an infinite extension of it. We solve it

405: for cyclic groups and also for metabelian groups. 

406: Some remarks, computational

407: results and open problems are given in Section \ref{Concl}.

408: 

409: In a subsequent paper \cite{GLu} we apply similar methods to study linear

410: representations of the mapping class group.

411: 

412: {\it Acknowledgement:} The authors are grateful to A. Rapinchuk for a useful

413: conversation which led to the results about abelian groups in Section

414: \ref{SL}. They thank the Minerva-Landau Center for its support and especially the 

415: Institute for Advanced Study in Princeton, where this

416: work was done, for its hospitality in the academic year 2005/2006. The second

417: author thanks the NSF, the Monell Foundation 

418: and the Ellentuck Fund for support. 

419: 

420: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

421: \section{Relation modules and representations of subgroups of finite index in

422: $\A(F_n)$ }\label{Two}

423: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

424: 

425: In this section we collect some results on the structure of the relation

426: modules of finite groups.  We also consider 

427: integrality properties of the linear reprentations of

428: subgroups of finite index in $\A(F_n)$ introduced in the introduction.

429: 

430: 

431: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

432: \subsection{The rational group ring}\label{sura}

433: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

434: 

435: Here $G$ is a finite group and $\BQ[G]$ is its rational group ring. 

436: Our basic reference for the structural results we need is the book of Curtis

437: and Reiner \cite{CR} (see also \cite{S}), in particular 

438: Sections IV and XI therein.  

439: 

440: We start to set up some notation. 

441: Let $N$ be an irreducible left $\BQ[G]$-module. We write

442: $$

443: {\bf I}_N(M)

444: $$

445: for the $N$-isotypic component of the left $\BQ[G]$-module $M$. We also choose

446: representatives $N_1,\ldots , N_\ell$ for the isomorphism classes of 

447: irreducible left $\BQ[G]$-modules and abreviate 

448: $$

449: {\bf I}_i(M):={\bf I}_{N_i}(M)\qquad (i=1,\ldots,\ell)

450: $$ 

451: for a left $\BQ[G]$-module $M$. We take $N_1$ to be the one dimensional

452: trivial module.

453: 

454: Viewing $\BQ[G]$ as a left $\BQ[G]$-module we obtain a direct product

455: decomposition 

456: \begin{equation}\label{ri}

457: \BQ[G]=\prod_{i=1}^\ell\, {\bf I}_i(\BQ[G]).

458: \end{equation}

459: In this case every ${\bf I}_i(\BQ[G])$ is also a right submodule of 

460: $\BQ[G]$. Furthermore every ${\bf I}_i(\BQ[G])$ is an ideal of $\BQ[G]$ having

461: its own unity.

462: The decomposition

463: (\ref{ri}) is a decomposition of $\BQ[G]$ as a direct product of 

464: $\BQ$-algebras.

465: 

466: We let $P_i\in \BQ[G]$ ($i=1,\ldots,\ell$) be the right projector of 

467: $\BQ[G]$ onto ${\bf I}_i(\BQ[G])$, that is we have

468: \begin{itemize}

469: \item[{\bf RP}:]

470: $AP_i=A$ for all $A\in {\bf I}_i(\BQ[G])$       

471: and $AP_i=0$

472: if $A$ is in the direct product of the subrings

473: ${\bf I}_j(\BQ[G])$ with ${\bf I}_i(\BQ[G])$ excluded.

474: \end{itemize} 

475: Notice that we also have 

476: \begin{itemize}

477: \item[{\bf LP}:]

478: $P_iA=A$ for all $A\in {\bf I}_i(\BQ[G])$ and $P_iA=0$ 

479: if $A$ is in the direct product of the subrings

480: ${\bf I}_j(\BQ[G])$ with ${\bf I}_i(\BQ[G])$ excluded.

481: \end{itemize}

482: In fact we have $P_i\in {\bf I}_i(\BQ[G])$ for $i=1,\ldots ,\ell$

483: and $P_i$ can be thought of as the

484: unit element in ${\bf I}_i(\BQ[G])$.

485: 

486: Given two left $\BQ[G]$-modules $M_1$, $M_2$ we define 

487: ${\rm Hom}_G(M_1,M_2)$ to be the vector space of $\BQ[G]$-module homomorphisms

488: from $M_1$ to $M_2$. The $\BQ$-algebra of $\BQ[G]$-module endomorphisms of 

489: $M_1$ is denoted by ${\rm End}_G(M_1)$.

490: 

491: Let us write $S^{\rm op}$ for the {\it opposite ring} 

492: of an associative ring $S$. 

493: Mapping an element  $A\in {\bf I}_i(\BQ[G])$ to the 

494: multiplication from the right by $A$ defines $\BQ$-algebra isomorphisms

495: \begin{equation}\label{isoppp}

496: {\bf I}_i(\BQ[G])^{\rm op}\to 

497: {\rm End}_G({\bf I}_i(\BQ[G])) \qquad (i=1,\ldots ,\ell).

498: \end{equation}

499: The existence of the isomorphisms (\ref{isoppp}) implies that

500: \begin{equation}\label{dim17}

501: {\rm dim}_{\bb Q}({\rm Hom}_G({\bf I}_i(\BQ[G])^f,{\bf I}_i(\BQ[G]))

502: =f{\rm dim}_{\bb Q}({\bf I}_i(\BQ[G]))

503: \end{equation}

504: for all $f\in\BN$ and $i=1,\ldots,\ell$.

505: 

506: The algebras 

507: $$

508: D_i:={\rm End}_G(N_i)\qquad (i=1,\ldots,\ell)

509: $$

510: of left $\BQ[G]$-module endomorphism of the irreducible modules $N_i$ are 

511: finite dimensional division algebras with cyclotomic number fields as center.

512: We set 

513: $$h_i:={\rm dim}_{D_i}(N_i)\qquad (i=1,\ldots,\ell).$$ 

514: Then the algebra ${\bf I}_i(\BQ[G])$

515: is isomorphic to a $h_i\times h_i$ matrix algebra $M(h_i,D_i)$ over $D_i$,

516: that is

517: \begin{equation}\label{ra1}

518: {\bf I}_i(\BQ[G])\cong M(h_i,D_i)\qquad (i=1,\ldots ,\ell).

519: \end{equation}

520: Using (\ref{isoppp}) we get isomorphisms of algebras

521: \begin{equation}\label{ra1oppp}

522:  {\rm End}_G({\bf I}_i(\BQ[G])) \cong M(h_i,D_i^{\rm op})

523: \qquad (i=1,\ldots ,\ell).

524: \end{equation}

525: We shall further discuss integral versions of the isomorphisms (\ref{ri}) and

526: (\ref{ra1}). Our reference here is \cite{CR} Section XI or \cite{D} 

527: Section XX. 

528: 

529: Let $K$ be a number field with ring of integers ${\cal O}$ 

530: and $S$ a finite dimensional 

531: $K$-algebra. A subring ${\cal R}$ of $S$ is called an {\it order} if 

532: it contains a $K$-basis of $S$ and if there is a subring ${\cal O}_0$ of

533: finite index in ${\cal O}$ such that ${\cal R}$ is a finitely generated

534: ${\cal O}_0$-module

535: (subrings are always assumed to contain the identity element of the bigger

536: ring). Let $h$ be a natural number and $S=M(h,D)$ where $D$ is a 

537: finite dimensional division algebra over $K$ with center $K$. We write

538: ${\rm End}_S(S^m)$ ($m\in\BN$) for the algebra of $S$-left module

539: endomorphisms of the $S$-left module $S^m$. We obtain an algebra isomorphism 

540: \begin{equation}\label{ord}

541: \Theta: {\rm End}_S(S^m)\to M(m,S)^\op=M(m,S^\op)=M(m,M(h,D^\op)). 

542: \end{equation}

543: Here we identify $M(m,S)^\op$ in the usual way with $M(m,S^\op)$ and then with

544: $M(m,M(h,D^\op))$.

545: We call a subgroup $\Lambda$ of the additive group of $S^m$ a {\it lattice} if 

546: $\Lambda$ contains a $K$-basis of $S^m$ and if 

547: there is a subring ${\cal O}_0$ of finite index in ${\cal O}$ such that

548: $\Lambda$ is a finitely generated ${\cal O}_0$-module.

549: We write

550: ${\rm End}_S(S^m)_{\Lambda}$ for the ring of those endomorphisms stabilising

551: $\Lambda$. The following is well known (see \cite{CR}, \cite{D}).

552: \begin{lemma}\label{repkon} Let $\Lambda\le S^m$ ($m\in \BN$) be a lattice.

553: There is a sublattice $\tilde\Lambda\le \Lambda$ (of finite index in

554:  $\Lambda$) and an order

555:  ${\cal R}$ in $D$ such that

556: $$\Theta({\rm End}_S(S^m)_{\tilde\Lambda})\subset M(m,M(h,{\cal R}^\op)). $$

557: \end{lemma}

558: Let ${\rm Aut}_S(S^m)_\Lambda$ be the group of invertible elements in 

559: ${\rm End}_S(S^m)_\Lambda$. Note that 

560: ${\rm Aut}_S(S^m)_{\tilde\Lambda}$ is commensurable with 

561: ${\rm Aut}_S(S^m)_{\Lambda}$ in particular 

562: ${\rm Aut}_S(S^m)_{\Lambda}\cap {\rm Aut}_S(S^m)_{\tilde\Lambda}$

563: has finite index in ${\rm Aut}_S(S^m)_{\Lambda}$ and we obtain a

564: representation 

565: \begin{equation}\label{ord1}

566: \Theta: {\rm Aut}_S(S^m)_{\Lambda}\cap {\rm Aut}_S(S^m)_{\tilde\Lambda}

567: \to \GL(m,M(h,{\cal R}^\op))=\GL(mh,{\cal R}^\op).

568: \end{equation}

569: 

570: 

571: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

572: \subsection{A result of Gasch\"utz}\label{suga}

573: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

574: 

575: Next we need to report on a result of Gasch\"utz (see \cite{G}, \cite{Gr} and

576: \cite{JR}). 

577: For this, let $G$ be a finite group and $\pi : F_n\to G$ a 

578: surjective homomorphism

579: of a rank $n$ free group $F_n$ onto $G$. Let $R$ be the kernel of 

580: $\pi$ and $\bar R:=R/R'$ the corresponding relation module with its action  

581: (from the left) of $G$. The result of Gasch\"utz says that there is 

582: a $\BQ[G]$-module isomorphism

583: \begin{equation}\label{gasch}

584: \kappa:\BQ\oplus \BQ[G]^{n-1}\to  \BQ\otimes_{\bb Z}\bar R.

585: \end{equation}

586: where $\BQ$ is to be the trivial one dimensional $\BQ[G]$-module. 

587: We fix a $\BQ[G]$-module isomorphism $\kappa$ once for all and identify

588: $\BQ\otimes_\BZ\bar R$ with $\BQ\oplus \BQ[G]^{n-1}$. In

589: particular we obtain  ${\rm dim}_\BQ(\BQ\otimes_\BZ\bar{R})=1+|G|(n-1)$,

590: which is also clear from the formula of Reidemeister and Schreier.

591: We deduce 

592: \begin{equation}\label{gaschi1}

593: {\bf I}_1(\BQ\otimes_{\bb Z}\bar R)={\bf I}_1(\BQ[G])^{n}=\BQ^n,

594: \end{equation}

595: \begin{equation}\label{gaschi}

596: {\bf I}_i(\BQ\otimes_{\bb Z}\bar R)={\bf I}_i(\BQ[G])^{n-1}

597: \qquad (i=2,\ldots,\ell).

598: \end{equation}

599: 

600: From (\ref{dim17}) we get

601: \begin{lemma}\label{dim19}

602: Let $G$ be a finite group and $\pi : F_n\to G$ a 

603: surjective homomorphism. Let $\bar R:=R/R'$ be the corresponding 

604: relation module. Then we have:

605: $${\rm dim}_{\bb Q}({\rm Hom}_G({\bf I}_1(\BQ\otimes_{\bb Z}\bar{R}), 

606: {\bf I}_1(\BQ[G])))=n$$

607: $${\rm dim}_{\bb Q}(

608: {\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G])))=

609: (n-1){\rm dim}_{\bb Q}({\bf I}_i(\BQ[G])$$

610: \end{lemma}

611: 

612: The result of Gasch\"utz enables us further to decompose the algebraic group

613: ${\cal G}_{G,\pi}$ and to describe the components. 

614: We define for $i=1,\ldots,\ell$:

615: \begin{equation}\label{stabi}

616: {\cal G}_{G,\pi,i}:=\left\{\, \phi\in {\cal G}_{G,\pi}

617: \Mid \phi |_{{\bb C}\otimes{\bf I}_j({\bb Q}\otimes_{\bb Z}\bar R)}=

618: {\rm Id}_{{\bb C}\otimes{\bf I}_j({\bb Q}\otimes_{\bb Z}\bar R)}\ 

619: {\rm for}\ j\ne i\, \right\}.

620: \end{equation}

621: Note that the groups ${\cal G}_{G,\pi,i}$ centralise each other and

622: have pairwise trivial intersection. It is also easy to see that we have an

623: internal product decomposition:

624: $$

625: {\cal G}_{G,\pi}=\prod_{i=1}^\ell\, {\cal G}_{G,\pi,i}

626: $$

627: of $\BQ$-defined algebraic groups.

628: 

629: We write $\GL(d,S)$ ($d\in \BN$) for the group of 

630: invertible $d\times d$-matrices with entries in an associative ring $S$.

631: Following the discussion of the previous section we infer that the

632: identification made using  

633: (\ref{gasch}) defines isomorphisms

634: $$

635: \Sigma_i : {\cal G}_{G,\pi,i}(\BQ)\to \GL(n-1,{\bf I}_i(\BQ[G])^{\rm op})

636: =\GL((n-1)h_i,D_i^{\rm op})

637: $$

638: for $i=2,\dots \ell$ and also an isomorphism

639: $$

640: \Sigma_1 : {\cal G}_{G,\pi,1}(\BQ)\to \GL(n,{\bf I}_1(\BQ[G])^{\rm op})=

641: \GL(n,\BQ).

642: $$

643: Together they define an isomorphism

644: $$

645: \Sigma : {\cal G}_{G,\pi}(\BQ)\to \GL(n,\BQ)\times \prod_{i=2}^\ell 

646: \GL((n-1)h_i,D_i^{\rm op}).

647: $$

648: 

649: Suppose that $i$ is one of the numbers $1,\ldots,\ell$ and let 

650: $K_i$ be the center of ${\bf I}_i(\BQ[G])^{\rm op}$. As already

651: mentioned, this is a cyclotomic number field. The group 

652: $\GL((n-1)h_i,D_i^{\rm op})$ is the group of $K_i$-rational points of a

653: $K_i$-defined linear algebraic group ${\cal H}_i$. The above discussion shows 

654: that the $\BQ$-defined algebraic group ${\cal G}_{G,\pi,i}$ 

655: is equal to the base

656: field restriction of ${\cal H}_i$ from $K_i$ to $\BQ$. For this concept see

657: \cite{PLR}. We shall add a description of the 

658: groups ${\cal G}_{G,\pi}^1$ introduced in the introduction.

659: Every element $g$ of $\GL((n-1)h_i,D_i^{\rm op})$ ($i=1,\ldots,\ell$) 

660: acts by multiplication on $D_i^{(n-1)h_i}$. Taking a $K_i$-vector space

661: identification $D_i=K_i^{d_i}$ we associate to $g$ a square matrix of size 

662: $(n-1)h_id_i$. The determinant of this matrix is called the reduced norm 

663: ${\rm Nrd}(g)$ of $g$. The group of elements of reduced norm $1$ is

664: conventionally denoted by $\SL((n-1)h_i,D_i^{\rm op})$.

665: We have

666: $$

667: {\cal G}^1_{G,\pi,i}(\BQ)=\{\, g\in {\cal G}_{G,\pi,i}(\BQ)\Mid 

668: {\rm Nrd}(\Sigma_i(g))=1\,\}.

669: $$

670: 

671: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

672: \subsection{Representations of subgroups of finite index in

673:   $\A(F_n)$}\label{repsfini}

674: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

675: 

676: In this section we describe a matrix version of the representation (\ref{I5}).

677: To do this we use Lemma \ref{repkon} which constructs an integral 

678: matrix representation of a subgroup of finite index in $\Gamma(G,\pi)$ which

679: is compatible with the representations

680: $\Sigma_i$ ($i=1,\ldots,\ell$) and $\Sigma$

681: from the previous section. 

682: 

683: Let $G$ be a finite group and $\pi : F_n\to G$ a surjective homomorphism

684: of the free group $F_n$ onto $G$. Let $R$ be the kernel of 

685: $\pi$ and $\bar R$ the corresponding relation module. Let further

686: $\Gamma(G,\pi)$ be the subgroup of $\A(F_n)$ defined in 

687: (\ref{I1}). In the sequel we shall often use the following fact without

688: further notice.

689: \begin{lemma} An element $\varphi\in\Gamma(R)$ is in $\Gamma(G,\pi)$ if and

690: only if the map $\bar\varphi :\bar R\to \bar R$ induced by $\varphi$ is

691: $G$-equivariant. 

692: \end{lemma}

693: \begin{proof} If $\varphi$ is in $\Gamma(G,\pi)$ then  

694: for every $w\in F_n$ we have a $r_w\in R$ with $\varphi(w)=wr_w$. Hence

695: $\varphi(wrw^{-1})=wr_w\varphi(r)r_w^{-1}w^{-1}$ holds for every $r\in R$.

696: The latter is equal to $ w\varphi(r)w^{-1}$ modulo $R'$ which 

697: implies that $\bar\varphi :\bar R\to \bar R$ is

698: $G$-equivariant.

699: 

700: If $\bar\varphi :\bar R\to \bar R$ is $G$-equivariant then for every 

701: $w\in F_n$ the congruence

702: $\varphi(w)\varphi(r)\varphi(w)^{-1}\equiv w \varphi(r) w^{-1}$ 

703: holds modulo

704: $R'$ for every $r\in R$. This means that $w$ and $\varphi(w)$ act the same on

705: $\bar R$. The result of Gasch\"utz implies that the action of $G$ on $\bar R$

706: is faithful, that is no element except the identity of $G$ acts as the

707: identity. We conclude that $\varphi(w)w^{-1}$ is in $R$.

708: \end{proof}

709: 

710: Keeping the notations set up in subsections \ref{sura}, \ref{suga} we define  

711: $$

712: \Lambda_i:=\bar R\cap {\bf I}_i(\BQ\otimes_{\bb Z} \bar R)

713: \quad (i=1,\ldots ,\ell).

714: $$

715: Each $\Lambda_i$ is a $\BZ[G]$-submodule of 

716: $\bar R\cap {\bf I}_i(\BQ\otimes_{\bb Z} \bar R)$ and 

717: $\prod_{i=1}^\ell \Lambda_i\le \bar R$

718: is a finite index $\BZ[G]$-submodule.

719: Notice that the $\Lambda_i$ is a sublattice of 

720: ${\bf I}_i(\BQ\otimes_{\bb Z} \bar R)=M(h_i,D_i)^{n-1}$ 

721: ($i=2,\ldots,\ell$) in the

722: sense of the previous section, so is 

723: $\Lambda_1$ in ${\bf I}_1(\BQ\otimes_{\bb Z} \bar R)=\BQ^{n}$. Applying Lemma

724: \ref{repkon} we choose orders ${\cal R}_i\subset D_i$ and finite index

725: additive subgroups $\tilde\Lambda_i\le \Lambda_i$ ($i=1,\ldots,\ell$). 

726: We define 

727: \begin{equation}\label{hut}

728: \hat\Gamma(G,\pi)\le \Gamma(G,\pi)

729: \end{equation}

730: to be the simultaneous stabiliser of all the 

731: $\tilde\Lambda_i\le \Lambda_i$ ($i=1,\ldots,\ell$).

