1: \documentclass[a4paper]{article}

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

3: \usepackage{latexsym}

4: 

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: 

36: % Fritz compatibility macros 

37: 

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

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

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

41: 

42: 

43: \def\g{{\Gamma}}

44: \def\a{{\Aut}}

45: \hsize= 15cm

46: \vsize= 20.5cm

47: \font\Bbb=msbm10

48: \font\Bbbb=msbm7

49: %\def\Z{{\bf Z}}

50: \def\I{{\bf I}}

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

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

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

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

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

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

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

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

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

60: %\def\Z{{\bf Z}}

61: \def\I{{\bf I}}

62: \def\g{{\bf g}}

63: % \def\S{{\cal S}}

64: % \def\A{{\rm Aut}}     % sorry 

65: % \def\a{{\lie Aut}}

66: \def\r{{\rm rk}}

67: \def\R{{\bf R}}

68: \def\Ad{{\rm Ad}}

69: \def\g{{\Gamma}}

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

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

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

73: %\def\IQ{{\hbox{\Bbbb Q}}}

74: %\def\IZ{{\hbox{\Bbbb Z}}}

75: %%\def\A{{\rm Aut}}

76: %%\def\a{{\lie Aut}}

77: %\def\aq{{\lie Aut}_\BQ}

78: %\def\u{{U(G)\backslash G}}

79: %\def\bB{{\mathbf{B}}} 

80: %\font\ber=cmbx12 scaled \magstep1

81: %\font\uber=cmbx12 scaled \magstep2

82: 

83: %\input macros

84: \baselineskip= 13pt

85: \parindent= 0pt

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

87: 

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

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

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

91: \font \msbm= msbm10

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

93: 

94: % End of Fritz compatibility 

95: 

96: % Oft benutze Symbole

97: \newcommand{\AGF}{A_{\Gamma \mid F}}

98: \newcommand{\tAGF}{A_{\tGam \mid \tF}}

99: 

100: \newcommand{\cAS}{\cA_\bS}

101: \newcommand{\cASid}{\cA_\bS^1}

102: 

103: \newcommand{\GLHC}{\GL\big(\H^*(\Gamma, \bbC)\big)} 

104: \newcommand{\KGone}{K(\Gamma,1)} 

105: 

106: 

107: \title{ Automorphism groups of polycyclic-by-finite groups 

108: and arithmetic groups\newline}

109: 

110: 

111: %\bigskip

112: %\bigskip

113: %\bigskip

114: %\bigskip

115: \author{Oliver Baues \thanks{e-mail: baues@math.uni-karlsruhe.de} \\

116: Mathematisches Institut II\\ 

117: Universit\"at Karlsruhe\\ 

118: D-76128 Karlsruhe% \vspace*{1ex} 

119: \and 

120: Fritz Grunewald \thanks{e-mail: fritz@math.uni-duesseldorf.de}\\

121: Mathematisches Institut\\

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

123: D-40225 D\"usseldorf}

124: 

125: \date{May 18, 2005}     % ver 1.0 

126: 

127: \bigskip

128: \bigskip

129: \bigskip

130: \bigskip

131: 

132: 

133: \begin{document}

134: \maketitle 

135: \begin{abstract} 

136: We show that the outer automorphism group of a 

137: polycyclic-by-finite group is an arithmetic group. 

138: This result follows from a detailed structural

139: analysis of the automorphism groups of such groups.

140: We use an extended version of the theory of the algebraic 

141: hull functor initiated by Mostow.

142: We thus make applicable refined methods from the theory of algebraic 

143: and arithmetic groups.

144: We also construct examples of polycyclic-by-finite 

145: groups which have an automorphism group which

146: does not contain an arithmetic group of finite index.

147: Finally we discuss applications of our results to the groups of homotopy

148: self-equivalences of $\KGone$-spaces and  obtain an extension 

149: of arithmeticity results of Sullivan in rational homotopy theory.

150: \newline

151: \hrule 

152: 

153: \hspace{1cm} \\

154: 

155: \noindent 

156: 2000 Mathematics Subject Classification: Primary 20F28, 20G30; \\

157: Secondary 11F06, 14L27, 20F16,  20F34, 22E40, 55P10 

158: \end{abstract}

159: 

160: 

161: \medskip

162: 

163: 

164: 

165: \bigskip

166: \bigskip

167: \bigskip

168: %\centerline{\bf Table of contents}

169: \newpage

170: \tableofcontents 

171: \medskip

172: 

173: \bigskip

174: \bigskip

175: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

176: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

177: 

178: 

179: \newpage 

180: \section{Introduction} 

181: 

182: \subsection{The main results}

183: 

184: We write $\Aut(\Gamma)$ for the group of automorphisms of a group $\Gamma$.

185: The subgroup consisting of automorphisms induced by conjugations with elements

186: of $\Gamma$ is denoted by $\Inn_\Gamma$. It is normal in $\Aut(\Gamma)$ and

187: the quotient group 

188: $$ \Out(\Gamma):=\Aut(\Gamma)\big/\Inn_\Gamma$$ 

189: is called the

190: outer automorphism group of $\Gamma$. This paper is devoted to the detailed

191: study of the groups $\Aut(\Gamma)$ and $\Out(\Gamma)$ in case $\Gamma$ is

192: polycyclic-by-finite. Here we say that a group $\Gamma$ 

193: is $\cal E$-by-finite whenever $\cal E$ is a property of groups and when 

194: $\Gamma$ has a subgroup of finite index having property 

195: $\cal E$.  

196: As one of our main results we prove:

197: 

198: \begin{theorem}\label{teoa} 

199: For any polycyclic-by-finite group 

200: $\Gamma$, $\Out(\Gamma)$ 

201: is an arithmetic group.

202: \end{theorem} 

203: 

204: To explain the concept of an arithmetic group we recall that a 

205: {\it $\BQ$-defined linear algebraic group} $G$ is a subgroup $G \le

206: \GL(n,\BC)$

207: $(n\in\BN)$ which is also an affine algebraic set defined by polynomials with

208: rational coefficients in the natural coordinates of $\GL(n,\BC)$. If $R$ is a

209: subring of $\BC$ we put $G(R):=G\cap \GL(n,R)$.

210: We have $G=G(\BC)$.

211: Let $G$ be

212: a $\BQ$-defined linear algebraic group. A subgroup

213: $\Gamma

214: \le G(\BQ)$ is called an {\it arithmetic subgroup} of $G$ if $\Gamma$

215: is commensurable with $G(\BZ )$.

216: An abstract group $\Delta$ is called {\it arithmetic} if it is isomorphic to

217: an arithmetic subgroup of a $\BQ$-defined linear algebraic group.

218: Two subgroups $\Gamma_1,\, \Gamma_2$ of $\GL(n,\BC)$ are called

219: {\it commensurable}

220: if their intersection $\Gamma_1\cap \Gamma_2$ has finite index both in

221: $\Gamma_1$ and $\Gamma_2$. These definitions are taken from \cite{PR}. 

222: 

223: In case $\Gamma$ is a finitely generated nilpotent group,

224: Segal \cite{Segal2} observed that $\Out(\Gamma)$ contains an

225: arithmetic subgroup of finite index. Such a group need itself not be

226: arithmetic, see \cite{GP3} for examples. Our Theorem \ref{teoa} is a

227: strengthening of Segal's result even in this restricted case. 

228: The results of Sullivan \cite{Sullivan} 

229: imply for finitely generated nilpotent groups a

230: still weaker result on their outer automorphism groups. We shall come back to

231: this in Section \ref{applih}. 

232: A related result, for any polycyclic-by-finite group, is proved 

233: by Wehrfritz in \cite{Wehrfritz}. 

234: He shows that $\Out(\Gamma)$ admits a faithful representation

235: into $\GL(n,\bbZ)$, for some $n$,  in this general case.

236: 

237: A role model for these results  

238: is the case $\Gamma=\BZ^n$ ($n\in\BN$). Here we have

239: $$\Aut(\BZ^n)=\Out(\BZ^n)=\GL(n,\BZ)$$

240: and both $\Aut(\BZ^n)$ and $\Out(\BZ^n)$ are arithmetic groups. 

241: More generally, $\Aut(\Gamma)$ is arithmetic for every finitely generated

242: nilpotent group $\Gamma$, see \cite[Corollary 9, Chapter 6]{Segal}. This

243: result was obtained for torsion-free finitely generated

244: nilpotent groups by Auslander and Baumslag, see  \cite{AusBaumslag,Baumslag,Baumslag1}. 

245: We generalize it to the case of finitely generated

246: nilpotent-by-finite groups in Corollary \ref{coronilfi}. 

247: But this is not the general picture. 

248: We now describe a polycyclic group so that 

249: $\Aut(\Gamma)$ does not contain an arithmetic subgroup of finite index. To do

250: this we choose $d\in\BN$ not a square,  and let $K=\BQ(\sqrt d)$ be the

251: corresponding real quadratic number field and write $x\mapsto \bar x$ for the

252: non-trivial element of the Galois group  of $K$ over $\BQ$. 

253: Consider the subring ${\cal O}=\BZ+\BZ\sqrt d\le K$. By Dirichlet's unit

254: theorem we may choose a unit $\epsilon\in {\cal O}^*$ which is of infinite

255: order and satisfies $\epsilon \bar\epsilon=1$. Define 

256: \begin{equation}\label{dihe}

257: D_\infty:= \; \langle \, A,\,\tau \Mid \tau^2=(A\tau)^2=1 \, \rangle

258: \end{equation}

259: to be the infinite dihedral group. It is easy to see that $F={\cal O}\times

260: \BZ$ obtains the structure of a $\Gamma$-module by defining

261: \begin{equation}\label{modu}

262: A\cdot (m,n) =(\epsilon m,n),\quad \tau\cdot (m,n)=(\bar m,-n)\qquad

263: (m\in{\cal O},\, n\in\BZ).

264: \end{equation}   

265: We denote the corresponding split extension by

266: $$\Gamma(\epsilon):= \; F\rfish D_\infty.$$

267: The group $\Gamma(\epsilon)$ is polycyclic and even an arithmetic group, but

268: we have:

269: \begin{theorem}\label{teob} 

270: The automorphism group of\/ $\Gamma(\epsilon)$ does not

271: contain an arithmetic group of finite index.

272: \end{theorem} 

273: 

274: The statement of Theorem \ref{teob} is stronger than just saying that 

275: the automorphism group of $\Gamma(\epsilon)$ is not arithmetic, since there are

276: groups (see \cite{GP3}) which are not arithmetic but which contain a subgroup

277: of finite index which has this property.

278: Our examples show that the structure of the automorphism group varies

279: dramatically when $\Gamma$ is replaced by one of its subgroups of 

280: finite index. It is for example easy to see (also from results in

281: Subsection \ref{outline}) that the automorphism group of the subgroup of

282: $\Gamma(\epsilon)$ which is generated by $F$ and $A$ is an arithmetic group. 

283: More generally, as we will explain below, every polycyclic-by-finite group has

284: a finite index subgroup with an arithmetic automorphism group. 

285: 

286: Theorem \ref{teob} is complemented by a result of Merzljakov 

287: \cite{Merzljakov} who has proved,

288: for any polycyclic group $\Gamma$,  that 

289: $\Aut(\Gamma)$ has a faithful representation into $\GL(n,\bbZ)$ for some 

290: $n\in\BN$. This generalizes also to the case of 

291: polycyclic-by-finite groups, see \cite{Wehrfritz2}. 

292: 

293: Let $F= \Fitt(\Gamma)$ be the Fitting subgroup

294: of $\Gamma$. This is the maximal nilpotent normal subgroup of $\Gamma$. 

295: Since $F$ is characteristic in $\Gamma$ (that is normalized by every

296: automorphism of $\Gamma$),  we may define 

297: \begin{equation}\label{defiagf}

298: \A_{\Gamma |F} :=  \; \{\,  \phi \in \Aut(\Gamma) \Mid 

299: \phi|_{\, \Gamma/F} = \id_{\Gamma/F} 

300: \}. 

301: \end{equation}

302: 

303: 

304: As our main structural result on the automorphism group of a general 

305:  polycyclic-by-finite group we show:

306: \begin{theorem}\label{teoc} 

307: Let $\Gamma$ be a polycyclic-by-finite group. Then $\AGF$ is a normal 

308: subgroup of $\Aut(\Gamma)$ and $\AGF$ is an arithmetic group. 

309: There exists a finitely generated nilpotent group 

310: $B \leq \Aut(\Gamma)$ 

311: which consists of inner automorphisms 

312: such  that $$  \A_{\Gamma|F} \cdot B  \; \, \leq \; \Aut(\Gamma) $$

313: is a subgroup of finite index in the automorphism group of\/ $\Gamma$.

314: \end{theorem}

315: 

316: This theorem is a more precise version of \cite[Theorem 2, Chapter 8]{Segal}

317: and of certain similar theorems contained in \cite{Ausl1}. It has many

318: finiteness properties of $\Aut(\Gamma)$ as a consequence, 

319: see \cite[Chapter 8]{Segal}. The fact that $\Aut(\Gamma)$ is finitely 

320: presented was first proved by Auslander in \cite{Ausl1}. 

321: 

322: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

323: 

324: \subsection{Outline of the proofs and more results}\label{outline}

325: 

326: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

327: 

328: We shall describe now the general strategy in our proofs and more results on

329: the structure of the automorphism group $\Aut(\Gamma)$ of a general

330: polycyclic-by-finite group $\Gamma$.

331: We take the basic theory of polycyclic-by-finite groups for granted.

332: As a reference for our notation and results we use the book \cite{Segal}. 

333: 

334: A polycyclic-by-finite group $\Gamma$ has a maximal finite normal subgroup 

335: $\tau_\Gamma$. We say that $\Gamma$ is a {\it wfn-group} if

336: $\tau_\Gamma=\{1\}$. We first consider the case of 

337: polycyclic-by-finite groups which are wfn-groups and later reduce to this

338: case. 

339: 

340: Let now $\Gamma$ be a polycyclic-by-finite group which is a wfn-group. 

341: We use the construction from \cite{Baues2} of a $\BQ$-defined  solvable-by-finite 

342: linear algebraic group ${\bf H}_\Gamma$ which contains $\Gamma$ in its group of $\BQ$-rational 

343: points. The study of the functorial construction $\Gamma \mapsto \bH_\Gamma$ traces back to

344: work of Mostow \cite{Mostow1, Mostow2}. 

345: The algebraic group ${\bf H}_\Gamma$ has

346: special features which allow us to

347: identify the group of algebraic automorphisms of 

348: ${\bf H}_\Gamma$, which we call $\Aut_a({\bf H}_\Gamma)$, with a 

349: $\BQ$-defined linear algebraic group $\cA_\Gamma$. In general, the group of 

350: algebraic automorphisms of a $\BQ$-defined algebraic group is an extension of

351: a linear algebraic by an arithmetic group, see \cite{BS}.

352: The functoriality of ${\bf H}_\Gamma$ 

353: leads to a natural embedding

354: \begin{equation}

355: \Aut(\Gamma) \, \le \, \Aut_a({\bf H}_\Gamma)=\cA_\Gamma.

356: \end{equation}

357: These steps are carried out in detail in Sections \ref{section:auta}, 

358: \ref{sechull}.

359: 

360: After preparations in Sections 

361: \ref{secthick}, \ref{secshad} we prove the following in Section \ref{secautoauto}.  

362: \begin{theorem}\label{teoagf}

363: Let $\Gamma$ be a polycyclic-by-finite 

364: group which is a wfn-group. 

365: Then $\Aut(\Gamma)$ is contained in $\cA_\Gamma(\BQ)$.

366: A subgroup of finite index in $\Aut(\Gamma)$ is contained in

367: $\cA_\Gamma(\BZ)$.

368: The group $\A_{\Gamma|F}$ is an arithmetic 

369: subgroup in its Zariski-closure in  $\cA_\Gamma$.

370: \end{theorem}

371:  

372: 

373: Theorem \ref{teoagf} is the main step in the proof of Theorem

374: \ref{teoc} which is given in Section \ref{generalpofi}. 

375: 

376: In Section \ref{subsecari} 

377: we develop, starting from results in \cite{GP3}, 

378: a theory of proving arithmeticity for linear groups which are related to

379: arithmetic groups in certain ways. This shows that the situation 

380: build up in Theorems \ref{teoagf}, \ref{teoc} 

381: implies that $\Out(\Gamma)$ is an arithmetic group for 

382: polycyclic-by-finite groups which are wfn-groups:

383: To construct a $\BQ$-defined algebraic group ${\cal O}_\Gamma$ which contains 

384: $\Out(\Gamma)$ as an arithmetic subgroup we start off by considering the

385: Zariski-closure ${\cal B}$ of $\A_{\Gamma|F}$ in $\cA_\Gamma$. We then take

386: the quotient group ${\cal B}/{\cal C}$ where ${\cal C}$ is the   

387: Zariski-closure of $\Inn_F$ in $\cA_\Gamma$. 

388: The image of $\A_{\Gamma|F}$ in ${\cal B}/{\cal C}$ is an arithmetic 

389: subgroup.

390: We show that 

391: $\Out(\Gamma)$ is a finite extension group of the 

392: quotient $\A_{\Gamma|F} \big/ \, \Inn_F$ by its central subgroup 

393: $D= (\A_{\Gamma|F}\cap \Inn_\Gamma) \big/\, \Inn_F$. 

394: We modify 

395: the algebraic group ${\cal B}\big/{\cal C}$ to obtain 

396: a $\BQ$-defined linear algebraic group ${\cal E}$ such that

397: the group $D$ is unipotent-by-finite, and $\cE$

398: still contains an isomorphic copy of $\A_{\Gamma|F} \big/\, \Inn_F$ as 

399: an arithmetic subgroup. The construction ensures that 

400: $(\A_{\Gamma|F}\cap \Inn_\Gamma) \big/ \, \Inn_F$ 

401: is arithmetic in its Zariski-closure $\cD$ in ${\cal E}$ and 

402: we can prove that     

403: $\Out(\Gamma)$ is arithmetic in a finite extension group 

404: ${\cal O}_\Gamma$ of the quotient ${\cal E}/{\cD}$. The final definition of 

405: ${\cal O}_\Gamma$ is contained in Section \ref{asec1}

406: 

407: To prepare for later applications we introduce some more notation. 

408: The group of inner automorphisms $\Inn_{{\bf H}_\Gamma}$ is 

409: by our constructions a $\BQ$-closed subgroup of the group of algebraic

410: automorphisms 

411: $\cA_\Gamma=\Aut_a({\bf H}_\Gamma)$. We call the quotient 

412: \begin{equation}

413: \Out_a({\bf H}_\Gamma):=\cA_\Gamma\big/\, \Inn_{{\bf H}_\Gamma}

414: \end{equation}

415: the algebraic outer automorphism group of ${\bf H}_\Gamma$.

416: It is again a 

417: $\BQ$-defined linear algebraic group, and we obtain a group homomorphism

418: \begin{equation}

419: \pi_\Gamma :\Out(\Gamma)\to \Out_a({\bf H}_\Gamma)

420: \end{equation}

421: by restricting the quotient homomorphism $\cA_\Gamma \ra  

422: \Out_a({\bf H}_\Gamma)$ to the subgroup $\Aut(\Gamma)\le\cA_\Gamma$.  

423: We then prove in Section \ref{asec1}:

424: 

425: \begin{theorem}\label{ftop} 

426: Let $\Gamma$ be a polycyclic-by-finite 

427: group which is a wfn-group. Then there is a $\BQ$-defined 

428: linear algebraic group ${\cal O}_\Gamma$ which contains $\Out(\Gamma)$ 

429: as an arithmetic subgroup and a $\BQ$-defined homomorphism 

430: $\pi_{\cO_\Gamma}:{\cal O}_\Gamma\to \Out_a({\bf H}_\Gamma)$ such that the

431: diagram

432: \begin{equation}

433: \xymatrix{

434: \Out(\Gamma) \ar[r]^{} \ar[d]_{\pi_\Gamma} 

435: & {\cal O}_\Gamma \ar[ld]^{\pi_{\cO_\Gamma}} \\

436: \Out_a({\bf H}_\Gamma) & 

437: }

438: \end{equation}

439: is commutative.

440: The kernel of $\pi_\Gamma$ is 

441: finitely generated, abelian-by-finite and is centralized by a finite

442: index subgroup of $\Out(\Gamma)$. If\/ $\Gamma$ is nilpotent-by-finite

443: then the kernel of $\pi_\Gamma$ is finite. 

444: \end{theorem}

445: 

446: We shall start the discussion of the general situation now. Let $\Gamma$ 

447: be a polycyclic-by-finite group, possibly with $\tau_\Gamma$ non-trivial.

448: Building on an idea of  \cite[Chapter 6, exercise 10 ]{Segal} for nilpotent

449: groups, we show in Section \ref{generalpofi}:

450: \begin{proposition}\label{equivi}  

451: Let $\Gamma$ be polycyclic-by-finite group and 

452: let $\tau_\Gamma$ denote its maximal finite normal subgroup. 

453: Then the groups $\Aut(\Gamma)$ and $\Aut(\Gamma /\tau_\Gamma)$ 

454: are commensurable. 

455: If $\Aut(\Gamma/\tau_\Gamma)$ is an arithmetic group 

456: then $\Aut(\Gamma)$ is arithmetic too. If $\Aut(\Gamma)$

457: is an arithmetic group then 

458: $\Aut(\Gamma/\tau_\Gamma)$ contains a subgroup of finite index which 

459: is an arithmetic group. 

460: \end{proposition}

461: Recall that two abstract groups $G$ and $G'$ are called \emph{commensurable}

462: if they contain finite index subgroups $G_0 \leq G$ and $G'_0 \leq G'$  

463: which are isomorphic. The examples in \cite{GP3} show that a 

464: group can contain an arithmetic subgroup of

465: finite index without being an arithmetic group.

466: We do not know whether this

467: phenomenon can arise in the situation of Proposition \ref{equivi}. 

468: As proved in \cite{GP3}, all arithmetic subgroups in algebraic groups

469: which do not have a quotient isomorphic to ${\rm PSL}(2)$ have only arithmetic

470: groups as finite extensions.  

471: 

472: 

473:    

474: 

475: The groups $\Out(\Gamma)$ and $\Out(\Gamma/\tau_\Gamma)$ satisfy a weaker

476: equivalence relation (they are called \emph{S-commensurable}, see Section \ref{applih}). 

477: \begin{proposition}\label{finiin}

478: The natural homomorphism 

479: $\Out(\Gamma) \ra \Out(\tGam)$ has finite kernel and

480: maps $\Out(\Gamma)$ onto a finite index subgroup 

481: of\/  $\Out(\Gamma/\tau_\Gamma)$.

482: \end{proposition}

483: 

484: Clearly, $\Gamma/\tau_\Gamma$ is a wfn-group. From Theorem 

485: \ref{ftop} we already know that $\Out(\Gamma/\tau_\Gamma)$ is an arithmetic

486: group. We further know from \cite{Wehrfritz} 

487: that $\Out(\Gamma)$ is isomorphic to a subgroup of $\GL(n,\BZ)$, for some $n\in

488: \BN$, and hence is residually finite. Recall that a group $G$ is called {\it

489:   residually finite} if for every $g\in G$ with $g\ne 1$ there is a subgroup

490: of finite index in $G$ not containing $g$.

491: We complete the proof of Theorem \ref{teoa} in the general case by the following

492: result.

493: \begin{proposition}\label{finker} 

494: Let $A$ be a residually finite group and $E$ a finite normal subgroup in $A$.

495: If $A/E$ is arithmetic then $A$ is an arithmetic group. 

496: \end{proposition}

497: In \cite{Deligne} Deligne constructs lattices (in non-linear) Lie groups which

498: map onto arithmetic groups with finite kernel but which are not residually finite. 

499: The above gives a strong converse to this result.

500: 

501: To contrast the examples from Theorem \ref{teob} we formulate 

502: now a simple condition on a polycyclic-by-finite group which

503: implies that $\Aut(\Gamma)$ is an arithmetic group. We write 

504: $\lambda: \Gamma\to \tilde\Gamma=\Gamma/\tau_\Gamma$ for the projection

505: homomorphism. We define 

506: \begin{equation}

507: \tilde F(\Gamma):= \lambda^{-1}(\Fitt(\tilde\Gamma)).

508: \end{equation}

509: From general theory we infer that $\tilde F(\Gamma)$ is normal in $\Gamma$ and

510: that $\Gamma / \tilde F(\Gamma)$ is an abelian-by-finite group. Suppose that

511: $\Sigma \le \Gamma $ is a normal subgroup of finite index with 

512: $\tilde F(\Gamma)\le \Sigma$ and the additional property that 

513: $\Sigma /\tilde F(\Gamma)$ is abelian. Then the finite group 

514: $\mu(\Gamma,\Sigma)=\Gamma/\Sigma$ acts through conjugation on the abelian

515: group $\Sigma /\tilde F(\Gamma)$. We prove:

516: 

517: \begin{theorem}\label{teosep} 

518: Let $\Gamma$ be a polycyclic-by-finite group. 

519: Assume that there is a 

520: normal subgroup  $\Sigma\le\Gamma$ which is of finite index, and such that 

521: $\Sigma /\tilde F(\Gamma)$ is abelian and $\mu(\Gamma,\Sigma)=\Gamma/\Sigma$ 

522: acts trivially on $\Sigma /\tilde F(\Gamma)$.

523: Then

524: $\Aut(\Gamma)$ is an arithmetic group. 

525: \end{theorem}

526: 

527: Theorem \ref{teosep} has the following immediate corollaries.

528: \begin{corollary}\label{coronil} 

529: Let $\Gamma$ be a polycyclic-by-finite group.

530: If\/ $\Gamma\big/\Fitt(\Gamma)$ is nilpotent then $\Aut(\Gamma)$

531: is an arithmetic group.

532: \end{corollary}\label{coronilfi}

533: 

534: \begin{corollary} The automorphism group of a finitely generated and 

535: nilpotent-by-finite group is an arithmetic group. 

536: \end{corollary}

537: 

538: \begin{corollary}

539: Every polycylic-by-finite group has a finite index 

540: subgroup whose automorphism group is arithmetic.

541: \end{corollary}

542: 

543: Analogues of Theorem \ref{teosep} and Corollary \ref{coronil} have been proved in 

544: \cite{GP1} for arithmetic polycyclic-by-finite groups. The book \cite{Segal} contains in 

545: Chapter 8 various results which also show the

546: arithmeticity of $\Aut(\Gamma)$, but the hypotheses on $\Gamma$ are a lot

547: stronger than ours.  

548: 

549: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

550: 

551: \subsection{Cohomological representations}

552: \label{topram}

553: 

554: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

555: 

556: Let $\Gamma$ be a torsion-free 

557: polycyclic-by-finite group, and $R$ a commutative ring.  Let 

558: $\H^*(\Gamma,R)$ denote the cohomology

559: of $\Gamma$ with (trivial) $R$-coefficients. 

560: Since inner automorphisms

561: act trivially on cohomology, the outer automorphism group 

562: $\Out(\Gamma)$ is naturally represented on the cohomology 

563: space $\H^*(\Gamma,R)$. 

564: The finite dimensional complex vector 

565: space $\H^*(\Gamma,\BC)$ comes with an $\BZ$-structure

566: (and a resulting $\BQ$-structure)

567: given by the image of the base change homomorphism 

568: $\H^*(\Gamma,\bbZ) \ra  \H^*(\Gamma, \bbC)$.

569: This allows us to identify $\GLHC$ with a $\BQ$-defined

570: linear algebraic group.

571: The representation 

572: \begin{equation}

573: \rho: \Out(\Gamma) \to \GLHC

574: \end{equation}

575: is integral with respect to the above $\BZ$-structure.  

576: Moreover, we define in Section \ref{geometrc3}, 

577: building on geometric methods developed in \cite{Baues2}, a 

578: $\BQ$-defined homomorphism

579: \begin{equation}

580: \eta: \Out_a({\bf H}_\Gamma) \to \GLHC

581: \end{equation}

582: which extends the homomorphism $\rho$ via our homomorphism $\pi_{\cO_\Gamma}$

583: from Proposition \ref{ftop}. 

584: By composition, we obtain a $\BQ$-defined homomorphism

585: \begin{equation}

586: \rho_{\cO_\Gamma}:=\eta\circ\pi_{\cO_\Gamma}  :\  {\cal O}_\Gamma\to \GLHC. 

587: \end{equation}

588: We collect all of this together in the following theorem: 

589: 

590: \begin{theorem}\label{topolo1} 

591: Let $\Gamma$ be a torsion-free polycyclic-by-finite group.

592: Then there is a $\BQ$-defined homomorphism

593: $\rho_{\cO_\Gamma}: {\cal O}_\Gamma\to \GLHC$ such that the diagram

594: \begin{equation}

595: \xymatrix{

596: \Out(\Gamma) \ar[r]  \ar[d]_{\rho} 

597: & {\cal O}_\Gamma \ar[ld]^{\rho_{\cO_\Gamma}} \\

598: \GLHC&   

599: }

600: \end{equation}

601: is commutative.

602: \end{theorem}

603: Thus, the Zariski-closure of the image 

604: of $\Out(\Gamma)$ in $ \GLHC $

605: is a $\bbQ$-closed  

606: subgroup and we have: 

607: \begin{theorem}\label{topolo2} 

608: Let $\Gamma$ be a torsion-free polycyclic-by-finite group. Then

609: the cohomology image $\rho\big(\Out(\Gamma)\big)$

610: is an arithmetic subgroup of its Zariski-closure in 

611: $\GLHC$. The kernel of $\rho$ is a finitely generated

612: subgroup of $\Out(\Gamma)$.

613: If\/  $\Gamma$ is in addition nilpotent then the

614: kernel of $\rho$ is nilpotent by-finite.

615: \end{theorem}

616: 

617: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

618: 

619: \subsection{Applications to the groups of homotopy 

620: self-equi\-va\-lences of spaces}\label{applih}

621: 

622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

623: 

624: Let $\Gamma$ be a group. The motivation to study 

625: the structure of $\Out(\Gamma)$

626: comes partially from topology since $\Out(\Gamma)$, for example, 

627: is naturally isomorphic to the group of homotopy classes of homotopy

628: self-equivalences of any $\KGone$-space (see \cite{Rutt}). 

629: A substantial theory 

630: using this tie between algebra and topology has been

631: developed starting from the work of Sullivan \cite{Sullivan}.  

632: We shall explain here the 

633: connections of our work with Sullivan's theory 

634: as described in \cite{Sullivan} and we also mention the additions 

635: we can make to Sullivan's theory. 