732: Notice further that $\hat\Gamma(G,\pi)$

733: is of finite index in  $\Gamma(G,\pi)$. Using the various representations

734: $\Theta$ from Lemma \ref{repkon} we obtain representations 

735: $$

736: \sigma_{G,\pi,1}: \hat\Gamma (G,\pi) \to \GL(n,\BZ),\qquad

737: \sigma_{G,\pi,i}: \hat\Gamma (G,\pi) \to 

738: \GL((n-1)h_i,{\cal R}_i^{\rm op})

739: $$

740: ($i=2,\ldots, \ell$), 

741: which come by projection from the analogously defined representation 

742: $$

743: \sigma_{G,\pi}: \hat\Gamma(G,\pi) \to 

744: \GL(n,\BZ)\times \prod_{i=2}^\ell \GL(n-1,M(h_i,{\cal R}_i^{\rm op})).

745: $$

746: Notice that the diagram 

747: $$

748: \xymatrix{

749: \hat\Gamma(G,\pi) \ar[r]^{} \ar[d]^{\sigma_{G,\pi}}

750: & {\cal G}_{G,\pi}(\BQ) \ar[ld]^{\Sigma} \\

751: \GL(n,\BQ)\times \prod_{i=2}^\ell \GL(n-1,M(h_i,D_i^{\rm op})) & 

752: }

753: $$

754: is commutative.

755: 

756: 

757: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

758: \section{Derivations and the relation module}\label{Drei}

759: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

760: 

761: This section contains a brief discussion of the 

762: correspondences between Fox-derivatives, the

763: relation module of a finite group and cohomology. We apply these results to

764: obtain a certain canonical system of right module generators of 

765: ${\rm Hom}_G(\BQ\otimes_{\bb Z} \bar R,\BQ[G])$ 

766: where $G$ is a finite group and 

767: $\bar R$ is the relation module coming from a presentaion $\pi :F_n\to G$.

768: 

769: 

770: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

771: \subsection{Fox calculus and cohomology}

772: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

773: 

774: For $n\in\BN$, let $F_n$ be the free group with basis $x_1,\ldots ,x_n$. We

775: write $\BZ[F_n]$, $\BQ[F_n]$ for the group rings of $F_n$ over the integers or

776: the rational numbers, respectively.

777: Let 

778: $$

779: \frac{\partial}{\partial x_i}:F_n\to \BZ[F_n]\qquad (i=1,\ldots n)

780: $$

781: be the Fox-derivatives (see \cite{Bi} for a more elaborate exposition). 

782: They are defined by the rules

783: \begin{equation}\label{F1}

784: \frac{\partial x_j}{\partial x_i}=\delta_{ij},\qquad \ 

785: \frac{\partial w_1w_2}{\partial x_i}=\frac{\partial w_1}{\partial x_i}

786: +w_1\frac{\partial w_2}{\partial x_i}

787: \end{equation}

788: for $i,\, j=1,\ldots n$ and $\ w_1,\, w_2\in F_n$. As usual,

789: $\delta_{ij}$ denoting the Kronecker symbol ($\delta_{ij}=1$ if $i=j$ and 

790: $\delta_{ij}=0$ if $i\ne j$). For later use we note some consequences of 

791: (\ref{F1}). For $i=1,\ldots , n$ and $e\in \BN$ we have 

792: \begin{equation}\label{F2}

793: \frac{\partial x_i^e}{\partial x_i}=\sum_{k=0}^{e-1}x_i^k,\qquad\quad 

794: \frac{\partial x_i^{-e}}{\partial x_i}=-\sum_{k=1}^{e}x_i^{-k}.

795: \end{equation}

796: Let further $w:=v_1x_i^{e_1} v_2x_i^{e_2}\cdot 

797: \ldots \cdot v_kx_i^{e_k}v_{k+1}$ ($e_1,\ldots ,e_k\in\BZ$)

798: be an element of $F_n$ with the subwords 

799: $v_j$ ($j=1,\ldots ,k+1$) not involving $x_i$, then

800: \begin{equation}\label{F3}

801: \frac{\partial w}{\partial x_i}=v_1\frac{\partial x_i^{e_1}}{\partial x_i}+

802: v_1x_i^{e_1}v_2 \frac{\partial x_i^{e_2}}{\partial x_i}+\ldots +

803: v_1x_i^{e_1}\cdot v_2x_i^{e_2}\cdot \ldots \cdot v_k

804: \frac{\partial x_i^{e_k}}{\partial x_i}.

805: \end{equation}

806: Let $G$ be a finite group generated by its elements

807: $g_1,\ldots , g_n$ ($n\in \BN$) and let

808: $$

809: \pi: F_n\to G\qquad \pi(x_1)=g_1,\ldots ,\pi(x_n)=g_n

810: $$

811: be the corresponding surjective group homomorphism.  

812: The homomorphism $\pi:F_n\to G$ extends by linearity to 

813: surjective ring homomorphisms:

814: $$

815: \pi: \BZ [F_n]\to \BZ[G]\qquad \pi: \BQ [F_n]\to \BQ[G].

816: $$

817: Let $R\le F_n$ be the kernel of $\pi$ and set $\bar{R}:=R/R'$

818: to be the commutator factor group of $R$. Let $\bar{} :R\to R/R'$ stand for

819: the quotient homomorphism.

820: The free group $F_n$ acts by

821: conjugation on $R$. Since its subgroup $R$ acts trivially on 

822: $\bar{R}$ we get an induced action of $G$ on $\bar{R}$ which

823: we denote by $g\cdot u$ ($g\in G$, $u\in \bar{R}$). We generally write the

824: composition in $\bar{R}$ additively.   

825: The free abelian group $\bar{R}$ then obtains the structure of a left

826: $\BZ[G]$-module which extends to a left $\BQ[G]$-module structure on

827: $\BQ\otimes_\BZ\bar{R}$. We have

828: \begin{lemma}\label{leF0}

829: Let $G$ be a group and $\pi :F_n\to G$ a homomorphism. 

830: Let $R$ be the kernel of $\pi$. Then

831: $$\pi\left( \frac{\partial r}{\partial x_i}\right)=0$$

832: for all $r\in R'$ and $i=1,\ldots n$.

833: \end{lemma}

834: \begin{proof}

835: Note first that (\ref{F1}) implies 

836: $\frac{\partial w^{-1}}{\partial x_i}=-w^{-1}

837: \frac{\partial w}{\partial x_i}$ for all $w\in F_n$ and $i=1,\ldots n$.

838: It is enough to show the result for single commutators 

839: $r=r_1r_2r_1^{-1}r_2^{-1}$. A repeated application of (\ref{F1}) implies that 

840: $$\frac{\partial r}{\partial x_i}= 

841: \frac{\partial r_1}{\partial x_i}+

842: r_1\frac{\partial r_2}{\partial x_i}-

843: r_1r_2r_1^{-1}\frac{\partial r_1}{\partial x_i}-

844: r_1r_2r_1^{-1}r_2^{-1}\frac{\partial r_2}{\partial x_i}$$

845: for all $i=1,\ldots n$. As $\pi(r_1)=\pi(r_2)=1$ the result follows.

846: \end{proof}

847: 

848: The Fox-derivatives  

849: $\frac{\partial}{\partial x_i}$ now induce maps

850: $$

851: \partial_i: \bar{R}\to \BZ[G]\qquad (i=1,\ldots ,n)

852: $$

853: defined as follows.

854: For every $u\in\bar{R}$ choose $w_u\in R$ with $ \bar w_u=u$ and

855: define 

856: \begin{equation}\label{deldef} 

857: \partial_i(u):=\pi\left( \frac{\partial w_u}{\partial x_i}\right)\qquad

858: (i=1,\ldots ,n). 

859: \end{equation}

860: We have

861: \begin{lemma}\label{leF}

862: The maps $\partial_1,\ldots \partial_n$ do not depend on the choices of the 

863: preimages $w_u$ ($u\in\bar{R}$). They are $\BZ[G]$-module

864: homomorphisms when $\BZ[G]$ acts on itself from the left.

865: \end{lemma}

866: \begin{proof}

867: The first statement follows from Lemma \ref{leF0}. For the second note that

868: $$\frac{\partial grg^{-1}}{\partial x_i}=

869: \frac{\partial g}{\partial x_i}+g \frac{\partial r}{\partial x_i}-

870: grg^{-1}\frac{\partial g}{\partial x_i}$$

871: for all $g,\, r\in F_n$ and all $i=1,\ldots n$. 

872: When $r$ is in $R$, $\pi(r)=1$ and the projection of the right hand side into

873: $\BZ[G]$ is equal to $\pi(g) \pi(\frac{\partial r}{\partial x_i})$.

874: \end{proof}

875: We also write

876: $\partial_i: \BQ\otimes_{\bb Z}\bar{R}\to \BQ[G]$

877: for the induced $\BQ[G]$-module homomorphisms.

878: 

879: Let $t$ be the $\BZ$-rank of the free abelian group $\bar{R}$.

880: Note that $t=1+|G|(n-1)$ and also

881: $t={\rm dim}_{\bb Q}(\BQ\otimes_{\bb Z}\bar{R})$. 

882: Let

883: $B:=(u_1,\ldots, u_t)$ be a $\BQ$-basis of $\BQ\otimes_{\bb Z}\bar{R}$. 

884: We consider here

885: \begin{equation}\label{pm}

886: J_B:= \left( \begin{array}{ccc}

887:                 \partial_1(u_1) & \ldots & \partial_n(u_1)\\

888:                 \vdots &  & \vdots \\

889:                 \partial_1(u_t) & \ldots & \partial_n(u_t)

890: \end{array} \right),\qquad

891: \end{equation}

892: which is a $t\times n$ matrix with entries in $\BQ[G]$.

893: 

894: Given a left $\BQ[G]$-module $M$ and $s\in\BN$ we write

895: $$M^s:=\left\{ 

896: \left(

897: \begin{array}{c}

898: m_1 \\ \vdots \\ m_s

899: \end{array} \right) \Mid m_1,\ldots, m_s\in M\right\}

900: $$

901: for the corresponding $\BQ[G]$-module of column vectors.

902: The matrix $J_B$ induces a $\BQ$-linear map

903: $J_B:M^n\to M^t$. Notice for later use that given a $\BQ[G]$-submodule

904: $N\le M$ we have $J_B(N^n)\subseteq N^t$.

905: 

906: Given the left $\BQ[G]$-module $M$ we write

907: $$

908: {\rm Der}(G,M):=\{\, d:G\to M\Mid d(gh)=gd(h)+d(g) \ {\rm for\ all}\ g,h\in

909: G\,\} 

910: $$

911: for the vector space of derivations from $G$ to $M$.

912: Using the generators $g_1,\ldots ,g_n$ of our finite group $G$ we define 

913: for $d\in {\rm Der}(G,M)$ 

914: $$

915: L(d):=

916: \left(\begin{array}{c}

917: d(g_1) \\ \vdots \\ d(g_n)

918: \end{array}\right)\in M^n.

919: $$

920: We have

921: \begin{lemma}\label{coholemm}

922: The map $L: {\rm Der}(G,M)\to M^n$ is injective and its image is equal to

923: ${\rm Ker}(J_B)$. Hence $L$ defines a $\BQ$-vector space isomorphism

924: between ${\rm Der}(G,M)$ and ${\rm Ker}(J_B)$.

925: \end{lemma}

926: \begin{proof} 

927: Let $d$ be a derivation in the kernel of $L$. We infer that

928: $d(g_1)=\ldots=d(g_n)=0$. The derivation rule and the fact that the 

929: $g_1,\ldots ,g_n$ generate $G$ implies that $d$ is identically zero.

930: This shows that $L$ is injective.

931: 

932: Let $D:F_n\to M$ be a derivation where $F_n$ acts on $M$ through the

933: homomorphism $\pi : F_n\to G$. Since both sides are derivations and agree on

934: the $x_i$'s we find 

935: \begin{equation}\label{deriv}

936: D(w)=\frac{\partial w}{\partial x_1}D(x_1)+\ldots

937: +\frac{\partial w}{\partial x_n}D(x_n)

938: \end{equation}

939: for all $w\in F_n$. 

940: 

941: Given a derivation $d: G\to M$, define an induced derivation 

942: $D: F_n\to M$ by setting $D(w):=d(\pi(w))$. Applying (\ref{deriv}) we find

943: for every $r\in R$

944: $$0=d(1)=D(r)=\sum_{i=1}^n \frac{\partial r}{\partial x_i}D(x_i)=

945: \sum_{i=1}^n \partial_i (r) d(g_i).$$

946: Applying this to $r=u_1,\ldots, u_t$ shows that the image of $L$ is contained

947: in the kernel of $J_B$.

948: 

949: Let now 

950: $$m:=\left(\begin{array}{c}

951: m_1 \\ \vdots \\ m_n

952: \end{array}\right)\in M^n

953: $$

954: be an $n$-tuple of elements of $M$. It is well known that there is a

955: derivation $D :F_n\to M$ with $D(x_1)=m_1,\ldots, D(x_n)=m_n$. 

956: If $m$ is an element of the kernel of $J_B$ then (\ref{deriv}) implies that

957: $D(r)=0$ for all $r\in R$. The derivation $D$ then induces a derivation

958: $d: G\to M$ with the property $d(g_1)=m_1,\ldots, d(g_n)=m_n$. This shows that

959: every element of the kernel of $J_B$ is in the image of $L$.

960: \end{proof}

961: \begin{corollary}\label{cohocor} 

962: Let $M$ be a $\BQ[G]$-module then the kernel of the map

963: $J_B :M^n\to M^t$ is zero if $M$ is a trivial module and has $\BQ$-vector

964: space dimension equal to ${\rm dim}_{\bb Q}(M)$ if M does not have a trivial

965: submodule. 

966: \end{corollary}

967: \begin{proof} 

968: Given an element $m\in M$ we let $d_m :G\to M$ be the corresponding inner

969: derivation $d_m(g)=(g-1)m$. Let  

970: $$

971: {\rm IDer}(G,M):=\{\, d_m \Mid m\in M\,\}\subseteq  {\rm Der}(G,M)

972: $$

973: be the space of inner derivations. It is well known that

974: the first cohomology group $H^1(G,M)={\rm Der}(G,M)/{\rm IDer}(G,M)$

975: is zero ($G$ is a finite group). Note also that the map 

976: $\mu : M\to {\rm IDer}(G,M)$, 

977: defined by $\mu(m):=d_m$ ($m\in M$) is surjective.

978: 

979: If $M$ is a trivial module then $\mu$ is identically zero, that implies  

980: ${\rm IDer}(G,M)=\{0\}$ and hence also ${\rm Der}(G,M)=\{0\}$. 

981: We find from Lemma \ref{coholemm} that 

982: ${\rm dim}_{\bb Q}({\rm ker}(J_B))=0.$

983: 

984: If $M$ does not contain the trivial module then 

985: $\mu$ is injective, that implies

986: ${\rm dim}_{\bb Q}({\rm IDer}(G,M))={\rm dim}_{\bb Q}(M)$. We find from 

987: Lemma \ref{coholemm} that 

988: ${\rm dim}_{\bb Q}({\rm ker}(J_B))={\rm dim}_{\bb Q}({\rm Der}(G,M))={\rm

989:   dim}_{\bb Q}(M).$

990: \end{proof}

991: 

992: The results of this section will be applied in the next section in case of the

993: relation modules.

994: 

995: 

996: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

997: \subsection{${\rm Hom}_G(\BQ\otimes_{\bb Z} \bar R,\BQ[G])$ 

998: as right $\BQ[G]$-module}\label{Hom}

999: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1000: 

1001: We keep here our notation from before, that is $G$ is a finite group 

1002: generated by its elements $g_1,\ldots , g_n$ ($n\in \BN$) and 

1003: $\pi: F_n\to G$ is the homomorphism defined by 

1004: $\pi(x_1)=g_1,\ldots ,\pi(x_n)=g_n$. Furthermore $R$ is the kernel of $\pi$

1005: and $\bar R$ is the corresponding relation module.

1006: As the main result we shall show that the 

1007: $\partial_1,\ldots,\partial_n\in {\rm Hom}_G(\BQ\otimes_\BZ \bar R,\BQ[G])$,

1008: which were constructed in the previous subsection generate  

1009: ${\rm Hom}_G(\BQ\otimes_{\bb Z} \bar R,\BQ[G])$ if this 

1010: space is considered as a

1011: $\BQ[G]$-module from the right in the following way. 

1012: 

1013: Let $M$ be a left $\BQ[G]$-module. For 

1014: $f\in {\rm Hom}_G(M ,\BQ[G])$

1015: and $A\in \BQ[G]$ we define $f^A\in {\rm Hom}_G(M ,\BQ[G])$ by 

1016: \begin{equation}

1017: f^A(x):=f(x)A\qquad (x\in M).

1018: \end{equation}

1019: Thus the $\BQ$-vector space ${\rm Hom}_G(M ,\BQ[G])$

1020: becomes a right $\BQ[G]$-module.

1021: 

1022: We shall prove here:

1023: \begin{proposition}\label{wich} The maps $\partial_1,\ldots, \partial_n$ 

1024: (defined in (\ref{deldef})) generate 

1025: ${\rm Hom}_G(\BQ\otimes_{\bb Z} \bar{R},\BQ[G])$ 

1026: as right $\BQ[G]$-module. 

1027: \end{proposition}

1028: 

1029: To prepare for the proof we decompose the matrix $J_B$ into submatrices. 

1030: To do this we choose

1031: the basis $B$ adapted to the decomposition

1032: $$

1033: \BQ\otimes_{\bb Z}\bar{R}=\prod_{i=1}^\ell\, 

1034: {\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}).

1035: $$

1036: That is we choose bases

1037: $B_i=(u_{i1},\ldots u_{it_i})$ 

1038: $(t_i:={\rm dim}_{\bb Q}({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}))

1039: $

1040: for $i=1,\ldots,\ell$ and put them together to obtain a 

1041: $\BQ$-vector space basis 

1042: $$

1043: B=(u_1=u_{11},\ldots,u_t=u_{\ell t_\ell}) 

1044: $$

1045: of $\BQ\otimes_{\bb Z}\bar{R}$.

1046: 

1047: Consider the matrices 

1048: \begin{equation}\label{pmi}

1049: J_B^{P_i}:= \left( \begin{array}{ccc}

1050:                 \partial_1(u_1)P_i & \ldots & \partial_n(u_1)P_i\\

1051:                 \vdots &  & \vdots \\

1052:                 \partial_1(u_t)P_i & \ldots & \partial_n(u_t)P_i

1053: \end{array} \right)=

1054: \left( \begin{array}{ccc}

1055:                 \partial_1^{P_i}(u_1) & \ldots & \partial_n^{P_i}(u_1)\\

1056:                 \vdots &  & \vdots \\

1057:                 \partial_1^{P_i}(u_t) & \ldots & \partial_n^{P_i}(u_t)

1058: \end{array} \right)

1059: \end{equation}

1060: for $i=1,\ldots,\ell$. Notice that according to 

1061: property {\bf LP} of the right projector $P_i$ (see section \ref{sura})

1062: the maps

1063: $$J_B,\, J_B^{P_i}: {\bf I}_i(\BQ[G])^n\to {\bf I}_i(\BQ[G])^t$$

1064: are the same. Furthermore we have 

1065: $\partial_j(u)\in {\bf I}_i(\BQ[G])$ ($j=1,\ldots ,n$)

1066: for $u\in {\bf I}_i(\BQ\otimes_{\bb Z}\bar{R})$. 

1067: This follows since each $\partial_j$ is a $\BQ[G]$-module homomorphism which

1068: has to respect isotypic components.

1069: Thus property {\bf RP} of the right projector $P_i$

1070: implies that all entries of $J_B^{P_i}$ are equal to zero except

1071: those in the submatrix

1072: \begin{equation}\label{pmis}

1073: J_{B_i}:= \left(\begin{array}{ccc}

1074:                 \partial_1(u_{i1})P_i & \ldots & \partial_n(u_{i1})P_i\\

1075:                 \vdots &  & \vdots \\

1076:                 \partial_1(u_{i t_i})P_i & \ldots & \partial_n(u_{i t_i})P_i

1077: \end{array} \right)

1078: \end{equation}

1079: hence the kernels of the two maps 

1080: $J_B : {\bf I}_i(\BQ[G])^n\to {\bf I}_i(\BQ[G])^t$ and

1081: $J_{B_i}:{\bf I}_i(\BQ[G])^n\to {\bf I}_i(\BQ[G])^{t_i}$

1082: are the same. Note finally that we have 

1083: $$

1084: J_{B_i}=\left(\begin{array}{ccc}

1085:            \partial_1^{P_i}(u_{i1}) & \ldots & \partial_n^{P_i}(u_{i1})\\

1086:            \vdots &  & \vdots \\

1087:            \partial_1^{P_i}(u_{i t_i}) & \ldots & \partial_n^{P_i}(u_{i t_i})

1088: \end{array} \right)

1089: $$

1090: for all $i=1,\ldots,\ell$.