636: 

637: Sullivan considers spaces $X$ which have a nilpotent homotopy system

638: of finite type, that is, the homotopy groups 

639: $$(\pi_1(X);\pi_2(X),\pi_3(X),\ldots)$$

640: are all finitely generated, $\pi_1(X)$ is torsion-free nilpotent 

641: and acts nilpotently on

642: all $\pi_i(X)$ ($i\ge 2$). 

643: Sullivan associates to such a space $X$ 

644: the minimal model of the $\BQ$-polynomial forms of some complex representing

645: $X$. This is a finitely generated nilpotent differential algebra defined over

646: $\BQ$, it is called ${\cal X}$.

647: Under the further assumption that $X$ is either a finite complex or that the

648: homotopy groups $\pi_i(X)$ are trivial, for almost all $i\in \BN$,

649: he uses this construction to prove 

650: important results on the group $\Aut(X)$ of classes of

651: homotopy self-equivalences of the space $X$, cf.\ \cite[Theorem 10.3]{Sullivan}. 

652:  

653: Let us specialize, for a moment, to the case where $X$ is a $\KGone$-space,

654:  $\Gamma$ a torsion-free finitely generated nilpotent group. The results of

655: Sullivan can then be related to what we prove. 

656: The following table contains a dictionary

657: between the objects defined by Sullivan and those appearing in our theory.

658: \begin{center}

659: \begin{tabular}{|c|c|}

660: \hline

661: Sullivan  & algebraic theory \\

662: \hline

663: $\Aut(X)$ & $\Out(\Gamma)$ \\

664: ${\cal X}$ & ${\lie g}$ \\

665: $\Aut({\cal X})$ & $\Aut({\lie g})$ \\

666: $\Aut_\BQ(X)=\Aut({\cal X})/{\rm inner\ automorphisms}$ 

667: & ${\cal O}_\Gamma \xrightarrow{\pi_{\cO_\Gamma}}  \Aut({\lie g})\big/\Inn_{\lie g}$ \\

668: \hline

669: \end{tabular}

670: \end{center}

671: Here ${\lie g}$ is the Lie algebra of the Malcev-completion of 

672: $\pi_1(X)=\Gamma$. The associated DGA $\cal X$  

673: coincides with the Koszul-complex of $\lie{g}$.  The $\BQ$-defined linear

674: algebraic group ${\cal O}_\Gamma$ is defined in the previous subsection. It

675: admits a $\BQ$-homomorphism $\pi_{\cO_\Gamma}$ with finite kernel onto 

676: the group $\Aut({\lie g})/\Inn_{\lie g}$ 

677: of Lie algebra automorphisms of ${\lie g}$ modulo inner automorphisms.

678:   

679: We mention the concept of 

680: {\it S-commensurability} of groups appearing in \cite{Sullivan}. 

681: This is the equivalence relation amongst groups which is generated by the 

682: operations of taking quotients with finite kernel and finite index subgroups. 

683: 

684: We shall now go through the four statements about $\Aut(X)$ which

685: are contained in \cite[Theorem 10.3]{Sullivan}, specialized to a 

686: $\KGone$ space $X$, and rephrase them in our context:

687: 

688: \medskip 

689: In Theorem 10.3 (i), it is stated that

690: {\it $\Aut(X)$ is S-commensurable with a full arithmetic subgroup of 

691: $\Aut_\BQ(X)$.} We prove (see Theorem \ref{ftop}), quite equivalently, that there is

692: a  homomorphism 

693: $\pi_{\Gamma} :\Out(\Gamma) \to  \Aut({\lie g})\big/\Inn_{\lie g}$ which has finite 

694: kernel and an arithmetic subgroup as image. We additionally  show that $\Out(\Gamma)$ is isomorphic

695: to an arithmetic subgroup in a $\BQ$-defined linear algebraic group ${\cal O}_\Gamma$ 

696: which is a finite extension group of $ \Aut({\lie g})\big/\Inn_{\lie g}$. 

697: (Examples (\cite{Deligne}, \cite{GP3}) show that the class of arithmetic

698: groups is not closed under any of the two operations used to define

699: S-commensurability. Hence

700: the information on $\Aut(X)=\Out(\Gamma)$ 

701: contained in  \cite[Theorem 10.3 i)]{Sullivan},  for our restricted case,  is

702: weaker then what follows from our results. 

703: Most dramatically there are groups which can be mapped onto an arithmetic

704: group with a finite kernel which are not even residually finite. As Serre

705: \cite{Serre} remarks, localization results in \cite{Sullivan2} imply that

706: the group $\Aut(X)$ is residually finite.  We wonder, whether the general theory 

707: can be extended similarly to our results.)

708: 

709: \medskip 

710: In Theorem 10.3 (ii) it is proved that

711: {\it the natural action of $\Aut(X)$ on the integral homology is compatible

712:   with an algebraic matrix representation of $\Aut_\BQ(X)$ on the vector

713:   spaces of rational homology.} We consider here cohomology instead of 

714:   homology and we obtain a similar result. We prove that the natural action 

715:   of $\Out(\Gamma)=\Aut(X)$ on the image of integral cohomology is compatible with an algebraic matrix 

716: representation of ${\cal O}_\Gamma$ on the vector

717: spaces of rational cohomology (see Theorem \ref{ftop} and Section

718: \ref{topram}). This has the important consequence that the stabilizers

719: of cohomology classes are arithmetic groups. 

720: 

721: \medskip 

722: By Theorem 10.3 (iii),

723: {\it the reductive part of $\Aut_\BQ(X)$ is faithfully represented on the

724:   natural subspace of homology generated by maps of spheres into $X$.}

725: We recall  (see Proposition \ref{cohom1}) that the reductive part of 

726:  $ \Aut({\lie g})\big/\Inn_{\lie g}$ is faithfully represented on the cohomology vector spaces.

727: Of course, Sullivan's result implies that the representation is already

728: faithful on the first homology space. We obtain a similar result for the

729: action of ${\cal O}_\Gamma$ on the first cohomology space. 

730: 

731: \medskip 

732: In Theorem 10.3 (iv) it is proved that

733: {\it as we vary $X$ through finite complexes, $\Aut(X)$ runs through every

734: commensurability class of groups containing arithmetic groups.}

735: Here our results give something different and new since

736: Sullivan uses simply connected spaces ($\pi_1(X)=\{1\}$)

737: which have non-vanishing higher homotopy groups to show his existence result:

738: 

739: Let \emph{N-commensurability} be the equivalence relation for groups which 

740: is generated by the 

741: operations of taking quotients 

742: with finitely generated nilpotent-by-finite groups as kernel 

743: and by taking finite index subgroups. We can prove: 

744: \begin{proposition}\label{exinili}

745: As we vary $X$ through all finite $\KGone$ complexes, $\Gamma$ a

746: finitely generated nilpotent group, $\Aut(X)$ runs through every

747: N-commensur\-ability class of groups containing an arithmetic group.

748: \end{proposition}

749: The result just mentioned follows

750: from our theory of cohomological

751: representations and from the existence theorems in \cite{BS}. We do not

752: provide details here. 

753: \medskip

754:  

755: Having discussed the relation of our results to Sullivan's we can mention

756: the following generalization for spaces $X$ 

757: which are $\KGone$ spaces for torsion-free

758: polycyclic-by-finite groups. In fact,  we can collect the results described in

759: Sections \ref{outline}, \ref{topram} to show: 

760: 

761: \begin{theorem}\label{teotop}

762: Let $\Gamma$ be a torsion-free polycyclic-by-finite group and let $X$ a

763: $\KGone$ space. Let $\Aut(X)$ be the group of classes of homotopy

764: self-equivalences of $X$ then the following hold: 

765: \begin{itemize}

766: 

767: \item[i)] $\Aut(X)$ is an arithmetic group.

768: 

769: \item[ii)] The action of $\Aut(X)$ on integral cohomology is 

770: compatible with an algebraic matrix 

771: representation of the $\BQ$-defined linear algebraic ${\cal O}_\Gamma$ on 

772: the vector spaces of cohomology.

773: 

774: \item[iii)] The stabilizer of an integral (rational) cohomology class is

775: arithmetic.  

776: \end{itemize} 

777: \end{theorem}

778: Again, we wonder whether our Theorem \ref{teotop} has an extension which does not

779: assume the vanishing of the higher homotopy groups.

780: 

781: As a corollary,  using well known finiteness results for arithmetic groups we

782: get:

783: \begin{corollary}\label{corotop}

784: Let $\Gamma$ be a torsion-free polycyclic-by-finite group and let $X$ a

785: $\KGone$-space. Let $\Aut(X)$ be the group of classes of homotopy

786: self-equivalences of $X$. Then the following hold:

787: \begin{itemize}

788: \item[i)] The group $\Aut(X)$ is residually finite.

789: \item[ii)] The group $\Aut(X)$ is finitely presented.

790: \item[iii)] The group $\Aut(X)$ contains only finitely many conjugacy classes of

791: finite subgroups.

792: \item[iv)]  Let $\mu$ be a finite group acting by group automorphisms on 

793: $\Aut(X)$, then the cohomology set ${\rm H}^1(\mu,\Aut(X))$ is finite.

794: \item[v)] The group $\Aut(X)$ is of type $WF$.

795: \end{itemize} 

796: \end{corollary}

797: For the definition and discussion of cohomology sets, see Section

798: \ref{grouplem}. The properties i),...,v) all follow from the fact that 

799: $\Aut(X)$ is an arithmetic group. Property i) is well known to hold for

800: finitely generated linear groups, 

801: for ii), iii) see \cite{Borel1}, iv) is proved in \cite{GP3} (for a weaker

802: form see \cite{BS}). 

803: The last item, property $WF$, means that every torsion-free subgroup $\Delta$ of 

804: finite index in $\Aut(\Gamma)$ appears as the fundamental group of a finite $K(\Delta,1)$.

805: See \cite{Serre} for discussion of this property of arithmetic groups and further references.

806: It is not known whether groups which are finite extensions  of  a

807: $WF$-group inherit the property $WF$, see \cite{Serre2, Brown}.

808: Note also that properties ii), iii) are compatible with

809: S-commensurability whereas i), iv) are not. 

810: 

811: 

812: Let $X$ be a $\KGone$-space for a group $\Gamma$. 

813: The automorphism group of

814: $\Gamma$ can be naturally identified with the group of classes of

815: pointed homotopy

816: self-equivalences $\Aut^*(X)$ (see \cite{Rutt}). In case $\Gamma$ is

817: finitely generated torsion-free nilpotent it was well known that    

818: $\Aut^*(X)$ is an arithmetic group. Our results from Theorem \ref{teosep}

819: extend this to a much larger class of groups. 

820: Our examples constructed in Section \ref{exnona}, Theorem

821: \ref{teob} give rise to $\KGone$-spaces $X$, where $\pi_1(X)$ is nice and

822:  arithmetic whereas

823: $\Aut^*(X)$ is far from being an arithmetic group.

824: 

825: 

826: \medskip

827: {\em Acknowledgement } \ The authors wish to thank Wilhem Singhof

828: for helpful conversations concerning the contents of this article. 

829: 

830: 

831: 

832: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

833: 

834: \section{Prerequisites on linear algebraic groups and arithmetic groups}

835: \label{prereq}

836: 

837: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

838: 

839: In the sequel we shall use a certain amount of the theory of 

840: linear algebraic groups and also of the theory of arithmetic groups. We 

841: briefly review here what we need and also add certain consequences of the

842: general theory. 

843: Our basic references are \cite{Borel}, \cite{BT}, \cite{PR}. 

844: 

845: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

846: 

847: \subsection{The general theory}

848: 

849: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

850: 

851: We use the usual terminology of Zariski-topology. Thus a 

852: linear algebraic group $\cA$ is a Zariski-closed subgroup of 

853: $\GL(n,\BC)$, for some $n\in \BN$. It is $\BQ$-defined if it is $\BQ$-closed.

854: We use the shorter term {\it $\BQ$-closed}

855: for

856: what should be called closed in the Zariski-topology with closed subsets

857: being those affine algebraic sets defined by polynomials with coefficients in

858: $\BQ$. If $R$ is a subring of $\bbC$, we put $\cA(R) = \cA \cap \GL(n,R)$.

859: We denote the connected component of the identity in 

860: a linear algebraic group $\cA$ by $\cA^\circ$. 

861: The connected component $\cA^\circ$ always has finite index in $\cA$ and 

862: $\cA$ is called \emph{connected} if $\cA=\cA^\circ$. If $\cA$ is defined

863: over $\bbQ$, its group of $\bbQ$-points $\cA(\bbQ)$ is Zariski-dense in $\cA$.

864: 

865: A {\it homomorphism of algebraic groups} 

866: is a morphism of the underlying affine algebraic varieties which is also

867: a group homomorphism. A homomorphism is {\it defined over $\BQ$} 

868: (or {\it $\BQ$-defined})

869: if the corresponding morphism of algebraic varieties is defined over $\BQ$.

870: It is called a {\it $\BQ$-defined isomorphism} if its inverse

871: exists and is also a homomorphism defined over $\BQ$.

872: An {\it automorphism} is an isomorphism of a linear algebraic group to itself.

873: We also use the more abstract concept of a $\BQ$-defined

874: linear algebraic group. 

875: It is well known that a $\BQ$-defined affine variety equipped with a

876: group structure given by $\BQ$-defined morphisms is $\BQ$-isomorphic with a 

877: $\BQ$-closed subgroup of 

878: $\GL(n,\BC)$, for some $n\in \BN$. As usually we write $\Ga=\BC$ 

879: for the additive

880: and $\Gm=\BC^*$ for the multiplicative group.

881: Quotients exist in the category of $\BQ$-defined linear algebraic groups. That

882: is given a $\BQ$-closed normal subgroup ${\cal N}$ of a 

883: $\BQ$-defined linear algebraic group $\cA$ the quotient group $\cA/{\cal N}$

884: is a $\BQ$-defined linear algebraic group and the natural map 

885: $\cA\to\cA/{\cal N}$ is a $\BQ$-defined homomorphism. We say that 

886: $\cA$ is the {\it almost direct product} of two Zariski-closed subgroups

887: ${\cal B},\,{\cal C}$ if ${\cal B}$ and ${\cal C}$ centralize each other, 

888: their intersection ${\cal B}\cap{\cal C}$ is finite and if  

889: $\cA={\cal B}\cdot{\cal C}$ holds. 

890: 

891: Let now $\cA$ be a $\BQ$-defined linear algebraic group, we write 

892: $\bU_{\cA}$ for its {\it unipotent radical}. This is the largest Zariski-closed

893: unipotent normal subgroup in $\cA$, it is $\BQ$-closed. 

894: The algebraic group $\cA$ is called {\it reductive} if $\bU_{\cA}=\{1\}$.

895: In particular, the quotient ${\cA}^{\rm red}  = \cA \big/ \bU_{\cA}$ is reductive,

896: for all linear algebraic groups $\cA$.

897: We write 

898: ${\cA}^{\rm sol}$ for the solvable radical of $\cA$. This is the largest 

899: connected Zariski-closed

900: solvable normal subgroup in $\cA$, it is $\BQ$-closed. 

901: A connected linear algebraic group $\cA$ is called \emph{semisimple} 

902: if ${\cA}^{\rm sol}  =\{1\}$. 

903: 

904: A linear algebraic group is called a {\it d-group} if

905: it consists of semisimple elements only. 

906: A Zariski-closed subgroup which is a

907: d-group is called a {\it d-subgroup}.

908: A d-group is reductive and

909: abelian-by-finite. A {\it torus} is a linear algebraic group isomorphic to 

910: $\Gm^n$,  for some $n\in\BN$. If ${\cal S}$ is a d-group then 

911: ${\cal S}^\circ$ is a torus.

912:  

913: Let $\cA$ be a $\BQ$-defined linear algebraic group. Then the following hold

914: (see \cite{Borel,BS,BT, PR}).

915: \begin{itemize}

916: \item[{\bf AG1:}] 

917: There exists a 

918: $\BQ$-closed reductive complement $\cR \leq \cA$, that is, 

919: $\cA=\bU_{\cA}\cdot \cR$ is a direct product of subgroups,

920: and $\cR  \cong \cA^{\rm red}$. In particular,  we also have 

921: $\cA \cong \bU_{\cA}\rfish \cA^{\rm red}$. Here the symbol $\cong$ indicates $\bbQ$-isomorphism

922: of linear algebraic groups,  and $\rfish$ indicates a semi-direct product.

923: \item[{\bf AG2:}]  All $\BQ$-closed reductive complements are conjugate by

924: elements of $\bU_{\cA}(\BQ)$.

925: \item[{\bf AG3:}] Let $\cA$ be a $\BQ$-defined connected and reductive 

926: linear algebraic group.  Then $\cA$ is the almost direct product of its 

927: center, which is a $\BQ$-closed d-subgroup, 

928: and the commutator subgroup $[\cA,\cA]$, which is

929: $\BQ$-closed, connected and semisimple. 

930: \item[{\bf AG4:}] Let $\cA$ be a $\BQ$-defined connected and semisimple 

931: linear algebraic group. Then its group $\Aut_a(\cA)$ of algebraic

932: automorphisms is  a $\BQ$-defined linear algebraic group and $\Inn_\cA$ is its

933: connected component.

934: \item[{\bf AG5:}] Let $\cA$ be a $\BQ$-defined connected and reductive 

935: linear algebraic group. Let ${\cal B}$ be $\BQ$-closed normal subgroup  of

936: $\cA$. Then there is a $\BQ$-closed  subgroup ${\cal C}$ of

937: $\cA$ which centralizes ${\cal B}$, satisfies $\cA={\cal B}\cdot {\cal C}$

938: such that ${\cal B}\cap {\cal C}$ is finite. That is, $\cA$ is the almost

939: direct product of ${\cal B}$ and ${\cal C}$. 

940: \item[{\bf AG6:}] Let ${\cal S}$ be a commutative d-group and $G$ 

941: a finite group of $\BQ$-defined

942: automorphisms of ${\cal S}$. Let ${\cal S}_1\le {\cal S}$ be a $G$-invariant

943: $\BQ$-closed subgroup. Then there is a $G$-invariant $\BQ$-closed subgroup 

944:  ${\cal S}_2\le {\cal S}$ such that ${\cal S}$ is the almost direct product of

945:  ${\cal S}_1$ and ${\cal S}_2$.

946:  \item[{\bf AG7:}]  Let $\cT \leq \cA$ be a torus. Then the centralizer

947:  $\Z_\cA(\cT)$ of $\cT$ in $\cA$ has finite index in the normalizer $\N_\cA(\cT)$.

948: \end{itemize}

949: 

950: Statement {AG6} is proved by using 

951: the category equivalence between the category of

952: $\BQ$-defined commutative 

953: d-groups and the category of continuous, $\BZ$-finitely

954: generated modules for the absolute Galois group of $\BQ$ 

955: (see \cite[\S 8]{Borel}) together with Maschke's theorem. 

956: The last fact AG7 is called the \emph{rigidity of tori}, see

957: \cite[Corollary 2 of III.8]{Borel} for a proof.

958: 

959: We shall add some remarks concerning the structure of 

960: solvable-by-finite groups. 

961: A linear algebraic group $\bH$ is solvable-by-finite if

962: its identity component $\bH^\circ$ is solvable.

963: A Cartan subgroup $\bC$ of  

964: $\bH$ is by definition 

965: the normalizer of a maximal torus $\bT$ in $\bH$.

966: 

967: 

968: For a $\BQ$-defined, solvable-by-finite linear algebraic group $\bH$ the

969: following hold:

970: 

971: \begin{itemize}

972: \item[{\bf SG1:}] There are maximal tori $\bT\le \bH$ which are $\BQ$-closed.

973: \item[{\bf SG2:}] Let $\bT$ be a maximal torus of 

974: $\bH$ and $\bC=\N_{\bH}(\bT)$. Then 

975: $\bC^\circ =  \N_{\bH^\circ}(\bT)$ 

976: equals the centralizer of $\bT$ in

977: $\bH^\circ$ and is a connected nilpotent group.    

978: Moreover, $\bC$ contains a maximal d-subgroup $\bS$ 

979: with $\bS^\circ = \bT$. 

980: \item[{\bf SG3:}] Let $\bT\le \bH$ be a maximal torus 

981: which is $\BQ$-closed. Then 

982: $\bC:=\N_{\bH}(\bT)$ is $\BQ$-closed and contains a maximal d-subgroup 

983: $\bS$ of

984: $\bH$ which is $\BQ$-closed and which satisfies $\bS^\circ = \bT$

985: \item[{\bf SG4:}] Let $\bS$ be a maximal d-subgroup of $\bH$. Then 

986: $\bH={\bf U}_{\bH}\cdot \bS={\bf U}_{\bH}\rfish \bS$.

987: \item[{\bf SG5:}] All maximal 

988: tori and also all maximal d-subgroups in $\bH$ are conjugate by elements 

989: of $[\bH^\circ, \bH^\circ] \leq \bU_{\bH}$. 

990: \item[{\bf SG6:}] All $\BQ$-closed maximal tori 

991: and all $\BQ$-closed maximal d-subgroups 

992: are conjugate by elements 

993: of $[\bH^\circ, \bH^\circ](\bbQ)$.

994: \item[{\bf SG7:}]

995: Let $\bC$ be a Cartan subgroup of $\bH$ and $\bF\le \bU_\bH$ a

996: normal subgroup of $\bH$ which contains 

997: $[\bH^\circ, \bH^\circ]$ then we have $\bH = \bF\cdot \bC$. 

998: \end{itemize}

999: 

1000: For almost all of this see \cite[Chapter III]{Borel}. 

1001: Statement SG6 is contained  in \cite{BS}, for a detailed 

1002: proof see \cite[Section 2]{GP1}.

1003: 

1004: 

1005: We have defined the notion of an arithmetic subgroup of a $\BQ$-defined linear

1006: algebraic group and that of an arithmetic group in the beginning of the

1007: introduction. We shall often have to use the behaviour of arithmetic subgroups

1008: under $\BQ$-defined homomorphisms. To describe this, let 

1009: $\rho :\cA_1\to\cA_2$ be a $\BQ$-defined homomorphism between 

1010: $\BQ$-defined linear algebraic groups $\cA_1,\, \cA_2$ and $A\le \cA_1$ a

1011: subgroup. Furthermore we suppose that $\rho$ is surjective. We have:  

1012: \begin{itemize}

1013: \item[{\bf AR1}:] If $A$ is an arithmetic subgroup of $\cA_1$ then     

1014: $\rho(A)$ is an arithmetic subgroup of $\cA_2$.

1015: \item[{\bf AR2}:] Suppose that $\rho$ is injective, then $A$ 

1016: is an arithmetic subgroup of $\cA_1$ if and only if $\rho(A)$ is commensurable

1017: with $\cA_2(\BZ)$.

1018: \item[{\bf AR3}:] Every abelian subgroup of an arithmetic group is finitely

1019: generated. 

1020: \item[{\bf AR4}:] Let $\cA$ be a $\bbQ$-defined 

1021: group and  $\cA_1$, $\cA_2$ $\bbQ$-defined subgroups

1022: such that $\cA = \cA_1 \rtimes \cA_2$ is a semi-direct product. 

1023: Let $A  \leq \cA(\BQ)$ be a subgroup. Assume there exists subgroups 

1024: $A_1 $, $A_2$ of $A$ such that $A = A_1 \rtimes \A_2$ is a semi-direct

1025: product. If  $A_1$ and $A_2$ are arithmetic subgroups in $\cA_1$ and $\cA_2$, respectively, then $A$ is an arithmetic subgroup of $\cA$. 

1026: \end{itemize}

1027: Statement AR1 is proved in \cite{Borel2} and AR2 is a consequence of it. 

1028: Statement AR2 shows that the notion of an arithmetic subgroup does not

1029: depend on the particular embedding of the ambient algebraic group into

1030: $\GL(n,\BC)$. Statement AR3 follows by consideration 

1031: of the Zariski-closure of

1032: the given abelian subgroup.  Statement AR4 is a consequence of AR1.

1033: 

1034: 

1035: 

1036: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1037: 

1038: \subsection{Algebraic groups of automorphisms}

1039: 

1040: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1041: 

1042: Let $G$ be a group acting by automorphisms on a group $A$ and let $B\le A$ be

1043: a subgroup of $A$. We write

1044: $$\N_G(B):=\{\, g\in G\ |\ g(B)=B\,\},\quad  

1045: {\rm Z}_G(B):=\{\, g\in G\ |\ g(b)=b\ {\rm  for\ all}\ b\in B\,\}$$

1046: for the normalizer and centralizer of $B$ in $G$.

1047: 

1048: \begin{definition}\label{algauto} 

1049: Let ${\cal A}$ be a $\BQ$-defined linear algebraic group 

1050: and let $G$ be a group which acts by $\BQ$-defined automorphisms on $\cA$. 

1051: Then $G$ normalizes the unipotent radical $\bU_\cA$ and we

1052: say that $G$ acts as an {\it algebraic group of automorphisms} on $\cA$ if 

1053: ${\rm Z}_G({\cal T})$ has finite index in $\N_G({\cal T})$ for every

1054: $\bbQ$-defined torus  ${\cal T}$ of ${\cA}^{\rm red} = \cA\big/\bU_\cA$.   

1055: \end{definition}

1056: 

1057: Note that $G=\GL(n,\BZ)$ acts by $\BQ$-defined automorphisms on the torus

1058: $\cA=\Gm^n$. But for $n\ge 2$ this is not an algebraic group of automorphisms

1059: on $\cA$. Positive examples are given in the next lemma. 

1060: 

1061: \begin{lemma}\label{algautoex}

1062: Let ${\cal B}$ be a $\BQ$-defined linear algebraic group and 

1063: $\cA\le{\cal B}$ a 

1064: $\BQ$-closed subgroup. Let further $G\le {\cal B}(\BQ)$ be a subgroup which

1065: normalizes $\cA$. Then $G$ acts by conjugation on $\cA$ as an algebraic group

1066: of automorphisms.

1067: \end{lemma}

1068: \begin{proof}

1069: Replacing ${\cal B}$ by the Zariski-closure of $\cA \cdot G$, 

1070: we may assume that $\cA$ is

1071: normal in ${\cal B}$. Then $\bU_\cA$ is also normal in ${\cal B}$ and the image

1072: of $G$ in ${\cal B}\big/\bU_\cA$ acts by conjugation on $\cA\big/\bU_\cA$. 

1073: Thus  the lemma follows from the rigidity of tori  (AG7). 

1074: \end{proof}

1075: 

1076: 

1077: \begin{lemma}\label{algformal1} 

1078: Let ${\cal A}$ be a $\BQ$-defined linear algebraic group. 

1079: Let $G$ be an algebraic  group of automorphisms on $\cA$.

1080: Then 

1081: ${\rm Z}_G({\cal S})$ has finite index in $\N_G({\cal S})$ for every

1082: $\bbQ$-closed commutative d-subgroup $\cS$ of $\cA\big/\bU_\cA$. 

1083: \end{lemma}

1084: \begin{proof}

1085: To prove the lemma, we may replace $\cA$ by $\cA\big/\bU_\cA$. 

1086: We consider a $\BQ$-closed commutative

1087: d-subgroup ${\cal S}$ of $\cA$. 

1088: Its connected component ${\cal S}^\circ$ is a

1089: $\BQ$-defined torus in $\cA^\circ$.  

1090: We have 

1091: $\N_G({\cal S})\le \N_G({\cal S}^\circ)$.

1092: Our hypothesis implies that

1093: $Z_1=\N_G({\cal S})\cap{\rm Z}_G({\cal S}^\circ)$ has finite index in 

1094: $\N_G({\cal S})$. 

1095: 

1096: There is a finite Zariski-closed

1097: subgroup ${\cal E}\le{\cal S}$ such that ${\cal S}={\cal S}^\circ\cdot  

1098: {\cal E}$. Let  ${\cal E}_0$ be the kernel of the homomorphism from ${\cal S}$ to  

1099: ${\cal S}$ which sends $x\in {\cal S}$ to $x^{|{\cal E}|}$.

1100: It is finite, contains ${\cal E}$ and is normalized by 

1101: $\N_G({\cal S})$. 

1102: Trivially $Z_2=\N_G({\cal S})\cap{\rm Z}_G({\cal E}_0)$

1103: has finite index in $\N_G({\cal S})$. Since $Z_1\cap Z_2$ is contained in 

1104: ${\rm Z}_G({\cal S})$, this  proves that ${\Z}_G(\cS)$ is of finite index

1105: in ${\N}_G({\cal S})$.

1106: \end{proof}

1107: 

1108: The following lemmata describe natural operations which can be perfomed with

1109: algebraic groups of automorphisms. 

1110: 

1111: 

1112: \begin{lemma}\label{algformal2} 

1113: Let ${\cal A}$ be a $\BQ$-defined linear algebraic group and ${\cal B}$ a

1114: $\BQ$-closed normal subgroup in $\cA$. 

1115: Let $G$ be an algebraic group of automorphisms of $\cA$ which normalizes

1116: ${\cal B}$.

1117: Then $G$ acts as a group of algebraic automorphisms on ${\cal B}$ and also on 

1118: $\cA\big/{\cal B}$. 

1119: \end{lemma}

1120: \begin{proof} This is clear for $\cB$, since $\cB$ is normal in $\cA$.

1121: To prove that $G$ acts as an algebraic group of

1122: automorphisms on $\cA\big/\cB$, we use a 

1123: $G$-invariant decomposition AG5.  We leave the details to

1124: the reader. 

1125: \end{proof}

1126: 

1127: Remark that, in general,  the converse of the statements of Lemma \ref{algformal2} 

1128: is not true. Nevertheless, we have: 

1129: 

1130: 

1131: \begin{lemma}\label{algformalproducts} 

1132: Let ${\cal A}$ and $\cB$ be $\BQ$-defined linear algebraic groups, and let 

1133: $G$ act as an algebraic group of automorphisms on $\cA$ and on $\cB$. Then

1134: $G$ acts as an algebraic group of automorphisms on the product 

1135: $\cC = \cA \times \cB$.  \end{lemma}

1136: \begin{proof} Let $G$ act on $\cC$ by the product action. 

1137: Let $\cT$ be a torus in $\cC^{\rm red} =  {\cA}^{\rm red} \times {\cB}^{\rm red} $ 

1138: and put  $G' = N_G(\cT)$. The torus $\cT$ is contained in the direct product  

1139: of factors  $\cT_\cA \leq {\cA}^{\rm red} $, $\cT_\cB \leq {\cB}^{\rm red} $ which 

1140: are stabilized by $G'$. Hence,  both are centralized by a finite index subgroup of $G'$.