1091: 

1092: \medskip

1093: \begin{prof} {\it of Proposition \ref{wich}:}

1094: We have the decomposition

1095: $$

1096: {\rm Hom}_G(\BQ\otimes_{\bb Z}\bar{R}, \BQ[G])=

1097: \prod_{i=1}^\ell

1098: {\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))

1099: $$

1100: of ${\rm Hom}_G(\BQ\otimes_{\bb Z}\bar{R}, \BQ[G])$ as 

1101: a right $\BQ[G]$-module. 

1102: Since the $\partial_1^{P_i},\ldots ,\partial_n^{P_i}$ are zero on  

1103: ${\bf I}_j(\BQ\otimes_{\bb Z}\bar{R})$ for $j\ne i$ it is enough to show that

1104: they generate 

1105: ${\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))$ 

1106: as a right $\BQ[G]$-module.

1107: Consider the $\BQ$-linear map

1108: $\Phi_i : {\bf I}_i(\BQ[G])^n \to 

1109: {\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))$

1110: given by 

1111: $$

1112: \Phi_i\left(\left(

1113: \begin{array}{c}

1114: m_1 \\ \vdots \\ m_n

1115: \end{array} \right)\right):=

1116: \partial_1^{P_i m_1}+ \ldots +\partial_n^{P_i m_n}.

1117: $$

1118: Notice that ${\bf I}_i(\BQ[G])\cdot \BQ[G]={\bf I}_i(\BQ[G])$ and hence

1119: the image of $\Phi_i$ is the right $\BQ[G]$-submodule of 

1120: ${\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))$ 

1121: generated by

1122: the $\partial_1^{P_i},\ldots ,\partial_n^{P_i}$.

1123: The key observation now is 

1124: \begin{equation}\label{dime}

1125: {\rm kernel}(J_{B_i})={\rm kernel}(\Phi_i)

1126: \end{equation}

1127: which is obvious by what is explained above. Notice that in fact 

1128: $(m_1,\ldots,m_n)^t$ is in the kernel of $\Phi_i$ if and only if

1129: $\partial_1^{P_im_1}(u_{ik})+\ldots +\partial_n^{P_im_n}(u_{ik})=0$

1130: holds for all $k=1,\ldots,t_i$.

1131: 

1132: We can treat now the case $i=1$ (the case of the trivial module).

1133: Here we have by (\ref{gaschi1})  that  

1134: ${\rm dim}_{\bb Q}({\bf I}_1(\BQ\otimes_{\bb Z}\bar{R}))=n$ and the kernel 

1135: of $\Phi_1$ is trivial (by Corollary \ref{cohocor}). Since we have

1136: ${\rm dim}_{\bb Q}({\rm Hom}_G({\bf I}_1(\BQ\otimes_{\bb Z}\bar{R}), 

1137: {\bf I}_1(\BQ[G])))=n$ (by Lemma \ref{dim19}) the map $\Phi_1$ has to be

1138: surjective. 

1139: 

1140: In the cases $i >1$ we have

1141: $ {\rm dim}_{\bb Q}({\rm kernel}(\Phi_i))={\rm dim}_{\bb Q}({\bf I}_i(\BQ[G]))$

1142: by (\ref{dime}) and Corollary \ref{cohocor}. By Lemma \ref{dim19} and 

1143: (\ref{dim17}) we have

1144: ${\rm dim}_{\bb Q}(

1145: {\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))=

1146: (n-1){\rm dim}_{\bb Q}({\bf I}_i(\BQ[G]))$

1147: which implies that $\Phi_i$ has to be surjective.

1148: \end{prof}

1149: 

1150: As a corollary to Proposition \ref{wich} we note

1151: \begin{corollary}\label{wichcor} 

1152: The maps 

1153: $\partial_1^{P_i},\ldots ,\partial_n^{P_i}$  

1154: generate 

1155: ${\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))$ 

1156: as right ${\bf I}_i(\BQ[G])$-module for every $i=1,\ldots ,\ell$. 

1157: \end{corollary}

1158: We shall give now a reformulation of the last corollary which will be needed

1159: later. Let  

1160: ${\rm Id}_{i,k}\in {\bf I}_i(\BQ\otimes_{\bb Z}\bar{R})$ 

1161: ($k=1,\ldots,n-1$) be the images 

1162: under $\kappa_i$ (see (\ref{gaschi}), (\ref{gaschi1})) 

1163: of the unit elements of of the components of 

1164: ${\bf I}_i(\BQ\otimes_{\bb Z}\bar{R})^{n-1}$ respectively of

1165: ${\bf I}_1(\BQ\otimes_{\bb Z}\bar{R})^{n}$ with $k=1,\ldots,n$. 

1166: Using the identification (\ref{ra1oppp}) we think of  

1167: $f\in{\rm Hom}_G({\bf I}_i(\BQ\otimes_{\bb Z}\bar{R}), {\bf I}_i(\BQ[G]))$ 

1168: being identified with the collection of matrices 

1169: $f({\rm Id}_{i,1})=A_{i,1},\ldots ,

1170: f({\rm Id}_{i,n-1})=A_{i,n-1}\in M(h_i,D_i^\op)$ or 

1171: $f({\rm Id}_{1,1})=A_{1,1},\ldots ,

1172: f({\rm Id}_{i,n})=A_{1,n}\in M(h_1,D_1^\op)=\BQ$.

1173: 

1174: \begin{corollary}\label{wichcor1}

1175: Let $i$ be one of the numbers $2,\ldots,\ell$.

1176: Given $A_{1},\ldots ,A_{n-1}\in M(h_i,D_i^\op)$

1177: there are $B_1,\ldots,B_n\in \BQ[G]$ such that

1178: $$\partial_1^{P_iB_1}+\ldots +\partial_n^{P_iB_n}({\rm Id}_{i,k})=A_k

1179: \qquad (k=1,\ldots,n-1).$$

1180: Given $A_{1},\ldots ,A_{n}\in M(h_1,D_1^\op)$

1181: there are $B_1,\ldots,B_n\in \BQ[G]$ such that

1182: $$\partial_1^{P_1B_1}+\ldots +\partial_n^{P_1B_n}({\rm Id}_{1,k})=A_k

1183: \qquad (k=1,\ldots,n).$$

1184: \end{corollary}

1185: 

1186: {\it Remark:} We sketch here another (more conceptual) 

1187: proof of Proposition \ref{wich}. Features of the proof given above 

1188: will play a role in the sequel. 

1189:  

1190: Using $H^2(G,\BQ\otimes_{\bb Z}\bar R)=0$ we 

1191: obtain an injective homomorphism 

1192: $$\epsilon : F_n/R'\to (\BQ\otimes_{\bb Z}\bar R)\rfish G$$

1193: from $F_n/R'$ to the semidirect product on the right hand side. 

1194: On $R/R'\le F_n/R'$ the homomorphism $\epsilon$ induces the inclusion of 

1195: $R/R'$ into $\BQ\otimes_{\bb Z}\bar R\le (\BQ\otimes_{\bb Z}\bar R)\rfish G$.

1196: The second component of $\epsilon$ coincides with $\pi$.  

1197: Associate to $f\in{\rm Hom}_G(\BQ\otimes_{\bb Z}\bar R,\BQ[G])$ the 

1198: homomorphism $\tau_f: (\BQ\otimes_{\bb Z}\bar R)\rfish G\to \BQ[G]\rfish G$

1199: ($\tau_f((u,g))=(f(u),g)$). Composition with $\epsilon$ creates a homomorphism

1200: $$\lambda_f: F_n\to F_n/R'\to \BQ[G]\rfish G, \qquad

1201: \lambda_f(w)=(d_f(w),\pi(w)),\quad (w\in F_n)$$

1202: The map $d_f:F_n \to \BQ[G]$ is a derivation. Hence we have for $w\in F_n$

1203: $$d_f(w)=\frac{\partial w}{\partial x_1}d_f(x_1)+\ldots

1204: +\frac{\partial w}{\partial x_n}d_f(x_n).

1205: $$

1206: Restricting this formula to $R/R'$ we have expressed $f$ in the desired way.

1207: 

1208: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1209: \section{Redundant presentations}\label{Vier}

1210: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1211: 

1212: In this section, we present a special decomposition of the relation module

1213: $\bar R$ for redundant presentations of finite groups. We then describe the

1214: action of various Nielsen automorphisms of $F_n$ on $\bar R$ with respect to

1215: this decomposition. 

1216: 

1217: Let $F_{n}$ ($n\in\BN, n\ge 2$)

1218: be the free group with basis $x_1,\ldots ,x_{n-1},\, y$.

1219: Let $G$ be a finite group generated by its elements

1220: $g_1,\ldots , g_{n-1}$  and let

1221: \begin{equation}\label{redu1}

1222: \pi: F_{n}\to G\qquad \pi(x_1)=g_1,\ldots ,\pi(x_{n-1})=g_{n-1},\ \pi(y)=1

1223: \end{equation}

1224: be the surjective group homomorphism corresponding to these data. 

1225: We keep the notation introduced above of objects related to the presentation

1226: $\pi$ of $G$. That is $R$ is the kernel of $\pi$ and $\bar R=R/R'$ is the

1227: relation module, etc.. 

1228: We shall construct certain Nielsen-like elements in $\hat\Gamma(G,\pi)$ 

1229: (see \ref{hut}) which

1230: will allow us to find many unipotent elements in the image of the

1231: representation  

1232: \begin{equation}\label{rep12}

1233: \sigma_{G,\pi}: \hat\Gamma(G,\pi) \to 

1234: \GL(n,\BZ)\times \prod_{i=2}^\ell \GL(n-1,M(h_i,{\cal R}_i^\op))

1235: \end{equation}

1236: defined in Section \ref{repsfini}.

1237: 

1238: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1239: \subsection{Decomposition of the relation module}

1240: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1241: 

1242: With the notation fixed as indicated above we define

1243: $$

1244: R_{\rm old}:=\{\, x\in \langle x_1,\ldots ,x_{n-1}\rangle\Mid \pi(x)=1\,\}\ 

1245: {\rm and}\ \Ro:=R_{\rm old}/R'_{\rm old}.

1246: $$

1247: The injection $R_{\rm old}\to R$ defines a $\BZ[G]$-module homomorphism

1248: $\Ro\to \bar R$. Since $y\in R$, we can consider the $\BZ[G]$-submodule

1249: of $\bar R$ generated by $\bar y\in \bar R$. 

1250: 

1251: \begin{lemma} The homomorphism $\Ro\to\bar R$ is injective and so is the

1252: homomorphism

1253: $\BZ[G]\to \bar R$ defined by $A\mapsto A\cdot \bar y$.

1254: Identifying $\Ro$ with its image in $\bar R$ we obtain a direct

1255: sum decomposition

1256: $$

1257: \bar R=\bar R_{\rm old}\oplus \BZ[G]\cdot \bar y.

1258: $$

1259: as a $\BZ[G]$-module.

1260: \end{lemma}

1261: \begin{proof}

1262: Let us temporarily write $\hat R_{\rm old}$ for the image of $R_{\rm old}$ in

1263: $\bar R$. We first show that $\bar R=\hat R_{\rm old}+ \BZ[G]\cdot \bar y$.

1264: Given any $w\in F_n$, by looking at the image of $w$ modulo the normal closure

1265: of $y$ in $F_n$, we see that $w$ can be written in a unique way as 

1266: $w=w_1w_2$ with $w_1,\, w_2\in F_n$ such

1267: that $w_1$ does not involve the letter $y$ whereas $w_2$ is a product of

1268: conjugates of powers of $y$. If $w$ is an element of $R$ then $w_1$ has to be

1269: in $R_{\rm old}$. This implies $\bar R=\hat R_{\rm old}+ \BZ[G]\cdot \bar y$.

1270: 

1271: We infer that $\BQ\otimes_{\bb Z}\bar R=\BQ\otimes_{\bb Z}\hat R_{\rm old}

1272: + \BQ[G]\cdot \bar y$. We now count dimensions using the result of 

1273: Gasch\"utz. As $\bar R$ and $\bar R_{\rm old}$ are free over $\BZ$, 

1274: the lemma follows.

1275: \end{proof}

1276: 

1277: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1278: \subsection{Nielsen Automorphisms}

1279: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1280: 

1281: We study here the effect of certain Nielsen automorphisms on the relation

1282: module $\bar R$.

1283: We define elements $\alpha_i\in \A(F_n)$ ($i=1,\ldots ,n-1$) by

1284: \begin{equation}\label{al}

1285: \alpha_i(x_i):=yx_i.

1286: \end{equation}

1287: Our convention here is that values not given are identical to the argument,

1288: that is we assume  

1289: $\alpha_i(x_j):=x_j,\ {\rm for}\ j\ne i\ {\rm and}\ \alpha_i(y)=y.$

1290: 

1291: For $X\in \langle x_1,\ldots ,x_{n-1}\rangle$ we define elements 

1292: $\beta_X\in \A(F_n)$ by

1293: \begin{equation}\label{be}

1294: \beta_X(y):=XyX^{-1}

1295: \end{equation}

1296: and finally for $U\in  R_{\rm old}$ we define $\gamma_U\in \A(F_n)$ by

1297: \begin{equation}\label{ge}

1298: \gamma_U(y):=Uy.

1299: \end{equation}

1300: The $\alpha_i$, $\beta_X$, $\gamma_U$ extend to automorphisms of $F_{n}$.

1301: We have

1302: \begin{lemma}\label{NLe} 

1303: The $\alpha_i$ ($i=1,\ldots,n-1$), $\beta_X$

1304: ($X\in \langle x_1,\ldots ,x_{n-1}\rangle$) and $\gamma_U$ 

1305: ($U\in  R_{\rm old}$) are in 

1306: $\Gamma(G,\pi)$ and the following formulas hold for the 

1307: corresponding automorphisms induced on $\bar R$:

1308: \begin{equation}\label{alf}

1309: \bar\alpha_i(u)=u+\partial_i(u)\cdot \bar y\quad {\it for}\ 

1310: u\in \bar R_{\rm old}\ {\it and}\qquad

1311: \bar\alpha_i(\bar y)=\bar y.

1312: \end{equation}

1313: \begin{equation}\label{bef}

1314: \bar\beta_X(u)=u,\quad {\it for}\ u\in \bar R_{\rm old}\ {\it and}\qquad 

1315: \bar\beta_X(\bar y)=\pi(X)\cdot \bar y,

1316: \end{equation}

1317: \begin{equation}\label{gef}

1318: \bar\gamma_U(u)=u,\quad {\it for}\ u\in \bar R_{\rm old}\ {\it and}\qquad 

1319: \bar\gamma_U(\bar y)=\bar U+\bar y,

1320: \end{equation}

1321: \end{lemma}

1322: \begin{proof}

1323: It is first of all clear from our definitions that all the $\alpha_i$, 

1324: $\beta_X$, $\gamma_U$ are elements of\ 

1325: $\Gamma(R):=\{\, \varphi\in \A(F_n)\Mid \varphi(R)=R \, \}$. Moreover all of

1326: them induce the identity on $F_n/R$ and hence are even in $\Gamma(G,\pi)$. It

1327: follows (see Section \ref{repsfini}) that they act 

1328: as $\BZ[G]$-automorphisms on $\bar R$.

1329: 

1330: Next we prove formula (\ref{alf}). Fix $i\in\{1,\ldots ,n-1\}$ 

1331: and define for $e\in\BN$

1332: $$

1333: W_e:=y\cdot x_iyx_i^{-1}\cdot\ldots \cdot x_i^{e-1}yx_i^{1-e},\quad

1334: W_{-e}:=x_i^{-1}y^{-1}x_i\cdot\ldots \cdot x_i^{-e}y^{-1}x_i^{e}\in F_n 

1335: $$

1336: and add $W_0:=1$. We then have 

1337: $\alpha_i(x_i^e)=W_e\cdot x_i^e$

1338: for all $e\in \BZ$. For $u\in\Ro$ choose $w\in R_{\rm old}$ with $\bar w=u$.

1339: We decompose $w\in \langle x_1,\ldots ,x_n\rangle$ as

1340: $w:=v_1x_i^{e_1} v_2x_i^{e_2}\cdot 

1341: \ldots \cdot v_kx_i^{e_k}v_{k+1}$

1342: ($e_1,\ldots ,e_k\in\BZ$) such that the $v_j$ do not involve $x_i$. Define

1343: further 

1344: $$

1345: V_1:=v_1,\ V_2:=v_1x_i^{e_1}v_2,\ldots ,V_k:=v_1x_i^{e_1}v_2\ldots 

1346: x_i^{e_k}v_k.

1347: $$

1348: We then have

1349: $$

1350: \alpha_i(w)=V_1W_{e_1}V_1^{-1}\cdot V_2W_{e_2}V_2^{-1}

1351: \cdot\ldots\cdot 

1352: V_k W_{e_2} V_k^{-1}\cdot w.

1353: $$

1354: Switching to the additive notation and using formulas (\ref{F2}) we find

1355: $$

1356: \bar\alpha_i(u)=u+\pi\left(v_1\frac{\partial x_i^{e_1}}{\partial x_i}+

1357: v_1x_i^{e_1}v_2 \frac{\partial x_i^{e_2}}{\partial x_i}+\ldots +

1358: v_1x_i^{e_1} v_2x_i^{e_2}\cdot \ldots \cdot v_k

1359: \frac{\partial x_i^{e_k}}{\partial x_i}\right)\cdot \bar y

1360: $$

1361: and formula (\ref{alf}) follows from (\ref{F3}) and (\ref{deldef}). 

1362: 

1363: To prove formula (\ref{bef}) notice that $\beta_X$ is the identity on $\Ro$, 

1364: subsequently (\ref{bef}) is clear.  

1365:   

1366: All statements made about the automorphisms $\gamma_U$ ($U\in R_{\rm old}$)

1367: are evident. Notice that the map $\bar\gamma_U$ depends only on the image 

1368: $\bar U$ of $U$ in $\bar R_{old}$.

1369: \end{proof}

1370: For later use we introduce now a notation for some elements in 

1371: $\Gamma(G,\pi)$ obtained by composing the $\alpha_i$ and $\beta_X$. For

1372: every $g\in G$ we choose $X_g\in \langle x_1,\ldots ,x_{n-1}\rangle$ 

1373: with $\pi(X_g)=g$ and

1374: define $\beta_g:=\beta_{X_g}$. We infer from formulas (\ref{alf}) and 

1375: (\ref{bef})

1376: $$\bar\beta_g\bar\alpha_i^m\bar\beta_{g^{-1}}(u+A\cdot\bar y)=

1377: u+(\partial_i(u)mg+A)\cdot\bar y\qquad (u\in \Ro,\, A\in\BZ[G])

1378: $$

1379: for all $g\in G$, $m\in \BZ$.

1380: Given an element $B=m_1g_1+\ldots +m_kg_k\in \BZ[G]$ 

1381: ($m_1,\ldots ,m_k\in\BZ$, $g_1,\ldots ,g_k\in G$) we define

1382: $\eta_{B,i}\in\A(F_n)$ by

1383: \begin{equation}\label{eta}

1384: \eta_{B,i}:=\beta_{g_1}\alpha_i^{m_1}\beta_{g_1^{-1}}\circ\ldots \circ

1385: \beta_{g_k}\alpha_i^{m_k}\beta_{g_k^{-1}}.

1386: \end{equation}

1387: for $i=1,\ldots ,n-1$.

1388: By Lemma \ref{NLe} it follows that $\eta_{B,i}\in \Gamma(G,\pi)$ for

1389: $i=1,\ldots ,n$ and $B\in \BZ[G]$.