1141: Therefore, $\cT$ is centralized by this finite index subgroup of $G'$.

1142: \end{proof}

1143: 

1144: 

1145: \begin{lemma}\label{algformal3} 

1146: Let ${\cal A}$ be a $\BQ$-defined linear algebraic group and $G$ a group of

1147: $\BQ$-defined automorphisms of $\cA$. Suppose $L,\, H\le G$ 

1148: are subgroups such that $L$ is normal in $G$ and $G=L\cdot H$ holds. If

1149: both $L$ and $H$ are $\cA$-algebraic groups of automorphisms of $\cA$ then also $G$

1150: has this property.

1151: \end{lemma}

1152: \begin{proof} We consider $G$ as a group of

1153: $\BQ$-defined automorphisms of\ ${\cal B}=(\cA\big/\bU_\cA)^\circ$.

1154: We have the almost direct decomposition 

1155: ${\cal B}={\cal Z}\cdot [{\cal B},{\cal B}]$, where ${\cal Z}$ is the center of

1156: ${\cal B}$. The group $G$ normalizes both ${\cal Z}$ and 

1157: $[{\cal B},{\cal B}]$. Let ${\cal T}$ be a $\bbQ$-defined torus

1158: of ${\cal B}$. If ${\cal T}\le [{\cal B},{\cal B}]$ then 

1159: ${\rm Z}_G({\cal T})$ has finite index in   

1160: ${\rm N}_G({\cal T})$ since $G$ contains a subgroup of finite index which acts

1161: by inner automorphisms on ${[\cB,\cB]}$ (AG4). 

1162: Now assume $\cT \leq \cZ$.

1163: By our assumption on $L$ and $H$ it follows that 

1164: ${\rm Z}_G({\cal Z}^0)$ has finite index in   

1165: $G$  (since $G$ normalizes ${\cal Z}^0$), and, a fortiori,

1166: ${\rm Z}_G({\cal T})$ has finite index in   

1167: ${\rm N}_G({\cal T})$. 

1168: The case of a general ${\cal T}$ can be treated 

1169: by projecting $\cT$ on the factors of the almost direct

1170: product decomposition AG3.

1171: \end{proof}

1172: 

1173: 

1174: In Section \ref{subsecari} the following fact plays a crucial role.

1175: 

1176: \begin{proposition}\label{decozen}

1177: Let ${\cal A}$ be a connected\/  $\BQ$-defined linear algebraic group,  equipped 

1178: with an algebraic group of automorphisms $G$.

1179: Then there are: \begin{itemize}

1180: 

1181: \item[i)] a $\BQ$-closed, $G$-invariant central d-subgroup 

1182: ${\cal Z}_1$ in $\cA$ and 

1183: 

1184: \item[ii)] a $\BQ$-closed, $G$-invariant 

1185: subgroup ${\cal A}_1$ of $\cA$ with unipotent-by-finite

1186: center 

1187: \end{itemize}

1188: such that

1189: $\cA$ is decomposed as the almost direct product of $\cZ_1$ and $\cA_1$. 

1190: \end{proposition}

1191: \begin{proof}

1192: Let $\cA^{\rm red}$ be a maximal reductive subgroup of $\cA$ which

1193: is defined over $\bbQ$. Let ${\cal Z}$ be the center of $\cA^{\rm red}$

1194: and  let $\cZ_1$ be the maximal torus in the center of $\cA$. Thus $\cZ_1$

1195: is $\bbQ$-closed in $\cA$, it is 

1196: contained in $\cZ$, and it is normalized by $G$.

1197: 

1198: By Lemma \ref{algformal1}, the group $G_1$ of automorphisms of $\cZ$ 

1199: which is induced by the quotient action of $G$ on the isomorphic image 

1200: of $\cZ$ in $\cA\big/\bU_\cA$ is finite.  Hence, by AG6, we may choose a $\BQ$-closed and 

1201: $G_1$-invariant subgroup ${\cal Z}_2\le {\cal Z}$ such that

1202: ${\cal Z}$ is the almost

1203: direct product of ${\cal Z}_1$ and ${\cal Z}_2$.

1204: This shows that  ${\cal Z}_1$ and $\cA_1 ={\bf U}_\cA\cdot{\cal Z}_2\cdot [\cA^{\rm red},\cA^{\rm red}]$ satisfy the requirements of

1205: the lemma.

1206: \end{proof}

1207:  

1208: 

1209: 

1210: 

1211: 

1212: 

1213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1214: 

1215: \section{The group of automorphisms of a solvable-by-finite 

1216: linear algebraic group} \label{section:auta}

1217: 

1218: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1219: 

1220: Let $\cA$ be a linear algebraic group

1221: defined over $\bbQ$.     

1222: We let 

1223: \begin{equation} \Aut_{a}(\cA) \; , \; \Aut_{a,\bbQ}(\cA)

1224: \end{equation} 

1225: denote 

1226: the group of all automorphisms of $\cA$, respectively 

1227: the group of all $\bbQ$-defined automorphisms of $\cA$. 

1228: We also need the following concept. 

1229: \begin{definition}

1230: We say that a linear algebraic group $\cA$ has a {\it strong unipotent

1231:   radical} if the centralizer $\C_{\cA}(\bU_\cA)$ of its unipotent radical

1232: $\U_\cA$ is contained in $\U_\cA$. 

1233: \end{definition}

1234: Given a $\BQ$-defined solvable-by-finite linear algebraic group $\bH$  

1235: with a strong unipotent radical we will identify $\Aut_{a}(\bH)$ in a natural 

1236: way with a $\bbQ$-defined linear algebraic group. We obtain special features 

1237: of this identification which we will need later. 

1238: 

1239: For a general linear algebraic group $\cA$,

1240: the group $\Aut_{a}(\cA)$ is a group of type ALA, that is, 

1241: $\Aut_{a}(\cA)$ is an extension 

1242: of an affine algebraic group by 

1243: an arithmetic group, see \cite{BS}. 

1244: Our approach is a variation

1245: of corresponding results in \cite{BS}. A construction similar to ours, but in a more

1246: restricted situation, is contained in \cite{GP1}.   

1247:  

1248: 

1249: 

1250: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1251: 

1252: \subsection{The algebraic structure of $ \Aut_{a}(\bH)$} 

1253: \label{subsect:algebraicstructure}

1254: 

1255: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1256: 

1257: We will assume here that $\bH$ is a $\BQ$-defined solvable-by-finite

1258: linear algebraic group which has a 

1259: strong unipotent radical $\bU:=\bU_\bH$.

1260: Let $\lu$ denote the Lie-algebra of $\bU$. 

1261: The Lie-algebra $\lu$ is

1262: defined over $\bbQ$ and $\Aut(\lu) \leq \GL(\lu)$

1263: is a $\bbQ$-defined linear algebraic group. The exponential map

1264: \begin{equation} 

1265: \exp:  \lu \rightarrow  \bU

1266: \end{equation} 

1267: is a $\bbQ$-defined isomorphism of varieties. Thus, via the 

1268: bijective homomorphism  

1269: \begin{equation} \label{eq:exp}

1270:   \Aut_a(\bU) \ni \, \Phi \; \mapsto \; \exp^{-1} \circ \,

1271: \Phi \circ \exp  \; \, \in \Aut(\lu)\; 

1272: \end{equation}

1273: the group $\Aut_a(\bU)$ attains a natural structure of a linear 

1274: algebraic group which is defined over $\bbQ$.  

1275: 

1276: We let $\bS\le \bH$ be a maximal $\BQ$-closed d-subgroup and obtain the

1277: decomposition $\bH=\bU\cdot \bS$ described in SG4 of Section \ref{prereq}.

1278: We define

1279: \begin{equation}

1280: \Aut_{a}(\bH)_\bS \, := \, 

1281: \{\, \Phi \in  \Aut_{a}(\bH)  \mid \Phi(\bS) = \bS \} \; \leq 

1282: \Aut_a(\bH) . 

1283: \end{equation}

1284: 

1285: \begin{lemma}

1286: The restriction map $\Phi \mapsto \Phi|_{\bU}$

1287: identifies $ \Aut_{a}(\bH)_\bS$ with a 

1288: $\bbQ$-defined closed subgroup of $\Aut_a(\bU)$.

1289: \end{lemma} 

1290: \begin{proof}

1291: Since $\bH$ has a strong unipotent radical the restriction map is

1292: injective on  $\Aut_{a}(\bH)_\bS$. 

1293: Let $\Ad(\bS)$ denote the image of $\bS$ 

1294: under the restriction of the adjoint representation to $\lu$.

1295: Then  $\Ad(\bS)$ is a $\bbQ$-defined 

1296: subgroup of $\Aut(\lu)$. Note that the obvious homomorphism

1297: $\bS\to \Ad(\bS)$ is a $\BQ$-defined isomorphism since $\bH$ has a strong

1298: unipotent radical (see also \cite{GP1}, Lemma 2.3).

1299: It is clear that, via the 

1300: exponential map, the restriction of $ \Aut_{a}(\bH)_\bS$ to $\bU$ 

1301: is isomorphic to the normalizer 

1302: of $\Ad(\bS)$ in $\Aut(\lu)$. Since $\Ad(\bS)$ is a $\bbQ$-defined

1303: group, its normalizer and centralizer are Zariski-closed subgroups

1304: of $\Aut(\lu)$. By the Galois criterion for 

1305: rationality (see \cite[AG 14]{Borel})

1306: the normalizer and centralizer of $\Ad(\bS)$ are defined over $\bbQ$.

1307: In particular, $ \Aut_{a}(\bH)_\bS$ restricts to a $\bbQ$-defined subgroup of

1308: $\Aut_a(\bU)$. 

1309: \end{proof}

1310: 

1311: Thus the lemma furnishes a natural structure of $\bbQ$-defined 

1312: linear algebraic group on $ \Aut_{a}(\bH)_\bS$. 

1313: Since $ \Aut_{a}(\bH)_\bS$ acts on $\bU$ by

1314: $\bbQ$-defined morphisms, 

1315: the semi-direct product $\bU \rtimes \Aut_{a}(\bH)_\bS$ is

1316: an affine algebraic group over $\bbQ$, and thus is

1317: also equipped with the natural structure of 

1318: a $\bbQ$-defined linear algebraic group, see \cite{Borel}.

1319: Let $\Inn_u \in \Aut_a(\bH)$ denote the inner automorphism 

1320: corresponding to $u\in \bU$, that is,  $\Inn_u(h) = u h u^{-1}$, for

1321: all $h \in \bH$. We consider the  homomorphism 

1322: \begin{equation} \label{eq:Theta}

1323:  \Theta: \, \bU \rtimes   \Aut_{a}(\bH)_\bS  \longrightarrow  \Aut_{a}(\bH) 

1324: \,\; , \; \;  (u, \Phi) \mapsto \Inn_u \circ \Phi \; . 

1325: \end{equation}

1326: 

1327: 

1328: \begin{lemma} The homomorphism $\Theta$ is surjective. 

1329: \end{lemma}

1330: \begin{proof}

1331: The group $\bH$ 

1332: decomposes as a semi-direct

1333: product of $\bbQ$-defined algebraic groups

1334: $ \bH = \bU \cdot \bS  \cong  \bU \rtimes \bS$.  

1335: All maximal d-subgroups of $\bH$ are conjugate by elements of $\bU$, 

1336: and $\bbQ$-defined

1337: d-subgroups are conjugate by elements of $\bU(\bbQ)$ (see Section

1338: \ref{prereq}). 

1339: Therefore, the homomorphism $\Theta$ 

1340: is onto $\Aut_a(\bH)$.

1341: \end{proof}

1342: 

1343: 

1344: \begin{lemma}

1345: The kernel of the homomorphism $\Theta$ is a $\bbQ$-defined 

1346: unipotent subgroup of $\bU \rtimes  \Aut_{a}(\bH)_\bS$.

1347: \end{lemma}

1348: \begin{proof}

1349: In fact, $\ker \Theta = \{ (u, \Inn_u^{-1} ) \mid 

1350: \Inn_u \in  \Aut_{a}(\bH)_\bS \}$ is a unipotent subgroup 

1351: of  $\bU \rtimes  \Aut_{a}(\bH)_\bS$. Since the homomorphism 

1352: $\Theta$ corresponds to

1353: a $\bbQ$-defined action of  $\bU \rtimes  \Aut_{a}(\bH)_\bS$

1354: on the variety $\bU$, its kernel is a $\bbQ$-defined

1355: subgroup. 

1356: \end{proof}

1357: Let 

1358: \begin{equation}

1359:  \cA_\bH :=  

1360: \bU \rtimes  \Aut_{a}(\bH)_\bS \big/ \, \ker \Theta

1361: \end{equation}  

1362: denote the quotient group 

1363: of $ \bU \rtimes   \Aut_{a}(\bH)_\bS$ by the kernel of $\Theta$. 

1364: The group $\cA_\bH$ has a natural structure of

1365: a $\bbQ$-defined linear algebraic group, 

1366: such that

1367: \begin{equation}

1368:  \cA_\bH(\bbQ) = \bU \rtimes  \Aut_{a}(\bH)_\bS  \left(\bbQ\right) \big/ 

1369: \left(\ker \Theta\right)(\bbQ) \;. 

1370: \end{equation}

1371: Therefore, we have:

1372: 

1373: 

1374: \begin{theorem}

1375: The homomorphism 

1376: $   \cA_\bH  \rightarrow  \Aut_{a}(\bH) $

1377: which is induced by \eqref{eq:Theta} naturally identifies $\Aut_a(\bH)$ 

1378: with the complex points 

1379: of the $\bbQ$-defined linear algebraic group $\cA_\bH$.  

1380: \end{theorem}

1381: By arguments using the above setup, 

1382: the rational points of the corresponding affine algebraic group 

1383: are naturally interpreted in the following way: 

1384: 

1385: \begin{proposition} \label{rationalisq}

1386: Under the homomorphism $   \cA_\bH  \rightarrow  \Aut_{a}(\bH) $

1387: which is induced by \eqref{eq:Theta} 

1388: the group of rational points $\cA_\bH(\bbQ)$ of $\cA_\bH$ 

1389: corresponds to the group $\Aut_{a,\bbQ}(\bH)$ of\/

1390: $\bbQ$-defined automorphisms of\/ $\bH$. 

1391: \end{proposition}

1392: \begin{proof} Let $\Phi \in \cA_\bH(\bbQ)$. It is clear from

1393: our construction that $\Phi$ preserves the group of rational 

1394: points $\bH(\bbQ) \leq \bH$ which is Zariski-dense in $\bH$

1395: (cf. \cite[\S18.2]{Borel}). 

1396: By the Galois-criterion for rationality (compare \cite[AG 14]{Borel}), 

1397: $\Phi$ is defined over $\bbQ$.

1398: 

1399: Conversely, assume $\Phi \in \Aut_a(\bH)_\bbQ$ is a $\bbQ$-defined

1400: automorphism of $\bH$. Then, by AG2, there exists $v \in \bU(\bbQ)$ such that 

1401: $\Psi = \Inn_v \circ \Phi \in \Aut_a(\bH)_\bS$, and $\Psi$ is

1402: defined over $\bbQ$ as well.  The exponential correspondence shows

1403: that $\Psi \in \Aut_a(\bH)_\bS(\bbQ)$. It follows that 

1404: $\Phi \in   \Aut_a(\bH)(\bbQ)$. 

1405: \end{proof}

1406: Henceforth, we will identify $\Aut_a(\bH)$ with (the complex points of)

1407: the linear algebraic group $\cA_\bH$, and

1408: $\Theta: \bU \rtimes   \Aut_{a}(\bH)_\bS \rightarrow \cA_\bH = \Aut_a(\bH)$

1409: becomes a morphism of algebraic groups which is defined over $\bbQ$.

1410: 

1411: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1412: 

1413: \subsection{Arithmetic subgroups of $\Aut_a(\bH)$} \label{section:asubgroups}

1414: 

1415: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1416: 

1417: We keep the notation of the previous subsection. In particular, 

1418: $\bH$ denotes a $\BQ$-defined solvable-by-finite linear algebraic group with 

1419: a strong unipotent radical  $\bU=\bU_\bH$. 

1420: Let further $\bS$ be a maximal $\BQ$-defined d-subgroup

1421: of $\bH$.

1422: 

1423: Let $\theta \leq \bU({\bbQ})$ be a finitely generated 

1424: subgroup which is Zariski-dense in $\bU$. Then  $\theta$ is an

1425: arithmetic subgroup of the $\bbQ$-defined group $\bU$. 

1426: We explain now how the choice of $\theta$ gives rise

1427: to an arithmetic subgroup $A_\theta$ of  $\Aut_a(\bH)=\cA_\bH$.  

1428: 

1429: \begin{definition}\label{LThedef}

1430: We  define the following subgroups of $\Aut_a(\bH)$: 

1431: \begin{equation}

1432: \cA_\theta :=  \{\, \Phi \in \Aut_a(\bH)  \mid \Phi(\theta) = 

1433: \theta \, \} \; ,\qquad 

1434:  \Lambda_\theta \, := \,  \Aut_{a}(\bH)_\bS  \cap  \cA_\theta \; ,  

1435: \end{equation} 

1436: \begin{equation}

1437:  \A_\theta :=  \Theta\left(\theta \rtimes 

1438: \Lambda_\theta\right) \,  \leq \cA_\bH{(\bbQ)} \; .

1439: \end{equation}

1440: \end{definition}

1441: 

1442: \begin{lemma}   \label{lemma:Lambdatheta}

1443: Let $\theta \leq \bU({\bbQ})$ be a finitely generated 

1444: subgroup which is Zariski-dense in $\bU$. Then 

1445: the group $\Lambda_\theta$ is an arithmetic subgroup 

1446: of $ \Aut_{a}(\bH)_\bS$. 

1447: \end{lemma} 

1448: \begin{proof} As explained in Subsection \ref{subsect:algebraicstructure}, 

1449: the group $ \Aut_{a}(\bH)_\bS$ is $\bbQ$-isomorphic to the normalizer

1450: $\N_{\Aut(\lu)}(\Ad(\bS))$ via the isomorphism \eqref{eq:exp}.

1451: Embedding $\Aut(\theta)$ as a subgroup of  $\Aut_a(\bU)$, it

1452: corresponds to 

1453: $$ N_{\Aut(\lu)}\big( \log \theta\big) = 

1454: \{ \Psi \in \Aut(\lu) \mid \Psi (\log 

1455: \theta) \subseteq \log \theta \}\, \leq \Aut(\lu) \; . $$

1456: The group $N_{\Aut(\lu)}( \log \theta)$

1457: stabilizes the  lattice $L$ in $\lu$ 

1458: which is generated by the set $\log \theta$. 

1459: Moreover, $N_{\Aut(\lu)}( \log \theta)$ is arithmetic in $\Aut(\lu)$, 

1460: see  \cite[Chapter 6]{Segal}. 

1461: It follows that $\Lambda_\theta \leq  \Aut_{a}(\bH)_\bS$  

1462: corresponds to 

1463: $$ \N_{\Aut(\lu)}\big(\Ad(\bS)\big)  

1464: \cap N_{\Aut(\lu)}\big( \log \theta(\Gamma)\big) \; . $$ 

1465: Thus,

1466: under the natural linear representation of $ \Aut_{a}(\bH)_\bS$ in

1467: $\GL(\lu)$,  $\Lambda_\theta$ is commensurable to

1468: the stabilizer of a lattice $L  \subset \lu({\bbQ})$.    

1469: Therefore,  $\Lambda_\theta$ is arithmetic in $ \Aut_{a}(\bH)_\bS$.

1470: \end{proof}

1471: 

1472: Using AR4 in Section \ref{prereq} we deduce that $\theta \rtimes \Lambda_\theta $

1473: is arithmetic in $\bU \rtimes  \Aut_{a}(\bH)_\bS$.

1474: Since $A_\theta$ is the image of the arithmetic group 

1475: $\theta \rtimes \Lambda_\theta$ under the $\bbQ$-defined 

1476: surjective homomorphism

1477: $\Theta$, $A_\theta$ is arithmetic in $ \Aut_{a}(\bH)$. This proves:

1478: 

1479: \begin{proposition} \label{prop:Atheta}

1480: Let $\theta \leq \bU({\bbQ})$ be a finitely generated 

1481: subgroup which is Zariski-dense in $\bU$.  Then 

1482: the group $\A_\theta$ is an arithmetic subgroup of $ \Aut_{a}(\bH)$. 

1483: \end{proposition}

1484: 

1485: 

1486: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1487: 

1488: \subsection{Some closed subgroups of $\Aut_a(\bH)$}\label{section:subgroups}

1489: 

1490: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1491: 

1492: We stick to the conventions about $\bH$, $\bU$, $\bS$.

1493: In addition,  we introduce $\bF \leq \bU$ to be a 

1494: $\bbQ$-closed normal subgroup of $\bH$ 

1495: which contains the commutator group $[\bH^\circ, \bH^\circ]$. 

1496: By SG7 of Section \ref{prereq},  we have $\bH=\bF\cdot\bC$, for every

1497: $\BQ$-closed Cartan subgroup of $\bH$.

1498: As before

1499: $\bS$ denotes a maximal d-subgroup in $\bH$. 

1500: 

1501: Let $\N_\cA(\bF)$ denote  the subgroup of elements in 

1502: $\Aut_a(\bH)$ which preserve $\bF$. Additionally, we define: 

1503: $$   \cA_{\bH | \bF}   :=  \,    \{ \Phi \in \N_\cA(\bF)  \mid 

1504:   \Phi|_{{\bH /\bF}} =\id_{\bH / \bF} \}  \; , $$

1505: $$    \cAS   :=  \,  \{ \Phi \in \N_\cA(\bF)  \mid 

1506:   \Phi|_{{\bH^\circ /\bF}} = \id_{\bH^\circ  / \bF} \, , \, \Phi(\bS) = \bS \} \; , $$

1507: $$  \cASid   :=  \, \{ \Phi \in \cA_{\bS} 

1508: \mid \Phi|_\bS  = \id_\bS \} \; . $$

1509: It is easy to see that these groups are $\bbQ$-closed 

1510: subgroups of $\Aut_a(\bH)$. 

1511: 

1512: \begin{lemma}  \label{lemma:subgroups1}

1513: The following hold: \begin{itemize}

1514: 

1515: \item[i)] $\cASid \leq \cA_{\bH | \bF} $.

1516: 

1517: \item[ii)] Define

1518: $ \cA_{\bH | \bU} =  \{ \Phi \in \cA \mid \Phi|_{{\bH/ \bU }} 

1519: = \id_{\bH/\bU} \}  $, 

1520: then $ \cA_{\bH | \bU} \cap \cAS = \cASid$.

1521: 

1522: \item[iii)]  $\cA_{\bH | \bF} \cap \cAS = \cASid$. 

1523: \end{itemize}

1524: \end{lemma}

1525: \begin{proof} Let $\Phi \in \cASid$ and $u \in \bU$. 

1526: Since  $\Phi|_{{\bH^\circ /\bF}} = \id_{\bH^\circ / \bF}$,

1527: we have

1528: $\Phi(u) = u f_u$, where $f_u \in \bF$. Now let $h \in \bH$, 

1529: and write $h = s u$, where $s \in \bS$, $u \in \bU$. It follows

1530: that $\Phi(h) = h f_u$. Hence, $\Phi \in \cA_{\bH | \bF}$. This 

1531: proves i). 

1532: 

1533: Let $\Phi \in  \cA_{\bH | \bU} \cap \cAS$ and $s \in \bS$.

1534: Then $\Phi(s) = s u_s$, where $u_s  \in \bU$. Since $\Phi \in \cA_S$, 

1535: $\Phi(s) \in \bS$. It follows that

1536: $u_s = 1$. Therefore, $\Phi \in \cASid$. This proves

1537: ii). 

1538: 

1539: By i), $\cASid \subset  \cA_{\bH | \bF}$. Conversely, 

1540: $\cA_{\bH | \bF} \cap \cA_S \leq  \cA_{\bH | \bU}  \cap \cA_S$,

1541: and hence by ii),  $\cA_{\bH | \bF} \cap \cA_S \leq \cASid$.

1542: This proves iii). 

1543: \end{proof}

1544: 

1545: Before we proceed let us introduce some notation concerning inner

1546: automorphisms.

1547: 

1548: \begin{definition}\label{innn} 

1549: Let $A$ be a group and $B\le A$ a subgroup. Let $a \in A$. 

1550: We write $\Inn_a\in \Aut(A)$ for the inner

1551: automorphism of $A$ defined by $g \mapsto a g a^{-1}$, for all $g \in A$. 

1552: Given an element $b\in B$, 

1553: we set $\Inn_b^A\in \Aut(A)$ for the corresponding inner 

1554: automorphism of $A$ (to distinguish it from the induced inner 

1555: automorphism of $B$). We write $\Inn_B^A$ for the subgroup of $\Aut(A)$

1556: consisting of all elements  $\Inn_b^A$, $b \in B$.

1557: \end{definition}

1558: Let $\bC$ be Cartan subgroup of $\bH$ which contains $\bS$. 

1559: Thus $\bC = \bU_\bC\cdot \bS$. We define

1560: \begin{equation}  

1561: %\cA_\bC :=   \{ \Phi \in {\rm N}_\cA(\bF) \mid 

1562: %  \Phi|_{{\bH^\circ /\bF}} = \id_{\bH^\circ / \bF}  \; , 

1563: %\,  \Phi(\bC) \subset \bC \} \; .

1564: \cA_\bC :=   \{ \Phi \in \N_\cA(\bF)  \mid   \Phi|_{{\bH^\circ /\bF}} = \id_{\bH^\circ  / \bF} \, , \,  

1565: \Phi(\bC) \subset \bC \} \; .

1566: \end{equation}

1567: 

1568: If $\bC$ is defined over $\bbQ$ then $\cA_\bC$

1569: is a $\bbQ$-defined subgroup of $\Aut_a(\bH)$.

1570: 

1571: \begin{lemma} \label{lemma:subgroups2}

1572: The following hold:  \begin{itemize}

1573: 

1574: \item[i)] Let $u \in \bU_\bC$ such that $\Inn^\bH_u \in \cA_{\bH | \bF}$.

1575: Then there exists $v \in \bC \cap \bF$ 

1576: such that, for all $s \in \bS$,  

1577: $\Inn^\bH_u\,  (s) \;  = \;  \Inn^\bH_v \,  (s) $.

1578: 

1579: \item[ii)]  $ \cA_{\bH | \bF} \cap \cA_\bC = \Inn^\bH_{\bF \cap \bC}\cdot 

1580: \cASid$.   

1581: 

1582: \end{itemize}

1583: \end{lemma} 

1584: \begin{proof} 

1585: Let $s \in \bS$. Then $ \Inn^\bH_u (s) =  

1586: s f_s$, where $f_s  \in \bF \cap \bC$.  

1587: Moreover, $f_s =1$, for $s \in \bS^\circ$ since $u$ normalizes $\bS^\circ$. 

1588: By a standard argument (compare for example \cite{BS})

1589: the cocycle $s \mapsto f_s$ is of the form $f_s = s^{-1} v s v^{-1}$, for

1590: some $v \in \bF \cap \bC$. Thus, $\Inn^\bH_v (s) =   \Inn^\bH_u  (s)$, 

1591: for all $s \in \bS$.   

1592: 

1593: Let $\Phi \in  \cA_{\bH | \bF} \cap \cA_\bC$. Since 

1594: $\Phi(\bC) \leq \bC$, 

1595: there exists $v \in \U_\bC$ 

1596: such that $ v \bS v^{-1} = \Phi(\bS)$. Then 

1597: $\Psi = \Inn^\bH_{v^{-1}}\circ \Phi \in \cAS \cap \cA_{\bH|\bU}$ holds. 

1598: By Lemma \ref{lemma:subgroups1} (ii), $\Psi \in \cASid$ follows.

1599: In particular, $\Psi(s) = s$, for all $s \in \bS$. 

1600: Hence, $\Phi(s) = \Inn^\bH_v(s)$, for all $s \in \bS$. 

1601: Since we have $\Phi \in 

1602: \cA_{\bH | \bF}$,  this implies that $\Inn^\bH_v \in  \cA_{\bH | \bF}$ holds.

1603: By the first part,  there exists $w \in \bC \cap \bF$ such that 

1604: $\Inn^\bH_w (s) = \Inn^\bH_v (s)$ holds for all $s \in \bS$. 

1605: In particular we find 

1606: $\Inn^\bH_{w^{-1}}\circ \Phi \in \cAS \cap \cA_{\bH | \bF} = 

1607: \cASid$. 

1608: Hence, $\cA_{\bH | \bF} \cap \cA_\bC \subset \Inn^\bH_{\bF \cap \bC}

1609: \cdot \cASid$ holds.

1610: The lemma follows.

1611: \end{proof}

1612: 

1613: \begin{proposition} \label{cAHF}

1614: Let $\bH$ be a

1615: $\BQ$-defined solvable-by-finite linear algebraic group and 

1616: $\bF \leq \bU_\bH$ be a 

1617: $\bbQ$-closed subgroup  

1618: which contains $[\bH^\circ, \bH^\circ]$, then

1619: $$ \cA_{\bH | \bF} \, = \, \Inn^\bH_\bF  \cdot \cASid$$ holds.

1620: \end{proposition} 

1621: \begin{proof}  

1622: It is clear that $\Inn^\bH_\bF\cdot \cASid 

1623: \leq  \cA_{\bH | \bF}$.  

1624: Now let $\Phi \in  \cA_{\bH | \bF}$. Since $\bF$ contains 

1625: $[\bH^\circ, \bH^\circ]$, there exists $v \in \bF$ such

1626: that $ v \bC v^{-1} = \Phi(\bC)$. Therefore, $\Inn^\bH_{v^{-1}}\circ \Phi 

1627: \in \cA_\bC \cap  \cA_{\bH | \bF}$ holds. The proposition follows from 

1628: the previous lemma, part ii).  

1629: \end{proof} 

1630: 

1631: 

1632: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1633: 

1634: \section{The algebraic hull of a polycyclic-by-finite

1635: wfn-group}\label{sechull}

1636: 

1637: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1638: 

1639: Let $\Gamma$ be a polycyclic-by-finite group. Its maximal nilpotent

1640: normal subgroup $\Fitt(\Gamma)$ is called the Fitting subgroup of

1641: $\Gamma$. We assume that $\Fitt(\Gamma)$ is 

1642: torsion-free and $\C_\Gamma\left(\Fitt(\Gamma)\right) \leq \Fitt(\Gamma)$.