1390: Using the notation of Section \ref{Hom} we have

1391: $$

1392: \bar\eta_{B,i}(u+A\cdot\bar y)=u+(\partial_i^B(u)+A)\cdot\bar y

1393: \qquad (u\in \Ro,\, A\in\BZ[G])

1394: $$

1395: for $i=1,\ldots ,n-1$ and $B\in\BZ[G]$. This formula makes it clear that

1396: $$

1397: \bar\eta_{m_1B_1+m_2B_2,i}=\bar\eta_{B_1,i}^{m_1}\circ \bar\eta_{B_2,i}^{m_2}

1398: $$

1399: holds for all $i=1,\ldots ,n-1$ and $B_1, \, B_2\in\BZ[G]$ and $m_1,\, m_2\in

1400: \BZ$. 

1401: 

1402: Using the representation (\ref{furep}) we set

1403: $$

1404: \A^+(F_n):=\{\, \phi\in \A(F)\Mid {\rm det}(\rho_1(\phi))=1\, \}.

1405: $$

1406: The subgroup $\A^+(F_n)$ is of index $2$ in $\A(F_n)$. For $n\ge 3$ it is 

1407: a perfect group, that is $\A^+(F_n)$ is equal to its own commutator subgroup. 

1408: We note that all elements $\bar\eta_{B,i}$ are in $\A^+(F_n)$.

1409: 

1410: 

1411: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1412: \section{Arithmeticity in case of a redundant presentation}\label{Fuenf}

1413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1414: 

1415: In this section we prove Theorem \ref{teo}: The decomposition of $\bar R$ to

1416: its isotypic components shows that our

1417: main goal is to prove that the image of $\hat\Gamma(G,\pi)$ in (\ref{rep12})

1418: contains a finite index subgroup of each component 

1419: $$\SL(n-1,M(h_i,{\cal R}_i^\op))=\SL((n-1)h_i,{\cal R}_i^\op).$$ 

1420: To prove this we will show that this image contains sufficiently

1421: many unipotent elements with nonzero entries in the bottom strip of $h_i$ rows

1422: and also in the most right hand strip of $h_i$ columns of this  

1423: $\SL((n-1)h_i,{\cal R}_i^\op)$. Then we can appeal to a theorem of Vaserstein

1424: which ensures that the elementary matrices with entries from a non-zero

1425: two-sided ideal of ${\cal R}_i$ generate a finite index subgroup of the

1426: arithmetic group $\SL((n-1)h_i,{\cal R}_i^\op)$.

1427: 

1428: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1429: \subsection{Elementary matrices}\label{Fuenf1}

1430: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1431: 

1432: This section sets up certain notations and facts about elementary matrices.

1433: 

1434: Let $S$ be an associative ring with unity.  

1435: We denote by $\GL(n,S)$ 

1436: ($n\in\BN$) the group

1437: of invertible $n\times n$ matrices with entries in $S$. We write $I_n$ for the

1438: identity matrix. Given $1\le i,\, j\le n$ with $i\ne j$ and $s\in S$ we 

1439: define $E_{ij}(s)\in \GL(n,S)$ to be the corresponding elementary matrix, that

1440: is $E_{ij}(s)$ is equal to $I_n$ except for the $(i,\, j)$ position, where we

1441: put $s$. For $1\le i,\, j\le n$ with $i\ne j$, $1\le k,\, l \le n$ 

1442: with $k\ne l$ and $s_1,\, s_2\in S$

1443: the Steinberg relations between the elementary matrices are 

1444: \begin{equation}\label{stein}

1445: [E_{ij}(s_1),E_{k\ell}(s_2)]:=\left\{\begin{array}{ccc}

1446:                  1  &  {\rm for}\ j\ne k, & i\ne l, \\

1447:                  E_{il}(s_1s_2) &  {\rm for}\ j= k, & i\ne l, \\

1448:                  E_{kj}(-s_2s_1) &  {\rm for}\ j\ne k, & i= l, \\

1449: \end{array} \right.

1450: \end{equation}

1451: where 

1452: $[E_{ij}(s_1),E_{k l}(s_2)]=E_{ij}(s_1)E_{kl}(s_2)

1453: E_{ij}(s_1)^{-1}E_{kl}(s_2)^{-1}$. For every admissible pair $i,\, j$ we have

1454: $E_{ij}(s_1)E_{ij}(s_2)=E_{ij}(s_1+s_2)$ for all $s_1,\, s_2\in S$.

1455: 

1456: Let ${\lie a}\subset S$ be an ideal of $S$. Given natural numbers $m_1,\,

1457: m_2\in\BN$ we denote by $M(m_1,m_2;{\lie a})$ the space of 

1458: $m_1\times m_2$-matrices with entries in ${\lie a}$. 

1459: Suppose now that $m_1+m_2=n$. For 

1460: $A\in M(m_1,m_2;{\lie a})$ we define

1461: $$

1462: H(A):=\left(\begin{array}{cc}

1463:            I_{m_2} &  0\\

1464:            A &   I_{m_1}

1465: \end{array} \right),\qquad

1466: V(A):=\left(\begin{array}{cc}

1467:            I_{m_2} &  A^t\\

1468:            0 &   I_{m_1}

1469: \end{array} \right)

1470: $$

1471: Here $A^t$ stands for the transpose of the matrix $A$. Suppose that 

1472: $A:=(a_{ij})$ where $i=1,\ldots,m_1$, $j=1,\ldots,m_2$ with 

1473: $a_{ij}\in{\lie a}$, then we have 

1474: $$

1475: H(A)=\prod_{i=1}^{m_1}   \prod_{j=1}^{m_2}\,     E_{m_2+i\, j}(a_{ij}),

1476: \qquad 

1477: V(A)=\prod_{i=1}^{m_1}   \prod_{j=1}^{m_2}\,     E_{j\, m_2+i}(a_{ij}).

1478: $$

1479: Given $m_1,\, m_2\in\BN$ with $m_1+m_2=n$ and an 

1480: ideal $\lie{a}\le S$ we define 

1481: $$

1482: H(m_1,m_2;{\lie a}):=\{\, H(A)\Mid A\in M(m_1,m_2; {\lie a})\,\},

1483: $$

1484: $$

1485: V(m_1,m_2;{\lie a}):=\{\, V(A)\Mid A\in M(m_1,m_2; {\lie a})\,\}.

1486: $$

1487: Both $H(m_1,m_2;{\lie a})$ and $V(m_1,m_2;{\lie a})$ 

1488: are abelian subgroups of $\GL(n,S)$.

1489: 

1490: We apply this notation in case of the orders $M(h_i,{\cal R}_i)$ 

1491: of the simple

1492: factor rings ${\bf I}_i(\BQ[G])$ of the rational group ring $\BQ[G]$. 

1493: So let $K$ be a number field and $D$ a 

1494: finite dimensional division algebra over $K$ with center $K$. 

1495: Let ${\cal R}$ be a order in $D$. It is well

1496: known that every two-sided ideal $\lie{a}\le {\cal R}$ 

1497: is additively finitely

1498: generated and has finite index in ${\cal R}$ 

1499: if it is non-zero. We have

1500: 

1501: \begin{proposition}\label{wasser} 

1502: Let $n\in \BN$ with $n\ge 3$ and let $m_1,\, m_2\in\BN$ satisfy $m_1+m_2=n$.

1503: Let ${\lie a}$ be a non-zero two-sided ideal of ${\cal R}$

1504: Then $H(m_1,m_2;{\lie a})$ and $V(m_1,m_2;{\lie a})$ generate a subgroup of

1505: finite index in $\SL(n,{\cal R})$.  

1506: \end{proposition}

1507: \begin{proof} Let $U$ be the sugroup generated by 

1508: $H(m_1,m_2;{\lie a})$ and $V(m_1,m_2;{\lie a})$. Using 

1509: the Steinberg relations (\ref{stein}) we conclude

1510: that there is a non-zero two-sided ideal ${\lie b}$ ($={\lie a}^2$) in ${\cal

1511:   R}$ such that every $E_{ij}(b)$ ($b\in {\lie b}$, $i\ne j$) is in $U$. The

1512: main result of \cite{Vas} implies that $U$ has finite index in  

1513: $\SL(n,{\cal R})$.

1514: \end{proof}

1515: 

1516: 

1517: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1518: \subsection{Constructing elements of ${\cal G}_{G,\pi}^1(\BZ)$}

1519: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1520: 

1521: 

1522: Let $F_{n}$ ($n\in\BN,\, n\ge 3$)

1523: be the free group with basis $x_1,\ldots ,x_{n-1},\, y$, $G$ a finite group

1524: with generators

1525: $g_1,\ldots , g_{n-1}$  and let

1526: $$

1527: \pi: F_{n}\to G\qquad \pi(x_1)=g_1,\ldots ,\pi(x_{n-1})=g_{n-1},\ \pi(y)=1

1528: $$

1529: be the redundant presentation resulting from the data. Notice the assumption

1530: $n\ge 3$, if $n=2$ then $G$ is cyclic. We enclude this case into the next

1531: section. 

1532: 

1533: Given $i=2,\ldots ,\ell$ and an ideal ${\lie a}$ of the order 

1534: ${\cal R}_i^\op$,

1535: let us define 

1536: \begin{equation}\label{pde1} 

1537: H_i({\lie a}):=H_i(h_i,(n-2)h_i;{\lie a})\le \GL(n,\BZ)

1538: \times \prod_{i=2}^\ell \GL(n-1,M(h_i,{\cal R}_i^\op))

1539: \end{equation}

1540: to be 

1541: $H(h_i,(n-2)h_i;{\lie a})\le \GL((n-1)h_i,{\cal R}_i^\op)=

1542: \GL(n-1,M(h_i,{\cal R}_i^\op))$ 

1543: but considered as a subgroup of

1544: the big direct product on the right hand side of (\ref{pde1}) 

1545: (with all components of the factors different from the $i$-th factor equal to

1546: the identity matrix). In case $i=1$ we have ${\cal O}_1=\BZ$ 

1547: and we define 

1548: $H_1({\lie a})=H_1(1,n-1;{\lie a})$ 

1549: with $n-1$ replaced by $n$. Let us also

1550: define 

1551: \begin{equation}\label{pde2} 

1552: V_i({\lie a}):=V_i(h_i,(n-2)h_i;{\lie a})\le \GL(n,\BZ)

1553: \times \prod_{i=2}^\ell \GL(n-1,M(h_i,{\cal R}_i^\op))

1554: \end{equation}

1555: in a similar way. We have 

1556: \begin{lemma}\label{eli1}

1557: Let $i$ be one of the $1,\ldots ,\ell$ and $g\in H_i({\cal R}_i^\op)$. 

1558: There is an

1559: $e \in\BN$ such that $g^e$ is in $\sigma_{G,\pi}(\hat\Gamma(G,\pi))$

1560: where $\sigma_{G,\pi}$ is as in (\ref{rep12}). 

1561: \end{lemma}

1562: \begin{proof} We prove the statement for $i\ge 2$, for $i=1$ the proof is only

1563: different in notation.

1564: 

1565: Let $i\ge 2$ be fixed and consider $g\in H_i({\cal R}_i^\op)$. 

1566: There are matrices $A_1,\ldots,A_{n-2}\in M(h_i,{\cal R}_i^\op)$ such that

1567: the $i$-th component of $g_i$ of $g$ is equal to

1568: $$g_i=\left(\begin{array}{cccccc}

1569:            I_{h_i} &  0 & 0 & \ldots & 0 & 0\\

1570:            0 & I_{h_i} & 0 & \ldots & 0 & 0\\

1571:            \vdots &  \vdots &  \vdots &  \vdots &  \vdots &  \vdots  \\

1572:            0 & 0 & 0 & \ldots & I_{h_i} & 0\\

1573:            A_1 & A_2 & A_3 & \ldots & A_{n-2} &  I_{h_i}

1574: \end{array} \right),

1575: $$

1576: all other components being equal to the identity matrix in the corresponding

1577: component group.

1578: By Corollary \ref{wichcor1}  

1579: there are $B_1,\ldots,B_{n-1}\in \BQ[G]$ such that

1580: $$\partial_1^{P_iB_1}+\ldots +\partial_{n-1}^{P_iB_{n-1}}({\rm Id}_{i,k})=A_k$$

1581: for $k=1,\ldots, n-2$. Choose $e \in\BN$ to have sufficiently many divisors so

1582: that $eP_iB_1,\ldots ,eP_iB_{n-2}$ are in $\BZ[G]$. Put 

1583: $$\gamma:=\eta_{eP_iB_1,1}\circ \ldots \circ\eta_{eP_iB_{n-1},n-1} $$ 

1584: with the $\eta$ defined in (\ref{eta}).

1585: Increase now (if

1586: necessary) $e$ to ensure $\gamma\in \hat\Gamma(G,\pi)$ (see (\ref{hut})). 

1587: We have $\sigma_{G,\pi}(\gamma)=g^e$ and the lemma is proved.

1588: \end{proof}

1589: \begin{lemma}\label{eli2}

1590: Let $i$ be one of the $1,\ldots ,\ell$ and $g\in V_i({\cal R}_i^\op)$. 

1591: There is an

1592: $e \in\BN$ such that $g^e$ is in $\sigma_{G,\pi}(\hat\Gamma(G,\pi))$.  

1593: \end{lemma}

1594: \begin{proof}

1595: We prove the statement for $i\ge 2$, again for $i=1$ the proof is only

1596: different in notation.

1597: 

1598: Let $i\ge 2$ be fixed and consider $g\in V_i({\cal R}_i^\op)$.

1599: There are matrices $A_1,\ldots,A_{n-2}\in M(h_i,{\cal R}_i^\op)$ such that

1600: the $i$-th component of $g_i$ of $g$ is equal to

1601: $$g_i=\left(\begin{array}{cccccc}

1602:            I_{h_i} &  0 & 0 & \ldots & 0 & A_1\\

1603:            0 & I_{h_i} & 0 & \ldots & 0 & A_2\\

1604:            \vdots &  \vdots &  \vdots &  \vdots &  \vdots &  \vdots  \\

1605:            0 & 0 & 0 & \ldots & I_{h_i} & A_{n-2}\\

1606:            0 & 0 & 0 & \ldots &  0 &  I_{h_i}

1607: \end{array} \right),

1608: $$

1609: The matrices $A_1,\ldots,A_{n-2}$ can be considered as elements of 

1610: $M(h_i,{\cal R}_i)$, after all $M(h_i,{\cal R}_i^\op)$ and 

1611: $M(h_i,{\cal R}_i)$ are the same sets. Using the convention of Section

1612: \ref{repsfini} we consider $u:=(A_1,\ldots,A_{n-2})$ as an element of 

1613: $M(h_i,D_i)^{n-2}\subseteq \BQ\otimes_{\bb Z}\bar R_{\rm old}$ 

1614: (see (\ref{gaschi})).   

1615: Then there is  $e \in\BN$ such that $eu\in \bar R_{\rm old}$. 

1616: We choose $U\in R_{\rm old} $ with $\bar U=eu$. We 

1617: take $e \in\BN$ to have sufficiently many divisors so

1618: that $\bar\gamma_U$ (see (\ref{ge})) is in $\hat\Gamma(G,\pi)$. We have 

1619: $\bar\gamma_U=g^e$ and the lemma is proved.

1620: \end{proof}

1621: 

1622: Since every $H_i({\cal R}_i^\op)$ or $V_i({\cal R}_i^\op)$ ($i=1,\ldots,\ell$)

1623: is a finitely generated abelian group the two previous lemmas immediately 

1624: imply

1625: \begin{proposition}\label{prfi}

1626: There is $e\in \BN$ such that

1627: \begin{equation}\label{pde3}

1628: \prod_{i=1}^\ell H_i(e{\cal R}_i)\le \sigma_{G,\pi}(\hat\Gamma(G,\pi))

1629: \quad {and}\quad \prod_{i=1}^\ell V_i(e{\cal R}_i)\le 

1630: \sigma_{G,\pi}(\hat\Gamma(G,\pi)).

1631: \end{equation}

1632: \end{proposition}

1633: 

1634: Finally we can state 

1635: \begin{theorem}\label{prfit} Let $n\in \BN$ satisfy $n\ge 4$, then

1636: $$\sigma_{G,\pi}(\hat\Gamma(G,\pi))\cap \SL(n,\BZ) 

1637: \times \prod_{i=2}^\ell \SL((n-1)h_i,{\cal R}_i^\op))$$

1638: is of finite index in $\SL(n,\BZ) 

1639: \times \prod_{i=2}^\ell \SL((n-1)h_i,{\cal R}_i^\op))$.

1640: \end{theorem}

1641: \begin{proof} Just use Proposition \ref{prfi} together with Proposition 

1642: \ref{wasser}.

1643: \end{proof}

1644: 

1645: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1646: \section{Cyclic groups}\label{Choose}

1647: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1648: 

1649: 

1650: In this section we present a detailed study of the case when 

1651: $G$ is a cyclic group. We restrict ourselves to a particular 

1652: type of (redundant) presentation, but following the discussion in Section

1653: \ref{Concl3} this is in fact the general case. An important feature 

1654: is that we are able to present a set of

1655: generators of $\Gamma(G,\pi)$, in the case $n\ge 3$. For a general 

1656: group $G$ we are not

1657: able to do this. As a consequence we can determine the precise image of 

1658: $\Gamma(G,\pi)$ (not only up to commensurability) 

1659: acting on its invariant pieces of the relation module. 

1660: Some of the results here will be needed in Section 

1661: \ref{Proofs}, but we believe that this section is of independent interest.

1662: In fact, our method can be observed here in a very simple, almost elementary

1663: situation.   

1664: 

1665: Let us call the attention of the reader to the fact that our results

1666: concerning cyclic groups, in fact, cover the general case of $G$ being

1667: abelian. This follows since every irreducible representation of an abelian

1668: group factors through a cyclic quotient, see the discussion in Section

1669: \ref{SL3}.

1670: 

1671: Let the free group $F_n$ ($n\ge 2$) be generated 

1672: by $x,\, y_1,\ldots, y_{n-1}$. Let us introduce the following elements of

1673: $\A(F_n)$. Our convention here is again that values 

1674: not given are identical to the argument.   

1675: \begin{itemize}

1676: \item $\delta_i(x):=y_ix$  and $\epsilon_i(x):=xy_i$ for $i=1,\ldots ,n-1$,

1677: \item $\varphi_i(y_i):=xy_i$ and 

1678: $\psi_i(y_i):=y_ix$ for $i=1,\ldots ,n-1$,

1679: \item $\lambda_{ij}(y_i):=y_jy_i$ and $\nu_{ij}(y_i):=y_iy_j$ 

1680: for $i,\, j=1,\ldots ,n-1$ with $i\ne j$.

1681: \end{itemize}

1682: A theorem of Nielsen (see \cite{MKS}) asserts that these elements generate 

1683: $\A^+(F_n)$ where $\A^+(F_n)$ is the kernel of the homomorphism 

1684: $\rho_1: \A(F_n) \to \GL(n,\BZ)$ followed by the determinant. If 

1685: $\Gamma \le \A(F_n)$ is a subgroup, we define $\Gamma^+:=\Gamma\cap

1686: \A^+(F_n)$. Let us further introduce 

1687: \begin{itemize}

1688: \item $\kappa_{jk}(x):= x [y_j,y_k]$, $\kappa_{ijk}(y_i):= y_i [y_j,y_k]$

1689: for $1\le i,\, j,\, k\le n-1$ with $i\ne j$, $i\ne k$,

1690: \item $\tau_{ij}(y_i):= y_i [x,y_j]$ for $1\le i,\, j\le n-1$ 

1691: with $i\ne j$.

1692: \end{itemize}

1693: The set $T_n$ consisting of the $\kappa_{jk}$, $\kappa_{ijk}$, 

1694: $\tau_{ij}$ (with indices as above), the 

1695: $\delta_i^{-1}\circ \epsilon_i$, $\varphi_i^{-1}\circ \psi_i$ for

1696: $i=1,\ldots ,n-1$ and the $\lambda_{ij}\circ\nu_{ij}^{-1}$ for

1697: $1\le i,\, j\le n-1$ with $i\ne j$

1698: generates the group ${\rm IA}(F_n)$ by

1699: another theorem of Nielsen (see \cite{MKS}). Notice that $T_n$ is contained in

1700: $\Gamma(G,\pi)$ for every presentation $\pi$ of an abelian group $G$.

1701: 

1702: Let us also introduce the following notation concerning matrix groups. We

1703: write 

1704: $$\Gamma^1(n,m)\le \SL(n,\BZ)$$

1705: for the subgroup of $\SL(n,\BZ)$ consisting of those elements having a first

1706: row which is congruent to $(1,0,\ldots,0)$ modulo $m\in\BN$. We need

1707: \begin{lemma}\label{eli11} Let $n$ and $m$ be natural numbers with $n\ge

1708:   3$. The group $\Gamma^1(n,m)$ is generated by the elementary matrices 

1709: $$E_{ij}(1)\quad (i,\, j=1,\ldots,n, i\ne j, i\ne 1),\qquad E_{1j}(m) \quad

1710:  (j=2,\ldots,n).$$   

1711: \end{lemma}

1712: This lemma is an elementary exercise using the euclidean algorithm in $\BZ$.

1713: 

1714: 

1715: We first treat the cyclic group of order $2$ which we call $C_2$. We have

1716: singled out this case since it is completely elementary and also plays a

1717: special role in the proof of Corollary \ref{kaz}. 

1718: Let $g$ be a generator of $C_2$. We choose

1719: the presentation 

1720: $$

1721: \pi:F_n\to C_2,\quad \pi(x)=g,\ \pi(y_1)=\ldots=\pi(y_{n-1})=1.

1722: $$

1723: The corresponding relation module $\bar R$ has the following elements as a

1724: $\BZ$-basis. 

1725: $$\overline{x^2},\ \bar y_1,\ldots,\bar y_{n-1},\  

1726: \overline{xy_1x^{-1}}=g\bar y_1,\ldots,

1727: \overline{xy_{n-1}x^{-1}}=g\bar y_{n-1}.$$ 

1728: The $\BQ$-vector space $\BQ\otimes_{\bb Z} \bar R$ decomposes as  

1729: $\BQ\otimes_{\bb Z} \bar R=V_1\oplus V_{-1}$ where 

1730: $V_1$, $V_{-1}$ are the $\pm 1$ 

1731: eigenspaces of $g$ respectively.

1732: Set $\bar R_1:=\bar R\cap V_1$ and $\bar R_{-1}:=\bar R\cap V_{-1}$. 

1733: Notice that $\bar R_1+\bar R_{-1}$ is of finite index in $\bar R$. Introduce

1734: \begin{equation}\label{basi}

1735: v_i:=\bar y_i+\overline{xy_{i}x^{-1}},\quad 

1736: w_i:=\bar y_i-\overline{xy_{i}x^{-1}}\qquad (i=1,\ldots ,n-1).

1737: \end{equation}

1738: Then $\overline{x^2},v_1,\ldots,v_{n-1}$ is a $\BZ$-basis of $\bar R_1$ and

1739: $w_1,\ldots,w_{n-1}$ is a $\BZ$-basis of $\bar R_{-1}$. 

1740: Since $\Gamma(C_2,\pi)$

1741: leaves $\bar R_1$ and $\bar R_{-1}$ invariant we obtain, 

1742: with the above $\BZ$-bases being chosen, representations

1743: \begin{equation}\label{rep2}

1744: \sigma_1:\Gamma^+(C_2,\pi)\to \GL(n,\BZ),\qquad 

1745: \sigma_{-1}:\Gamma^+(C_2,\pi)\to \GL(n-1,\BZ).

1746: \end{equation}

1747: With all these data given we have: 

1748: \begin{proposition}\label{cyc2} Let $n\ge 2$ be a natural number. The

1749:   following hold.

1750:  

1751: {\rm (i)} The group $\Gamma^+(C_2,\pi)$ is generated by the 

1752: automorphisms $\delta_i$, $\epsilon_i$

1753: ($i=1,\ldots,n-1$), the $\lambda_{ij}$ and $\nu_{ij}$ 

1754: ($i,\, j=1,\ldots ,n-1$, $i\ne j$), the $\varphi_i^2$ ($i=1,\ldots,n-1$) and

1755: the elements of the set $T_n$ introduced above.

1756: 

1757: {\rm (ii)} The image of $\sigma_1$ is equal to $\Gamma^1(n,2)$.

1758: 

1759: {\rm (iii)} The index of $\Gamma^+(C_2,\pi)$ in $\A^+(F_n)$ is $2^{n}-1$.

1760: 

1761: {\rm (iv)} The representation $\sigma_{-1}$ is surjective onto $\GL(n-1,\BZ)$.

1762: 

1763: {\rm (v)} The group $\sigma_{-1}({\rm IA}(F_3))\le \GL(2,\BZ)$ is generated by

1764: the matrices

1765: $$\left(\begin{array}{cc}

1766:            -1 &  0\\

1767:             0  &  1

1768: \end{array} \right),\qquad 

1769: \left(\begin{array}{cc}

1770:            1 &  0\\

1771:            0 &  -1

1772: \end{array} \right),\qquad

1773: \left(\begin{array}{cc}

1774:            1 &  2\\

1775:            0 &  1

1776: \end{array} \right),\qquad

1777: \left(\begin{array}{cc}

1778:            1 &  0\\

1779:            2 &  1

1780: \end{array} \right).

1781: $$

1782: \end{proposition}

1783: 

1784: The proof of this proposition is elementary but repeated below 

1785: in the more general case, so we skip it here.

1786: Items (ii), (iv) and (v) follow

1787:  from formulas describing the action of the given

1788: generators of $\Gamma^+(C_2,\pi)$ on the bases in $\bar R_1$ and $\bar R_{-1}$.

1789: An important ingredient is Lemma \ref{eli11}. Note that $\Gamma^1(2,2)$ is

1790: generated by the elementary matrices  

1791: $$ E_{21}(1)=\left(\begin{array}{cc}

1792:            1 &  0\\

1793:            1 &  1

1794: \end{array} \right),\qquad 

1795: E_{12}(2)=\left(\begin{array}{cc}

1796:            1 &  2\\

1797:            0 &  1

1798: \end{array} \right).$$

1799: For $\Gamma^1(2,m)$ ($m\ge 3$) the analogous statement is not true.

1800: Proposition \ref{cyc2} already

1801: implies Corollary \ref{kaz}.  

1802: 

1803: We turn now to the case of a general cyclic group $C_m=\langle g\rangle$ 

1804: ($m\in \BN$). 

1805: We choose the presentation 

1806: $$

1807: \pi:F_n\to C_m,\quad \pi(x)=g,\ \pi(y_1)=\ldots=\pi(y_{n-1})=1.

1808: $$

1809: The corresponding relation module $\bar R$ has the following 

1810: $1+m(n-1)$ elements as a

1811: $\BZ$-basis 

1812: $$\overline{x^m},\   

1813: \overline{x^{k}y_1x^{-k}},

1814: \ldots, \overline{x^{k}y_{n-1}x^{-k}}\qquad (k=0,\ldots,m-1).$$

1815: The rational group ring of $C_m$ decomposes as

1816: \begin{equation}\label{grouri}

1817: \BQ[C_m]=\prod_{d|m}\BQ(\zeta_d).

1818: \end{equation}

1819: Here $\zeta_d\in\BC$ is a primitive $d$-th root of unity. We consider

1820: each of the $\BQ(\zeta_d)$ as a $C_m$-module letting $g$ act by multiplication

1821: with $\zeta_d$. The ring of integers $\BZ(\zeta_d)$ is then a 

1822: $\BZ[C_m]$-submodule of $\BQ(\zeta_d)$.

1823: We mimic the decomposition (\ref{grouri}) inside the relation module 

1824: by introducing

1825: \begin{equation}

1826: v_i(d,m):=\sum_{k=0}^{m-1}{\rm Tr}(\zeta_d^k)\, \overline{x^ky_ix^{-k}}

1827: \end{equation}

1828: for $i=1,\ldots,n-1$.

1829: Here ${\rm Tr}(\zeta_d^k)$ is the trace of $\zeta_d^k$ taken from 

1830: $\BQ(\zeta_d)$ to $\BQ$. Notice that the $v_i(d,m)$ are generalisations of the

1831: generators in (\ref{basi}).

1832: Let $\bar R_d$ be the $C_m$-submodule of $\bar R$ generated by the

1833: $v_i(d,m)$ for $i=1,\ldots,n-1$ if $d>1$ and define 

1834: $\bar R_1$ to be the $C_m$-submodule of $\bar R$ generated by the

1835: $v_i(1,m)$ for $i=1,\ldots,n-1$ together with $\overline{x^m}$. By setting 

1836: $$\Lambda_{d}(v_i(d,m)):=\left(0,\ldots,0,

1837: \sum_{k=0}^{m-1}{\rm Tr}(\zeta_d^k)\zeta_d^k,0,\ldots,0\right)

1838: \qquad (i=1,\ldots,n-1)$$

1839: (with the non-zero entry in the $i$-th component) for $d>1$ and extending 

1840: $C_m$-linearly we obtain well defined additive homomorphisms

1841: $$\Lambda_d:\bar R_d\to \BZ(\zeta_d)^{n-1}\qquad (d|m,\, d>1). $$

1842: The following are true 

1843: \begin{itemize}

1844: \item the homomorphism $\Lambda_d$ is $C_m$-equivariant and injective (onto an

1845:   ideal in $\BZ(\zeta_d)$),

1846: \item the $\BZ$-rank of $\bar R_d$ is $(n-1)\varphi(d)$ 

1847: for $d>1$ and $n$ for $d=1$,

1848: \item $\bar R_d\cap \bar R_{d'}=\langle 0\rangle$ for distinct divisors $d,\,

1849:   d'$ of $m$,

1850: \item the (direct) sum of all $\bar R_d$ ($d|m$) has finite index in $\bar R$,

1851: \item  each $\bar R_d$ ($d|m$) is left invariant by $\Gamma(C_m,\pi)$.  

1852: \end{itemize}

1853: As above we transport the action of $\Gamma(C_m,\pi)$ on $\bar R_d$ 

1854: to the image of $\lambda_d$ and obtain our representations

1855: $$\sigma_{d}: \Gamma(G,\pi)\to \GL(n-1,\BQ(\zeta_d))\qquad (d|m,\, d>1).$$  

1856: To control the image of $\sigma_{d}$ we first show:

1857: \begin{proposition}\label{genem}

1858: Let $n$ and $m$ be natural numbers with $n\ge 3$.  

1859: The group $\Gamma^+(C_m,\pi)$ is generated by the 

1860: automorphisms $\delta_i$, $\epsilon_i$

1861: ($i=1,\ldots,n-1$), the $\lambda_{ij}$ and $\nu_{ij}$ 

1862: ($i,\, j=1,\ldots ,n-1$, $i\ne j$) and the $\varphi_i^m$, 

1863: ($i=1,\ldots,n-1$) and

1864: the elements of the set $T_n$ introduced above.

1865: \end{proposition}

1866: \begin{proof}

1867: Since every element of $\Gamma(C_m,\pi)$ has 

1868: to leave the normal closure of

1869: $x^m$ in $F_n$ invariant 

1870: and has to fix $x$ modulo $R$

1871: we see that $\rho_1(\Gamma^+(C_m,\pi))$ has to be

1872: contained in $\Gamma^1(n,m)$ (for $\rho_1$ see (\ref{furep})). 

1873: Examining the action of elements in the

1874: statement of the proposition on $F_n/F_n'$ and using Lemma \ref{eli11} we find

1875: that $\rho_1(\Gamma^+(C_m,\pi))=\Gamma^1(n,m)$.

1876: 

1877: Now our set of generators contains enough elements to generate ${\rm IA}(F_n)$

1878: by the theorem of Nielsen and also enough elements to generate the image of 

1879: $\Gamma^+(G,\pi)$ in $\SL(n,\BZ)$ by Lemma \ref{eli11}. 

1880: The proposition follows.

1881: \end{proof}

1882: Finally we can state:

1883: \begin{proposition}\label{cycfin}

1884: Let $n$ and $m$ be natural numbers with $n\ge 2$.

1885: The image of the representation $\sigma_{d}$ ($d|m,\, d>1$) 

1886: is $\GL^+(n-1,\BZ(\zeta_d))$

1887: where $\GL^+(n-1,\BZ(\zeta_d))$ is the subgroup consisting of those elements in

1888: $\GL(n-1,\BZ(\zeta_d))$ which have a power of $\zeta_d$

1889: as determinant. 

1890: \end{proposition}

1891: \begin{proof}

1892: Assume first that $n\ge 3$.

1893: Evaluating $\sigma_{d}$ on the generators from Proposition

1894: \ref{genem} we see that $\sigma_{d}(\Gamma^+(C_m,\pi))$ is contained in

1895: $\GL^+(n-1,\BZ(\zeta_d))$.

1896: We now show that all elementary matrices $E_{ij}(\zeta_d^k)$

1897: ($i,\, j=1,\ldots n-1,\, k\in \BZ$) are in the image of $\sigma_{d}$. This is

1898: done by evaluating $\sigma_{d}$ on some of the generators from Proposition

1899: \ref{genem}. We report only a special case: Take $\tau_{ij}$ for a pair

1900: $(i,j)$ with $i\ne j$. We have

1901: $\tau_{ij} (y_i)=y_i[x,y_j]$. This leads to

1902: $$\bar\tau_{ij}(v_i(d,m))=v_i(d,m)+gv_j(d,m)-v_j(d,m)$$

1903: which in turn shows that $E_{ij}(\zeta_d-1)$ is in the image of 

1904: $\sigma_{d}$. Since $E_{ij}(1)$ is also present (use $\lambda_{ij}$) we have

1905: shown that $E_{ij}(\zeta_d)$ is contained in the image of 

1906: $\sigma_{d}$. 

1907: 

1908: After more elementary considerations like that we can show that all 

1909: the elementary matrices with entries in $\BZ(\zeta_d)$ are in the image of 

1910: $\sigma_{d}$. These matrices generate $\SL(n-1,\BZ(\zeta_d))$: For $n\ge 4$

1911: this follows from results in \cite{Vas}, for $n=3$ and $d\ge 5$ we apply the

1912: main result of \cite{Vas1}, in case $n=3$ and $d\le 4$ the ring $\BZ(\zeta_d)$

1913: is euclidean, in case $n=2$ nothing has to be proved.

1914: 

1915: As a last step we use the $\varphi_i^{-1}\circ \psi_i$ ($i=1,\ldots, n-1$) to

1916: show that the image of $\sigma_d$ contains a matrix with determinant equal to

1917: $\zeta_d$. This finishes the proof for $n\ge 3$.

1918: 

1919: Let us finally discuss the case $n=2$ which seems trivial, but in fact is

1920: not. Note that $\GL(1,\BZ(\zeta_d))$ contains elements of infinite order

1921: whenever $d\ge 5$, $d\ne 6$ holds. Since we do not have generators for 

1922: $\Gamma^+(2,m)$ ($m\ge 3$) at hand we cannot directly rule out the 

1923: possibility that such an element lies in the image of $\sigma_d$. We argue as

1924: follows. Let $\varphi\in \Gamma^+(2,m)$ be such that for some divisor $d$ of

1925: $m$ the image $\sigma_d(\varphi)$ has a determinant of infinite order. We

1926: extend $\varphi$ to an element of $\psi$ of $\Gamma^+(3,m)$ by setting

1927: $\psi(x):=\varphi(x)$, $\psi(y_1):=\varphi(y_1)$, $\psi(y_2):=y_2$. Since the

1928: determinants of $\sigma_d(\varphi)$ and $\sigma_d(\psi)$ coincide we have

1929: finished also the case $n=2$.  

1930: \end{proof}

1931: Proposition \ref{cycfin} allows us to conclude the following result concerning

1932: the abelianisations of the images of $\Gamma(C_m,\pi)$ under the

1933: representations $\sigma_d$ ($d|m,\, d>1$).  

1934: \begin{corollary}\label{cycfincor}

1935: Let $n$ and $m$ be natural numbers with $n\ge 2$.

1936: Then $\sigma_{d}(\Gamma(C_m,\pi))$ ($d|m,\, d>1$) has finite abelianisation.

1937: \end{corollary}

1938: \begin{proof}

1939: It is well known that $\GL^+(n-1,\BZ(\zeta_d))$ has finite abelianisation for

1940: all $d\in\BN$: For $n=2$ this fact is obvious,

1941: for $n=3$ and $d=2,\, 3,\, 4,\, 6$ a presentation of 

1942: $\SL(2,\BZ(\zeta_d))$ (see \cite{EGM}) can be used, for $d=5$ or $d\ge 7$ this

1943: fact is contained in \cite{Serresl}. 

1944: For other values of $n,\, d$ the group $\GL^+(n-1,\BZ(\zeta_d))$ has Kazdhan's

1945: property (T).  

1946: Since $\sigma_{d}(\Gamma(C_m,\pi))$ contains 

1947: $\sigma_{d}(\Gamma^+(C_m,\pi))$ as a subgroup of finte index the result

1948: follows.   \end{proof}

1949: 

1950: We finally remark that

1951: $$\sigma_{d} ({\rm IA}(F_n))\le \GL^+(n-1,\BZ(\zeta_d))\qquad (d|m,\, d>1) 

1952: $$

1953: is equal to the full congruence subgroup of $\GL^+(n-1,\BZ(\zeta_d))$ with

1954: respect to the ideal of $\BZ(\zeta_d)$ generated by $\zeta_d-1$. This can be

1955: shown by a more detailed analysis of the arguments involved in the proof of

1956: Proposition \ref{cycfin}.

1957: 

1958: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1959: \section{Completion of proofs}\label{Proofs}

1960: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1961: 

1962: In this section we complete the proofs of all the results promised in the

1963: introduction. 

1964: 

1965: 

1966: Theorem \ref{prfit} proves Theorem \ref{teo}. Both theorems were formulated

1967: only for $n\ge 4$ but our method applies also to the

1968: case $n=3$. The only obstacle for $n=3$ is that congruence elementary matrices

1969: in $\SL(2)=\SL(n-1)$ do not always generate a finite index subgroup, for

1970: example the matrices 

1971: $$ \left(\begin{array}{cc}

1972:            1 &  3\\

1973:            0 &  1

1974: \end{array} \right),\qquad 

1975: \left(\begin{array}{cc}

1976:            1 &  0\\

1977:            3 &  1

1978: \end{array} \right)$$

1979: do not generate a subgroup of finite index in $\SL(2,\BZ)$. By the result of

1980: Vaserstein \cite{Vas1} this is a real obstacle (in the case of commutative

1981: rings of endomorphisms) only for $\SL(2,\BZ)$ and $\SL(2)$ of a ring of

1982: integers in an imaginary quadratic number field. Thus our results still work

1983: for $n=3$ in many cases. For example

1984: if the 

1985: group algebra $\BQ[G]$ of the finite group $G$ has 

1986: (except for the trivial module) no irreducible module $N$ with

1987: ${\rm dim}_D(N)=1$ ($D={\rm End}_G(N)$). If there are such  modules $N$ but

1988: if in each case $D$ is a number field which is not $\BQ$ 

1989: or imaginary quadratic we are also fine. The other cases remain open.

1990: 

1991: \smallskip

1992: \begin{prof} {\it of Theorem \ref{teo4}:} $n=2$: In this case we 

1993: just consider the (surjective) representation

1994: $$\rho_1 : \A(F_2)\to \A(F_2/F'_2)\cong \GL(2,\BZ).$$

1995: It is well known that $\GL(2,\BZ)$ contains a free subgroup 

1996: of any rank ($\ge 2$) of finite index. Hence any finitely generated group

1997: is the image of a subgroup of finite index in $\A(F_2)$. Note that the

1998: groups appearing in Theorem \ref{teo4} as image groups are finitely generated

1999: since they are arithmetic groups.   

2000: 

2001: $n=3$: Here we use the representation 

2002: $\sigma_{-1}:\Gamma^+(C_2,\pi)\to \GL(2,\BZ)$ 

2003: from Section \ref{Choose} to find 

2004: for every $r\in \BN$ with $r\ge 2$ a subgroup of finite index in $\A(F_3)$

2005: which can be mapped onto the free group of rank $r$. We then argue as in the

2006: case $n=2$.