1643: These two conditions are equivalent

1644: to the requirement that $\Gamma$ has no non-trivial finite normal

1645: subgroups. We call a group with this property a {\it wfn-group}.

1646: Proofs of the following results may be found in \cite[Appendix A]{Baues2}.

1647:  

1648: \begin{theorem} \label{alghull} 

1649: Let $\Gamma$ be a polycyclic-by-finite wfn-group.  

1650: Then there exists a $\bbQ$-defined linear algebraic group $\bH_\Gamma$ and

1651: an injective group homomorphism $\psi: \Gamma \to 

1652: \bH_\Gamma({\bbQ})$ 

1653: such that: \begin{itemize}

1654: 

1655: \item[i)] $\psi(\Gamma)$ is Zariski-dense in $\bH_\Gamma$,

1656:  

1657: \item[ii)] $\bH_\Gamma$ has a strong unipotent radical $\bU=\bU_{\bH_\Gamma}$,

1658: 

1659: \item[iii)] $\dim \bU =  \rank \Gamma$. 

1660: 

1661:  \end{itemize}

1662:  

1663: Moreover, $\psi(\Gamma) \cap \bH_\Gamma({\bbZ})$ is of finite index in 

1664: $\Gamma$. 

1665: \end{theorem} 

1666: Here,  ${\rank}\Gamma$ denotes the number of

1667: infinite cyclic factors in a composition series of $\Gamma$.

1668: (This invariant is sometimes also called the Hirsch-rank of $\Gamma$.)

1669: 

1670: We remark that the group $\bH_\Gamma$ is determined by the conditions

1671: i)-iii) up to $\bbQ$-isomorphism of algebraic groups:

1672:  

1673: \begin{proposition} \label{prop:rigidity1}

1674: Let $\Gamma$ be a polycyclic-by-finite wfn-group. 

1675: Let $\bH'$ be a $\bbQ$-defined 

1676: linear algebraic group and $\psi': \Gamma \longrightarrow \bH'({\bbQ})$ 

1677: an injective homomorphism  which satisfies 

1678: i) to  iii) from above. Then there 

1679: exists a $\bbQ$-defined

1680: isomorphism $\Phi: \bH_\Gamma \rightarrow \bH'$ such that 

1681: $\psi' = \Phi \circ \psi$.   

1682: \end{proposition}

1683: 

1684: \begin{corollary} \label{cor:rigidity2}

1685: Let $\Gamma$ be a polycyclic-by-finite wfn-group. 

1686: The algebraic hull\/ $\bH_{\Gamma}$ 

1687: of\/ $\Gamma$ is unique up to $\bbQ$-isomorphism

1688: of algebraic groups. 

1689: In particular,  every automorphism $\phi$ of\/ $\Gamma$ 

1690: extends uniquely to a

1691: $\bbQ$-defined automorphism 

1692: $\Phi$ of\/ $\bH_{\Gamma}$. 

1693: \end{corollary}

1694: We call the $\bbQ$-defined linear algebraic group $\bH_\Gamma$ 

1695: the {\em algebraic hull for $\Gamma$}. We shall identify $\Gamma$ with the

1696: corresponding subgroup of its algebraic hull $\bH_\Gamma$. 

1697: 

1698: If $\Gamma$ is finitely generated torsion-free nilpotent then

1699: $\bH_\Gamma$ is unipotent and Theorem \ref{alghull} and Proposition

1700: \ref{prop:rigidity1} are essentially

1701: due to Malcev \cite{Malcev}. If $\Gamma$ is torsion-free polycyclic,

1702: Theorem \ref{alghull} is due to Mostow \cite{Mostow1} 

1703: (see also \cite[\S IV, p.74]{Raghunathan} for a different

1704: proof). 

1705: 

1706: \begin{proposition} \label{Fittu} 

1707: Let $\Gamma$ be a polycyclic-by-finite wfn-group.

1708: Let $\bH_\Gamma$ be the algebraic hull for $\Gamma$.

1709: Then $\Gamma\cap \bU_{\bH_\Gamma}={\rm Fitt}(\Gamma)$ holds.

1710: \end{proposition} 

1711: 

1712: \begin{definition} We define $\bF =\bF_\Gamma := 

1713: \ac{\Fitt(\Gamma)} \leq \bH_\Gamma$ as

1714: the Zariski-closure of the Fitting subgroup of $\Gamma$.

1715: \end{definition}

1716: 

1717: Thus, in particular, $\bF$ is a connected unipotent normal

1718: subgroup of $\bH_\Gamma$, and $\bF$ is defined over $\bbQ$.

1719: Moreover: 

1720: 

1721: \begin{proposition} The commutator subgroup 

1722: $[\bH_\Gamma^\circ, \bH_\Gamma^\circ]$ is 

1723: contained in $\bF$. Let $\bC$ be a Cartan subgroup 

1724: of\/  $\bH_\Gamma$.  Then

1725: there is a decomposition 

1726: \begin{equation} \label{eq:fcdecomp}

1727:  \bH_\Gamma \, = \, \bF \cdot \bC \;  

1728: \end{equation}

1729: \end{proposition}

1730: \begin{proof} Since $\Gamma$ is Zariski-dense, we see that 

1731: $[\bH_\Gamma^\circ, \bH_\Gamma^\circ] = 

1732: \ac{[\Gamma \cap \bH_\Gamma^\circ , \Gamma \cap \bH_\Gamma^\circ]} 

1733: \leq \ac{\Fitt(\Gamma)}= \bF$. The decomposition of $\bH_\Gamma$

1734: follows (see SG7 of Section \ref{prereq}).

1735: \end{proof}

1736: 

1737: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1738: 

1739: \subsection{A faithful rational representation of $\Aut(\Gamma)$}

1740: \label{auto111}

1741: 

1742: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1743: 

1744: Let $\Aut_a\!\left(\bH_\Gamma\right)$ 

1745: be the group of algebraic automorphisms of $\bH_\Gamma$ with 

1746: its natural structure of $\bbQ$-defined linear algebraic group. 

1747: We view $\Aut_a(\bH_\Gamma)$ as a $\bbQ$-defined closed subgroup

1748: of $\GL(n,\bbC)$, for some $n \geq 0$. 

1749: Then Proposition \ref{rationalisq} shows that the extension 

1750: \begin{equation} \label{eq:extensionofphi}

1751:  \Aut(\Gamma) \ni \; \; \phi \, \mapsto \, \Phi  \; \; \in \,\Aut_{a,\bbQ}\left(\bH_\Gamma\right) = \Aut_{a}\! \left(\bH_\Gamma\right) \left(\bbQ\right)

1752: \end{equation}

1753: which is  defined in Corollary \ref{cor:rigidity2} gives rise  

1754: to  a faithful representation of $\Aut(\Gamma)$ 

1755: into $\GL(n,\bbQ)$:  

1756: 

1757: \begin{corollary} The extension homomorphism \eqref{eq:extensionofphi} 

1758: is  a faithful homomorphism of $\Aut(\Gamma)$ into the 

1759: group of\/ $\bbQ$-points of the linear algebraic group $\Aut_a(\bH_\Gamma)$.  

1760: \end{corollary} 

1761: 

1762: Using this fact, if an embedding $\Gamma \leq \bH_\Gamma$ is

1763: fixed we identify $\Aut(\Gamma)$ with

1764: a subgroup of $\Aut_a(\bH_\Gamma)$.

1765: Remark though, that, in general, $\Aut(\Gamma)$ is not

1766: Zariski-dense in $\Aut_a(\bH_\Gamma)$ because the

1767: elements of $\Aut(\Gamma)$ preserve the Fitting subgroup

1768: $F$, and hence also the Zariski-closure $\bF$ of $F$. Thus, 

1769: with the conventions from Section \ref{section:subgroups}, 

1770: $\Aut(\Gamma)$ is contained in the subgroup ${\rm N}_\cA(\bF)$

1771: of $\Aut_a(\bH_\Gamma)$.

1772: 

1773: Let $\AGF\le \Aut(\Gamma)$ be the subgroup defined in (\ref{defiagf}).

1774: Since $\Gamma$ is Zariski-dense in $\bH_\Gamma$, the following is clear.

1775: \begin{lemma}\label{agfinAGF}

1776: We have $\AGF\le \cA_{\bH | \bF}(\BQ)$ under the extension 

1777: homomorphism \eqref{eq:extensionofphi}.

1778: \end{lemma}

1779: 

1780: 

1781: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 

1782:  

1783: \subsection{Thickenings of $\Gamma$ in $\bH_\Gamma$}

1784: 

1785: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1786: 

1787: Before introducing the thickening,

1788: we recall some results about finitely generated subgroups in

1789: unipotent algebraic groups.

1790: Let $\bU$ be a unipotent $\bbQ$-defined linear algebraic group, and

1791: let $ F \leq \bU({\bbQ})$ be a finitely generated subgroup. Then 

1792: $F$ is a torsion-free nilpotent group.

1793: Let $\bF:= \ac{F} \leq \bU$ be the Zariski-closure of $F$. 

1794: Then

1795: $\bF$ is $\bbQ$-defined, and $\dim \bF = \rank F$. The group 

1796: of $\bbQ$-points of $\bF$ is isomorphic to the Malcev 

1797: radicable hull of $F$, i.e., $\bF({\bbQ})$ is radicable, 

1798: and for every $x \in \bF({\bbQ})$ there

1799: exists $k \in \bbN$ such that $x^k \in F$. For $m \in \bbN$,

1800: we define 

1801: $$   F^{\frac{1}{m}} \, :=   

1802: \; \langle  x \in \bF({\bbQ}) \mid x^m \in F \rangle \; . $$   

1803: Then $F^{\frac{1}{m}} \leq \bF({\bbQ})$ is finitely generated,

1804: $F \leq F^{\frac{1}{m}}$ and 

1805: $|F^{\frac{1}{m}} : F| < \infty$.    

1806: Every finitely generated subgroup $G \leq \bF({\bbQ})$

1807: is contained in $F^{\frac{1}{m}}$, for some $m \in \bbN$.

1808: See \cite{Segal} for more details on all of this. 

1809: 

1810: Now let $\Gamma$ be a polycyclic-by-finite wfn-group,  

1811: let $F$ denote the Fitting subgroup of $\Gamma$,

1812: and let $\bF \leq \bH_\Gamma$ be 

1813: the Zariski-closure of $F$ in the algebraic hull 

1814: of $\Gamma$. Since $\bF$ is unipotent,  

1815: $F^{\frac{1}{m}}$ is defined as a subgroup of $\bF(\bbQ)$.  

1816: \begin{definition} \label{def:thickening}

1817: A subgroup $\tilde{\Gamma}$ of $\bH_\Gamma$

1818: which is of the form $\tilde{\Gamma} = F^{\frac{1}{m}}\cdot \Gamma$

1819: is called a \emph{thickening} of $\Gamma$. 

1820: \end{definition}

1821: Clearly thickenings $\tilde{\Gamma}$ exist, for every $m \in \bbN$.

1822: A thickening $\tilde{\Gamma}$ of $\Gamma$ 

1823: is a finitely generated subgroup of $\bH_{\Gamma}(\BQ)$ which

1824: is of finite index over $\Gamma$. We further remark 

1825: that $\Fitt(\tilde{\Gamma}) = F^{\frac{1}{m}}$.  

1826: The inclusion of $\tilde{\Gamma}$ into $\bH_{\Gamma}$ 

1827: shows that $\bH_{\Gamma}$ is an algebraic hull also for  

1828: the thickening $\tilde{\Gamma}$. 

1829: 

1830: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1831: 

1832: \subsection{The automorphism group of the thickening}

1833: 

1834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1835: 

1836: Let $\Gamma$ be a polycyclic-by-finite wfn-group and let $\tilde{\Gamma} \leq \bH_\Gamma(\bbQ)$ be a thickening of $\Gamma$.

1837: Let $\phi \in \Aut(\Gamma)$, and let  

1838: $\Phi: \bH_\Gamma  \rightarrow \bH_\Gamma$ 

1839: denote the extension of $\phi$ in $\Aut_a(\bH_\Gamma)$. 

1840: Since the automorphism $\Phi$ preserves  $\bH_\Gamma(\bbQ)$, it is

1841: clear that  $\Phi$ preserves $\tilde{\Gamma} \leq \bH_\Gamma(\bbQ)$

1842: as well. Restricting $\Phi$ to $\tilde{\Gamma}$ we thus obtain a 

1843: natural inclusion $\Aut(\Gamma) \hookrightarrow

1844: \Aut(\tilde{\Gamma})$. This shows that we may identify $\Aut(\Gamma)$ with 

1845: a finite index subgroup of $\Aut(\tilde{\Gamma})$ in a natural way:

1846: 

1847: \begin{proposition} \label{findex}

1848: Let $\Gamma$ be a polycyclic-by-finite wfn-group, and let $\tilde{\Gamma}$ be a

1849: thickening of $\Gamma$.  Then

1850: the group $\Aut(\Gamma) = 

1851: \{ \psi \in \Aut(\tilde{\Gamma}) \mid \psi(\Gamma) = \Gamma \, \} $ is a subgroup of finite

1852: index in $\Aut(\tilde{\Gamma})$.

1853: \end{proposition} 

1854: \begin{proof}

1855: Put $d = [ \tilde{\Gamma} : \Gamma ]$  for the index of $\Gamma$

1856: in $\tilde{\Gamma}$. Remark that there are only finitely many

1857: subgroups of $\tilde{\Gamma}$

1858: with index $d$, since $\tilde{\Gamma}$ is a finitely generated group. 

1859: The automorphism group  $\Aut(\tilde{\Gamma})$ acts on the set 

1860: of such subgroups 

1861: and the group $\Aut(\Gamma)$ 

1862: is the stabilizer of the subgroup $\Gamma$. Hence, we have

1863: $ [\Aut(\tilde{\Gamma}) : \Aut(\Gamma) ]  \, \leq \, \ell $, where $\ell$ 

1864: is the number of subgroups of index $d$.

1865: \end{proof}

1866: 

1867: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1868: 

1869: \section{Thickenings of $\Gamma$ admit a supplement}\label{secthick}

1870: 

1871: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1872: 

1873: We give here a short account of the construction

1874: of nilpotent-by-finite supplements in polycyclic-by-finite groups. 

1875: Similar results are contained 

1876: in the book \cite{Segal} where nilpotent supplements

1877: in polycyclic groups are considered. 

1878: 

1879: \begin{definition}

1880: Let $\Gamma$ be a polycyclic-by-finite group and

1881: let $C \leq \Gamma$ be a nilpotent-by-finite subgroup. 

1882: We call $C$ a {\em nilpotent-by-finite supplement in 

1883: $\Gamma$} if $\Gamma = \Fitt(\Gamma)\cdot C$.

1884: \end{definition}

1885: 

1886: Nilpotent-by-finite supplements do not exist for general 

1887: groups $\Gamma$.

1888: We will show below that a polycyclic-by-finite wfn-group 

1889: $\Gamma$ admits a thickening 

1890: which has a nilpotent-by-finite supplement.

1891: 

1892: As standing assumption for this section we have that $\Gamma$ is a 

1893: polycyclic-by-finite wfn-group, and we put $F = \Fitt(\Gamma)$. 

1894: We fix an inclusion $\Gamma \leq \bH_\Gamma(\bbQ)$ 

1895: of $\Gamma$ into its algebraic hull $\bH_\Gamma$. 

1896: We put $\bF$ for the Zariski-closure of $F$ in

1897: $\bH_\Gamma$, and  put $N= \bF(\bbQ)$.

1898:  

1899:  

1900: \begin{lemma}

1901: Let $\Gamma$ be a polycyclic-by-finite wfn-group and 

1902: let $\bC \leq \bH_\Gamma$ be a $\bbQ$-defined Cartan-subgroup.

1903: Then $\hat{C}= \Gamma N \cap \bC$ is a nilpotent-by-finite

1904: subgroup of $\Gamma \cdot N$ such that  $\Gamma \cdot N = \hat{C}\cdot N$ holds.  

1905: \end{lemma} 

1906: \begin{proof}

1907: The decomposition \eqref{eq:fcdecomp} induces a corresponding

1908: decomposition for the group of $\bbQ$-points of $\bH_\Gamma$,

1909: that is, $ \bH_{\Gamma}(\bbQ) \, = \, N \cdot  \bC(\bbQ) $.

1910: 

1911: Let $\gamma \in \Gamma$. 

1912: Since $\Gamma \leq \bH_{\Gamma}(\bbQ)$ holds, it follows that 

1913: $\gamma = n_\gamma c_\gamma$, where $n_\gamma \in N$,

1914: $c_\gamma \in \bC(\bbQ) \cap \Gamma N$. Hence, 

1915: $\Gamma \cdot N = \hat{C} \cdot N$ holds.   

1916: \end{proof}

1917: 

1918: We prove now the existence of supplements in

1919: a thickening $\tilde{\Gamma} = F^{\frac{1}{m}}\cdot\Gamma$.

1920: 

1921: \begin{proposition}\label{supsplits1}

1922: Let $\Gamma$ be a polycyclic-by-finite wfn-group and

1923: let $\bC \leq \bH_\Gamma$ be a $\bbQ$-defined Cartan subgroup.

1924: Then there exists $m \in \bbN$ such that  

1925: $$ F^{\frac{1}{m}}\cdot\Gamma = F^{1 \over m} \cdot C \, , \, \;

1926: \text{ where } C = (F^{\frac{1}{m}}\cdot \Gamma) \cap \bC \; .$$

1927: \end{proposition} 

1928: \begin{proof} 

1929: By the previous lemma, $\Gamma N = \hat{C} N$, 

1930: where $\hat{C} = \Gamma N \cap \bC$. 

1931: It follows that the natural map 

1932: $\hat{C} \rightarrow \Gamma N/ N$ is surjective.

1933: 

1934: Since $\Gamma N/ N$ is finitely generated, there exists  a

1935: finitely generated group $C \leq  \hat{C}$ so that 

1936: $C \, \rightarrow \, \Gamma N/ N $ 

1937: is surjective. 

1938: Let $c_1, \ldots, c_k$ be generators for $C$, $c_i= \gamma_i n_i$, where

1939: $\gamma_i \in \Gamma$ and $n_i \in N$. Choose $m \in \bbN$ such that 

1940: $n_i \in F^{\frac{1}{m}}$, $i= 1 \ldots k$. Then $C \leq  

1941: \tilde{\Gamma}= \Gamma F^{\frac{1}{m}}$, and, in particular,

1942: $ F^{\frac{1}{m}} C \leq \tilde{\Gamma}$. 

1943: 

1944: The surjectivity of the map $C \rightarrow \Gamma N/ N$ 

1945: shows that every $\gamma \in \Gamma$

1946: is of the form $\gamma= c n$, where $c \in C$, $n \in N \cap \tilde{\Gamma} = F^{\frac{1}{m}}$. 

1947: This shows that 

1948: $\tilde{\Gamma} = F^{\frac{1}{m}}\cdot C$. 

1949: \end{proof}

1950: A nilpotent-by-finite supplement is called \emph{maximal} if it 

1951: is a maximal element of the set of all nilpotent-by-finite supplements

1952: with respect to inclusion of subgroups. We show that the maximal

1953: supplements are those which arise by the construction of Proposition \ref{supsplits1}.   

1954: 

1955: \begin{proposition}\label{supsplits2}

1956: Let $\Gamma$ be a polycyclic-by-finite wfn-group. 

1957: Let $\tilde{\Gamma}= 

1958: F^{\frac{1}{m}}\cdot\Gamma$ be a thickening 

1959: which admits a maximal nilpotent-by-finite supplement $C$.

1960: Then there exists is a $\BQ$-defined 

1961: Cartan subgroup $\bC$

1962: of\/  $\bH_\Gamma$ such that $C=\tilde{\Gamma} \cap \bC$

1963: and  $C$ is Zariski-dense in $\bC$. 

1964: \end{proposition}

1965: \begin{proof}

1966: Let $C$ be any nilpotent-by-finite supplement in $\tilde{\Gamma}$. 

1967: Put $\cC = \ac{C}$ for the Zariski-closure of $C$. Since $C \leq 

1968: \bH_\Gamma(\bbQ)$,

1969: $\cC$ is defined over $\bbQ$. 

1970: Since  $\tilde{\Gamma}= F^{\frac{1}{m}}\cdot C$

1971: is Zariski-dense in $\bH_\Gamma$, we have $\bH_\Gamma = \bF\cdot \cC$. 

1972: Let $\bS$ be a maximal d-subgroup of $\cC$. 

1973: Then $\bS$ is also maximal in $\bH_\Gamma$. 

1974: In particular,

1975: $\cC$ contains a maximal $\bbQ$-defined torus

1976: $\bT$ of $\bH_\Gamma$. Since $\cC$ is nilpotent by-finite,

1977: $\bT$ is unique and normal in $\cC$. We let $\bC$ denote 

1978: the Cartan-subgroup corresponding to $\bT$. Then 

1979: $\cC \leq \bC$. It follows that $C \leq \tilde{\Gamma} \cap \bC$.

1980: In particular, if $C$ is maximal, then $C =  \tilde{\Gamma} \cap \bC$.

1981: 

1982: Let us prove now  that every

1983: maximal nilpotent by-finite supplement $C$ is Zariski-dense

1984: in the Cartan subgroup $\bC$ which contains $\cC$.  

1985: In fact, since $\bH \leq \bF\cdot \cC$,  $\bC = \U_\bC\cdot \bS$, 

1986: where $\bS \leq \bC$ is a maximal d-subgroup of $\cC$, and

1987: $ \U_\bC$ is a unipotent normal subgroup. Hence, $\bS $ is

1988: a maximal d-subgroup of $\bC$ as well. 

1989: Furthermore, we can write $u \in \U_\bC$

1990: as $u = f u_1$, where $u_1 \in \U_\cC$, $f \in \bF \cap \bC$.

1991: By the maximality of $C$, $F \cap \bC \leq C$.  Therefore,

1992: $C \cap \bF = F \cap \bC$ is Zariski-dense in $\bF \cap \bC$.

1993: This means, $\bF \cap \bC \leq \cC$. Hence, $f \in \cC$ and

1994: $u \in \cC$. It follows that $\cC = \bC$.

1995: \end{proof}

1996: 

1997: \begin{proposition}\label{subsplits3}

1998: Let $\Gamma$ be a polycyclic-by-finite wfn-group, and

1999: $\tilde{\Gamma}= 

2000: F^{\frac{1}{m}}\cdot\Gamma$ a thickening of $\Gamma$. 

2001: Then there  are at most finitely many $F^{\frac{1}{m}}$-conjugacy classes of  

2002: maximal nilpotent by-finite supplements in $\tilde{\Gamma}$.

2003: \end{proposition}

2004: \begin{proof}

2005: Let $C$ be a maximal nilpotent-by-finite

2006: supplement in $\tilde{\Gamma}$. By 

2007: Proposition \ref{supsplits2}, $C = \Gamma \cap \bC$, 

2008: where $\bC$ is a Cartan subgroup of $\bH$.  

2009: We consider 

2010: $ \tilde{\Gamma}_0 =\tilde{\Gamma} \cap \bH^\circ$,

2011: $C_0 = C \cap \tilde{\Gamma}_0 $. 

2012: Then $\tilde{\Gamma}_0$ is a polycyclic normal

2013: subgroup of $\tilde{\Gamma}$

2014: and $C_0 \, \ns C$.  Also 

2015: $C_0 = C \cap \tilde{\Gamma}_0 = \bC \cap \tilde{\Gamma}_0 

2016: = \bC^\circ \cap \tilde{\Gamma}_0$. Since $\bC^\circ$ is a Cartan subgroup  

2017: in $\bH^\circ$, $C_0$ is a maximal nilpotent supplement in $\tilde{\Gamma}_0$

2018: (see Proposition \ref{supsplits2}).

2019: Note further that $C =  \tilde{\Gamma} \cap \N_{\Gamma N}(C_0)$ is 

2020: uniquely determined by $C_0$.    

2021: By \cite[Chapter 3, Theorem 4]{Segal}, there are only 

2022: finitely many $F^{\frac{1}{m}}$-conjugacy classes of 

2023: maximal nilpotent supplements

2024: $C_0 \leq \tilde{\Gamma}_0$. This also implies that there are only 

2025: finitely many $F^{\frac{1}{m}}$-conjugacy classes of 

2026: maximal nilpotent by-finite supplements in $\tilde{\Gamma}$. 

2027: \end{proof}

2028: 

2029: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2030: 

2031: \section{Lemmas from group theory}\label{grouplem}

2032: 

2033: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2034: 

2035: We provide here some simple facts which shall be needed later.

2036: 

2037: We start off with a few remarks on group cohomology with

2038: non-abelian coefficients. Let $\mu$ be a group, and let $L$ be a group

2039: on which $\mu$ acts by automorphisms. If $s \in \mu$ 

2040: we write $v \mapsto v^s$, $v \in L$ , for the action of $s$ on $L$. 

2041: 

2042: The set ${\rm Z}^{1}(\mu, L) = \{ z: \mu \to L \mid

2043: z(s_1 s_2) = z(s_1) \, z(s_2)^{s_1} \} $

2044: is called the set of 1-cocycles. 

2045: Two 1-cocycles $z_1$ and $z_2$ are cohomologous if and

2046: only if there exists $v \in L$ such that 

2047: $z_1(s) = v^{-1} z_2(s)  v^s$.   

2048: Let ${\H}^{1}(\mu, F)$

2049: denote the set of equivalence classes of cocycles. It is 

2050: called the first cohomology set for $\mu$ with  

2051: coefficients in $L$. The following lemma is well

2052: known. 

2053: \begin{lemma} \label{lemma:cohomology}

2054: Let $\mu$ be a finite group, and $L$ a 

2055: finitely generated nilpotent

2056: group on which $\mu$ acts by automorphisms. 

2057: Then the cohomology set ${\H}^1(\mu,L)$

2058: is finite.  

2059: \end{lemma} 

2060: \begin{proof} Let $G=L\rfish\mu$ be the split extension corresponding to the

2061:  given action of $\mu$ on $L$. We consider $L$ as a normal subgroup in $G$.

2062: A 1-cocycle $z:\mu\to L$ gives rise to the

2063:  finite subgroup $\mu_z:=\{\, (z(s),s) \mid s\in \mu\,\}$ of $G$. 

2064: Two 1-cocycles $z_1$ and $z_2$ are cohomologous if and

2065: only if the corresponding subgroups $\mu_{z_1}$ and $\mu_{z_2}$ are conjugate

2066: by an element of $L$. Since $G$ is a finitely 

2067: finitely generated nilpotent-by-finite group we know 

2068: (see \cite{Segal}, Chapter 8, Theorem 5) that $G$ has only finitely many

2069: conjugacy classes of finite subgroups. Since $L$ has finite index in $G$, the

2070: lemma follows.  

2071: \end{proof}

2072: 

2073: We also need the following lemma.

2074: \begin{lemma}\label{InnLem} 

2075: Let $N$ be a group and $M \leq N$ a 

2076: finitely generated torsion-free abelian normal subgroup of finite index. 

2077: Define

2078: $$\Aut(N,M) = \{\, \phi \in \Aut(N) \mid \phi(M)=M,\ \phi|_{M} = \id_M \}$$

2079: then the inner automorphisms $\Inn_M^N$ form a subgroup of finite index in $\Aut(N,M)$. 

2080: \end{lemma} 

2081: \begin{proof} We briefly sketch the argument.  

2082: Remark first that is suffient to prove the 

2083: lemma in the case that the extension $M \leq N$

2084: is effective, that is, ${\rm Z}_N(M) \leq M$. We set $\mu = N/M$.     

2085: Assuming effectiveness, there exists a finite extension 

2086: group $N \leq L$ of $N$ which splits, that is, $L$ is

2087: a semi-direct product  

2088: $L = M_L \rtimes \mu$, where 

2089: $M_L\ge M$ is a torsion-free abelian group which contains $M$ as a

2090: subgroup of finite index. 

2091: Every automorphism of $N$ extends uniquely 

2092: to an automorphism of $L$ which preserves $M_L$. 

2093: Therefore, it is enough to show the lemma for $\Aut(L,M_L)$. 

2094: Now let $\phi \in \Aut(L,M_L)$, that is, $\Phi|_{M_L} = \id_{M_L}$,

2095: and assume additionally  that $\phi$ is the identity on the finite

2096: quotient $L/M_L$. The group of all such $\phi$ is isomorphic to 

2097: the group of 1-cocycles in $Z^1(\mu, M_L)$ with the inner 

2098: automorphisms corresponding to 1-coboundaries.  

2099: Since $\H^1(\mu, M_L)$ (see Lemma \ref{lemma:cohomology}) is finite, 

2100: $\Inn_{M_L}^L$ is of finite index in  $\Aut(L,M_L)$.

2101: \end{proof}

2102: The following can be deduced from \cite{Segal}, Section 6, we skip the proof.

2103: \begin{lemma}\label{uniple}

2104: Let $\bU$ be a unipotent $\BQ$-defined linear algebraic group. 

2105: The following hold: \begin{itemize}

2106: 

2107: \item[i)] Let $U_1\le U_2\le \bU(\BQ)$ be two finitely 

2108: generated subgroups and suppose that

2109: $U_1$ is Zariski-dense in $\bU$ then the index $[U_2:U_1]$ is finite.

2110: 

2111: \item[ii)]  Let $U \le \bU(\BQ)$ be a Zariki-dense finitely generated subgroup and let

2112: $d\in\BN$. Let $V \le \bU(\BQ)$ be a subgroup which contains $U$ and satisfies 

2113: $[V:U]\le d$. Then $V$ is contained in $U^{1 \over {d!}}$. In particular, 

2114: the set of all such subgroups $V$ is finite.

2115: \end{itemize}

2116: \end{lemma}

2117: 

2118: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2119: 

2120: \section{Unipotent shadows of $\Gamma$}\label{secshad}

2121: 

2122: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2123: 

2124: Let $\Gamma$ be a polycyclic-by-finite wfn-group. We set $F = \Fitt(\Gamma)$

2125: and write $\bF$ for its Zariski-closure in $\bH_\Gamma$. 

2126: Furthermore, we choose a thickening $\tilde\Gamma=F^{\frac{1}{m}}\cdot \Gamma$

2127: which has a (maximal) nilpotent-by-finite supplement $C\le\tilde\Gamma$.  

2128: We use this setup to 

2129: construct (in a controlled way, depending on $\Gamma$)

2130: a finitely generated nilpotent group $\theta \leq \bU(\bbQ)$

2131: which is Zariski-dense in $\bU$. We shall later use $\theta$ to find

2132: arithmetic subgroups in $\Aut(\Gamma)$.

2133: 

2134: Using the above data we start our construction. We shall use the results of

2135: Section \ref{secthick}.

2136: Let $\bC= \ac{C} \leq \bH_\Gamma$ denote the 

2137: Zariski-closure of $C$. Then $\bC$ is

2138: a $\bbQ$-defined Cartan-subgroup of $\bH_\Gamma$.

2139: We have $C=\tilde\Gamma\cap \bC$, by Proposition \ref{supsplits2}.

2140: Let $\bS \leq \bC$ be a maximal $\bbQ$-defined 

2141: d-subgroup in $\bC$. Then $\bS$ is

2142: a finite extension of the maximal 

2143: torus $\bS^\circ$. The torus $\bS^\circ$ 

2144: is central in $\bC^\circ$, and $\bC= \N_\bH(\bS^\circ)$.

2145: We set $C_0= C \cap \bC^\circ$. Then $C_0$ is a nilpotent

2146: finite index normal subgroup of $C$.

2147: 

2148: We consider the split decompositions  

2149: $\bC= \bU_\bC \cdot \bS$, $\bC^\circ= \bU_\bC \cdot \bS^\circ$,

2150: where $\bU_\bC$ is the

2151: unipotent radical of $\bC$. Every $c \in \bC({\bbQ})$

2152: can be  written uniquely as

2153: \begin{equation}\label{eledec}     

2154: c = u_c \cdot s_c  \; \; \text{ with } u_c \in \bU_\bC(\bbQ) ,  s_c \in \bS(\bbQ) \; .

2155: \end{equation}

2156: If $c \in \bC^\circ({\bbQ})$ holds, then $s_c\in \bS^\circ(\bbQ)$ 

2157: follows. We define

2158: \begin{equation} 

2159: U_{C,\bS} \, =  \, \langle\, u_c \mid c \in C\,\rangle \, ,  \quad 

2160: U_{C_0} \, =  \, \langle\, u_c \mid c \in C_0\, \rangle \; . 

2161: \end{equation}

2162: 

2163: \begin{lemma}\label{unipolem}

2164: The groups $U_{C_0}\le U_{C,\bS}$ are finitely generated,

2165: Zariski-dense subgroups of\/ $\bU_\bC$, and $U_{C_0}$ is of finite index

2166: in $U_{C,\bS}$. 

2167: Moreover, $U_{C,\bS}$ is normalized by $C$, and $U_{C_0}$

2168: is normalized by $C_0$.

2169: \end{lemma}

2170: \begin{proof} 

2171: Since $\bS^\circ \leq \bC^\circ$ is central in $\bC^\circ$, 

2172: the map $C_0 \ni c \, \mapsto u_c \in \bU_\bC$ is a homomorphism.

2173: Therefore, the group $U_{C_0}$

2174: is finitely generated. Moreover, since 

2175: $\bU_\bC = \bU_{\ac{C_0}}$, the group $U_{C_0}$ is 

2176: Zariski-dense in $\bU_\bC$. 

2177: Let $S \leq \bS$ denote the image of the homomorphism 

2178: $ C \rightarrow \bS \, , \; c \mapsto s_c$,  

2179: and $S_0 \leq S$ the corresponding image of

2180: $C_0$. The group $C$ acts on $C_0$ and on $\bU_\bC$ by conjugation.

2181: Since $\bS^\circ$ is central in $\bC^\circ$, this action factors over the finite group 

2182: $\mu= S/S_0$. 

2183: For all $c,d \in C$, we have the formula 

2184: \begin{equation}\label{compati}

2185: u_{c d} \, = \, u_c \, s_c u_d s_c^{-1} \; .

2186: \end{equation}

2187: This shows that the group 

2188: $U_{C,\bS}$ is generated by $U_{C_0}$ and a

2189: finite set $u_{c_1}, \ldots u_{c_l}$, where

2190: the $c_i \in C$

2191: represent generators for $C/C_0$.

2192: Therefore, $U_{C,\bS}$ is finitely generated. 

2193: The statement about the finite index follows from Lemma

2194: \ref{uniple}.  

2195: 

2196: Equation (\ref{compati}) also shows that the action of the finite

2197: group $\mu$ on $\bU_{\bC}$ preserves the subgroup

2198: $U_{C,\bS}$. This implies that $U_{C,\bS}$ is normalized by $C$. The second

2199: statement follows by similar reasoning. 

2200: \end{proof}

2201: 

2202: 

2203: \begin{definition}\label{unipodefi}

2204: We define  

2205: \begin{equation}

2206: \theta_{C,\bS} \, 

2207: := \; \langle\,  F^{\frac{1}{m}},\, U_{C,\bS}\, \rangle \, , \quad 

2208: \theta_{C_0} :=  \langle\,  F^{\frac{1}{m}},\, U_{C_0}\, \rangle \, .

2209: \end{equation} 

2210: The groups $\theta_{C,\bS}$ are called \emph{unipotent shadows} of $\Gamma$. 

2211: \end{definition}

2212: Since we have $\bU=\ac{F}\cdot \bC_\bU$ (see SG7 of Section \ref{prereq}),

2213: Lemma \ref{unipolem} shows that each unipotent shadow

2214: $\theta_{C,\bS}$ 

2215: is a finitely generated subgroup of $\bU(\bbQ)$ which is

2216: Zariski-dense in $\bU$, and it contains the

2217: group $\theta_{C_0}$ as a normal subgroup of finite index. 

2218: 

2219: \begin{definition}\label{godsha}

2220: We call $\theta_{C,\bS}$ a \emph{good unipotent shadow} 

2221: if the conditions

2222: \begin{equation} \label{eq:FTheta}

2223: \theta_{C_0} \cap \bF =\theta_{C,\bS} \cap \bF = 

2224: F^{\frac{1}{m}}=\Fitt(\tilde\Gamma) 

2225: \end{equation}

2226: are satisfied. 

2227: \end{definition}

2228: Good shadows may be obtained by further

2229: thickening of the Fitting-\-subgroup.

2230: 

2231: \begin{proposition}\label{supsplits4}

2232: Let $\Gamma$ be a polycyclic-by-finite wfn-group. Then there is a thickening 

2233: $\tilde\Gamma=F^{\frac{1}{m}}\cdot\Gamma$ with a nilpotent-by-finite

2234: supplement $C\le\tilde\Gamma$, such that, for every maximal $\BQ$-defined d-subgroup 

2235: $\bS\le\ac{C}$, $\theta_{C,\bS}$ is a good unipotent shadow. 

2236: \end{proposition} 

2237: \begin{proof} 

2238: Choose $\ell\in\BN$ such that the thickening

2239: $F^{\frac{1}{\ell}}\cdot\Gamma$ admits a nilpotent-by-finite supplement $C$.

2240: Let  $\bS$ be a maximal $\BQ$-defined d-subgroup in the Zariski-closure

2241: $\bC$ of $C$. Let $\theta_{C,\bS}$ be defined as in Definition \ref{unipodefi}. 

2242: 

2243: Since $\theta_{C,\bS}$ is finitely generated, 

2244: we may choose $m \in \bbN$, divisible by $\ell$, 

2245: such that $\theta_{C,\bS} \cap \bF 

2246: \leq F^{\frac{1}{m}}$.

2247: Now put $\tilde{\Gamma} = F^{\frac{1}{m}}\cdot\Gamma$, and remark that  

2248: $C$ is a nilpotent-by-finite supplement in $\tilde\Gamma$. 

2249: By Proposition \ref{supsplits2},

2250: $C_1= \bC \cap \tilde\Gamma$ is a maximal 

2251: nilpotent-by-finite supplement in $\tilde\Gamma$,

2252: and contains $C$. Since every element $c_1\in C_1$ can be expressed as 

2253: $c_1=fc$ with $c\in C$ and $f\in F^{\frac{1}{m}}$, we find,  going through the

2254: definitions of $\theta_{C,\bS}$ and $\theta_{C_1,\bS}$, 

2255: that $\theta_{C_1,\bS}\cap \bF\le 

2256: F^{\frac{1}{m}}\cdot\big(\theta_{C,\bS}\cap \bF\big)$. 

2257: This implies that $\theta_{C_1,\bS}\cap

2258: \bF=F^{\frac{1}{m}}=\Fitt(\tilde\Gamma)$. 

2259: Hence the requirements of Definition \ref{godsha} follow.

2260: \end{proof}

2261: 

2262: The following compatibility results are very important for our future

2263: constructions.  

2264: \begin{proposition}\label{supsplits5}

2265: Let $\Gamma$ be a polycyclic-by-finite wfn-group.  Let\/ 

2266: $\tilde\Gamma=F^{\frac{1}{m}}\cdot\Gamma$ be a thickening

2267: of\/ $\Gamma$ with a nilpotent-by-finite

2268: supplement $C\le\tilde\Gamma$. Let

2269: $\theta_{C,\bS}$ be a corresponding unipotent shadow.

2270: Then the following hold: \begin{itemize}

2271: 

2272: \item[i)] Let $\phi\in \Aut(\tilde\Gamma)$ be an automorphism which satisfies

2273: $\phi(C)=C$, and let $\Phi$ be its extension to an automorphism of\/

2274: $\bH_\Gamma$. Then we have $\Phi(\theta_{C_0})=\theta_{C_0}$.

2275: 

2276: \item[ii)] For a finite finite index subgroup of the group of all automorphisms $\phi \in \Aut(\tilde\Gamma)$ with 

2277: $\phi(C)=C$, the extension $\Phi$ satsifies

2278: $\Phi(\theta_{C,\bS})=\theta_{C,\bS}$. 

2279: 

2280: \item[iii)] The group $\tilde\Gamma$ normalizes $\theta_{C,\bS}$.

2281: \end{itemize}

2282: \end{proposition}

2283: \begin{proof} Since $\phi(C) = C$, we have

2284: $\Phi(\bC)=\bC$, hence also

2285: $\Phi(\bC^\circ)=\bC^\circ$ and $\Phi(\bS^\circ)=\bS^\circ$. The definition of 

2286: $U_{C_0}$ shows that $\Phi(U_{C_0})=U_{C_0}$. Since

2287: $\Phi$ also stabilizes $\Fitt(\tilde\Gamma)=F^{\frac{1}{m}}$, it stabilizes

2288: $\theta_{C_0}$. This proves i).

2289: 

2290: Since $\theta_{C_0}$ is of finite index in 

2291: $\theta_{C,\bS}$,  we can use i) together with part ii) of Lemma \ref{uniple} to

2292: prove ii).

2293: 

2294: By Lemma \ref{unipolem},  $U_{C,\bS}$ is normalized by $C$. Since $C$ also

2295: normalizes $\Fitt(\tilde\Gamma)=F^{\frac{1}{m}}$ it normalizes

2296: $\theta_{C,\bS}=F^{\frac{1}{m}}\cdot U_{C,\bS}$. The Fitting subgroup

2297: $F^{\frac{1}{m}}$ normalizes $\theta_{C,\bS}$ because it is contained in $\theta_{C,\bS}$.

2298: Hence, $\tilde\Gamma=F^{\frac{1}{m}}\cdot C$ normalizes $\theta_{C,\bS}$.

2299: Hence, iii) holds.

2300: \end{proof}

2301: 

2302: 

2303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2304: 

2305: \section{Arithmetic subgroups of $\Aut(\Gamma)$}\label{secarisag}

2306: 

2307: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2308: 

2309: Let $\Gamma$ be a polycyclic-by-finite wfn-group. As usually,  

2310: the group $\Gamma$ is considered as embedded in the $\bbQ$-points of its 

2311: algebraic hull $\bH_\Gamma$.  This also fixes an embedding of $\Aut(\Gamma)$ 

2312: in the $\bbQ$-points of the $\BQ$-defined linear algebraic group

2313: $\Aut_a(\bH_\Gamma)$.

2314: 

2315: We set $F = \Fitt(\Gamma)$

2316: and write $\bF$ for its Zariski-closure.

2317: We assume for this section that $\Gamma$ admits a nilpotent-by-finite supplement.

2318: Thus we may choose a nilpotent-by-finite subgroup $C$ of $\Gamma$ 

2319: such that $\Gamma=F\cdot C$ holds. We choose $C$ maximal with these

2320: properties. 

2321: We write $\bC=\ac{C}$ for its Zariski-closure. Then $C = \Gamma \cap \bC$. 

2322: We further choose a 

2323: $\BQ$-defined d-subgroup $\bS\le\ac{C}$. Associated with these

2324: data comes a unipotent shadow $\theta=\theta_{C,\bS}$,  as constructed in

2325: the previous section. We make the additional assumption that $\theta$ 

2326: is a  good unipotent shadow (see Definition \ref{godsha}).

2327: Our general philosophy is that we always can replace a general wfn-group 

2328: $\Gamma$ by one of its thickenings to enforce these assumptions.

2329: 

2330: 

2331: 

2332: We define $\bU$ to be the unipotent radical of $\bH_\Gamma$. The unipotent

2333: shadow $\theta\le \bU(\BQ)$ provides us with arithmetic subgroups of suitable 

2334: $\BQ$-closed subgroups of $\Aut_a(\bH_\Gamma)$ (compare 

2335: Section \ref{section:asubgroups}). 

2336: We then will find the

2337: position of $\Aut(\Gamma)\le \Aut_a(\bH_\Gamma)$ relative to them. 

2338: Given a subgroup $B\le \Aut_a(\bH_\Gamma)$ we define

2339: \begin{equation}

2340: B[\theta]:=\{\, \Phi\in B\mid \Phi(\theta)=\theta\,\}

2341: \end{equation}

2342: to be the stabilizer of $\theta$ in $B$. We have:  

2343: \begin{lemma}\label{construction} 

2344: Let $\cB\le \Aut_a(\bH_\Gamma)$ be a $\BQ$-closed subgroup which acts faithfully

2345: on $\bU$.  Then $\cB[\theta]$ is an arithmetic subgroup of $\cB$.

2346: \end{lemma}

2347: The lemma follows along the principles used in Section 

2348: \ref{section:asubgroups}, that is, by

2349: linearizing the action on $\bU$ via the exponential function to a linear

2350: action on the Lie algebra of $\bU$. 

2351: 

2352: As a first application of Lemma \ref{construction}, we obtain that

2353: $\cASid[\theta]$ is arithmetic in $\cASid$, (see

2354: Section \ref{section:subgroups} for the definition of $\cASid$).

2355: We deduce: 

2356: \begin{proposition}\label{construction1}

2357: Given the data $(\Gamma,C,\bS)$ as described above. Then 

2358: $$ \Inn_F^{\bH_\Gamma}\cdot \cASid[\theta]  \, \leq \, 

2359: \cA_{\bH_\Gamma | \bF} (\bbQ)$$   

2360: is an arithmetic subgroup of $\cA_{\bH_\Gamma | \bF}$.

2361: \end{proposition}

2362: \begin{proof}

2363: By our definitions, 

2364: $\cASid[\theta]$ normalizes both $\bF$ and $\theta$, and hence also

2365: $\bF\cap\theta$. We have $\bF\cap\theta=F$ since $\theta$ is a good shadow.

2366: It follows that 

2367: $$F\rfish\cASid[\theta] \, \le \, \bF\rfish\cASid$$

2368: is an arithmetic subgroup. We consider the natural $\BQ$-defined homomorphism

2369: $ \bF \rfish \cASid\to 

2370: \cA_{\bH_\Gamma |\bF}$ which is induced by \eqref{eq:Theta}.  By Proposition

2371: \ref{cAHF},  it is surjective. This implies the result. Of course we have also used that the image

2372: of an arithmetic group under a $\BQ$-defined homomorphism is arithmetic (see 

2373: AR1 of Section \ref{prereq}).

2374: \end{proof}

2375: 

2376: We turn now to the task of comparing $\Aut(\Gamma)$ to the above 

2377: arithmetic groups.

2378: We define, as in the introduction,

2379: $$   A_{\Gamma |F} :=  \{ \phi \in \Aut(\Gamma) 

2380: \mid \phi|_{\, \Gamma/F} = \id_{\Gamma/F} 

2381: \} \; .$$

2382: Clearly, $A_{\Gamma |F}$ is a characteristic subgroup of $\Aut(\Gamma)$.

2383: Set also 

2384: $$ A_{\Gamma |F}^C:=\{ \phi \in  A_{\Gamma |F} 

2385: \mid \phi(C)=C \}  \; .

2386: $$

2387: We obtain from Proposition \ref{subsplits3}:

2388: \begin{lemma}\label{ter1}

2389: Given the data $(\Gamma,C,\bS)$ described above, 

2390: $\Inn_F^\Gamma\cdot A_{\Gamma |F}^C$ has finite index in 

2391: $A_{\Gamma |F}$.

2392: \end{lemma}

2393: 

2394: Next we analyze the group $A_{\Gamma |F}^C[\theta]$. We obtain from 

2395: Proposition \ref{supsplits5} ii):

2396: \begin{lemma}\label{ter2}

2397: Given the data $(\Gamma,C,\bS)$ described above,

2398: $A_{\Gamma |F}^C[\theta]$ has finite index in $A_{\Gamma |F}$.

2399: \end{lemma} 

2400: 

2401: We have the semi-direct product decomposition $\bC=\bU_\bC\cdot \bS$. Relative to

2402: this decomposition we can consider the quotient homomorphism 

2403: $\pi_\bS :\bC\to \bS$,  and define

2404: \begin{equation}\label{esss}

2405: S:=\pi_\bS(C), \qquad S_0:=\pi_\bS(C\cap \bC^\circ)

2406: \end{equation}

2407: Notice that, by the

2408: constructions in Section \ref{secshad}, we have 

2409: $c\cdot \pi_\bS(c)^{-1}\in \theta$,  for every $c\in C$.

2410: We need the following technical observations:

2411: \begin{lemma}\label{ter3}

2412: Given the data $(\Gamma,C,\bS)$, described above and $S,\, S_0$ as

2413: defined in (\ref{esss}), we have \begin{itemize}

2414: 

2415: \item[i)] The group $S$ normalizes $F\cap C$ and $S_0$ centralizes it:

2416: 

2417: \item[ii)] Let $\Phi$ be the extension of the automorphism $\phi\in

2418: A_{\Gamma |F}^C[\theta]$ to an automorphism of\/ $\bH_\Gamma$. Then 

2419: $\Phi(s)\cdot s^{-1}\in F\cap C$ holds, for every $s\in S$. If $s\in S_0$ 

2420: then $\Phi(s)\cdot s^{-1}=1$.

2421: \end{itemize}

2422: \end{lemma} 

2423: \begin{proof}

2424: i): Let $s$ be in $S$, choose $c\in C$ with $\pi_\bS(c)=s$ and define 

2425: $v=c\cdot\pi_\bS(c)^{-1}$. As remarked above, we have $v\in \theta$.

2426: Since $c$ normalizes $F\cap C$, the following holds 

2427: $$s \left(F\cap C\right) s^{-1}=v^{-1}  \left(F\cap C\right) v\, .$$

2428: The right hand  side is in $\theta$ and in $\bF$, hence in $\theta\cap \bF=F$.

2429: This implies that the right hand  side is in $D=F\cap \bC$. The subgroup $D\le

2430: \Gamma$ is nilpotent-by-finite and normalized by $C$. Both $C$ and $D$ are

2431: contained in $\bC$. Hence $\langle \, C,\, D\,\rangle$ is a

2432: nilpotent-by-finite supplement in $\Gamma$. Since $C$ is maximal, we have 

2433: $D=F\cap \bC\le C$ and the first part of i) follows. For the second part,

2434: notice that $F\cap C\le \bC^\circ$.  

2435: 

2436: ii): Let $s$ be in $S$, choose $c\in C$ with $\pi_\bS(c)=s$ and define

2437: $v=c\cdot\pi_\bS(c)^{-1}$. Then

2438: $$\phi(c)c^{-1}=\Phi(v)\Phi(s)s^{-1} v^{-1}.$$  

2439: Since $\Phi$ is the identity modulo $\bF$, the right hand side is in $\bU$,

2440: whereas the left hand side is in $C$. Hence, the right hand side is in $C\cap

2441: \bU\le\theta$. Our assumptions imply $\Phi(s)s^{-1}\in\theta$ and then  

2442: $\Phi(s)s^{-1}\in\theta\cap \bF=F$. Here we have used that $\theta$ is a good

2443: unipotent shadow. Furthermore the above equation shows that 

2444: $\Phi(s)s^{-1}\in\bC$ holds. As under i), we finish the proof 

2445: of the first part of ii) by remarking

2446: that $F\cap \bC$ is contained in $C$. For the second notice that if $s\in

2447: \bS^\circ$ holds then  

2448: $\Phi(s)s^{-1}\in\bF\cap\bS^\circ=\{1\}$ follows.

2449: \end{proof}

2450: We proceed with the construction of subgroups in $\Aut(\Gamma)$. We define

2451: \begin{equation}

2452: A_{\Gamma |F}^C[\theta]^1:=\{ \, \phi\in A_{\Gamma |F}^C[\theta]

2453: \mid \Phi(\bS)=\bS,\ \Phi|_\bS={\rm id}_\bS\,\}.

2454: \end{equation}

2455: Here $\Phi$ is as always the extension of the automorphism

2456: $\phi\in\Aut(\Gamma)$ to an automorphism of $\bH_\Gamma$.

2457: We have:

2458: 

2459: \begin{lemma}\label{ter4}

2460: Given the data $(\Gamma,C,\bS)$, as above. Then

2461: $\Inn_{F\cap C}\cdot A_{\Gamma |F}^C[\theta]^1$ is of finite index in

2462: $A_{\Gamma |F}^C[\theta]$.

2463: \end{lemma}

2464: \begin{proof}

2465: Notice first that $\Inn_{F\cap C}$ is contained in $A_{\Gamma |F}^C[\theta]$, 

2466: as follows from the definitions. 

2467: Let $S,\, S_0\subset  \bS$ be the subgroups

2468: defined in (\ref{esss}). The quotient group $\mu=S/S_0$ is finite and acts by

2469: conjugation on $F\cap C$ (see Lemma \ref{ter3} i)). We let 

2470: ${\rm Z}^{1}(\mu, F \cap C)$ be the corresponding set of 1-cocycles and 

2471: ${\H}^1(\mu, F\cap C)$ the cohomology set (see Section \ref{grouplem} for

2472: definitions). This cohomology set is finite,  by Lemma \ref{lemma:cohomology}.

2473: 

2474: Given $\phi\in A_{\Gamma |F}^C[\theta]$ with extension $\Phi$,  we obtain,

2475: using Lemma \ref{ter3}, a map $D_\phi :S \to F\cap C$ by setting

2476: $D_\phi(s)=\Phi(s)\cdot s^{-1}$.

2477: The verification of the following is straightforward:

2478: \begin{itemize}

2479: \item the above $D$ induces a (well defined) map $D: A_{\Gamma|F}^C[\theta]\to 

2480: {\rm Z}^{1}(\mu, F \cap C)$,

2481: \item the map $D$ from the previous item induces a (well defined) map

2482: $$ \hat D: \; A_{\Gamma|F}^C[\theta] \Big/

2483: \Inn_{F\cap C}^\Gamma \cdot A_{\Gamma|F}^C[\theta]^1  \, 

2484: \longrightarrow \, {\H}^1(\mu, F\cap C) ,$$

2485: \item the map $\hat D$ is injective.

2486: \end{itemize}

2487: As remarked before, ${\H}^1(\mu, F\cap C)$ is finite and the lemma is proved. 

2488: \end{proof}

2489: 

2490: We put now Lemmas \ref{ter1}, \ref{ter2}, \ref{ter4} together and obtain:

2491: \begin{lemma}\label{ter5}

2492: Given the data $(\Gamma,C,\bS)$,  as above. Then

2493: the group

2494: $\Inn_{F}^\Gamma\cdot A_{\Gamma |F}^C[\theta]^1$ is of finite index in   

2495: $A_{\Gamma |F}$.

2496: \end{lemma}

2497: The link between Proposition \ref{construction1} and Lemma \ref{ter5} is given

2498: by:

2499: 

2500: \begin{proposition}\label{terfin}

2501: Given the data $(\Gamma,C,\bS)$,  as above. Then

2502: $A_{\Gamma |F}^C[\theta]^1=\cASid[\theta]$.

2503: \end{proposition}

2504: \begin{proof}

2505: By the definitions, 

2506: $A_{\Gamma |F}^C[\theta]^1\le\cASid[\theta]$ holds.

2507: Now let $\Phi$ be an element of $\cASid[\theta]$.

2508: We show that $\Phi$ is contained in $\Aut(\Gamma)$.

2509: First of all, we have $\Phi(F) = \Phi(\theta \cap \bF)

2510: = \Phi(\theta) \cap \Phi(\bF) = F$. Here we used that $\theta$ is a

2511: good unipotent shadow. 

2512: Since $\Phi|_{\bH^\circ/\bF} = \id_{\bH^\circ/\bF}$, it follows

2513: that 

2514: $ \Phi|_{{\theta/F}} \, = \, \id_{\theta/F}.$ 

2515: Let $c \in C$, $c= u s$ with $u \in \theta$ and $s \in \bS$. 

2516: By the above,  $\Phi(u) = f u$, for some $f \in F$.  Therefore, we get 

2517: $$  \Phi(c) \, =     \Phi(u) \Phi(s) = 

2518: \Phi(u) s = f u s= f c \, \in \Gamma \;  .$$

2519: Since $\Phi(\bC) = \bC$, it follows that $\Phi(c) \in \Gamma \cap \bC = C$. 

2520: This shows that $\Phi$ stabilizes $C$. 

2521: Hence, $ \Phi(\Gamma)= \Phi(F C) = F C =  \Gamma.$ 

2522: Thus,  $\Phi \in \Aut(\Gamma)$ holds. 

2523: The lemma follows.

2524: \end{proof}

2525: Putting together Lemma \ref{ter5}, Proposition \ref{terfin} and 

2526: Lemma \ref{agfinAGF}, we obtain:

2527: \begin{corollary} \label{AGFisa1} 

2528: The group $A_{\, \Gamma | F}$ is an 

2529: arithmetic subgroup of $\cA_{\bH_\Gamma|\bF}$. 

2530: \end{corollary}

2531: 

2532: 

2533: Finally, let us consider the arithmetic subgroup $\A_\theta$ 

2534: (see Definition \ref{LThedef}) of the group $\Aut_a(\bH_\Gamma)$.

2535: The group $A_{\tilde\Gamma |\tilde F}[\theta]$ is defined as 

2536: the stabilizer of $\theta$ in $\AGF$. We note: 

2537: 

2538: \begin{proposition}  \label{propcontat}

2539: A finite index subgroup of\/ 

2540: $\Inn_{\Gamma}\cdot \AGF[\theta]$ 

2541: is contained in $\A_\theta$.   

2542: \end{proposition}

2543: \begin{proof} 

2544: By construction, we have 

2545: $\AGF^C[\theta]^1 \le \A_{\theta}$. The group 

2546: $\Inn_{\Gamma}$ stabilizes $\theta$,  by Proposition \ref{supsplits5} iii).

2547: Obviously  $\Inn_{\tilde F}$ is 

2548: contained in $A_\theta$. We have $\Gamma=\tilde F\cdot C$.  

2549: Let $c$ be an element of $C$,  we write $c=v\cdot s$ with $v\in \theta$, $s\in\bS$. Then 

2550: $\Inn_v \in \A_\theta$,  and hence $\Inn_s$ stabilizes $\theta$. Clearly,

2551: $\Inn_s \in \Aut_a(\bH_\Gamma)_\bS$, therefore 

2552: $\Inn_s$ is also in $\A_\theta$. This shows that

2553: $\Inn_{\Gamma}$ is contained in $\A_\theta$.  

2554: Now $\Inn_{\Gamma}\cdot A_{\Gamma |\tilde F}[\theta]^1$ 

2555: is a finite index subgroup of 

2556: $\Inn_{\Gamma}\cdot \AGF[\theta]$, by Lemma \ref{ter5}. 

2557: \end{proof}

2558: 

2559: 

2560: 

2561: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2562: 

2563: \section{The automorphism group of $\Gamma$ as a subgroup of 

2564: $\Aut_a(\bH_{\Gamma})$}\label{secautoauto}

2565: 

2566: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2567: 

2568: This section contains the final proof of Theorem \ref{teoagf}. We

2569: also provide the input for the proofs of Theorem  \ref{teoc} and 

2570: Theorem \ref{teoa}. (These proofs will

2571: be given in Section \ref{secari}.)

2572: 

2573: Let $\Gamma$ be a polycyclic-by-finite wfn-group. We stick to our usual conventions. 

2574: Namely, $\Gamma$ is embedded in the $\bbQ$-points of its 

2575: algebraic hull $\bH_\Gamma$,  $\Aut(\Gamma) \leq \Aut_a(\bH_\Gamma)(\bbQ)$, 

2576: and  $\bU$ is the unipotent radical of $\bH_\Gamma$.

2577: We also fix,  as a reference, a thickening 

2578: $\tilde\Gamma=F^{\frac{1}{m}}\cdot\Gamma$ of $\Gamma$ in 

2579: $\bH_\Gamma$. We choose a thickening which satisfies the assumptions of Section \ref{secarisag}

2580: on the data $(\tilde{\Gamma},C, \bS)$. In particular, $C \leq \tilde{\Gamma}$ is a

2581: maximal nilpotent-by-finite supplement,  $\bC=\ac{C}$ its Zariski-closure, $\bS \leq \bC$ 

2582: a maximal $\BQ$-defined $d$-subgroup,  and

2583: $\theta = \theta_{\tilde{\Gamma}}$ is a good unipotent shadow for $\tilde{\Gamma}$.  

2584: Such a thickening exists,  by 

2585: Proposition \ref{supsplits4}.

2586: 

2587: \begin{proposition} \label{InnGAGFfi} 

2588: Let $\Gamma$ be 

2589: a polycyclic-by-finite wfn-group.

2590: Then the subgroup $\Inn_\Gamma\cdot A_{\Gamma |F}$ 

2591: has finite index in $\Aut(\Gamma)$.   

2592: \end{proposition}

2593: 

2594: \begin{proof} Recall that $A_{\Gamma |F} = \{ \phi \in \Aut(\Gamma) \mid \phi|_{\Gamma/F} = 

2595: \id_{\Gamma/F} \} $. We shall use that 

2596: \begin{equation}

2597: \AGF \, = \, \Aut(\Gamma) \cap \cA_{\bH_\Gamma|\bF} 

2598:  \, = \, \Aut(\Gamma) \cap \cA_{\bH_\Gamma|\bU}.

2599: \end{equation}

2600: This is  a straightforward consequence of Proposition \ref{Fittu}.