2007: 

2008: $n\ge 4$: We first consider the case $k=1$. For $h,\, m\in\BN$ we have to find

2009: a subgroup $\Gamma\le \A(F_n)$ of finite index and a representation 

2010: $$

2011: \rho : \Gamma \to \SL((n-1)h,\BQ(\zeta_{m}))^{m} 

2012: $$

2013: such that $\rho(\Gamma)$ is commensurable with 

2014: $\SL((n-1)h,\BZ(\zeta_{m}))^{m}$. To do this we take the group

2015: $$G:=S_{h+1}\times C_m\times C_m.$$ 

2016: As an input for Theorem \ref{prfit} we need  suitable $\BQ[G]$-modules. We let

2017: $PM_{h+1}$ be the standard $h+1$ dimensional $\BQ[S_{h+1}]$-permutation

2018: module and let $M_h$ be the kernel of the augmentation map from 

2019: $PM_{h+1}$ to $\BQ$. The $\BQ[S_{h+1}]$-module $M_h$ is 

2020: irreducible and has $\BQ$-dimension

2021: $h$. We put $N:=\BQ(\zeta_m)\otimes_\BQ M_h$ and consider $N$ as a 

2022: $\BQ$-vector space. Here, as before, $\zeta_m\in\BC$ is a primitive $m$-th

2023: root of unity. Let $\epsilon : C_m\times C_m\to \langle \zeta_m\rangle$ be a

2024: surjective homomorphism. We turn the $\BQ$-vector space 

2025: $N=\BQ(\zeta_m)\otimes_\BQ M_h$ into a $\BQ[G]$-module $N_\epsilon$ by

2026: letting $S_{h+1}$ act on $M_h$ and letting $g\in C_m\times C_m$ act by

2027: multiplication by $\epsilon(g)$ on $\BQ(\zeta_m)$. The following are clear

2028: for every surjective homomorphism 

2029: $\epsilon : C_m\times C_m\to \langle \zeta_m\rangle$.

2030: \begin{itemize}     

2031: \item ${\rm dim}_\BQ(N_\epsilon)=\varphi(m)h$,

2032: \item ${\rm End}_G(N_\epsilon)=\BQ(\zeta_m)$,

2033: \item ${\rm dim}_{\BQ(\zeta_m)}(N_\epsilon)=h$,

2034: \item if $\epsilon_1,\,\epsilon_2:  C_m\times C_m\to \langle \zeta_m\rangle$ 

2035: are two surjective homomorphisms with distinct kernels then $N_{\epsilon_1}$

2036: and $N_{\epsilon_2}$ are not isomorphic as $\BQ[G]$-modules. 

2037: \end{itemize}     

2038: Our group $G$ can be generated by $3$ elements: take $g_1,\, g_2,\, g_3\in 

2039: S_{h+1}$ such that $g_3$ and the commutator of $g_1,\, g_2$ generate 

2040: $S_{h+1}$ and consider the elements 

2041: $(g_1,1,0)$, $(g_2,0,1)$, $(g_3,0,0)$ in $G$. Let $\pi :F_n\to G$ be a

2042: redundant presentation of $G$ (remember the assumption is $n\ge 4$). 

2043: Since the number of surjective homomorphisms 

2044: $\epsilon : C_m\times C_m\to \langle \zeta_m\rangle$ with pairwise distinct

2045: kernels is bigger or equal to $m$ we may apply Theorem \ref{prfit} to obtain

2046: a subgroup $\Gamma \le \A(F_n)$ and a representation 

2047: $$\sigma :\Gamma\to   \GL((n-1)h,\BQ(\zeta_m))^m$$

2048: such that $\sigma(\Gamma)\cap \SL((n-1)h,\BZ(\zeta_m))^m$ is of finite index 

2049: in $\SL((n-1)h,\BZ(\zeta_m))^m$. Here we do not know whether 

2050: $\sigma(\Gamma)\cap \SL((n-1)h,\BZ(\zeta_m))^m$ is of finite index in   

2051: $\sigma(\Gamma)$. However,

2052: there is a subgroup $\Delta\le \GL((n-1)h,\BZ(\zeta_m))^m$ of finite index

2053: such that $ \SL((n-1)h,\BZ(\zeta_m))^m\le \Delta$ and such that

2054: $\Delta\le Z\cdot \SL((n-1)h,\BZ(\zeta_m))^m$ with a subgroup 

2055: $Z\le \GL((n-1)h,\BZ(\zeta_m))^m$ which is central and satisfies 

2056: $Z\cap \SL((n-1)h,\BZ(\zeta_m))^m=\langle 1\rangle$. Let $\Gamma_0\le \Gamma$

2057: be the inverse image of $\Delta$ under $\sigma$. The representation 

2058: $$\sigma_0:  \Gamma_0\to   \Delta/Z=\SL((n-1)h,\BZ(\zeta_m))^m$$

2059: satisfies the requirements for $k=1$. Another way of looking at the last

2060: construction is to project a finite index subgroup of the image of 

2061: $\sigma(\Gamma)$ into ${\rm PGL}(n-1,\BZ(\zeta_m))$ which contains 

2062: ${\rm PSL}(n-1,\BZ(\zeta_m))$ as a finite index subgroup. We then intersect

2063: this subgroup in ${\rm PGL}(n-1,\BZ(\zeta_m))$ with 

2064: ${\rm PSL}(n-1,\BZ(\zeta_m))$ and pull the resulting subgroup of 

2065: ${\rm PSL}(n-1,\BZ(\zeta_m))$ back to obtain a finite index subgroup of

2066: $\Gamma$ which is mapped into ${\rm SL}(n-1,\BZ(\zeta_m))$.

2067: 

2068: We turn now to the general case. Let $\Gamma_1,\ldots,\Gamma_k\le \A(F_n)$

2069: be the subgroups of finite index and 

2070: $$\sigma_i:  \Gamma_i\to   \SL((n-1)h_i,\BZ(\zeta_{m_i}))^{m_i}\qquad

2071: (i=1,\ldots,k)$$

2072: be the representations constructed above. Set 

2073: $\Gamma:=\Gamma_1\cap\ldots\cap\Gamma_k$ and consider the representation

2074: $$\sigma:  \Gamma\to   \prod_{i=1}^k\SL((n-1)h_i,\BZ(\zeta_{m_i}))^{m_i}$$

2075: which maps an element $\gamma\in\Gamma$ to the tuple

2076: $(\sigma_1(\gamma),\ldots,\sigma_k(\gamma))$. The image of $\Gamma$ projects

2077: onto an arithmetic subgroup in each of the factors of the direct product.

2078: We finish the proof by applying the following lemma.

2079: \begin{lemma}\label{Margu}

2080: Let ${\cal H}_1,\ldots ,{\cal H}_k$ be pairwise non-isogeneous

2081: $\BQ$-defined simple linear algebraic groups which all have real rank greater

2082: or equal to $2$. Let

2083: $$\Gamma\le \prod_{i=1}^k{\cal H}_i^{m_i}(\BQ)$$ 

2084: be a subgroup such that all projections into the factors ${\cal

2085:   H}_i^{m_i}(\BQ)$ are arithmetic groups. Then $\Gamma$ is an arithmetic

2086: subgroup of $\prod_{i=1}^k{\cal H}_i^{m_i}$, that is $\Gamma$ is commensurable

2087: with $\prod_{i=1}^k{\cal H}_i^{m_i}(\BZ)$.

2088: \end{lemma}

2089: This lemma is proved using Margulis-superrigidity. See \cite{L} item (iv) on

2090: pages 328-329 or

2091: \cite{GP} Section 2. 

2092: \end{prof}

2093: 

2094: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2095: \section{(SL) or not (SL)?}\label{SL}

2096: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2097: 

2098: In this section we show some results concerning

2099: the question whether $\rho_{G,\pi}(\Gamma(G,\pi))\cap{\cal G}^1_{G,\pi}$ is of

2100: finite index in $\rho_{G,\pi}(\Gamma(G,\pi))$ or not (for the notation see the

2101: introduction). The relevance of this 

2102: question and the notation are explained in the introduction. We change 

2103: here from

2104: the consideration of the rational relation module to its complex version. 

2105: This brings certain technical advantages and still 

2106: reflects the problem.  

2107: 

2108: Let us set up some notation for this section. Let $G$ be a finite group and

2109: $\pi:F_n\to G$ an epimorphism with kernel $R$. We define $\Gamma(G,\pi)$, 

2110: ${\cal G}_{G,\pi}$ and ${\cal G}^1_{G,\pi}$ as in the introduction and

2111: consider the action of $\Gamma(G,\pi)$ now on the complex relation module

2112: $\BC\otimes_\BZ\bar R$. The result of Gasch\"utz now says that

2113: $$\BC\otimes_{\bb Z}\bar R=\BC\oplus \BC[G]^{n-1}$$ 

2114: as $\BC[G]$-modules.

2115: 

2116: If $Q$ is an irreducible $\BC[G]$-module and 

2117: $M$ is an arbitrary $\BC[G]$-module we keep the notation ${\bf I}_Q(M)$ 

2118: for the $Q$-isotypic component inside $M$. We hope that this creates no

2119: confusion with the ${\bf I}_N(M)$ of Section \ref{sura} where $N,\, M$ are 

2120: $\BQ[G]$-modules.

2121: 

2122: Let $Q$ be an non-trivial irreducible $\BC[G]$-module of complex dimension 

2123: $h_Q$. Let ${\cal G}_{G,\pi,Q}$ be the stabiliser of 

2124: $I_Q(\BC\otimes_\BZ\bar R)$ inside ${\cal G}_{G,\pi}$ (see \ref{stabi}).

2125: By our usual procedure we obtain a representation 

2126: \begin{equation}

2127: \rho_{G,\pi,Q}:\Gamma(G,\pi)\to {\cal G}_{G,\pi,Q}=\GL(n-1,M(h_Q,\BC))=

2128: \GL((n-1)h_Q,\BC).

2129: \end{equation}

2130: We call the pair $(G,Q)$ of {\it type (SL)} if 

2131: $\rho_{G,\pi,Q}(\Gamma(G,\pi))\cap  \SL((n-1)h_Q,\BC)$ is of finite index in 

2132: $\rho_{G,\pi,Q}(\Gamma(G,\pi))$ for every $n\in \BN$ and every epimorphism

2133: $\pi:F_n\to G$. We say the finite group $G$ is of {\it type (SL)}

2134: if for all non-trivial irreducible $\BC[G]$-module $Q$ the pair 

2135: $(G,Q)$ is of type (SL). 

2136: 

2137: Note that, since $\Gamma(G,\pi)$ is finitely generated, 

2138: $\rho_{G,\pi,Q}(\Gamma(G,\pi))\cap  \SL((n-1)h_Q,\BC)$ is of finite index in 

2139: $\rho_{G,\pi,Q}(\Gamma(G,\pi))$ if and only if $\rho_{G,\pi,Q}(\Gamma(G,\pi))$

2140: contains no element with a determinant of infinite order.

2141: 

2142: Note that the above implies that once we find a finite group $G$

2143: which is not of type (SL), we have found a subgroup $\Gamma$ of finite 

2144: index in some $\A(F_n)$ which has an epimorphism

2145: onto $\BZ$.

2146: 

2147: Let $G$ be a finite group $G$ and $N_2,\ldots,N_\ell$ its non-trivial

2148: irreducible $\BQ[G]$-modules. Let $\pi:F_n\to G$ be a surjective homomorphism.

2149: The following are equivalent: 

2150: \begin{itemize}

2151: \item For all $i=2,\ldots,\ell$ the intersection  

2152: $$\sigma_{G,\pi,i}(\Gamma(G,\pi))\cap \SL((n-1)h_i,{\cal R}_i^{\rm op})$$

2153: has finite index in  

2154: $\sigma_{G,\pi,i}(\Gamma(G,\pi))\le \GL((n-1)h_i,{\cal R}_i^{\rm op})$.

2155: \item For all non-trivial  

2156: irreducible $\BC[G]$-modules $Q$ the intersection 

2157: $$\rho_{G,\pi,Q}(\Gamma(G,\pi))\cap \SL((n-1)h_Q,\BC)$$

2158:  is of finite index in $\rho_{G,\pi,Q}(\Gamma(G,\pi))$.

2159: \end{itemize}

2160: This is seen by introducing the decomposition of the

2161: complexifications $\BC\otimes_\BQ N_i$ into irreducible $\BC[G]$-modules.  

2162: 

2163: There are some non-trivial irreducible $\BC[G]$-modules $Q$ which are

2164: obviously of type (SL):

2165: \begin{proposition}\label{aal}

2166: Let $G$ be a finite group and $Q$ a non-trivial irreducible $\BC[G]$-module.

2167: If there is an (irreducible) $\BQ[G]$-module $N$ with $Q=\BC\otimes_\BQ N$

2168: then $(G,Q)$ is of type (SL).

2169: \end{proposition}

2170: \begin{proof}

2171: We note that under the above assumptions the endomorphism algebra 

2172: $D:={\rm End}_G(N)$ is a division algebra 

2173: over the rational numbers with $\BQ$ as

2174:   center. In this case $\SL(m,{\cal R})$ (i.e. the group of elements of

2175:   reduced norm $1$ in $M(m,{\cal R})$) is of finite index in 

2176: $\SL(m,{\cal R})$ for any order ${\cal R}\le D$ and any $m\in \BN$.

2177: \end{proof}

2178: In case of the symmetric groups $S_m$ ($m\in \BN$) it is known, that every

2179: $\BC[S_m]$-module satisfies the assumptions of Proposition \ref{aal}, see

2180: \cite{S}, Section 13.  This implies:

2181: \begin{corollary}

2182: All the symmetric groups $S_m$ are of type (SL).

2183: \end{corollary}

2184: In case of the alternating group $A_5$ Proposition \ref{aal} applies to

2185: the irreducible $\BC[A_5]$-modules of dimensions $4$ and $6$ but not to the

2186: two $3$-dimensional irreducible $\BC[A_5]$-modules. They fit together to form

2187: a $6$-dimensional irreducible $\BQ[A_5]$-module $N$ with 

2188: ${\rm End}_G(N)=\BQ(\sqrt{5})$.

2189: 

2190: 

2191: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2192: \subsection{Reduction Steps}\label{SL2}

2193: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2194: 

2195: After the definitions above we discuss some reduction steps which will allow 

2196: us to recognise pairs $(G,Q)$ of type (SL).

2197: 

2198: \begin{lemma}\label{redu01}

2199: Let $G$ be a finite group with normal subgroup $H_0$ and quotient $H=G/H_0$. 

2200: Let $Q$ be a 

2201: non-trivial irreducible $\BC[G]$-module on which $H_0$ acts trivially. If 

2202: $(H,Q)$ is of type (SL) then $(G,Q)$ also has this property.

2203: \end{lemma}

2204: \begin{proof}  Let $\pi:F_n\to G$ ($n\ge 2$) be any epimorphism 

2205: and let $R_G$ be the kernel of $\pi$. Let $\pi_0 :F_n\to G\to H$ be the

2206: resulting presentation of $H$ and $R_H$ its kernel. Obviously $R_G$ is a

2207: subgroup of finite index in $R_H$. It is also easily seen that $\Gamma(G,\pi)$

2208: is a subgroup of finite index of $\Gamma(H,\pi_0)$. The inclusion of   

2209: $R_G$ into $R_H$ gives rise to a surjective $\BC$-linear map

2210: $$\Theta: \BC\otimes_{\bb Z} \bar R_G\to \BC\otimes_{\bb Z} \bar R_H.$$

2211: This linear map is equivariant for the induced actions of $G$ and 

2212: $\Gamma(G,\pi)$.

2213: By restriction we obtain a map

2214: \begin{equation}\label{fro1}

2215: \Theta_Q: {\bf I}_Q(\BC\otimes_{\bb Z} R_G)\to{\bf I}_Q(\BC\otimes_{\bb Z} R_H)

2216: \end{equation}

2217: which is surjective and $H$-equivariant. 

2218: Gasch\"utz' result implies that $\Theta_Q$ is in

2219: fact an ($\Gamma(G,\pi)$-equivariant) isomorphism.  

2220: 

2221: We have to show that ${\rm det}(\rho_{G,\pi,Q}(\varphi))$ 

2222: is of finite order

2223: for any $\varphi\in\Gamma(G,\pi)$. We compute this determinant on the right

2224: hand side of (\ref{fro1}) where our assumption implies this property.

2225: \end{proof}

2226: 

2227: \begin{lemma}\label{redu2}

2228: Let $G$ be a finite group with  normal subgroup $H$. Let $Q$ be a 

2229: non-trivial irreducible $\BC[G]$-module which is induced from $H$, that is,

2230: $Q$ is of the form $Q={\rm Ind}_H^G(Q_0)$ where $Q_0$ is an irreducible

2231: $\BC[H]$-module. If 

2232: $(H,Q_0)$ is of type (SL) then $(G,Q)$ also has this property.

2233: \end{lemma}

2234: \begin{proof} Let $\pi:F_n\to G$ ($n\ge 2$) be any epimorphism 

2235: and let $R$ be the kernel of $\pi$. The subgroup 

2236: $\pi^{-1}(H)\le F_n$ is a free group on $m=1+[G:H](n-1)$ generators, 

2237: we denote it

2238: by $F_m$. We then have a commutative diagram  

2239: $$\xymatrix{

2240: 1 \ar[r] & R  \ar[r]  & F_n \ar[r]^{\pi} & G \ar[r] & 1\\                      

2241: 1 \ar[r] & R  \ar[u]^{=}\ar[r] & F_m \ar[u]\ar[r]^{\pi_0} & H\ar[u]\ar[r] & 1\\

2242: }

2243: $$

2244: where $\pi_0$ is the restriction of $\pi$ to $F_m$. Note that 

2245: $R={\rm ker}(\pi)={\rm ker}(\pi_0)$.  

2246: 

2247: An element $\varphi\in\Gamma(G,\pi)$ stabilises the subgroup $F_m\le F_n$ and

2248: as an automorphism of $F_m$ it lies in $\Gamma(H,\pi_0)$. On the vector space 

2249: $\BC\otimes_\BZ\bar R$ the linear maps induced by $\varphi$ as an element of 

2250: $\Gamma(G,\pi)$ and by $\varphi$ as an element of $\Gamma(H,\pi_0)$ coincide.

2251: We denote both by $\bar\varphi$. 

2252: 

2253: The complex relation module $\BC\times_\BZ\bar R$ now has a structure as

2254: a $\BC[G]$-module and a structure of a $\BC[H]$-module. In fact the second is

2255: just the restriction of the first. 

2256: 

2257: Let $Q=Q_0\oplus Q_1\oplus\ldots \oplus Q_r$ be the decomposition of $Q$ into

2258: irreducible $\BC[H]$-modules 

2259: We now prove that 

2260: \begin{equation}\label{frob}

2261: {\bf I}_Q(\BC\otimes_{\bb Z}\bar R)=

2262: {\bf I}_{Q_0}(\BC\otimes_{\bb Z}\bar R)\oplus

2263: {\bf I}_{Q_1}(\BC\otimes_{\bb Z}\bar R)\oplus\ldots \oplus 

2264: {\bf I}_{Q_r}(\BC\otimes_{\bb Z}\bar R).

2265: \end{equation}

2266: On the left hand side the isotypic component of the $\BC[G]$-module $Q$ is

2267: taken whereas on the right hand side we see the the isotypic component of the 

2268: corresponding $\BC[H]$-modules. Clearly the left hand side of (\ref{frob})

2269: is contained in the right hand side.

2270: Note that, since $H$ is normal in $G$ the module $Q$ is also induced from any

2271: of the $Q_0,\ldots, Q_r$. 

2272: The Frobenius reciprocity law (\cite{S}, chapter 7) 

2273: implies that none of the $Q_0,\ldots, Q_r$ can occur as an irreducible

2274: constituent of the restriction of an irreducible $\BC[G]$-module different

2275: from $Q$. 

2276: 

2277: We have to show that ${\rm det}(\rho_{G,\pi,Q}(\varphi))$ is of finite order

2278: for any $\varphi\in\Gamma(G,\pi)$. We compute this determinant on the right

2279: hand side of (\ref{frob}) where our assumption implies this property.

2280: \end{proof}

2281: 

2282: 

2283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2284: \subsection{Abelian and metabelian groups}\label{SL3}

2285: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2286: 

2287: This section contains results which describe the image of $\rho_{G,\pi}$ for

2288: abelian and then for metabelian groups $G$ up to commensurability. In Section  

2289: \ref{Choose} we have proved a very precise result in this direction for cyclic

2290: groups $G$. This allowed us to conclude

2291: for $n\ge 2$ that the image of $\Gamma(G,\pi)$ 

2292: (under certain linear representations) has finite abelianisation. 