2601: 

2602: Let us put  $N =  \pi_{\bU}(\Gamma)$, and $M= \pi_{\bU}(\Gamma_0)$,

2603: where $\Gamma_0 = \Gamma \cap \bH^\circ$. Define 

2604: $\hat{\bS} = \bH_\Gamma \big/ \bU$ and note that $\hat{\bS}$ is a

2605: $\BQ$-defined d-group.

2606: 

2607: Let $\phi \in \Aut(\Gamma)$ and let $\Phi\in\Aut_a(\bH_\Gamma)$ be its

2608: extension to $\bH_\Gamma$. The $\BQ$-defined automorphism $\Phi$

2609: induces a $\BQ$-isomorphism 

2610: $\Phi_{\hat{\bS}}$ of $\hat{\bS}$, 

2611: which preserves $N$ and $M$. 

2612: The restriction

2613: of $\Phi_{\hat{\bS}}$ to $N$ will be denoted by

2614: $\phi_N$. By the rigidity of tori (AG7), 

2615: for all $\phi$ in a finite index subgroup of $\Aut(\Gamma)$,

2616: $\phi_N$ is the identity on $M$, that is $\phi_N  \in \Aut(N,M)$. 

2617: Thus, by Lemma \ref{InnLem}, $\phi_N \in \Inn^N_M$ holds

2618: in a finite index subgroup of $\Aut(\Gamma)$. 

2619: If $\phi_N  \in \Inn_M^N$,  there exists 

2620: $c \in \Gamma_0$ 

2621: such that $\left(\Inn_c^\Gamma \circ \phi\right)_N = \id_N$.

2622: Since $N$ is Zariski-dense in

2623: $\hat{\bS}$, this implies 

2624: $\Inn_c^\Gamma\circ \phi \in \Aut(\Gamma) \cap \cA_{\bH_\Gamma| \bU} 

2625: = \A_{\Gamma|F}$.

2626: Therefore, $\Inn_{\Gamma_0}^\Gamma \cdot \A_{\Gamma|F}$ is of finite

2627: index in $\Aut(\Gamma)$. 

2628: \end{proof}

2629: 

2630: 

2631: \begin{prf}{Proof of Theorem \ref{teoagf}.}

2632: By the results  proved in Section \ref{auto111},  $\Aut(\Gamma)$ is

2633: contained in the $\BQ$-points of $\Aut_a(\bH_\Gamma)$. 

2634: We come now to the statement about $\AGF$. We use that 

2635: $\Aut(\Gamma)$ is naturally contained in 

2636: $\Aut(\tilde\Gamma)$ as a subgroup of finite index, see Proposition \ref{findex}. 

2637: The 

2638: obvious fact that

2639: $\tilde F=\Fitt(\tilde\Gamma)$ satisfies $\tilde F\cap\Gamma=F$ implies that

2640: $\AGF = A_{\tilde\Gamma |\tilde F} \cap \Aut(\Gamma)$ 

2641: is of finite index in $A_{\tilde\Gamma |\tilde F}$. Hence, by Corollary

2642: \ref{AGFisa1},  $\AGF$ is an arithmetic subgroup of

2643: $\cA_{\bH_\Gamma |\bF}$.

2644: 

2645: Now we prove that $\Aut(\Gamma)$ is,  up to finite index, contained in the

2646: arithmetic group $\A_\theta$  defined relative

2647: to the good unipotent shadow $\theta = \theta_{\tilde{\Gamma}}$. 

2648: (For the construction of $\A_\theta$ refer to  Definition \ref{LThedef}.)  

2649: Since $\Aut(\Gamma)$ is with finite index naturally contained in 

2650: $\Aut(\tilde\Gamma)$ (see Proposition \ref{findex}) it is enough to prove 

2651: that a finite index subgroup of  $\Aut(\tilde\Gamma)$ is contained 

2652: in $\A_\theta$. The latter is implied  by 

2653: Proposition \ref{propcontat} and Proposition \ref{InnGAGFfi} .

2654: \end{prf}

2655: 

2656: \begin{prf}{Proof of Theorem \ref{teoc} for wfn-groups.}

2657: By Proposition \ref{InnGAGFfi},  we have that

2658: $\Inn_\Gamma \cdot  \AGF$ is of finite index in $\Aut(\Gamma)$.

2659: Moreover,  $\tilde\Gamma =\tilde F\cdot C$ contains 

2660: $\Gamma$ as a subgroup of finite index. Hence, 

2661: $ F \cdot (\Gamma \cap C)$ is of finite index in $\Gamma$. 

2662: We have $\Inn_{ F}$ is in 

2663: $\AGF$. Therefore, $\Inn_C^\Gamma \cdot \AGF$ is of finite

2664: index in $\Aut(\Gamma)$.  Now choose a finite index invariant 

2665: nilpotent subgroup $B$ of $\Inn_C^\Gamma$ to obtain

2666: the result. 

2667: \end{prf}

2668: 

2669: The usual induction procedure gives the following 

2670: immediate corollary of Theorem \ref{teoagf}:

2671: \begin{corollary}\label{imBett} 

2672: Let $\Gamma$ be 

2673: a polycyclic-by-finite wfn-group.

2674: Then there exists a faithful 

2675: representation of $\Aut(\Gamma)$ into 

2676: $\GL(n, \bbZ)$, for some $n \in\BN$. 

2677: \end{corollary}

2678: 

2679: 

2680: 

2681: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2682: 

2683: \section{Extensions and quotients of arithmetic groups}\label{subsecari}

2684: 

2685: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2686: 

2687: In the following we shall accumulate some results about extensions and

2688: quotients of arithmetic groups which will be necessary for the proofs in the next

2689: subsection. To formulate these results we need the following 

2690: concepts. 

2691: 

2692: \begin{definition} Let $\cA$ be a $\BQ$-defined 

2693: linear algebraic group and $A\le \cA$ a

2694: subgroup. An automorphism $\phi$ of $A$ is said 

2695: to be \emph{$\cA$-rational} if there is a

2696: $\BQ$-defined automorphism of $\cA$ which normalizes $A$ and coincides with

2697: $\phi$ on $A$. Moreover, a homomorphism $\rho: A \ra \cG(\bbQ)$ 

2698: into a $\bbQ$-defined linear algebraic group $\cG$ is called 

2699: \emph{$\cA$-rational} if $\rho$ extends to a $\bbQ$-homomorphism 

2700: $\rho_\cA: \cA \ra \cG$.   

2701: \end{definition}

2702: 

2703: \begin{definition}

2704: Let $G$ be a group of automorphisms of $A$.

2705: If the action of $G$ on $A$ extends to an algebraic group of automorphisms on $\cA$  (see Definition \ref{algauto}) then $G$ is said to be an  \emph{$\cA$-algebraic group of automorphisms} of $A$. 

2706: \end{definition}

2707: 

2708: As explained in the introduction, a finite extension group of an arithmetic

2709: group need not be arithmetic.  We shall give now a slight generalization of a criterion

2710: from \cite{GP3} which allows to show that certain finite 

2711: extension groups of arithmetic groups are again arithmetic. 

2712: 

2713: \begin{lemma}\label{crit}

2714: Let $\cA$ be a $\BQ$-defined linear algebraic group and $A\le \cA$ a

2715: Zariski-dense

2716: arithmetic subgroup. Let $B\ge A$ be a group containing $A$ as a 

2717: normal subgroup of finite index. Suppose that conjugation of $B$ on

2718: $A$ is  $\cA$-rational. Then the following hold:  

2719: 

2720: \begin{itemize}

2721: \item[i)] The inclusion of $A$ into $\cA$ can be extended to an embedding of

2722: the group $B$ as an arithmetic subgroup into a 

2723: $\BQ$-defined linear algebraic group ${\cal B}$ which contains $\cA$ as a

2724: subgroup of finite index. 

2725: 

2726: \item[ii)] Every automorphism of $B$ which

2727: normalizes $A$ and which induces an $\cA$-rational automorphism of $A$   

2728: is ${\cal B}$-rational.

2729: 

2730: \item[iii)] Let $G$ be a group which acts by automorphisms 

2731: on $B$ which normalize $A$. If $G$ acts

2732: as an $\cA$-algebraic group of automorphisms on $A$

2733: then $G$ is a ${\cal B}$-algebraic group of automorphisms of $B$.  

2734: \item[iv)] Let $\rho:  B \ra \cG(\bbQ)$ be a representation of $B$ which

2735: restricts to an $\cA$-rational representation of $A$. Then $\rho$ is

2736: $\cB$-rational. 

2737: \end{itemize}

2738: \end{lemma}

2739: \begin{proof}  

2740: Statement i) is an application  of \cite[Proposition 2.2]{GP3}. 

2741: The $\BQ$-defined linear algebraic group ${\cal B}$ is

2742: constructed through the usual induction procedure, and 

2743: if $R = \{ r_1,\ldots,r_n \} \subset B$ 

2744: is a complete set of coset representatives for $A$ 

2745: in $B$ then $R$ also forms a complete set of 

2746: coset representatives for $\cA$ in ${\cal B}$. 

2747: There exists thus $r_{ij} \in R$ and $a_{ij} \in A$ such that

2748: $ r_i r_j =  r_{ij} a_{ij}$.

2749: 

2750: We show iv).  Let $\rho:B\to \cG$ be a homomorphism of $B$ whose

2751: restriction to  $A$ is $\cA$-rational. That

2752: is, there is a $\BQ$-defined homomorphism $\rho_\cA :\cA\to \cG$ 

2753: with $\rho_\cA(a)=\rho(a)$

2754: for all $a\in A$. We shall show now that $\rho$ is ${\cal B}$-rational.

2755: Note first that, for all $b\in B$ and all $a\in\cA$, we have

2756: $$ \rho_\cA(b^{-1} a b)=\rho(b)^{-1} \rho_\cA(a)\rho(b) \; , $$

2757: since this identity is valid on the

2758: Zariski-dense subgroup $A$ of $\cA$. 

2759: Define now a map $f:{\cal B}\to {\cal G}$ by 

2760: $$f(r_ia):= \rho(r_i) \rho_\cA(a)\qquad (i=1,\ldots,n,\ a\in\cA).$$

2761: Clearly, $f$ is a $\BQ$-defined morphism of varieties.  A straightforward 

2762: computation using the above mentioned identity shows that $f$ is a

2763: homomorphism of groups. It follows that $f$ is a $\bbQ$-defined 

2764: homomorphism of linear algebraic groups. 

2765: It is clear that $f$ coincides with $\rho$ on $B$.

2766: This proves iv).

2767: 

2768: Note that  ii) is an immediate consequence of  iv). 

2769: 

2770: To prove iii) use ii), and note that the condition of being an algebraic group of automorphisms on $\cB$ depends only on the connected component  

2771: ${\cal B}^\circ = \cA^\circ$. 

2772: \end{proof}

2773: 

2774: The following remark is evident from the definition of an

2775: arithmetic group.

2776: \begin{lemma}\label{trivilem} 

2777: Let $\cB$ be a $\BQ$-defined linear algebraic group and $A\le B\le \cB(\BQ)$

2778: be subgroups. Assume that $A$ is of finite index in $B$ and is an

2779: arithmetic subgroup of its Zariski-closure. Then $B$ is an 

2780: arithmetic subgroup of its Zariski-closure in $\cB$.

2781: \end{lemma}

2782: We apply the lemmas just proved to show Proposition \ref{finker}.  

2783: 

2784: \begin{prf} {Proof of Proposition \ref{finker}.}

2785: Since $A$ is residually finite we can choose a subgroup $C$ of finite index in

2786: $A$ which is normal in $A$ and satisfies $ E\cap C=\{1\}$.

2787: To obtain a normal such $C$, take a finite index subgroup of $A$ with this 

2788: property, and then intersect over (the finite number of) all

2789: subgroups of the same index.

2790: 

2791: Choose a $\BQ$-defined linear algebraic group $\cA$ and a group

2792: homomorphism $\rho:A\to \cA$ with kernel $E$ such that the image is a

2793: Zariski-dense arithmetic subgroup of $\cA$. The homomorphism $\rho$ is injective 

2794: on $C$ and $\rho(C)$ is an arithmetic subgroup of $\cA$.

2795: Clearly, the  conjugations by elements of $A$ induce 

2796: $\cA$-rational automorphisms of $\rho(C)$. We may 

2797: now finish the proof by using Lemma \ref{crit}.

2798: \end{prf}

2799: 

2800: Next we study certain arithmetic quotients of arithmetic

2801: groups.

2802: 

2803: \begin{proposition}\label{propquot}

2804: Let $\cA$ be a $\bbQ$-defined linear algebraic group and 

2805: $A\le \cA$ a Zariski-dense arithmetic subgroup. Let

2806: $N \leq A$ be a normal subgroup of $A$ and let ${\cal N}$ denote the

2807: Zariski-closure of $N$ in $\cA$.  Assume furthermore that $N$ has finite 

2808: index in ${\cal N}\cap A$. Then:

2809: \begin{itemize} \item[i)] 

2810: The group $A/N$ embeds  an 

2811: arithmetic subgroup into a 

2812: $\bbQ$-defined linear algebraic group ${\cal D}$.

2813:  \item[ii)] 

2814: Every $\cA$-rational automorphism of $A$ which normalizes $N$ induces a

2815: ${\cal D}$-rational automorphism of $A/N$. 

2816: \item[iii)]  Every $\cA$-algebraic group of  automorphisms on $A$ which normalizes

2817:  $N$ induces a ${\cal D}$-algebraic group of 

2818: automorphisms of $A/N$.

2819: \item[iv)] Let $\rho$ be an $\cA$-rational representation of $A$ with $N \leq \ker \rho$.

2820: Then the induced quotient representation of $A/N$ is $\cD$-rational. 

2821: \end{itemize}

2822: \end{proposition}

2823: 

2824: For the proof of Proposition \ref{propquot} we need the following

2825: lemma.

2826: \begin{lemma} \label{lemmapropquot}

2827: Under the assumptions of Proposition \ref{propquot} there exists

2828: a finite index subgroup $C \leq A$ such that $\cN \cap C \leq N$. 

2829: \end{lemma}

2830: \begin{proof}

2831: To prove the lemma, we may clearly assume that the

2832: group $\cA$ is connected. 

2833: Let $\bU_\cA$ be the unipotent radical of $\cA$.

2834: The unipotent radical 

2835: $\bU_\cN$ of $\cN$ is contained as a normal subgroup in $\bU_\cA$. 

2836: We may now choose a reductive complement 

2837: $\cA^{{\rm red}}$ for $\bU_\cA$, such that 

2838: ${\cal N}^{\rm red} =\cN \cap  \cA^{{\rm red}}$ is

2839: a reductive complement for $\bU_\cN$ 

2840: in ${\cal N}$. In particular, $ {\cal N}^{\rm red}$

2841: is normal in $\cA^{{\rm red}}$. Thus,  by AG5,

2842: we may choose an almost direct 

2843: complement ${\cal H}$ for ${\cal N}^{\rm red}$ in 

2844: $ \cA^{{\rm red}}$.  That is, ${\cal H}$ is a $\bbQ$-defined subgroup of   

2845: $ \cA^{{\rm red}}$ which centralizes  ${\cal N}^{\rm red}$, satisfies 

2846: $\cA^{{\rm red}}= {\cal N}^{\rm red}\cdot {\cal H}$ and has finite

2847: intersection ${\cal N}^{\rm red}\cap {\cal H}$. 

2848:  

2849: We put $N_{\bU}={ \bU_\cN }\cap N$, and

2850: $N_1={\cal N}^{\rm red}\cap N$. 

2851: Since $N$ is arithmetic in $\cN$,  $N_{{\bU}} \cdot N_1$ has finite

2852: index in $N$. Since $N_{\bU}$ is 

2853: arithmetic in the unipotent group ${\bU_\cN }$, there is a congruence subgroup 

2854: $G$ of $A$ with the property ${\bU_\cN } \cap G \le N_{\bU}$ (see 

2855: \cite{Segal}, Chapter 4, Theorem 5). This congruence subgroup

2856: may be chosen torsion-free as well. We now set 

2857: $$G_{\bU} =\bU_\cA \cap G,\quad G_{1} =

2858: {\cal N}^{\rm red} \cap G,\quad G_\cH ={\cal H} \cap G\; .$$

2859: Since $G$ is an arithmetic subgroup of $\cA $, the product 

2860: $G_{\bU} \cdot  G_{1} \cdot G_\cH$ is of finite index in $G$. 

2861: Both $N_1$ and $G_1$ are arithmetic subgroups of 

2862: ${\cal N}^{\rm red}$.  Hence, $C_1= N_1\cap G_1$ has finite index in  

2863: $N_1$ and $G_1$. Therefore, 

2864:  $$ C=G_\bU\cdot C_1\cdot  G_\cH $$ has finite index in $G$ and $A$.

2865: Since $G$ is torsion-free, $\cN \cap C$ is contained in 

2866: $G_{\bU} \cdot C_1$. Now we find that 

2867: $${\cal N}\cap C  \le  G_{\bU} \cdot C_1

2868: \le \big( {\bU_\cA} \cap C \big) \cdot \left({\cal N}^{\rm red}\cap G \right)

2869: \le  N_{\bU} \cdot N_1 \leq N \; .$$  

2870: This finishes the proof of the lemma. 

2871: \end{proof}

2872: 

2873: 

2874: \begin{prf}{Proof of Proposition \ref{propquot}.}

2875: Replacing the subgroup $C$  constructed in Lemma  \ref{lemmapropquot} 

2876: above by one of its subgroups of

2877: finite index and also possibly by $C\cdot N$, the following can be arranged:

2878: \begin{itemize}

2879: \item $C$ is a normal subgroup of finite index in $A$,

2880: \item $C\cap{\cal N}= N$,

2881: \item $C$ is normalized by every automorphism of $A$ which normalizes $N$.

2882: \end{itemize}

2883: Let ${\cal C}$ denote the Zariski-closure of $C$. This is a $\BQ$-closed subgroup

2884: (of finite index) in $\cA$ which contains ${\cal N}$. Also $C/N$ is contained as a

2885: Zariski-dense subgroup in the 

2886: $\BQ$-defined linear algebraic group ${\cal C}/{\cal N}$. Moreover, $C/N$ is an

2887: arithmetic subgroup of ${\cal C}/{\cal N}$, by AR1. 

2888: 

2889: Now $B= A/N$ is a finite extension group of $C/N$. Clearly, conjugations by

2890: elements of $A$ give rise to ${\cal C}\big/{\cal N}$-rational automorphisms of

2891: $C/N$ and so does every $\cA$-rational 

2892: automorphism of $A$ which normalizes $N$. Therefore,  the group ${\cal D}$ may be 

2893: constructed by application of Lemma \ref{crit}, thus proving i), ii), iii). 

2894: 

2895: To prove iv), let $\rho_\cA: \cA \ra \cG$ denote the algebraic extension of $\rho$, 

2896: and let $\rho_{A/N}:  A/N \ra \cG(\bbQ)$ be the quotient 

2897: representation induced by $\rho$. Clearly, its restriction to $C/N$ is $\cC\big/ \cN$-rational,

2898: since $\rho_\cA$ factors over $\cC\big/ \cN$.  Thus, part iv) of Lemma

2899: \ref{crit} shows that  $\rho_{A/N}$ is  $\cD$-rational.

2900: \end{prf}

2901: 

2902: In the following we deal with the fact that a group which is isomorphic to an

2903: arithmetic group may admit essentially different arithmetic embeddings  

2904:  into linear algebraic groups. This phenomenon plays a role in our 

2905: arithmeticity proofs. At this point we will

2906: also need the full strength of the assumption that $G$ acts as an 

2907: algebraic group of automorphisms on $A \leq \cA$, in order to

2908: extend the action of $G$  to a modification of the ambient group $\cA$. 

2909: 

2910: \begin{proposition}\label{changerep}

2911: Let  $A$ be an arithmetic subgroup of  a $\bbQ$-defined linear algebraic group 

2912: $\cA$ and $G$ an $\cA$-algebraic group of automorphisms of $A$.

2913: Let $D$ be a normal subgroup of $A$ which is contained in the

2914: center of $A$ and which is normalized by $G$. 

2915: Then the group $A$ can be embedded as an arithmetic subgroup

2916: into a $\bbQ$-defined linear algebraic group ${\cE}$ such that:

2917: \begin{itemize}

2918: \item[i)] The subgroup $D$ of $A$ is unipotent-by-finite in $\cE$.

2919: %\item[ii)] The automorphisms of $A$ 

2920: %coming from $G$  are ${\cal D}$-rational. 

2921: \item[ii)] The group $G$ acts as an $\cE$-algebraic group of automorphisms of $A$.

2922: \item[iii)] If $\rho: A \ra \cG(\bbQ)$ is an $\cA$-rational representation 

2923: which satisfies $D \leq \ker \rho$ then $\rho$ is  $\cE$-rational. 

2924: \end{itemize}

2925: 

2926: \end{proposition}

2927: \begin{proof}

2928: Clearly, there is no harm in assuming that $A$ is Zariski-dense in $\cA$.

2929: We choose a subgroup $C\leq A$, subject to the following conditions:

2930: \begin{itemize}

2931: \item  $C$ is torsion-free and a normal subgroup of finite index in $A$,

2932: \item  $C$ is contained in the connected component $\cA^\circ$ of $\cA$, 

2933: \item  $C$ is normalized by  $G$.

2934: \end{itemize}

2935: Note that $C$ is Zariski-dense in $\cA^\circ$.  

2936: We define $G_1 = \Inn_A\cdot G$ 

2937: to be the subgroup of the automorphism group of $A$ 

2938: which is generated by $\Inn_A$ and $G$. Then $G_1$ acts by $\bbQ$-defined 

2939: automorphisms on $\cA$.  Now Lemma \ref{algautoex} and Lemma \ref{algformal3} show 

2940: that $G_1$ is an $\cA$-algebraic group of automorphisms of $A$ (and also of $C$).  

2941: Hence, according to Proposition \ref{decozen},  there exists a  

2942: $G_1$-invariant almost direct product decomposition 

2943: $$\cA^\circ={\cal Z}_1 \cdot \cA_1 \; \; , $$ 

2944: where ${\cal Z}_1$ is the $\BQ$-closed central d-subgroup consisting

2945: precisely of the semisimple elements contained in the center of $\cA^\circ$,

2946: and  $\cA_1$ is a $\BQ$-closed normal subgroup of $\cA$ 

2947: with unipotent by-finite center. 

2948: 

2949: Next we define 

2950: $$Z_1 =C\cap {\cal Z}_1 \; , \; C_1 =C\cap \cA_1 \; , \;  C_2 = Z_1  \cdot  C_1 \; .$$

2951: Observe that $Z_1$ is an arithmetic subgroup of ${\cal Z}_1$ and that 

2952: $C_1$ is arithmetic in $\cA_1$. It follows that $C_2$ is

2953: arithmetic in $\cA$ and of finite index in $C$.  

2954: Since $C$ is torsion-free, we have $Z_1 \cap C_1=\{1\}$,  and 

2955: therefore $C_2$ is isomorphic to the direct product   

2956: $Z_1 \times C_1$. The action of $G_1$ as an $\cA$-algebraic group 

2957: of automorphisms of $C$

2958: stabilizes the factors $Z_1$ and $C_1$, and hence also $C_2$. 

2959: 

2960: Now put $\cD$ for the Zariski-closure of $D$.  Since $D$ is

2961: central in $A$, $\cD \leq \Z(\cA)$. It follows that the maximal

2962: $d$-subgroup $\cS_\cD$ of $\cD$ is contained in $\cZ_1$.

2963: Since $\cS_\cD$ is invariant in $\cD$, there exists, by virtue of AG6,

2964: an almost direct product decomposition $\cZ_1 = \cS_\cD \cdot \cS_2$

2965: which is respected by $G_1$. We define $Z_\cD = Z_1 \cap \cS_\cD $

2966: and $Z_2 = Z_1 \cap \cS_2$. By the arithmeticity of the factors

2967: $Z_1$ and $Z_2$, the product $Z_\cD\cdot  Z_2$ is of finite index 

2968: in $Z_1$. Also this decomposition is preserved by $G_1$. 

2969: 

2970: Now define $C_3= Z_\cD \cdot Z_2 \cdot C_1 \leq \cS_\cD \cdot \cS_2 \cdot \cA_1$.

2971: This group is an arithmetic subgroup, $G_1$-invariant and of finite index in $A$. 

2972: Since it is torsion-free

2973: it is also a direct product of its factors. Let

2974: us put $\cA_2 =  \cS_2 \cdot \cA_1$  and

2975:  $\bar{\cA_2} = \cA_2 \big/  (\cA_2 \cap \cS_\cD)$.  

2976: Then we have an induced arithmetic and Zariski-dense embedding  

2977: $$    C_3 =  Z_\cD \times  \left(Z_2 \times  C_1\right) \, \leq \, \cS_\cD \times \bar{\cA_2}  \; . $$

2978: Since $Z_\cD$ is isomorphic 

2979: to $\BZ^n$, for some $n \geq 0$,  we may embed this group as an

2980: arithmetic subgroup into $\Ga^n$. This gives rise to an arithmetic and

2981: Zariski-dense embedding

2982: $$  C_3 =   Z_\cD \times  \left(Z_2 \times  C_1\right)  \, \leq \, \Ga^n \times \bar{\cA_2}  \;  $$

2983: which has the property that  $D \cap C_3$ is unipotent by-finite in 

2984: $\Ga^n\times \bar{\cA_2}$.

2985:  

2986: Note that $G_1$ induces 

2987: an $\bar{\cA_2}$-algebraic group of automorphisms of $C_1$.

2988: We consider now the arithmetic embedding of $Z_\cD$ 

2989: into the unipotent group $\Ga^n$.  The action of $G_1$ 

2990: on $Z_\cD$ extends to an action by $\bbQ$-defined 

2991: automorphisms of $\Ga^n$. This turns $G_1$ into

2992: a $\Ga^n$-algebraic group of automorphisms of $Z_\cD$.

2993: We infer from Lemma \ref{algformalproducts} that the

2994: product action of $G_1$ on $\Ga^n\times \bar{\cA}_2$ turns

2995: $G_1$ into an algebraic group of automorphisms of $\Ga^n\times \bar{\cA}_2$.

2996: This, in particular, turns $G_1$ into an $\Ga^n\times \bar{\cA}_2$-algebraic group 

2997: of automorphisms of $C_2$. 

2998: 

2999: Now let $\rho$ be an $\cA$-rational representation of $A$ which contains

3000: $D$ in its kernel. In particular, it satisfies $\cD \leq \ker \rho_\cA$, where

3001: $\rho_\cA$ denotes the extension of $\rho$ to $\cA$. The restriction of 

3002: $\rho_\cA$ to $\cA_2$ gives rise to a $\bbQ$-defined homomorphism

3003: $\bar{\rho}: \Ga^n \times \bar{\cA_2} \ra \cG$ which has the subgroup $\Ga^n$ in its kernel. 

3004: We contend that $\bar{\rho}$ extends the representation $\rho$ on $C_3$.

3005: This is easily verified. Thus $\rho: C_3 \ra \cG$ is  $\Ga^n \times \bar{\cA_2}$-rational. 

3006: 

3007: Since $C_3$ is normal and of finite index in $A$, this 

3008: allows, by application of Lemma \ref{crit},  to embed the

3009: group $A$ as an arithmetic subgroup in a finite extension ${\cE}$ of 

3010: $\Ga^n\times \bar{\cA}_2$ such that $G$ acts as a ${\cal E}$-algebraic group

3011: of automorphisms of $A$.  This embedding has the property that 

3012: the finite index subgroup $C_3 \cap D \leq D$ is unipotent. Hence, $D$ is

3013: unipotent-by-finite under the embedding of $A$ into ${\cal E}$. 

3014: This proves i) and ii). The last statement of Lemma \ref{crit}

3015: asserts that $\rho$ is $\cE$-rational, since the restriction of $\rho$

3016: to $C_3$ is $\Ga^n\times \bar{\cA}_2$-rational.

3017:  \end{proof}

3018: 

3019: Part of Proposition \ref{changerep} is reminiscent of Corollary 3.5 from \cite{GP1}

3020: and of Proposition 3.3 from \cite{GP3}, but it is stronger since no passages

3021: to subgroups of finite index are required. Our ultimate arithmeticity result

3022: is contained in the next proposition.

3023: 

3024: \begin{proposition}\label{endprop}

3025: Let $\cA$ be a  $\bbQ$-defined linear algebraic group and let 

3026: $A\le \cA(\bbQ)$ be a Zariski-dense subgroup. 

3027: Assume $N$,$B$,$C$ are normal subgroups of $A$ such that

3028: the following hold: \begin{itemize} 

3029: 

3030: \item[i)] $N\cdot B$ has finite index in $A$,

3031: 

3032: \item[ii)] $B$ is an arithmetic subgroup in its Zariski-closure ${\cal B}$, 

3033: 

3034: \item[iii)] $C$ is an arithmetic subgroup in its Zariski-closure ${\cal C}$,

3035: 

3036: \item[iv)] $C\le N\cap B$ and $D=  (N\cap B)\big/C$ is in the center of $B/C$.

3037: \end{itemize}

3038: Then $A/N$ is an arithmetic group.

3039: 

3040: Moreover, there exists an arithmetic embedding of  $A/N$ into a $\bbQ$-defined 

3041: linear algebraic group $\cA_N$ which has the following property: For any

3042: $\cA$-rational representation of $A$ with $N \leq \ker \rho$, the induced

3043: quotient representation of $A/N$ is  an  $\cA_N$-rational representation. 

3044: \end{proposition}

3045: \begin{proof} The group $A$ induces by conjugation an ${\cB}$-algebraic

3046: group of automorphisms of $B$ which we call $G$. 

3047: Note that ${\cal B}$ and ${\cal C}$ are normal in $\cA$, and preserved 

3048: by $G$ as well.  Since $B$ and $C$ are 

3049: arithmetic subgroups of their respective Zariski-closures, we find

3050: that $C$ has finite index in ${\cal C}\cap B$. We may hence use Proposition

3051: \ref{propquot} to embed the group $B/C$ as an arithmetic subgroup into a 

3052: $\bbQ$-defined linear algebraic group ${\cal D}$. This embedding has the

3053: property that the group $G$ of automorphisms of $B/C$ is ${\cal D}$-algebraic.

3054: 

3055: We consider now the subgroup $D= (N\cap B)\big/C$ in $B/C$. 

3056: By our assumption iv),  $D$ is central in $B/C$. Since $G$ is ${\cal D}$-algebraic,

3057: we may, by Proposition \ref{changerep}, change the arithmetic 

3058: embedding of $B/C$ in $\cD$ to an arithmetic embedding of $B/C$ into  

3059: a $\bbQ$-defined linear algebraic group ${\cal E}$ such that 

3060: $D$ is unipotent-by-finite in ${\cal E}$. Moreover, 

3061: the group $G$ acts as an ${\cal E}$-algebraic group of

3062: automorphisms. 

3063: 

3064: Let us  consider now the Zariski-closure ${\cal D}_1$ of $D$ in ${\cal E}$.

3065: Since $D$ has a unipotent finite index subgroup, 

3066: $D$ is an arithmetic subgroup of the 

3067: $\bbQ$-defined algebraic group ${\cal D}_1$. (For a proof consult 

3068: \cite[Chapter 8]{Segal}).

3069: Since $D$ is an arithmetic subgroup it has finite index in ${\cal D}_1 \cap B\big/C$.

3070: Therefore, we may apply Proposition

3071: \ref{propquot} to embed the quotient $(B/C)\big/D$ as an arithmetic

3072: subgroup into a $\bbQ$-defined linear algebraic group ${\cal B}_1$ such that $G$ acts by

3073: ${\cal B}_1$-rational automorphisms on $(B/C)\big/D$.

3074: 

3075: By assumption i), $A/N$ is isomorphic to a finite extension group of 

3076: $$(B/C)\big/D\cong B\big/B\cap N\cong \left(N\cdot B\right) \big/N.$$ 

3077: The elements of $A/N$ act on $(B/C)\big/D$ as ${\cal B}_1$-rational automorphisms

3078: since the elements of $G$ have this property.   

3079: Finally,  we apply Lemma

3080: \ref{crit} to find that $A/N$ is arithmetic 

3081: in a $\bbQ$-defined linear algebraic group .  

3082: 

3083: To prove that the restriction of $\rho$ to $A/N$ is 

3084: $\cA_N$-rational, we have to carry over the rationality of $\rho$ in

3085: each of the construction steps above. The details are easily verified. 

3086: \end{proof}

3087: 

3088: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3089: 

3090: \section{The arithmeticity of $\Out(\Gamma)$} \label{secari}

3091: 

3092: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3093: 

3094: 

3095: This section contains the complete proof of Theorem

3096: \ref{teoa}  which proceeds in two steps. These steps are carried out

3097: in Section \ref{asec1} and in Section \ref{generalpofi}. On our way,

3098: we provide (respectively, finish) the proof of  Theorem \ref{ftop} in 

3099: Section  \ref{asec1}, as well as the proofs of

3100: Theorem \ref{teoc} and of Theorem \ref{teosep} 

3101: in Section \ref{generalpofi}.

3102: 

3103: Throughout this section, $\Gamma$ denotes a polycyclic-by-finite

3104: group. We also stick to the notation introduced in Sections 2 to

3105: 6.  In particular,  $F\le \Gamma$ denotes the Fitting subgroup of $\Gamma$. 

3106: If in addition $\Gamma$ is a wfn-group,  $\bH_\Gamma$ denotes the

3107: algebraic hull of $\Gamma$,

3108: and $\bF$ the Zariski-closure of $F$. 

3109: 

3110: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3111: 

3112: \subsection{The case of polycyclic-by-finite wfn-groups} \label{asec1}

3113: 

3114: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3115: 

3116: The purpose

3117: of this subsection is to prove Theorem \ref{ftop} of the introduction. 

3118: Let us therefore assume here that $\Gamma$ is a wfn-group. 

3119: The arithmeticity of  $\Out(\Gamma)$, 

3120: in the case that $\Gamma$ is a wfn-group,  

3121: is an immediate consequence of the

3122: structural properties of the embedding $\Aut(\Gamma) \leq \Aut_a({\bf H_\Gamma})(\bbQ)$ 

3123: together with Proposition \ref{endprop}. To see this let us put now 

3124: \begin{equation}\label{festsetz}

3125: A=\Aut(\Gamma) \, ,\quad N=\Inn_\Gamma\, ,\quad B=\AGF \, , \quad C=\Inn_F \, .

3126: \end{equation}

3127: Then we have: 

3128: 

3129: \begin{proposition}  Let $\Gamma$ be a wfn-group. 

3130: Then the subgroups 

3131: $N,B,C\le A$ defined in \eqref{festsetz} 

3132: satisfy the hypotheses of Proposition \ref{endprop}

3133: with respect to the Zariski-closure $\cA$

3134: of $\Aut(\Gamma)$ in $\Aut_a({\bf H_\Gamma})$.

3135: \end{proposition} 

3136: 

3137: \begin{proof}

3138: Condition i) requires that  $N\cdot B = \Inn_\Gamma \cdot \AGF$

3139: has finite index in $A = \Aut(\Gamma)$. This is contained in  

3140: Proposition \ref{InnGAGFfi}.  

3141: Theorem \ref{teoagf} says that $B=\AGF$ 

3142: is arithmetic in its Zariski-closure ${\cal B}$ in $\cA$.

3143: This implies condition ii). 

3144: 

3145: The construction of the algebraic structure on $\Aut_a({\bf H_\Gamma})$ 

3146: (see Subsection \ref{subsect:algebraicstructure}) shows that 

3147: the group $\Inn_\bF$ is a Zariski-closed subgroup 

3148: of the unipotent radical of 

3149: $\Aut_a({\bf H_\Gamma})$. Moreover, $\Inn_\bF$

3150: contains the finitely generated group $\Inn_F$ as a Zariski-dense 

3151: subgroup of rational points. 

3152: In particular,  $C=\Inn_F$ is arithmetic in its Zariski-closure.

3153: This implies condition iii).

3154:  

3155: We shall finally verify the conditions iv) of Proposition \ref{endprop}.

3156: Clearly we have $\Inn_F \leq \Inn_\Gamma \cap \AGF$, that is, 

3157: $C\le N\cap B$.  

3158: Now let $\Phi \in \AGF$ and $\gamma\in\Gamma$ with $\Inn_\gamma \in \AGF$. Then

3159: $\Phi(\gamma) = \gamma f$, where $f \in F$. It follows that 

3160: $$\Phi\circ \Inn_\gamma\circ \Phi^{-1} = 

3161: \Inn_{\Phi(\gamma)}= \Inn_\gamma \, \Inn_f.$$

3162: This shows that $\Inn_\Gamma \cap \AGF = N\cap B$ projects onto

3163: a central subgroup of $B/C=\AGF /\, \Inn_F^\Gamma$. Hence, iv) holds.

3164: \end{proof}

3165: 

3166: \begin{prf} {\it Proof of Theorem \ref{ftop}.}

3167: We may now apply Proposition \ref{endprop} which asserts that 

3168: there exists a $\BQ$-defined linear algebraic group

3169: \begin{equation*} %\label{impodefi}

3170: {\cal O}_\Gamma =\cA_{N}=\cA_{\Inn_\Gamma}  

3171: \end{equation*}

3172: which contains an isomorphic copy of the group 

3173: $\Out(\Gamma)=A/N$ as an arithmetic subgroup. 

3174: This already establishes the arithmeticity of $\Out(\Gamma)$.

3175: 

3176: Consider next the algebraic outer automorphism group  

3177: $$ \Out_a(\bH_\Gamma)=    \Aut_a(\bH_\Gamma)   \big/\, \Inn_{\bH_\Gamma} \, , $$

3178: and let $\pi_\Gamma:  \Out(\Gamma)\ra \Out_a(\bH_\Gamma)$

3179: be the homomorphism induced on $\Out(\Gamma)$.

3180: Since the natural map 

3181: $ \Aut_a(\bH_\Gamma) \ra   

3182: \Out_a(\bH_\Gamma)$

3183: is a $\BQ$-defined homomorphism, it

3184: induces a $\bbQ$-defined homomorphism $\cA \ra \Out_a(\bH_\Gamma)$.

3185: Since $\pi_\Gamma$ contains $N = \Inn_\Gamma$ in its kernel,  

3186: Proposition \ref{endprop} asserts that the

3187: homomorphism $\pi_\Gamma :\Out(\Gamma)\to \Out_a(\bH_\Gamma)$ 

3188: can be extended

3189: to a $\BQ$-defined homomorphism 

3190: $$ \pi_{\cO_\Gamma} : \, {\cal O}_\Gamma\to \Out_a(\bH_\Gamma) \; .$$

3191: This proves the first part of Theorem \ref{ftop}.

3192: 

3193: To show the statements about the kernel of $\pi_{\Gamma}$, we define:

3194: \begin{equation*}

3195: K=\Inn_\bH\cap\Aut(\Gamma),\qquad K_F=\Inn_\bH\cap \AGF,\qquad

3196: E_F = \Inn_{\bf F}^{\bH_\Gamma}   \cap \AGF   \; . 

3197: \end{equation*}

3198: 

3199: \begin{lemma}\label{kerra}

3200: With the above notation the following hold: \begin{itemize}

3201: 

3202: \item[i)] $\Inn^\Gamma_F$ has finite index in $E_F$.

3203: 

3204: \item[ii)] There is a finite index normal subgroup $T\le \AGF$ 

3205: such that $T\cap E_F\le \Inn^\Gamma_F$.

3206: 

3207: \item[iii)]  The commutator group $[\AGF, K_F]$ is contained in $E_F$. 

3208: \end{itemize}

3209: \end{lemma}

3210: \begin{proof} Note that

3211: $\Inn_F^\bH$ is Zariski-dense and arithmetic 

3212: in the unipotent group $\Inn_{\bf F}^\bH$.

3213: Since $\AGF$ is an arithmetic subgroup of

3214: its Zariski-closure in $\cA$, $E_F$ is arithmetic

3215: in $\Inn_{\bf F}^\bH$ as well. This implies i). 

3216: 

3217: Now ii) follows from i) together with the congruence 

3218: subgroup property for $\Inn^\Gamma_F$.

3219: (Compare the proof of Lemma \ref{lemmapropquot}.)

3220: 

3221: For iii),  let $\Phi$ be the extension of 

3222: $\phi\in \AGF$ to an automorphism of $\bH_\Gamma$.  

3223: Let $h\in\bH$ such that $\Inn_h=\psi \in K_F$. 

3224: Since $\Phi \in  \cA_{\H_\Gamma| \bF}$

3225: $$\Phi\circ \Inn_h\circ\Phi^{-1}=\Inn_{\Phi(h)}=\Inn_h\circ \Inn_{f_h}.$$ 

3226: This in turn gives $\psi^{-1}\circ\phi\circ\psi\circ \phi^{-1}\in

3227: \Inn_{\bf F}^{\bH_\Gamma}  \cap \AGF = E_F$, proving iii).  

3228: \end{proof}

3229: 

3230: The kernel of $\pi_\Gamma$ is the image of $K= \Inn_{\bH_\Gamma}  \cap \Aut(\Gamma)$

3231: in $\Out(\Gamma)$. Let $\bar K_F$ and $\bar E_F$ be the images of $K_F$, $E_F$ in 

3232: $\AGF\big/\, \Inn_F^\Gamma$. Let $\bar T$ be the corresponding image of the finite index 

3233: subgroup $T\le\AGF$ as in Lemma \ref{kerra} ii).  

3234: Lemma \ref{kerra} i) shows that $\bar E_F$ is finite. By ii) and  iii) of the same 

3235: Lemma, $\bar K_F$ is centralized by the finite index subgroup $\bar{T}

3236: \leq \AGF\big/\, \Inn_F^\Gamma$. In particular, 

3237: $\bar K_F$ is abelian-by-finite. Consider now the commutative diagram

3238: $$

3239: \xymatrix{

3240: \AGF\big/\, \Inn_F^\Gamma \ar[r]^{ } \ar[d]_{ } 

3241: & \Out(\Gamma) = \Aut(\Gamma)\big/\, \Inn_\Gamma \ar[ld]^{ } \\

3242: \Out_a({\bf H}_\Gamma) & 

3243: }

3244: $$

3245: of natural homomorphisms.  By AR3,

3246: every abelian subgroup of an arithmetic group is finitely generated. 

3247: Hence, the image of $\bar K_F$ in $\Out(\Gamma)$ is so. 

3248: We may also infer that $\bar K_F$ is finitely generated.

3249: 

3250: Since $\Inn_\Gamma \cdot \AGF$ has finite index in $\Aut(\Gamma)$, 

3251: the normal subgroup $\bar K_F$ maps onto a finite index subgroup of the image of $K$ in 

3252: $\Aut(\Gamma)\big/\, \Inn_\Gamma$. This proves that $\ker \pi_{\Gamma}$

3253: is finitely generated, abelian-by-finite and centralized by a finite index 

3254: subgroup of $\Out(\Gamma)$.  If $\Gamma$ is nilpotent-by-finite then $E_F$ is of finite

3255: index in  $K_F$, and hence $\ker \pi_{\Gamma}$ is finite. 

3256: This finishes the proof of Theorem \ref{ftop}. 

3257: \end{prf}

3258: 

3259: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3260: 

3261: \subsection{The case of a general polycyclic-by-finite group}

3262: \label{generalpofi}

3263: 

3264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3265: 

3266: In this subsection, we explain the transfer of our arithmeticity results from 

3267: the case of wfn-groups to general polycyclic-by-finite groups. Thereby, 

3268: we provide the final step in the proofs of Theorem \ref{teoa} and Theorem  \ref{teoc}. 

3269: We also prove Proposition \ref{equivi} and Theorem \ref{teosep}.  

3270: 

3271: Let $\Gamma$ be a polycyclic-by-finite group. Let $\tau_\Gamma$ denote the maximal finite normal 

3272: subgroup of $\Gamma$.  Note that $\tau_\Gamma$ is

3273: characteristic in $\Gamma$ that is,

3274: it is normalized by every automorphism of $\Gamma$.

3275: The quotient group $$ \tGam := \, \Gamma\big/ \tau_\Gamma$$

3276: is a wfn-group. We let ${\mathsf j}: \Gamma \ra \tGam$ denote the

3277: quotient homomorphism. By Theorem \ref{ftop}, the group  

3278: $\Out(\tGam)$ is an arithmetic group.  We 

3279: shall show that $\Out(\Gamma)$ has the same property. 

3280: 

3281: Let $\Gamma_{0} \leq \Gamma$ be a characteristic

3282: finite index subgroup with 

3283: $\Gamma_{0} \cap \tau_\Gamma = \{ 1\}$. We may suppose that 

3284: the image $\tGam_0 \leq \tGam$ of $\Gamma_0$ in

3285: $\tGam$ is also characteristic. (To obtain such a subgroup,  let $n\in\BN$ be the

3286: index of a torsion-free subgroup of finite index in $\Gamma$. Take $\Gamma_0$ to

3287: be the subgroup generated by all $\gamma^{n!}$, $\gamma\in\Gamma$). 

3288:  

3289: Let us put  $$ \mu: = \, \Gamma/ \Gamma_{0} \; . $$

3290: The quotient homomorphism ${\mathsf j}$ and the projection to $\mu$

3291: induce injective homomorphisms  

3292: $$  {\mathsf j}_\mu: \, \Gamma \rightarrow \tGam \times \mu  \, , \, \; \text{  and } \; \;

3293:  {\mathsf k}_\mu: \, \Aut(\Gamma)  \rightarrow \Aut(\tGam) \times\Aut( \mu) \; . $$

3294:  

3295: We define finite index subgroups of $\Aut(\Gamma)$ and $\Aut(\tGam)$,

3296: respectively:

3297: \begin{equation*}

3298: A_{0}: =  \;  \{\, \phi \in \Aut(\Gamma) \mid 

3299: \phi_{\Gamma/ \Gamma_{0}} = \id_{\Gamma/ \Gamma_{0}}\,  \} \;  \leq \Aut(\Gamma) \; , 

3300: \end{equation*}

3301: \begin{equation*}

3302: \tilde{A}_{0}: = \; \{\, \phi \in \Aut(\tGam) \mid 

3303: \phi_{\tGam/ \tGam_{0}} = \id_{\tGam/ \tGam_{0}}\, \} \; \leq \Aut(\tGam) \; . 

3304: \end{equation*}

3305:  

3306: 

3307: 

3308: \begin{lemma} 

3309: \label{lemma:findex} 

3310: With the above notation the following hold: 

3311: \begin{itemize} 

3312: \item[i)]  The group ${\mathsf j}_\mu(\Gamma)$ is of finite index in $\tGam \times \mu$.

3313: \item[ii)]  Let $F = \Fitt(\Gamma)$. Then ${\mathsf j}(F)$  is of finite index in $\Fitt(\tGam)$.

3314: \item[iii)]   The induced homomorphism ${\mathsf k}: \Aut(\Gamma) \ra \Aut(\tGam)$ maps the

3315: group $A_0$ isomorphically onto $\tilde{A}_{0}$.  In particular, ${\mathsf k}(\Aut(\Gamma))$

3316: is of finite index in $\Aut(\tGam)$.

3317: \item[iv)]   The subgroup $\AGF \leq \Aut(\Gamma)$ is mapped by ${\mathsf k}$ onto

3318: a finite index subgroup of $\tAGF$.  

3319: \end{itemize} 

3320: \end{lemma}

3321: \begin{proof} Part i) is clear. For ii), note first that ${\mathsf j}(F)$

3322: is a nilpotent ideal in $\tGam$, and hence ${\mathsf j}(F) \leq \Fitt(\tGam)$. 

3323: Then $F_0 = {\mathsf j}^{-1}(\Fitt(\tGam) \cap \Gamma_0)$ is a nilpotent 

3324: normal subgroup of $\Gamma$ which is of finite

3325: index in the preimage ${\mathsf j}^{-1}(\Fitt(\tGam))$. Moreover, 

3326: $F_0 \leq F$. Hence $F$ is of finite index in this preimage.

3327: This implies ii). 

3328: 

3329: To prove iii), we show that 

3330: ${\mathsf k}_\mu(A_{0}) =  \tilde{A}_{0} \times \{ 1 \} $.

3331: Clearly,  ${\mathsf k}_\mu(A_{0})$ is contained in $\tilde{A}_{0} \times \{ 1 \}$. 

3332: Let $\psi \in  \tilde{A}_{0} $. We show that there exists

3333: $\phi \in A_{0}$ such that  $\psi =\mathsf{k}({\phi})$ is induced by $\phi$.  

3334: If $\gamma \in \Gamma$, we let $\tilde{\gamma} = \mathsf{j}(\gamma)$

3335: denote its projection into $\tilde{\Gamma}$.

3336: Since the projection ${\mathsf j}$ maps $\Gamma_0$ isomorphically

3337: onto the invariant subgroup $\tGam_0$, the automorphism 

3338: $\psi$ defines $\psi_0 \in \Aut(\Gamma_0)$ uniquely with the property that 

3339: \[   \mathsf{j} (\phi_0( \gamma) )   =   \psi( \tilde{\gamma})   \;  \; \;  (\gamma \in\Gamma_0 ) \; . \]

3340: Let 

3341: $\Gamma\big / \Gamma_0 = \bigcup_i \gamma_i \Gamma_0$ be the

3342: coset decomposition.

3343: Now we can write 

3344: $\psi(\tilde{\gamma}_i) =  \tilde{\gamma}_i  \epsilon_i$, $\epsilon_i \in \tGam_0$.

3345: There exist unique $ \delta_i \in \Gamma_0$ such that  $\epsilon_i = \tilde{\delta}_i$. 

3346: We declare now 

3347: $$  \phi(\gamma_i s) = \tilde{\gamma}_i \delta_i \phi_0(s)    \;  \; \;  (s \in\Gamma_0 )  \; . $$

3348: It is easy to verify that  this actually defines 

3349: an automorphism $\phi \in \Aut(\Gamma)$.  This

3350: $\phi$ is clearly a lift of  $\psi$.

3351: 

3352: For iv) consider the quotient homomorphism $\Gamma/ F \ra \tGam/\tF$. 

3353: By ii), this homomorphism has finite kernel. Since it is surjective, the group

3354: $\AGF$ is mapped into $\tAGF$ by ${\mathsf k}$. Let $A_1$ be the

3355: preimage of $\tAGF$ in $\Aut(\Gamma)$. Now applying the

3356: reasoning of iii) to the above quotient homomorphism with finite kernel, 

3357: we deduce that finite index subgroup of $A_1$ acts as the identity on $\tGam/\tF$.

3358: That is, $\AGF$ has finite index in $A_1$. This shows iv). 

3359: \end{proof}

3360: 

3361: Part iii) of the above lemma immediately implies Proposition \ref{equivi}:

3362:  

3363: \begin{prf} {\it Proof of Proposition \ref{equivi}:}

3364: The groups $\Aut(\Gamma)$ and $\Aut(\tGam)$ are 

3365: commensurable, since they have isomorphic finite index

3366: subgroups $A_0$ and $\tilde{A}_{0}$.  

3367: If  $\Aut(\tGam)$ is arithmetic, then the product  

3368: $\Aut(\tGam) \times\Aut( \mu)$ is arithmetic. 

3369: Since $\Aut(\Gamma)$ embeds as a subgroup of 

3370: finite index in the latter product,  

3371: $\Aut(\Gamma)$  is an arithmetic group as well.  

3372: Conversely, if $\Aut(\Gamma)$ is arithmetic, 

3373: the finite index subgroup  $A_0 \leq \Aut(\Gamma)$ is arithmetic too. 

3374: Therefore, the subgroup $\tilde{A}_0 \leq \Aut(\tGam)$ is arithmetic.

3375: \end{prf}

3376: 

3377: A related result is: 

3378: \begin{proposition} The subgroup $\AGF$ of $\Aut(\Gamma)$ is arithmetic. \label{AGFisa}

3379: \end{proposition} 

3380: \begin{proof} By iv) of  Lemma \ref{lemma:findex}, the injection ${\mathsf j}_\mu$ 

3381: maps $\AGF$ onto a finite index subgroup of $\tAGF \times \Aut(\mu)$. 

3382: Since $\tAGF$ is arithmetic, by Corallary \ref{AGFisa1}, we can infer that 

3383: $\AGF$ is arithmetic.  

3384: \end{proof}

3385: 

3386: \begin{prf} {\it Proof of Theorem \ref{teoc} in the general case.}

3387: As remarked above, the projection $\mathsf k$ maps $\AGF$ onto a finite

3388: index subgroup of $\tAGF$.  Let $B \leq \Gamma \cap \Gamma_0$ be a 

3389: nilpotent subgroup such that $\tAGF \cdot {\mathsf j}(B)$ is of finite

3390: index in $\Aut(\tGam)$ (see Section \ref{secautoauto}). Then $\AGF \cdot B$ is of finite index 

3391: in $\Aut(\Gamma)$.  Together with Proposition \ref{AGFisa}, this

3392: proves the required decomposition of $\Aut(\Gamma)$. 

3393: \end{prf}

3394: 

3395: As another consequence of Lemma \ref{lemma:findex}, we 

3396: infer that $\Out(\Gamma)$ and $\Out(\tGam)$ are S-commensurable: 

3397: 

3398: \begin{prf}{Proof of Proposition \ref{finiin}.}

3399: Since ${\mathsf k}(\Inn_{\Gamma_0}) \leq \Inn_\tGam$ is

3400: of finite index in $\Inn_\tGam$ and ${\mathsf k}: A_0 \ra \tilde{A}_0$ is an isomorphism, 

3401: $\A_0 \cap \Inn_\Gamma$ is a subgroup of finite index in  ${\mathsf k}^{-1}(\tA_{0} \cap \Inn_\tGam)$. 

3402: We consider the map on quotients 

3403: $$   A_{0} \big/ A_{0} \cap \Inn_\Gamma 

3404: \xrightarrow{{\mathsf k}^*} \tA_{0 }\big/ \tA_{0} \cap \Inn_\tGam$$

3405: which is induced by $\mathsf k$.

3406: The above implies that ${\mathsf k}^*$ has finite kernel. 

3407: Since the left hand side 

3408: is a finite index subgroup of $\Out(\Gamma)$, the corollary follows.

3409: \end{prf} 

3410: 

3411: \begin{prf} {\it Proof of Theorem \ref{teoa} in the general case.}

3412: Let $\Gamma$ be a polycyclic-by-finite group.  Then we know that $\Out(\Gamma)$

3413: is residually finite (by \cite{Wehrfritz}). By Corollary

3414: \ref{finiin},  it projects with finite kernel onto a finite index subgroup of

3415: the arithmetic group $\Out(\tGam)$. Thus Proposition \ref{finker} 

3416: implies that $\Out(\Gamma)$ is arithmetic.  

3417: \end{prf}

3418:    

3419: \begin{prf} {\it Proof of Theorem \ref{teosep}.}

3420: By Proposition \ref{equivi} we reduce to the case of $\tilde\Gamma$.

3421: Using Proposition \ref{InnGAGFfi} 

3422: we infer that  

3423: $\A_{\tilde\Gamma|F}$ is of finite index in $\Aut(\tilde\Gamma)$.

3424: We finally use Theorem \ref{teoagf}.

3425: \end{prf}

3426: 

3427: 

3428: 

3429: 

3430: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3431: 

3432: \section{Polycyclic groups with non-arithmetic automorphism groups}

3433: 

3434: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3435: 

3436: We present examples of polycyclic groups 

3437: whose automorphism groups are not isomorphic 

3438: to any arithmetic group. In particular, we shall prove Theorem \ref{teob}.

3439: 

3440: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3441: 

3442: \subsection{Automorphism groups of semi-direct products}\label{splitext}

3443: 

3444: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3445: 

3446: Here are some remarks concerning the automorphism group of groups $\Gamma$

3447: which are  semi-direct products $F\rfish D$ where 

3448: $D$ is a group and $F$ is a (commutative) $D$-module. 

3449: 

3450: We write the group product in $F$ additively, and 

3451: for $h \in D$, we write $f \mapsto h \cdot f$, $f \in F$, 

3452: to denote the action of the element $h$ on $F$.

3453: Let $\Theta$  be a subgroup of $\Gamma$. We write ${\rm Inn}_\Theta$ for the

3454: subgroup of $\Aut(\Gamma)$ consisting of the inner automorphisms defined by

3455: the elements of $\Theta$.

3456: Similarly as before,  we put

3457: $$\AGF:=\{\, \phi \in \Aut(\Gamma)\Mid \phi(F)=F,\, 

3458: \phi |_{\Gamma /F}=  \id_{\Gamma /F} \,\}.$$

3459: There are two constructions for auto\-morphisms in $\AGF$. For the first, let 

3460: ${\rm Der}(D,F) = \{ d: D \to F \mid d(h_1 h_2) = z(h_1)+ h_1 \cdot z(h_2) \}$ 

3461: be the group of derivations from $D$ into $F$. The group of

3462: derivations naturally obtains the structure of a $D$-module by setting

3463: $$g*d\; (h):=g\cdot d(g^{-1}hg)\qquad (g,\, h\in D,\, d\in {\rm Der}(D,F)).$$

3464: A derivation $d\in {\rm Der}(D,F)$ gives rise to 

3465: an automorphism $\phi_d :\Gamma\to \Gamma$ by 

3466: $$\phi_d((m,g)):=(m+d(g),g)\qquad\qquad (m\in F,\, g\in D) \; .$$

3467: We write $\Aut^{\rm d}(\Gamma)$ for the (abelian) subgroup of $\AGF$

3468: consisting of these automorphisms. We remark that the homomorphism 

3469: $${\rm Der}(D,F)\to \AGF,\qquad d\mapsto \phi_d \quad (d\in {\rm

3470:   Der}(D,F))$$

3471: is $D$-equivariant with respect to the above $D$-action on ${\rm Der}(D,F)$ 

3472: and conjugation by elements of  $\Inn_D$ on $\AGF$.

3473: 

3474: Let us define $\Aut_D(F)$ to be the group of $D$-equivariant automorphisms of $F$.

3475: Given a $D$-equivariant automorphism $\rho :F\to F$, we define

3476: an automorphism $\phi_\rho :\Gamma\to \Gamma$ by

3477: $$\phi_\rho((m,g)):=(\rho(m),g)\qquad\qquad (m\in F,\, g\in D) \; .$$

3478: We write $\Aut^{\rm a}(\Gamma)$ for the subgroup of $\AGF$

3479: consisting of these automorphisms.

3480: 

3481: \begin{proposition}\label{splita} Let $\Gamma=F\rfish D$ be a semi-direct

3482: product of a abelian $D$-module $F$ by the group $D$. We then have:

3483: \begin{itemize}

3484: \item[i)]  $\AGF=\Aut^{\rm d}(\Gamma)\cdot \Aut^{\rm a}(\Gamma)$.

3485: 

3486: \item[ii)] ${\rm Inn}_\Gamma\cdot\AGF=

3487: \left(\Aut^{\rm d}(\Gamma)\cdot \Aut^{\rm a}(\Gamma)\right)\cdot {\rm Inn}_D$.

3488: 

3489: \item[iii)] $\Aut^{\rm a}(\Gamma)$ centralizes $\Inn_D$.

3490: 

3491: \item[iv)] $\Aut^{\rm d}(\Gamma)\cap \Aut^{\rm a}(\Gamma)=\{1\}$.

3492: 

3493: \item[v)] $\Aut^{\rm d}(\Gamma)$ is an abelian normal subgroup in 

3494: ${\rm Inn}_\Gamma\cdot \AGF$.  

3495: \end{itemize}

3496: \end{proposition}

3497: 

3498: The proof of this proposition is straightforward, we skip it.

3499: 

3500: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3501: 

3502: \subsection{Examples}\label{exnona}

3503: 

3504: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3505: 

3506: In order to show that certain groups are not arithmetic we use the following

3507: simple criterion.

3508: 

3509: \begin{proposition}\label{splitprop}

3510: For a matrix $A\in\GL(n,\BZ)$, let $\Gamma(A)=\BZ^n\rfish \langle

3511: A \rangle$ be the split extension of $\BZ^n$ by the cyclic group generated by

3512: $A$. If $\Gamma(A)$ is an arithmetic group then either $A$ is of finite order

3513: or a power of $A$ is unipotent or $A$ is semisimple.

3514: \end{proposition}

3515: 

3516: In the first two cases of Proposition \ref{splitprop} that is if  

3517: either $A$ is of finite order or a power of $A$ is unipotent the group 

3518: $\Gamma(A)$ is arithmetic. In case $A$ is semisimple $\Gamma(A)$ can be

3519: arithmetic but examples in \cite{GP4} show that it need not have this

3520: property. 

3521: 

3522: \begin{proof}

3523: Suppose that $\Gamma(A)$ is an arithmetic group and that $A$ is 

3524: not of finite order nor a power of $A$ is unipotent. In this case, we have

3525: $\Fitt(\Gamma(A))=\BZ^n$. Assume further that $\Gamma(A)$ is an arithmetic group.