2293: We present here the following much stronger result for $n\ge 3$. This result

2294: is due to A. Rapinchuk (it appears in a somewhat hidden form in \cite{RA},

2295: page 150), we thank him for the permission to include it. 

2296: \begin{proposition}\label{rapi}

2297: Let $n\ge 3$ be a natural number, $G$ a finite abelian group $G$ and let

2298: $\pi:F_n\to G$ be any surjective homomorphism. Then $\Gamma(G,\pi)$ has finite

2299: abelianisation. 

2300: \end{proposition}

2301: \begin{proof}

2302: Since $G$ is abelian we have $F_n'\le R$ where $R$ is, as always, the kernel

2303: of $\pi$. Consequently we have ${\rm IA}(F_n)\le \Gamma(G,\pi)$. Writing

2304: $\tilde\Gamma(G,\pi):=\rho_1(\Gamma(G,\pi))\le \GL(n,\BZ)$ for the image of 

2305: $\rho_1(\Gamma(G,\pi))$ in $\GL(n,\BZ)$ we have the exact sequence of groups

2306: $$1\to {\rm IA}(F_n)\to \Gamma(G,\pi)\to \tilde\Gamma(G,\pi)\to 1$$

2307: Note that $\tilde\Gamma(G,\pi)$ is a subgroup of finite index in 

2308: $\GL(n,\BZ)$. Since $\Gamma(G,\pi)$ is a finitely generated group we need only

2309: show that 

2310: $$H^1(\Gamma(G,\pi),\BQ)={\rm Hom}(\Gamma(G,\pi),\BQ)=0$$

2311: We shall do that by applying the edge-term sequence of the Hochschild-Serre

2312: spectral sequence for the above short exact sequence of groups. We shall use

2313: that there is an exact sequence of $\BQ$-vector spaces

2314: \begin{equation}\label{Hose}

2315: H^1(\tilde\Gamma(G,\pi),\BQ)\to H^1(\Gamma(G,\pi),\BQ)\to 

2316: H^1({\rm IA}(F_n),\BQ)^{\tilde\Gamma(G,\pi)}

2317: \end{equation}

2318: First of all the left hand side of (\ref{Hose}) is $0$ since $n\ge 3$ and

2319: $\GL(n,\BZ)$ has Kazdhan's property (T).

2320: 

2321: Next we evaluate the right hand side of  (\ref{Hose}). By a theorem of

2322: Bachmuth (\cite{Ba} and also \cite{F}) there is a $\A(F_n)$-equivariant

2323: isomorphism 

2324: $$H^1({\rm IA}(F_n),\BZ)\to {\rm Hom}_{\bb Z}(F_n/F_n',F_n'/\gamma_2(F_n))$$

2325: where $\gamma_2(F_n):=[[F_n,F_n],F_n]$ is the third term of the lower central

2326: series of $F_n$. Let $\GL^\pm (n,\BC)$ be the subgroup of   

2327: $\GL (n,\BC)$ consisting of those elements having determinant $\pm 1$. Let

2328: $\BC^n$ be the standard $\GL^\pm (n,\BC)$-module of dimension $n$ and

2329: $\Lambda^2(\BC^n)$ be its outer square. Consider $\GL^\pm (n,\BZ)$ as a

2330: subgroup of $\GL^\pm (n,\BC)$. By identifying $F_n'/\gamma_2(F_n))$ with

2331: $\Lambda^2(F_n/F_n')=\Lambda^2(\BZ^n)$ we obtain an

2332: $\GL^\pm (n,\BZ)$-equivariant

2333: isomorphism  

2334: $$\BC\otimes_{\bb Z} {\rm Hom}_{\bb Z}(F_n/F_n',F_n'/\gamma_2(F_n))\to

2335: {\rm Hom}(\BC^n,\Lambda^2(\BC^n)).$$

2336: It is well known (see \cite{F}, page 427)

2337: that the right hand side splits as a $\GL^\pm (n,\BC)$-module

2338: as 

2339: $${\rm Hom}(\BC^n,\Lambda^2(\BC^n))=\BC^n\oplus V_n$$

2340: where $V_n$ is a $\GL^\pm (n,\BC)$-module of dimension 

2341: ${\rm dim}_\BC(V_n)=n(n+1)(n-2)/2$ which is irreducible even as 

2342: $\SL(n,\BC)$-module. Hence the space of $\SL(n,\BC)$-invariants in   

2343: ${\rm Hom}(\BC^n,\Lambda^2(\BC^n))$ is $0$. Since the Zariski closure of

2344: $\tilde\Gamma(G,\pi)$ contains $\SL(n,\BC)$ the group 

2345: ${\rm Hom}_\BZ(F_n/F_n',F_n'/\gamma_2(F_n))^{\tilde\Gamma(G,\pi)}$ is $0$.

2346: This shows that the right hand side of (\ref{Hose}) is $0$. Together with the

2347: above the proposition is proved.

2348: \end{proof}

2349: Proposition \ref{rapi} is clearly not true for $n=2$, since the left-most term

2350: in (\ref{Hose}) contributes to the $H^1(\Gamma (G,\pi),\BQ)$. See also the

2351: tables in Section \ref{Comp}. 

2352: As an immediate corollary from Proposition \ref{rapi} we get:

2353: \begin{corollary}

2354: All finite abelian groups are of type (SL).

2355: \end{corollary}

2356: This corollary can also be deduced from Corollary \ref{cycfincor} (in fact

2357: this corollary is needed in the case $n=2$) and  Lemma \ref{redu01}.

2358: But the statement

2359: incorporated in Proposition \ref{rapi} (for $n\ge 3$) is in fact stronger.

2360: 

2361: 

2362: A metabelian group is a group $G$ with abelian commutator subgroup $G'$.

2363: We prove:

2364: \begin{proposition}

2365: All metabelian finite groups $G$ are of type (SL)

2366: \end{proposition}

2367: \begin{proof}

2368: We use induction on the order of $G$ and the fact proved above that all finite

2369: abelian groups have property (SL).

2370: 

2371: Let $Q$ be a non-trivial irreducible $\BC[G]$-module.

2372: We may assume by induction and Lemma \ref{redu01} that $G$ acts faithfully on

2373: $Q$. 

2374: 

2375: {Case 1:} the commutator subgroup is not contained in the center ${\rm Z}(G)$ 

2376: of $G$. The restriction of the module $Q$ to $G'$ cannot be

2377: isotypic. Otherwise the abelian subgroup $G'$ would act by scalar matrices on

2378: $Q$ and $G'$ would have to be contained in the center of $G$. 

2379: Now Proposition 24 of \cite{S} says that $Q$ is either isotypic or induced.

2380: It follows that

2381: $Q$ is induced from a proper subgroup $H$ of $G$ containing $G'$.

2382: 

2383: By our construction $H$ is normal in $G$. 

2384: Hence this case is finished by induction and

2385: application of Lemma \ref{redu2}.   

2386: 

2387: {Case 2:} the commutator subgroup is contained in the center ${\rm Z}(G)$ 

2388: of $G$. In this case $G$ is nilpotent of class $2$. Unless $G$ is abelian (the

2389: case treated above) we find $g\in G$ with $g\notin {\rm Z}(G)$. The subgroup

2390: $A:=\langle G',g\rangle$ generated by $G'$ and $G$ is abelian, normal in $G$

2391: and not contained in ${\rm Z}(G)$. We then proceed as in case 1. 

2392: \end{proof}

2393: 

2394: 

2395: 

2396: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2397: \section{Concluding remarks and suggestions for further research}\label{Concl}

2398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2399: 

2400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2401: \subsection{Dependence on the presentation}\label{Concl3}

2402: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2403: 

2404: The results of the current paper can be considered as a first step toward a

2405: systematic study of the relation module of a finite group as a

2406: $\Gamma(G,\pi)$-module, i.e. an equivariant Gasch\"utz's theory. Here we gave

2407: a quite satisfactory answer in the case of redundant presentations. It is not

2408: clear to what extend this represents the general case. 

2409: 

2410: For a fixed finite group $G$ let us look at all the possible epimorphisms from

2411: the free group $F_n$ to $G$ and ${\bf R}(n,G)$ the set of their kernels. The

2412: automorphism group $\A(F_n)$ acts on this set. It is easy to see that the

2413: equivariant Gasch\"utz theory that we are trying to develop here depends only

2414: on the orbits of  $\A(F_n)$ on ${\bf R}(n,G)$ and not on the actual

2415: presentation. Various authors studied the transitivity properties 

2416: of the action of $\A(F_n)$ on ${\bf R}(n,G)$. It is not transitive in general 

2417: (a first example was given by B. H. Neumann \cite{Neu})

2418: but it is in some interesting cases. For example, it is transitive if $G$ is a

2419: cyclic group and $n\ge 2$. Thus our results in Section \ref{Choose} give the

2420: full picture for all relation modules of cyclic groups. It is also

2421: transitive if $n>2{\rm log}_2(|G|)$ or if $G$ is solvable and $n>d(G)$, where

2422: $d(G)$ denotes the minimal number of generators of $G$ (see \cite{DU}). An

2423: old conjecture of Wiegold predicts transitivity for finite simple groups if

2424: $n\ge 3$, in which case again $n>d(G)$ holds. For some partial results see

2425: \cite{Gi}, \cite{Ev} and the references in \cite{Pak}. 

2426: It was even suggested in \cite{Pak}

2427: that for every finite group $G$ it is transitive if $n>d(G)$. 

2428: 

2429: In \cite{Ev} it is shown that if a finite group $G$ 

2430: has spread 2 (i.e. for every

2431: $g_1,\, g_2\in G$ with $g_1\ne 1 \ne g_2$

2432: there is  $h\in G$ such that $\langle g_1,\, h\rangle=    

2433: \langle g_2,\, h\rangle=G$) then $\A(F_n)$ acts transitively on the set of

2434: kernels of all redundant presentations. Guralnick and Shalev  

2435: \cite{GS} proved that almost all

2436: finite simple groups have spread 2.

2437: 

2438: Theorem 1.4 is not true in general if the presentation is not redundant, at

2439: least when $n=2$, see the $A_5$-example in Section \ref{Comp}.

2440: The story for $n\ge 3$ may be different as $\A(F_2)$ behaves

2441: in many ways differently from $\A(F_n)$ for $n\ge 3$.

2442: 

2443: 

2444: 

2445: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2446: \subsection{Representations of IA($F_n$)}\label{Concl1}

2447: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2448: 

2449: A well known theorem of Formanek and Procesi (see \cite{FP}) asserts 

2450: that the automorphism group $\A(F_n)$ has, for $n\ge 3$, no faithful

2451: linear representation over any field. Their proof suggests a stronger  

2452: 

2453: {\it Conjecture: For every linear representation $\rho$ of $\A(F_n)$ the image 

2454: $\rho({\rm Inn}(F_n))$ is virtually solvable.}

2455: 

2456: As far as we know, in all previously constructed cases of linear

2457: representations even 

2458: $\rho({\rm IA}(F_n))$ had this property. 

2459: It should be mentioned that in all the representations $\rho$ studied in this

2460: paper, $\rho({\rm Inn}(F_n))$ is finite but $\rho({\rm IA}(F_n))$ is far from

2461: being virtually solvable and, in fact, our results show that 

2462: $\rho({\rm IA}(F_n))$ can be Zariski dense in very large semi-simple groups. 

2463: Our reprentations seem to be the first known with this property. 

2464: We have:   

2465: \begin{proposition}\label{teo3}

2466: Let $n\ge 3$, $k\ge 1$, $h_1<\ldots <h_k$, $m_1,\ldots ,m_k$ be 

2467: natural numbers.  Let $\BQ(\zeta_{m_i})$ be the field of 

2468: $m_i$-th roots of unity and

2469: $\BZ(\zeta_{m_i})$ its ring of integers. There is a 

2470: subgroup $\Gamma\le {\rm IA}(F_n)$ of finite index and a representation 

2471: $$

2472: \rho : \Gamma \to \prod_{i=1}^k\SL((n-1)h_i,\BQ(\zeta_{m_i}))^{m_i} 

2473: $$

2474: such that $\rho(\Gamma)$ is commensurable with 

2475: $\prod_{i=1}^k\SL((n-1)h_i,\BZ(\zeta_{m_i}))^{m_i}$. 

2476: \end{proposition}

2477: \begin{proof}

2478: The proof of Theorem \ref{teo4} shows that there is a subgroup of finite index

2479: $\Gamma_0\le\A(F_n)$ and a representation 

2480: \begin{equation}\label{Mech}

2481: \rho_0 : \Gamma_0 \to \SL(n,\BZ)\times

2482: \prod_{i=1}^k\SL((n-1)h_i,\BQ(\zeta_{m_i}))^{m_i} 

2483: \end{equation}

2484: such that the image $\rho_0(\Gamma_0)$ is the internal direct product of its

2485: intersections with the factors in (\ref{Mech}). 

2486: These intersections in turn have

2487: finite index in the corresponding factor. Using the fact that a normal

2488: subgroup of a subgroup of finite index in $\SL(n,\BZ)$ ($n\ge 3$) is either

2489: finite or has finite index we conclude the proof of the prposition.

2490: \end{proof}

2491: 

2492: Another easy consequence of Theorem \ref{teo4} concerns representations of the

2493: the outer automorphism group ${\rm Out}(F_n)$. We have:

2494: \begin{proposition}\label{teoo}

2495: Let $n\ge 2$, $k\ge 1$, $h_1<\ldots <h_k$, $m_1,\ldots ,m_k$ be 

2496: natural numbers.  Let $\BQ(\zeta_{m_i})$ be the field of 

2497: $m_i$-th roots of unity and

2498: $\BZ(\zeta_{m_i})$ its ring of integers. There is a 

2499: subgroup $\Gamma\le {\rm Out}(F_n)$ of finite index and a representation 

2500: $$

2501: \rho : \Gamma \to \prod_{i=1}^k\SL((n-1)h_i,\BQ(\zeta_{m_i}))^{m_i} 

2502: $$

2503: such that $\rho(\Gamma)$ is commensurable with 

2504: $\prod_{i=1}^k\SL((n-1)h_i,\BZ(\zeta_{m_i}))^{m_i}$. 

2505: \end{proposition}

2506: Here we only have to note that the image of the inner automorphisms under

2507: any of the representations from Theorem \ref{teo4} is finite. This finite

2508: subgroup of the image can be avoided by going to a subgroup of finite index in

2509: $\Gamma$. 

2510: 

2511: As a consequence of Propositions \ref{teo3}, \ref{teoo} we get:

2512: \begin{corollary}\label{coro8}

2513: The groups ${\rm IA}(F_3)$ and ${\rm Out}(F_3)$ are large.

2514: \end{corollary}

2515: 

2516: Our result from Theorem \ref{teo4} can 

2517: conviniently be summarised in the language 

2518: of the pro-algebraic completion ${\cal A}(\A(F_n))$ of $\A(F_n)$. 

2519: For a definition and

2520: discussion of properties of the pro-algebraic completion (also called the

2521: Hochschild-Mostow group) of a group see

2522: \cite{LM} and \cite {BLMM}. We have 

2523: \begin{proposition}\label{teo9} 

2524: Let $n\ge 2$ be a natural number and  ${\cal S}(\A(F_n))$ be

2525: the semisimple part of the connected component of the identity of the 

2526: pro-algebraic completion ${\cal A}(\A(F_n))$. Then for every $h\in \BN$, the

2527: group $\SL((n-1)h,\BC)$ appears infinitely many times as a factor 

2528: in ${\cal S}(\A(F_n))$. That is, there is a surjective homomorphism

2529: $$

2530: {\cal S}(\A(F_n))\to \prod_{h=1}^\infty \prod_{i=1}^\infty \SL((n-1)h,\BC).

2531: $$

2532: \end{proposition}

2533: Propositions \ref{teo3}, \ref{teoo} imply similar results for the groups

2534: ${\rm IA}(F_n)$, ${\rm Out}(F_n)$ ($n\ge 3$).

2535: 

2536: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2537: \subsection{Representations of $\Gamma(G,\pi)$ into affine groups}\label{Lini}

2538: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2539: 

2540: In this section we give another construction of a linear representation

2541: of subgroups of finite index in $\A(F_n)$. This representation takes values in 

2542: affine groups (which have a non-trivial unipotent radical). The construction

2543: originates from the proof of Gasch\"utz' result as presented in \cite{JR}.

2544: Besides giving new types of image groups, it is of importance for the

2545: computer algorithms which we have used to create the computational results

2546: described in the next subsection.  

2547: 

2548: Let $G$ be a finite group and $\pi : F_n\to G$ a surjective homomorphism

2549: of the free group $F_n$ onto $G$. Let $R$ be the kernel of 

2550: $\pi$ and $\bar R$ the corresponding relation module. Let further

2551: $\Gamma(G,\pi)$ be the subgroup of $\A(F_n)$ defined in 

2552: (\ref{I1}). Given a group $H$ and a commutative ring $S$ we write $S[H]$ for

2553: the corresponding group ring and ${\cal I}(S[H])$ for its augmentation ideal.

2554: If $H_0\le H$ is a subgroup we define ${\cal I}(S[H],H_0)$ to be the two-sided

2555: ideal of $S[H]$ generated by the $h-1$ for $h\in H_0$.  

2556: 

2557: 

2558: 

2559: Gasch\"utz' result (\ref{gasch}) is proved (in \cite{JR}) 

2560: by considering the exact sequence

2561: \begin{equation}\label{gashpr}

2562: 0\to \BQ\otimes_\BZ \bar R\to {\cal I}(\BQ[F_n])/({\cal I}(\BQ[F_n],R)

2563: \cdot{\cal I}(\BQ[F_n]))

2564: \to {\cal I}(\BQ[G])\to 0

2565: \end{equation}

2566: of $G$-equivariant homomorphisms.

2567: The right hand map is induced by $\pi : F_n\to G$ while the left hand map

2568: comes from the map from $R$ to ${\cal I}(\BQ[F_n])$ which sends $r\in R$ to

2569: $r-1$. The free group $F_n$ acts on the middle term in (\ref{gashpr}) by

2570: multiplication from the left. This leads to an action of the 

2571: finite group $G$ on this term. 

2572: 

2573: We now let $\Gamma(G,\pi)$ act on the left hand term as before, 

2574: on the middle term

2575: by its action on $F_n$ and trivially on $\BQ[G]$.

2576: It is straightforward to see that the sequence

2577: (\ref{gashpr}) is then $\Gamma(G,\pi)$-equivariant. We obtain a

2578: representation

2579: \begin{equation}\label{gashprep}

2580: \eta_{G,\pi}:\Gamma(G,\pi)\to 

2581: {\rm Hom}_G({\cal I}(\BQ[G]),\BQ\otimes_{\bb Z} \bar R)\rfish {\cal

2582:   G}_{G,\pi}(\BQ).

2583: \end{equation}

2584: The semi-direct product in (\ref{gashprep}) is formed with respect to the

2585: action of ${\cal G}_{G,\pi}(\BQ)$ on $\BQ\otimes_\BZ \bar R$.

2586: 

2587: Our methods show:

2588: \begin{theorem}\label{teo16} Assume $n$ is a natural number with $n\ge 4$.

2589: Let $\pi : F_n\to G$ be a redundant presentation of the finite group $G$. Then

2590: $$\eta_{G,\pi}(\Gamma(G,\pi))\cap 

2591: {\rm Hom}_G({\cal I}(\BQ[G]),\BQ\otimes_\BZ \bar R)\rfish 

2592: {\cal G}^1_{G,\pi}(\BQ)$$ is 

2593: of finite index in the arithmetic group 

2594: ${\rm Hom}_G({\cal I}(\BZ[G]),\bar R)\rfish{\cal G}_{G,\pi}^1(\BZ)$.

2595: \end{theorem}

2596: Theorem \ref{teo16} can then be used to prove

2597: \begin{theorem}\label{teo17}

2598: Let $n\ge 2$, $h$, $m$ be 

2599: natural numbers.  Let $\BQ(\zeta_{m})$ be the field of 

2600: $m$-th roots of unity and

2601: $\BZ(\zeta_{m})$ its ring of integers. There is a 

2602: subgroup $\Gamma\le \A(F_n)$ of finite index and a representation 

2603: $$

2604: \eta : \Gamma \to \left(\BQ(\zeta_m)^{(n-1)h}\rfish

2605: \SL((n-1)h,\BQ(\zeta_{m}))\right)^{m} 

2606: $$

2607: such that $\eta(\Gamma)$ is commensurable with 

2608: $(\BZ(\zeta_m)^{(n-1)h}\rfish  \SL((n-1)h,\BZ(\zeta_{m})))^{m}$. 