3526: We can find (compare \cite[Theorem 3.4]{GP1})

3527: a solvable $\BQ$-defined linear

3528: algebraic group ${\bf H}$, having a strong unipotent radical 

3529: so that there is an isomorphism 

3530: $\psi: \Gamma(A)\to \Gamma$

3531: where $\Gamma$ is a Zariski-dense arithmetic

3532: subgroup of ${\bf H}(\BQ)$. Let $\bU$ be the unipotent radical of ${\bf H}$ and

3533: let $\lu$ denote the Lie-algebra of $\bU$. 

3534: The exponential map

3535: $ \exp:  \lu \rightarrow  \bU $ 

3536: is a $\bbQ$-defined isomorphism of varieties.

3537: The adjoint representation leads to a $\BQ$-defined rational

3538: representation $\alpha_{\bf H}: {\bf H}\to \Aut({\lu})$ which is defined by 

3539: $$\alpha_{\bf H}(g)\; (x)=\; \exp^{-1}(g\exp(x)g^{-1})\quad (g\in {\bf H},\, 

3540: x\in {\lu}).$$

3541: Since $\alpha_\bH$

3542: is $\bbQ$-defined,  the image $\alpha_{\bf H}(\Gamma)$ is a Zariski-dense and

3543: arithmetic (by AR1) subgroup of $\alpha_{\bf H}({\bf H})$.

3544: The kernel of $\alpha_{\bf H}$ is equal to $\bU$. 

3545: Taking the image under $\exp^{-1}$ of the standard basis in $\BZ^n$ we obtain a

3546: $\BQ$-basis of $\lu(\BQ)$. Expressed in this basis $\psi(A)$ acts

3547: by the matrix $A$ on the subspace $\lu_0$ spanned by these elements. 

3548: Also ${\bf H}_0$ stabilizes $\lu_0$.

3549: Hence the cyclic subgroup generated by $A$ is arithmetic and Zariski-dense in

3550: $\bH_1 = \alpha_{\bf H}(\bH)$.

3551: 

3552: Let $A=SJ$ be the Jordan-decomposition of $A$, that is $J\in {\bf H}_1(\BQ)$ is

3553: unipotent and $S\in {\bf H}_1(\BQ)$ is semisimple and $JS=SJ$ holds. 

3554: There is a $n\in\BN$

3555: such that $J^n\in {\bf H}_1(\BZ)$,  and hence a $m\in\BN$

3556: such that $J^m\in \langle A\rangle$. This implies 

3557: $S^m\in \langle A\rangle$. We infer that $J=1$. 

3558: \end{proof}

3559:    

3560: We shall discuss now the example from the introduction. 

3561: That is, we choose $d\in\BN$ not a square, 

3562:  set $\omega=\sqrt d$ and

3563: let $K=\BQ(\omega)$ be the

3564: corresponding real quadratic number field. We write $x\mapsto \bar x$ for the

3565: non-trivial element of the Galois group  of $K$ over $\BQ$. 

3566: We consider the subring ${\cal O}=\BZ+\BZ\omega\subset K$ and choose 

3567: a unit $\epsilon=a+b\omega$ of ${\cal O}$ which is of infinite

3568: order and satisfies $\epsilon \bar\epsilon=1$. 

3569: 

3570: Let  $D_\infty$ be the infinite dihedral group as in (\ref{dihe}).

3571: We further take $F={\cal O}\times \BZ$ with the $D_\infty$-module structure

3572: defined as in (\ref{modu}).

3573: As done in the introduction, we put 

3574: $$\Gamma(\epsilon):=F\rfish D_\infty.$$

3575: 

3576: We describe four derivations $d_1,\ldots, d_4$

3577: in  ${\rm Der}(D_\infty,F)$ by specifying their values on 

3578: the generators $A,\, \tau$ of $D_\infty$. 

3579: We define $l$ to be the greatest common factor of $a+1$ and $bd$.

3580: Now put:

3581: $$d_1(A)=(0,1),\ d_1(\tau)=(0,0)\, ; \quad d_2(A)=(0,0),\ d_2(\tau)=(0,1) \, ;$$

3582: $$d_3(A)=(\omega,0),\ d_3(\tau)=(\omega,0) \, ;

3583: \quad d_4(A)=\left(\frac{(\epsilon+1)\omega}{l},0\right),\ d_4(\tau)=(0,0).$$

3584: Each of the above pairs of values defines a

3585: derivation by extension. 

3586: 

3587: We also define

3588: \begin{equation}\label{mamat}

3589: \hat A:=\left(\begin{array}{cccc}

3590: 1 & -2 & 0 & 0 \\

3591: 0 & 1  & 0 & 0 \\

3592: 0 & 0 & -1 & \frac{-2(a+1)}{l} \\

3593: 0 & 0 &  g & 2a+1 

3594: \end{array} \right).

3595: \end{equation}

3596: The structure of $\Aut\big(\Gamma(\epsilon)\big)$ is described in the following

3597: proposition. 

3598: 

3599: \begin{proposition}\label{specialgr} The following hold in

3600: $\Aut(\Gamma(\epsilon))$:  \begin{itemize}

3601: \item[i)] $\Aut_{D_\infty}(F)$ is finite.

3602: 

3603: \item[ii)] The derivations $d_1,\ldots, d_4$ are a $\BZ$-basis of 

3604: ${\rm Der}(D_\infty,F)$.

3605: 

3606: \item[iii)] The action of\/ $\Inn_A$ on ${\rm Der}(D_\infty,F)$ 

3607: expressed relative  to the basis  $d_1,\ldots, d_4$ is given by the matrix $\hat A$.

3608: 

3609: \item[iv)] $\Aut\big(\Gamma(\epsilon)\big)$ contains a subgroup of finite index which is

3610:    isomorphic to $\Gamma(\hat A)$.

3611: \end{itemize}   

3612: \end{proposition}

3613: 

3614: \begin{proof} Items i), ii), iii) are proved by some 

3615: straightforward computations

3616: which we skip. Remark that $F$ is the Fitting-subgroup

3617: of $\Gamma(\epsilon)$. Setting $\Gamma=\Gamma(\epsilon)$ 

3618: we know from Proposition \ref{InnGAGFfi} 

3619: that ${\Inn}_\Gamma\cdot \AGF$ has finite index  

3620: in $\Aut\big(\Gamma(\epsilon)\big)$. The rest follows from Proposition \ref{splita}.

3621: \end{proof}

3622: 

3623: We are now ready for the proof of Theorem \ref{teob}.

3624: 

3625: \begin{prf} {\it Proof of Theorem \ref{teob}} Suppose that

3626: $\Aut\big(\Gamma(\epsilon)\big)$ contains a subgroup of finite index which is an 

3627: arithmetic group. We infer from iv) in Proposition \ref{specialgr} that    

3628: $\Gamma(\hat A)$ is an 

3629: arithmetic group, where $\hat A$ is as in \eqref{mamat}. 

3630: We finish by the remark that $\hat A$ 

3631: does not satisfy the necessary properties 

3632: in Proposition \ref{splitprop}.

3633: \end{prf}

3634: 

3635: Building on the above method it is possible to construct many more

3636: examples of polycyclic groups $\Gamma$ with an

3637: automorphism group $\Aut(\Gamma)$ which does not contain an arithmetic

3638: subgroup of finite index. For example, as a slight variation of the above

3639: groups $\Gamma(\epsilon)$, we may replace the dihedral group

3640: $D_\infty$ by the non-trivial semi-direct product $D_1$ of $\bbZ$ with

3641: itself, and let $D_1$ act on $F= \cO \times \bbZ$ 

3642: via its natural homomorphism to $D_\infty$. 

3643: We thus obtain a torsion-free, arithmetic polycyclic group $\Gamma_1(\epsilon)$

3644: of rank five with non-arithmetic automorphism group. 

3645: Another interesting class of examples may be constructed 

3646: by starting with the (non-arithmetic) polycyclic

3647: groups constructed in  \cite{GP4}. For these examples

3648: the failure of arithmeticity is of rather different nature  

3649: than in the groups  $\Gamma(\epsilon)$. 

3650: 

3651: 

3652: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3653: 

3654: \section{Cohomology representations of $\Out(\Gamma)$}\label{topram1}

3655: 

3656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3657: 

3658: %Let $\Gamma$ be a torsion-free 

3659: %polycyclic-by-finite group, and $R$ a ring. Let 

3660: %$\H^*(\Gamma,R)$ denote the cohomology

3661: %of $\Gamma$ with $R$-coefficients. 

3662: In this section we study the representation of $\Aut(\Gamma)$, $\Gamma$ a 

3663: torsion-free polycyclic-by-finite group, on the cohomology groups 

3664: $\H^*(\Gamma,R)$, where $R = \bbZ, \bbQ, \bbC$.

3665: Since inner automorphisms act trivially on the cohomology of $\Gamma$,  

3666: the outer automorphism group 

3667: $\Out(\Gamma)$ is represented on the cohomology 

3668: ring $\H^*(\Gamma,R)$. 

3669: Considering the special case $R=\BC$, we find that the 

3670: complex vector space $\H^*(\Gamma,\BC)$ comes with a natural 

3671: $\BZ$-structure which is given by the image of the base change homomorphism 

3672: $\H^*(\Gamma,\bbZ) \ra  \H^*(\Gamma, \bbC)$. Recall that this image is a

3673: finitely generated subgroup containing a basis of $\H^*(\Gamma,\BC)$.

3674: We fix here this $\BZ$-structure and its resulting $\BQ$-structure on 

3675: $\H^*(\Gamma,\BC)$.

3676: The representation of  $\Out(\Gamma)$ 

3677: is an  \emph{integral representation} on $\H^*(\Gamma, \bbC)$, that is, 

3678: $\Out(\Gamma)$ normalizes the

3679: $\BZ$-lattice in $\H^*(\Gamma, \bbC)$ just described.

3680: The $\BQ$-structure on  

3681: $\H^*(\Gamma, \bbC)$ allows us to identify 

3682: the group of invertible linear maps $\GLHC$ 

3683: with a $\BQ$-defined linear algebraic group.

3684: The Zariski-closure of the image 

3685: of $\Out(\Gamma)$ in $\GLHC$ is a $\bbQ$-closed 

3686: subgroup. We will show

3687: that the representation of $\Out(\Gamma)$ on 

3688: $\H^*(\Gamma, \bbC)$ 

3689: is  an \emph{arithmetic representation}, 

3690: that is, the image of $\Out(\Gamma)$ in $\GLHC$

3691: is an arithmetic subgroup in its Zariski-closure.

3692: In particular, this establishes 

3693: our main results of Section \ref{topram}.

3694: 

3695: To carry over the information from the embedding of $\Out(\Gamma)$

3696: into a linear algebraic group to 

3697: topology and to the study of the cohomology $\H^*(\Gamma,R)$, 

3698: we apply geometric methods originating from \cite{Baues2}.

3699: 

3700: 

3701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3702: 

3703: \subsection{Automorphisms of Lie algebra cohomology}\label{liecohom}

3704: 

3705: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3706: 

3707: An important special case in our theory is that of a finitely generated

3708: torsion-free nilpotent group $\Gamma$. In this case, the cohomology of 

3709: $\Gamma$ is intimately related to the Lie algebra cohomology of the Lie

3710: algebra of the Malcev completion of $\Gamma$, see Section \ref{geometrc}.

3711: We add here some well known facts about Lie algebra cohomology.  

3712: 

3713: Let $\lg$ denote a Lie algebra. 

3714: The Lie product

3715: of $\lg$ is expressed by a map $\varphi: \lg \wedge \lg \ra \lg$ which

3716: satisfies the Jacobi-identity. The cohomology 

3717: ring $\H(\lg)$ of $\lg$ is defined as the cohomology 

3718: of the \emph{Koszul-complex} $\cK$ of $\lg$, cf.\

3719: \cite{Koszul}. The complex $\cK$ has the structure 

3720: of a differential graded algebra. 

3721: As a graded algebra $\cK = \bigwedge \lg^*$ is the

3722: exterior algebra of  the dual of $\lg$. The 

3723: differential $d$ of $\cK$ is determined in degree one,

3724: where $d:\lg^* \ra \bigwedge^2 \lg^*$ is defined as the 

3725: dual of the Lie product $\varphi$. In particular, 

3726: the cohomology of $\lg$

3727: in degree one is computed as $\H^1(\lg) =

3728: Z^1(\lg) =  [\lg, \lg]^\perp$. Note furthermore

3729: that, via the duality, the automorphism group  

3730: of the differential graded algebra $\cK$ 

3731: identifies with the group of Lie algebra

3732: automorphisms $\Aut(\lg)$. The automorphism 

3733: group $\Aut(\lg)$ acts on the cohomology 

3734: $H(\lg)$ with the inner automorphisms, generated 

3735: by the exponentials of inner derivations  

3736: of $\lg$ acting trivially. 

3737:  

3738: Assume now that $\lg$ is nilpotent. We consider the

3739: descending central series of $\lg$ which is defined by $\lg^0 = \lg$,

3740: $\lg^{i+1} = [\lg, \lg^i]$. Since $\lg$ is nilpotent,

3741: $\lg^{k} = \{ 0\}$, for some (minimal) $k \in \bbN$.

3742: Dualizing the descending central series,

3743: we obtain a filtration $\lg_0 = \{0 \} \subset \lg_1 \ldots 

3744: \subset \lg_k = \lg^*$, where $\lg_i = \left(\lg^i\right)^\perp$,

3745: and $d \lg_{i} \subset \bigwedge^2 \lg_{i-1}$.

3746:  

3747: \begin{lemma} Let $\Phi$ be a semi-simple automorphism of the 

3748: Koszul-complex of the nilpotent Lie algebra $\lg$. 

3749: If $\Phi$ induces the identity on $\H^1(\lg)$ then 

3750: $\Phi = \id$.

3751: \end{lemma}  

3752: \begin{proof} Since $\Phi$ is the identity on $\H^1$, it

3753: is the identity on $\lg_1$. We prove by induction that

3754: $\Phi$ is the identity on the subalgebra $\cK_j$, generated 

3755: by $\lg_j$, $j>1$. Now $\lg_j$ is obtained 

3756: from $\lg_{j-1}$ by 

3757: adding finitely many generators $x \in \lg_j$. 

3758: Since $\Phi$ is semisimple, $x$ may be chosen in a $\Phi$-invariant 

3759: complement $W$ of $\lg_{j-1}$ in $\lg_j$. 

3760: Since $d x \in  \cK_{i-1}$,

3761: $d \Phi x =  \Phi dx = dx$ and

3762: $d \left(\Phi x -x\right) = 0$. Since $W$ 

3763: has no intersection with $\ker d^1 = \lg_1$,

3764: this implies $\Phi x = x$. Therefore,

3765: $\Phi$ is the identity on $\lg_{j}$, and hence 

3766: on $\cK_j$.  

3767: \end{proof}

3768: 

3769: We thus obtain the following result:  

3770: 

3771: \begin{proposition}\label{cohom1} Let $\lg$ be a nilpotent Lie algebra. 

3772: Then the kernel of the natural representation of $\Aut(\lg)$ 

3773: on the cohomology\/  $\H(\lg)$ is unipotent. In particular,

3774: any reductive subgroup of $\Aut(\lg)$ acts faithfully

3775: on $\H(\lg)$, even on $\H^1(\lg)$. 

3776: \end{proposition}

3777: 

3778: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3779: 

3780: \subsection{Computation of $\H^*(\Gamma,\bbC)$ via geometry

3781: and $\Aut(\Gamma)$-actions}\label{geometrc} 

3782: 

3783: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3784: 

3785: Before treating the general case,

3786: we start by recalling some known facts which allow us to

3787: compute the complex cohomology of a finitely generated torsion-free 

3788: nilpotent group in terms of Lie algebra cohomology.

3789: Let $\Theta$ be a finitely generated torsion-free nilpotent group, and

3790: $\bU$ the complex Malcev-completion of $\Theta$,  

3791: $\lu$ the Lie algebra of $\bU$. Thus $\Theta \leq \bU(\bbQ)$ and 

3792: $M_{\Theta}= \Theta \backslash \bU({\bbR})$ is a smooth manifold  

3793: which is an Eilenberg-Mac Lane

3794: space of type $K(\Theta,1)$. In particular, there 

3795: is a natural identification

3796: of $\H^*(\Theta,\bbZ)$ with the singular cohomology group 

3797: $\H^*(M_{\Theta},\bbZ)$, see the discussion later in this section. 

3798: By de Rham's theorem,

3799: the singular cohomology ring $\H^*(M_{\Theta},\bbC)$

3800: of the smooth manifold $M_{\Theta}$ is isomorphic to the

3801: cohomology $\H^*_{\rm DR}(M_{\Theta},\bbC)$ of

3802: complex valued

3803: $C^{\infty}$-differential forms on $M_{\Theta}$. 

3804: In this situation, Nomizu \cite{Nomizu} proved that 

3805: the natural map from $\lu$ into the differential forms 

3806: on $\bU(\bbR)$  induces an  isomorphism of cohomology rings 

3807: \begin{equation}\label{enili101}

3808: {\mathsf n}: \,  \H^*(\lu, \bbC) \xrightarrow{\cong}  

3809: \, \H^*_{\rm DR}(M_{\Theta},\BC) \; .

3810: \end{equation}

3811: Composing this map with 

3812: the natural isomorphisms 

3813: $$ \H^*_{\rm DR}(M_{\Theta},\BC)\to \H^*(M_{\Theta},\BC)\to

3814: \H^*({\Theta},\bbC) $$ 

3815: gives thus a linear isomorphism

3816: \begin{equation}\label{enili102} 

3817: {\it n_{\Theta}}: \H^*(\lu, \bbC) \xrightarrow{\cong}

3818: \H^*({\Theta},\bbC) \; .

3819: \end{equation}

3820: Since $\Aut(\Theta)$ acts on 

3821: $\bU(\bbR)$ by algebraic automorphisms, it also acts 

3822: through smooth maps on $M_\Theta$.  Moreover, the

3823: isomorphisms ${\it n}$ and ${\it n}_{\Theta}$ are compatible with 

3824: the induced cohomology actions of $\Aut(\Theta)$ on  $ \H^*(\lu, \bbC) $, 

3825: $\H^*_{\rm DR}(M_{\Theta},\BC)$ and $\H^*(\Theta,\bbC)$. 

3826: %This line of argument

3827: %may be used to show (see Proposition \ref{prop:cequivariance} below) 

3828: %that the representation of $\Out(\Theta)$ on $\H^*(\Theta,\bbC)$ is

3829: %arithmetic. Furthermore the last statement of Theorem \ref{topolo2} follows by

3830: %application of Proposition \ref{cohom1}.   

3831: 

3832: A similar picture carries over to

3833: our general situation where we start with a

3834: torsion-free polycyclic-by-finite group $\Gamma$. 

3835: We explain now some geometric constructions which 

3836: extend the above picture from the 

3837: case of torsion-free nilpotent groups 

3838: to the more general situation. These constructions are closely 

3839: connected with the algebraic setup discussed

3840: so far in this paper. 

3841:  

3842: Let $\bH_{\Gamma}$ be the algebraic hull of 

3843: $\Gamma$, $\bS$ a maximal $\BQ$-closed d-subgroup, 

3844: and $\bU$ the unipotent radical of $\bH_\Gamma$. We have 

3845: $\bH_{\Gamma}=\bU\cdot \bS$. 

3846: We report from \cite{Baues2} the construction of the standard

3847: $\Gamma$-manifold $M_\Gamma$. To construct this manifold we write a 

3848: $\gamma\in\Gamma$ (uniquely) as 

3849: $\gamma=u s$ with $u\in \bU(\BQ),\, s\in \bS(\BQ)$ 

3850: and set

3851: \begin{equation}\label{op12}

3852: \gamma * x:= u s x s^{-1}=\gamma x s^{-1}\qquad ( x\in \bU(\BR) ).

3853: \end{equation}

3854: As noted in  \cite{Baues2}, this establishes a fixed-point-free, differentiable 

3855: and properly discontinuous action of $\Gamma$ on $\bU(\BR)$. Moreover, 

3856: the quotient space  

3857: $$M_\Gamma=\Gamma \big \backslash \bU(\BR)$$

3858: is a compact 

3859: ${\rm C}^\infty$-manifold and 

3860: an Eilenberg-Mac Lane space of type $\KGone$.

3861: 

3862: We now explain how to calculate the complex 

3863: cohomology of $M_\Gamma$, and hence the cohomology 

3864: of $\Gamma$. 

3865: Let $\lu$ denote the Lie algebra of $\bU$, and let 

3866: $\cK_{\lu}$ be the Koszul-complex of $\lu$ (see Section \ref{liecohom}).

3867: We let $\bS$ act by conjugation on $\bU$ and by the adjoint action on $\lu$

3868: and $\cK_{\lu}$.

3869: Let $\cK_{\lu}^\bS \subset \cK_{\lu}$ denote the

3870: differential subcomplex of invariants for $\bS$. 

3871: %The induced map  $\H^*(\cK_{\lu}^\bS,\BC) \ra   \H^*(\lu, \bbC)^\bS$

3872: %is an isomorphism. 

3873: Now, as is proved  \cite[\S 3]{Baues2},  the obvious map 

3874: \begin{equation}\label{epili101}

3875: {\mathsf  n}:   \H^*(\cK_{\lu},\bbC)^\bS   \ra  \H^*_{\rm DR}(M_{\Gamma},\BC).

3876: \end{equation}  

3877: is an isomorphism of cohomology rings.

3878: 

3879: We explain next how $\Aut(\Gamma)$ acts on the above 

3880: cohomology spaces. 

3881: Let $\phi\in \Aut(\Gamma)$ and $\Phi$ be its extension to an algebraic

3882: automorphism of $\bH_{\Gamma}$. 

3883: We choose a $w_\phi\in\bU(\BQ)$ with the

3884: property that $\Phi(\bS)= w_\phi\bS w_\phi^{-1}$ and set

3885: \begin{equation}\label{op13}

3886: X_\phi(x) =\Phi(x)w_\phi\qquad (x\in \bU(\BR)) \; .    

3887: \end{equation}

3888: This defines  a ${\rm C}^\infty$-map $X_\phi :\bU(\BR)\to \bU(\BR)$. A

3889: straighforward computation yields that

3890: \begin{equation}\label{op14} 

3891: X_\phi(\gamma * x)=\phi(\gamma) * X_\phi(x).

3892: \end{equation}

3893: This shows that the map $X_\phi$ descends to a map 

3894: $\bar X_\phi:M_\Gamma\to M_\Gamma$. 

3895: 

3896: Let

3897: $${\rm Z}_\bU(\bS):=\{\, v\in \bU\Mid vsv^{-1}=s\ {\rm for \ all}\ s\in

3898: \bS\,\}$$

3899: be the centralizer of $\bS$ in $\bU$. This is a $\BQ$-closed subgroup of

3900: $\bU$. Note that the similarly defined normalizer 

3901: of $\bS$ in $\bU$ is in fact equal to ${\rm Z}_\bU(\bS)$. Multiplication from

3902: the right defines an action of ${\rm Z}_\bU(\bS)(\BR)$ on $\bU(\BR)$ which 

3903: commutes with the action of $\Gamma$ on $\bU(\BR)$ defined in (\ref{op12}). 

3904: Since ${\rm Z}_\bU(\bS)(\BR)$ is connected, this action is homotopically

3905: trivial, and so is the induced action on $M_\Gamma$.

3906: 

3907: Let $\phi,\, \psi\in \Aut(\Gamma)$ be automorphisms. A straighforward

3908: computation shows that $X_\phi\circ X_\psi$ differs from    

3909: $X_{\phi\circ\psi}$ by an element of ${\rm Z}_\bU(\bS)(\BR)$ acting on 

3910: $\bU(\BR)$. This shows that $X_\phi\circ X_\psi$ and $X_{\phi\circ\psi}$

3911: are homotopic maps, 

3912: as well as the maps $\bar X_\phi\circ\bar X_\psi$ and $\bar X_{\phi\circ\psi}$. 

3913: In particular, this implies that,   via the maps $\bar X_\phi$, 

3914: $\phi\in \Aut(\Gamma)$, we obtain 

3915: an action of the group $\Aut(\Gamma)$ on the cohomology spaces  

3916: $\H^*_{\rm DR}(M_{\Gamma},\BC)$ and $\H^*(M_{\Gamma},\BC)$. 

3917: The de Rham isomorphism 

3918: \begin{equation}\label{deRa}

3919: {\mathsf I}^*: \; \H^*_{\rm DR}(M_{\Gamma},\BC)\to\H^*(M_{\Gamma},\BC)

3920: \end{equation} 

3921: is obviously equivariant. Since $M_\Gamma$ is a $\KGone$, 

3922: there is an isomorphism (compare

3923:  \cite[Theorem 11.5]{Macl}), 

3924: \begin{equation}\label{Macl} 

3925: {\mathsf l} :\;  \H^*(M_{\Gamma},\BC)\to \H^*(\Gamma,\BC) \; .

3926: \end{equation}

3927: The isomorphism ${\mathsf l}$ is natural with respect to the pairs

3928: $(\bar X_\phi,\phi)$. (For this, property (\ref{op14}) is essential,

3929: as is explained in \cite{Macl}.) In particular, ${\mathsf l}$

3930: is $\Aut(\Gamma)$-equivariant. 

3931:  

3932: 

3933: The group $\Aut_{a}(\bH_{\Gamma})$ of algebraic automorphisms of 

3934: $\bH_{\Gamma}$ stabilizes the unipotent radical $\bU$, and hence it 

3935: acts on $\lu$ and $\cK_{\lu}$. Since the group of inner automorphisms 

3936: $\Inn_\bU$ acts

3937: trivially on $\H^*(\cK_{\lu},\bbC)$,  we obtain an action of 

3938: $\Aut_{a}(\bH_{\Gamma})= \Inn_\bU \cdot \Aut_a(\bH_{\Gamma})_\bS$ on 

3939:  $\H^*(\cK_{\lu},\bbC)^\bS$. Here $\Aut_a(\bH_{\Gamma})_\bS$ stands for the

3940: stabilizer of $\bS$ in $\Aut_{a}(\bH_{\Gamma})$.  

3941: In particular, identifying $\Aut(\Gamma)$ as usually

3942: with a subgroup of $\Aut_a(\bH_{\Gamma})$, 

3943: this constructs a representation 

3944: of $\Aut(\Gamma)$ on $\H^*(\cK_{\lu},\bbC)^\bS$. 

3945: The isomorphism \eqref{epili101} 

3946: can then be easily  seen to be 

3947: $\Aut(\Gamma)$-equivariant. 

3948: 

3949: Let us define

3950: \begin{equation}\label{epili102}

3951: \mathsf{n}_{\Gamma} = {\mathsf l}  \circ {\mathsf I}^*  \circ {\mathsf n}: \;   

3952: \H^*(\cK_{\lu},\bbC)^\bS   \, \ra  \, \H^*({\Gamma}, \bbC) \; .

3953: \end{equation}

3954: We have proved:

3955: \begin{proposition}\label{equivi12}

3956: Let $\Gamma$ be a torsion-free polycyclic-by-finite group. The isomorphism

3957: ${\mathsf n}_\Gamma: \H^*(\cK_{\lu},\bbC)^\bS  \rightarrow  \H^*({\Gamma}, \bbC)$ is equivariant 

3958: with respect to the action of $\Aut(\Gamma)$ 

3959: on $\H^*(\cK_{\lu},\bbC)^\bS$ (as defined above) and 

3960: the natural action on $\H^*({\Gamma},\bbC)$. 

3961: \end{proposition}  

3962:  

3963: \subsection{Rational action of $\Out_a(\bH_\Gamma)$ on $\H^*({\Gamma},

3964: \bbC)$} \label{geometrc3}

3965: 

3966: As remarked before, the cohomology $\H^*({\Gamma},

3967: \bbC)$ carries a natural $\bbQ$-structure 

3968: induced by the coefficient homomorphism 

3969: $\H^*(\Gamma,\bbQ) \ra \H^*(\Gamma,\bbC)$.

3970: Thus, in particular, the group $\GLHC$ 

3971: attains the natural structure of a $\bbQ$-defined group.

3972: We discuss now the naturally defined $\bbQ$-structure on $\H^*(\cK_{\lu},\bbC)^\bS$.

3973: 

3974: Note first that the Lie algebra $\lu$ is defined over $\bbQ$. This means, there

3975: exists a $\bbQ$-Lie algebra $\lu_{\bbQ}$ such that 

3976: $\lu = \lu_{\bbQ} \otimes \BC$ is 

3977: the scalar extension of $\lu_{\bbQ}$. 

3978: The $\bbQ$-subalgebra $\lu_{\bbQ}$ is called a $\bbQ$-structure on $\lu$.

3979: It is  induced by the $\bbQ$-strucure on $\bH_{\Gamma}$ (or,  equivalently,

3980: by the unipotent shadow $\Theta$ of $\Gamma$) on $\lu$. (For related details

3981: concerning $\bbQ$-structures on nilpotent Lie algebras and unipotent

3982: groups one may consult  \cite{Segal,GO}.)   

3983: Since $\lu$ is defined over $\BQ$,  we obtain a rational structure for the

3984: vector space $\cK_{\lu}$ of the Koszul-complex of $\lu$, and 

3985: all differentials are defined over $\BQ$.

3986: Since $\bS$ is $\bbQ$-closed in $\bH_\Gamma$, 

3987: it follows that the complex $\cK_{\lu}^\bS$ and its 

3988: cohomology vector spaces $\H^*(\cK_{\lu}^\bS,\bbC)$

3989: inherit a natural $\bbQ$-structure from  $\cK_{\lu}$, 

3990: representing $\GLHC$  as a $\bbQ$-defined 

3991: linear algebraic group.

3992: 

3993: Recall the construction of the $\bbQ$-defined

3994: algebraic structure of $\Aut_{a}(\bH_{\Gamma})$ which

3995: is discussed in Section \ref{subsect:algebraicstructure}. 

3996: It is obtained by taking the natural

3997: quotient $\Inn_\bU \rtimes \Aut_a(\bH_{\Gamma})_\bS 

3998: \ra \Aut_{a}(\bH_{\Gamma})$. 

3999: Since, by definition of its algebraic structure, the natural 

4000: representation of   

4001: $\Aut_a(\bH_{\Gamma})_\bS$ on $\lu$ is defined over $\bbQ$,  the natural 

4002: representation of $\Aut_a(\bH_{\Gamma})$ on  

4003: $\H^*(\cK_{\lu