2609: \end{theorem}

2610: 

2611: 

2612: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2613: \subsection{Computational results}\label{Comp}

2614: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2615: 

2616: An important feature of our methods is that subgroups of quite high indices in

2617: $\A(F_n)$ may be computationally located. We describe here some of the results

2618: of our computer calculations. The computations where done using the computer

2619: algebra system MAGMA. 

2620: 

2621: We shall describe our approach now in detail for $n=2$. Let 

2622: $F_2:=\langle x,\, y\rangle$ be the free group. The automorphism group  

2623: $\A^+(F_2)$ is generated by $\alpha,\, \beta$ which are given by

2624: $$\alpha(x)=y^{-1},\quad \alpha(y)=x\qquad \beta(x)=x^{-1}y^{-1},\quad  

2625: \beta(y)=x.$$

2626: The group $\A^+(F_2)$ is finitely presented, a presentation is (see

2627: \cite{Neu1}) 

2628: \begin{equation}\label{neuma}

2629: \A^+(F_2)=\langle\, \alpha,\, \beta \Mid \alpha^4=\beta^3=

2630: \alpha^2\beta^2\alpha^2\beta\alpha\beta\alpha^2\beta^2\alpha=1\,\rangle.

2631: \end{equation}

2632: Let $\pi: F_2\to G$ be a presentation of a finite group. If $\varphi\in 

2633: \A^+(F_2)$ is given by its values on $x,\, y$ it is easy to check whether

2634: $\varphi$ is in $\Gamma(G,\pi)$ or not. Using a random process we generate

2635: a list of words $\varphi_1,\ldots, \varphi_k\in \Gamma(G,\pi)$ in the

2636: automorphisms $\alpha,\, \beta$. We then wait until

2637: $$\Delta:=\langle \, \varphi_1,\ldots, \varphi_k\,\rangle$$

2638: has finite index in $\A^+(F_2)$. This is checked using a Todd-Coxeter algorithm

2639: using the presentation (\ref{neuma}). By this approach we first of all find the

2640: following table. The first column contains the finite group $G$, which

2641: we always think of being given as a permutation group. The third and fourth

2642: column contain the images $x,\, y$ under the 

2643: surjective homomorphism $\pi:F_2\to G$.

2644: The fifth column gives the index of 

2645: $\Delta\le \Gamma(G,\pi)$ in $\A^+(F_2)$. We give

2646: the abelianised group $\Delta^{\rm ab}$ in the last column.   

2647: \bigskip

2648: \begin{center}

2649: \begin{tabular}{|c|c|c|c|c|c|}

2650: \hline

2651: $G$  & $|G|$ & $\pi(x)$ & $\pi(y)$ & $[\A^+(F_2):\Delta]$ & $\Delta^{\rm ab}$\\

2652: \hline

2653: $C_2$ & 2 & (1,2) & (1) & 3 & $\BZ^2\times C_2\times C_4$\\

2654: $C_3$ & 3 & (1,2,3) & (1) & 8 & $\BZ\times C_3^2$\\

2655: $C_4$ & 4 & (1,2,3,4) & (1) & 12 & $\BZ^2\times C_4$\\

2656: $C_2\times C_2$ & 4 & (1,2) & (3,4) & 6 & $\BZ^2\times C_2^3$\\

2657: $C_5$ & 5 & (1,2,3,4,5) & (1) & 24 & $\BZ^3\times C_5$\\

2658: $C_6$ & 6 & (1,2,3,4,5,6) & (1) & 24 & $\BZ^3\times C_6$\\

2659: $S_3$ & 6 & (1,2,3) & (1,2) & 18 & $\BZ^2\times C_2$\\

2660: $C_7$ & 7 & (1,2,3,4,5,6,7) & (1) & 48 & $\BZ^5\times C_7$\\

2661: $D_4$ & 8 & (1,2,3,4) & (1,4)(2,3) & 24 & $\BZ^3\times C_2$\\

2662: $Q_8$ & 8 & (1,7,2,8)(3,6,4,5) & (1,4,2,3)(5,7,6,8) & 24 & $\BZ^2\times C_4$\\

2663: $D_5$ & 10 & (1,2,3,4,5) & (1,5)(2,4) & 30 & $\BZ^2\times C_2$\\

2664: $A_4$ & 12 & (1,2,3) & (1,2)(3,4) & 96 & $\BZ^3$\\

2665: $S_3\times C_2$ & 12 & (1,3)(4,5) & (1,2) & 36 & $\BZ^3\times C_2$\\

2666: Sm($12,1$) & 12 & $\sigma_1$ & $\sigma_2$ & 72 & $\BZ^3\times C_2$\\

2667: $A_5$ & 60 & (1,2,3,4,5) & (1,2,3) & 1080 & $\BZ^{17}$\\

2668: \hline

2669: \end{tabular}

2670: \end{center}

2671: The notation for the finite groups is: $C_n$ is the cyclic group of order $n$,

2672: $D_n$ is the dihedral group of order $2n$,

2673: $Q_8$ is the quaternion group of order $8$

2674: and Sm($12,1$) is the non-abelian group of order $12$ not isomorphic to

2675: $A_4$ or $S_3\times C_2$. The two permutations $\sigma_1,\, \sigma_2$ are:

2676: $$\sigma_1:=(1, 8, 4, 11)(2, 9, 5, 12)(3, 7, 6, 10),\qquad

2677:     \sigma_2:=(1, 9, 4, 12)(2, 7, 5, 10)(3, 8, 6, 11).$$

2678: 

2679: Proceeding with our computer calculations, we then evaluated 

2680: the $\varphi_1,\ldots, \varphi_k$ (which now generate a

2681: subgroup of finite index in $\Gamma(G,\pi)$) on the middle term of the sequence

2682: (\ref{gashpr}). Locating the isotypic component of an irreducible

2683: $\BQ[G]$-module $N$ inside all three terms of (\ref{gashpr}) we were able

2684: to compute $\rho_{G,\pi,N}(\varphi_i)$ ($i=1,\ldots,k$) as matrices. Computing

2685: determinants we could decide whether 

2686: $\rho_{G,\pi}(\Gamma(G,\pi))\cap {\cal G}_{G,\pi}^1(\BZ)$ is of finite index

2687: in $\rho_{G,\pi}(\Gamma(G,\pi))$ or not. 

2688: 

2689: We have considered at least one presentation $\pi: F_2\to G$ for

2690: every group $G$ with $|G|\le 60$ which can be generated by 2 elements and

2691: have found that $\rho_{G,\pi}(\Gamma(G,\pi))\cap {\cal G}_{G,\pi}^1(\BZ)$ is

2692: of finite index in $\rho_{G,\pi}(\Gamma(G,\pi))$.   

2693: Of course we have run experiments on many more groups 

2694: (like $\PSL(2,7)$) but always found a

2695: similar result. 

2696: 

2697: Let us report some more on the particularly interesting case $G=A_5$.

2698: Let $N$ be the irreducible $\BQ[G]$-module of dimension 6. We have

2699: ${\rm End}_G(N)=\BQ(\sqrt{5})$. Its ring of integers is 

2700: ${\cal O}=\BZ((1+\sqrt{5})/2)$. 

2701: Identifying the isotypic component of $N$ in $\BQ[G]$ with $M(3,\BQ(\sqrt{5}))$

2702: we obtain a representation

2703: $$\rho: \Gamma(G,\pi)\to \GL(1,M(3,\BQ(\sqrt{5})))=\GL(3,\BQ(\sqrt{5})).$$

2704: Running the programm  described above we have found

2705: \begin{proposition}\label{compu18}

2706: For every presentation $\pi: F_2\to G=A_5$ there is a subgroup 

2707: $\Delta\le \Gamma(G,\pi)$

2708: of finite index such that $\rho_{A_5,\pi}(\Delta)$ 

2709: is contained up to change of bases in the subgroup of 

2710: $\SL(3,{\cal O})$ consisting of elements which have $(1,0,0)$ as a first row. 

2711: \end{proposition}

2712: Proposition \ref{compu18} shows that Theorem \ref{teo} is not true for $n=2$

2713: and 

2714: non-redundant presentations. It is not clear what Proposition \ref{compu18}

2715: suggests toward the general case. It may suggest that Theorem \ref{teo} is not

2716: true for non-redundant 

2717: presentations but it may also be that $n=2$ is exceptional.

2718: 

2719: We have developed similar programms also for the cases $n=3$ and $n=4$.

2720: They lead to the following tables. The notation is the same as in the above

2721: table for $n=2$. The presentations used for $n=4$ take the value $1$ on the

2722: fourth generator of the free group. 

2723: 

2724: \medskip

2725: \centerline{\it Subgroups in $\A(F_3)$}

2726: \begin{center}

2727: \begin{tabular}{|c|c|c|c|c|c|c|}

2728: \hline

2729: $G$  & $|G|$ & $\pi(x)$ & $\pi(y)$ & $\pi(z)$ & $[\A(F_3):\Delta]$ & 

2730: $\Delta^{\rm ab}$ \\

2731: \hline

2732: $C_2$ & 2 & (1,2) & (1) & (1) & 7 & $C_2^3$\\

2733: $C_3$ & 3 & (1,2,3) & (1) & (1) & 26 & $C_6$ \\

2734: $S_3$ & 6 & (1,2,3) & (1,2) & (1) & 168 & $C_2^3$\\

2735: $D_4$ & 8 & (1,3) & (1,4,3,2) & (1) & 336 & $C_2^8$ \\

2736: $Q_8$ & 8 & (1,7,2,8)(3,6,4,5) & (1,4,2,3)(5,7,6,8) & (1) & 336 & $C_2^7$\\

2737: $D_5$ & 10 & (1,2,3,4,5) & (1,5)(2,4) & (1) & 840 & $C_2^5$ \\

2738: $A_4$ & 12 & (1,2,3) & (1,2)( 3,4) & (1) & 1560 & $C_6$ \\

2739: $S_3\times C_2$ & 12 & (1,3)(4,5) & (1,2) & (1) & 1008 & $C_2^9$\\

2740: Sm($12,1$) & 12 & $\sigma_1$ & $\sigma_2$ & (1) & 1344 & $C_2^2\times C_4$\\

2741: $A_5$ & 60 & (1,2,3,4,5) & (1,2,3) & (1) & 200160 & ? \\

2742: \hline

2743: \end{tabular}

2744: \end{center}

2745: 

2746: \medskip

2747: \centerline{\it Subgroups in $\A(F_4)$}

2748: \begin{center}

2749: \begin{tabular}{|c|c|c|c|}

2750: \hline

2751: $G$  & $|G|$ & $[\A(F_4):\Delta]$ & $\Delta^{\rm ab}$ \\

2752: \hline

2753: $C_2$ & 2 & 15 & $C_2^2$\\

2754: $C_3$ & 3 & 80 & $C_6$\\

2755: $S_3$ & 6 & 80 & $C_6$ \\

2756: $D_4$ & 8 & 3360 & $C_2^4$ \\

2757: $Q_8$ & 8 & 840 & $C_2^6$ \\

2758: $D_5$ & 10 &  930 & $C_2^3$ \\

2759: $A_4$ & 12 & 1680 & $C_3\times C_6$ \\

2760: $S_3\times C_2$ & 12 & 2730 & $C_2^4$ \\

2761: Sm($12,1$) & 12 & 3120 & $C_2^2\times C_4$\\

2762: $A_5$ & 60 & 213098 & ? \\

2763: \hline

2764: \end{tabular}

2765: \end{center}

2766: We have run experiments on many finite groups $G$ and

2767: presentations $\pi: F_n\to G$, we have always found

2768: that $\rho_{G,\pi}(\Gamma(G,\pi))\cap {\cal G}_{G,\pi}^1(\BZ)$ is

2769: of finite index in $\rho_{G,\pi}(\Gamma(G,\pi))$.

2770: 

2771: 

2772: 

2773: 

2774: 

2775: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2776: \begin{thebibliography}{99}

2777: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2778: 

2779: \bibitem{Ba} S.\, Bachmuth,

2780: {\it Induced automorphisms of free groups and free metabelian groups.}

2781: Trans. Amer. Math. Soc. {\bf 122}, (1966), 1--17

2782: 

2783: \bibitem{BLMM} H. Bass, A. Lubotzky, A.R. Magid, S. Mozes,

2784: {\it The proalgebraic completion of rigid groups.}

2785: Geometriae Dedicata {\bf 95}, (2002), 19--58

2786: 

2787: \bibitem{Bi} J.\, Birman,

2788: {\it Braids, links and mapping class groups. \/},

2789: Ann. of Math. Stud. {\bf 82}, Princeton Univ. Press, Princeton (1974)

2790: 

2791: \bibitem{CR} C.W.\, Curtis, I. Reiner,

2792: {\it Representation Theory of finite Groups and associative Algebras.}

2793: Interscience Publishers (1962) 

2794: 

2795: \bibitem{D} M. Deuring,

2796: {\it Algebren.}

2797: Ergebnisse {\bf 41}, Springer Verlag, New York, Berlin, Heidelberg (1968)

2798: 

2799: \bibitem{DJ} D. Djokovi\'c, V. Platonov,

2800: {\it Low dimensional representations of $\A(F_2)$.}

2801: Manuscripta Math. {\bf 89}, (1996), 475--509

2802: 

2803: \bibitem{DU} M. Dunwoody,

2804: {\it Nielsen Transformations}. 

2805: In: Computational Problems in Abstract Algebra (1970), Pergamon, Oxford,

2806: 45--46

2807: 

2808: \bibitem{EGM} J. Elstrodt, F. Grunewald, J. Mennicke,

2809: {\it Groups acting on Hyperbolic Space: Harmonic Analysis

2810: and Number Theory.}

2811: Springer Monographs in Math., (1998), 524 pp 

2812: 

2813: \bibitem{Ev} M. Evans,

2814: {\it Presentations of groups involving more generators than are necessary.}

2815: Proc London Math. Soc. {\bf 67}, (1993), 106--126

2816: 

2817: \bibitem{Ev1} M. Evans,

2818: {\it T-systems of certain finite simple groups.}

2819: Math. Proc. Cambridge Philos. Soc. {\bf 113}, (1993), 9--22

2820: 

2821: \bibitem{F} E. Formanek,

2822: {\it Characterizing a free group in its automorphism group.}

2823: J. Algebra, {\bf 133}, (1990), 424--432

2824: 

2825: \bibitem{FP} E. Formanek, C. Procesi,

2826: {\it The automorphism group of a free group is not linear.}

2827: J. Algebra, {\bf 149}, (1992), 494--499

2828: 

2829: \bibitem{G} W. Gasch\"utz,

2830: {\it \"Uber modulare Darstellungen endlicher Gruppen, die von freien Gruppen

2831:   induziert werden.}

2832: Math. Z. {\bf 60}, (1954), 274--286

2833: 

2834: \bibitem{Gi} R. Gilman,

2835: {\it Finite quotients of the automorphism group of a free group.}

2836: Canad. J. Math. {\bf 29}, (1977), 541--551

2837: 

2838: \bibitem{Gr} K. Gruenberg,

2839: {\it Relation modules of finite groups.}

2840: CBMS {\bf 25}, Amer. Math. Soc. (1976)

2841: 

2842: \bibitem{GLu} F. Grunewald,, A. Lubotzky,

2843: {\it Linear representations of the mapping class group.}

2844: In preparation

2845: 

2846: \bibitem{GP} F. Grunewald, V. Platonov,

2847: {\it Rigidity results for groups with radical, cohomology of finite groups and

2848:   arithmeticity problems.}

2849: Duke Math. J. {\bf 100}, (1999), 321--358

2850: 

2851: \bibitem{GS} R. Guralnick, A. Shalev,

2852: {\it On the spread of finite simple groups.}

2853: Combinatorica {\bf 23}, (2003), 73--87 

2854: 

2855: \bibitem{JR} M. Jarden, J. Ritter, 

2856: {\it Normal Automorphisms of absolute Galois Groups of $\wp$-adic Fields.}

2857: Duke Math. J. {\bf 47}, (1980), 47--56

2858:  

2859: \bibitem{L} A. Lubotzky,

2860: {\it Torsion in profinite completions of torsion-free groups.}

2861: Quart. J. Math. Oxford Series {\bf 44}, (1993), 327--332

2862: 

2863: \bibitem{LM} A. Lubotzky, A.R. Magid, 

2864: {\it Varieties of representations of finitely generated groups.}

2865: Mem. Amer. Math. Soc. {\bf 58}, (1985)

2866: 

2867: \bibitem{LP} A. Lubotzky, I. Pak,

2868: {\it The product replacement algorithm and Kazhdan's property (T).}

2869: J. Amer. Math. Soc. {\bf 14}, (2001), 347--363 

2870: 

2871: \bibitem{MKS} W.\, Magnus, A. Karass, D. Solitar,

2872: {Combinatorial Group Theory.}

2873: Interscience Publishers (1966)

2874: 

2875: \bibitem{Neu1} B.H. Neumann,

2876: {\it Die Automorphismengruppe der freien Gruppen.}

2877: Math. Ann. {\bf 107}, (1933), 367--386

2878: 

2879: \bibitem{Neu} B.H. Neumann,

2880: {\it On a question of Gasch\"utz.}

2881: Archiv der Math. {\bf 7}, (1956), 87--90

2882: 

2883: \bibitem{N} J. Nielsen,

2884: {\it Die Isomorphismengruppe der freien Gruppen.}

2885: Math. Ann. {\bf 91}, (1924), 169--208

2886: 

2887: \bibitem{Pak} I. Pak,

2888: {\it What do we know about the product replacement algorithm?}

2889: Groups and Computation III (Columbus, Ohio, 1999), Ohio State

2890: Univ. Math. Res. Inst. Publ. {\bf 8}, 301--347, de Gruyter, Berlin, (2001)

2891: 

2892: \bibitem{PLR} V. Platonov, A. Rapinchuk,

2893: {\it Algebraic Groups and Number Theory.} Pure and Applied Mathematics {\bf

2894:   139}, Academic Press, (1994)

2895: 

2896: \bibitem{PORA} A. Potapchik, A. Rapinchuk,

2897: {\it Low-dimensional linear representations of $\A(F_n)$, $n\ge 3$.}

2898: Transactions of the Amer. Math. Soc. {\bf 352}, (1999), 1437--1451 

2899: 

2900: \bibitem{RA} A. Rapinchuk,

2901: {\it $n$-dimensional linear representations of $\A(F_n)$, and more.}

2902: J. of Algebra {\bf 197}, (1997), 146--152

2903: 

2904: \bibitem{Serresl} J.-P., Serre,

2905: {\it Le probl\'eme des groupes des congruence pour $\SL_2$.}

2906: Ann. of Math., {\bf 92}, (1970), 489--527

2907: 

2908: \bibitem{S} J.-P., Serre,

2909: {\it Linear Representations of Finite Groups.}

2910: Graduate Text in Math. {\bf 42},

2911: Springer Verlag, New York, Berlin, Heidelberg (1993)

2912: 

2913: \bibitem{Vas} L.\, Vaserstein,

2914: {\it The structure of classical arithmetic groups of rank greater than one.}

2915: Math. USSR Sbornik (English translation), {\bf 20}, (1973), 465--492 

2916: 

2917: \bibitem{Vas1} L.\, Vaserstein,

2918: {\it The group $\SL_2$ over Dedekind rings of arithmetic type.}

2919: Mat. Sb. {\bf 89}, (1972), 313--322

2920: 

2921: \bibitem{Venk} T.N.\, Venkataramana,

2922: {\it On systems of generators of arithmetic subgroups of higher rank groups.}

2923: Pacific Journal of Math., {\bf 166, 1}, (1994), 193--212

2924: 

2925: \end{thebibliography}

2926: 

2927: 

2928: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2929: \end{document}

2930: 

2931: 

2932: 

2933: 

2934: 

2935: 

2936: 

2937: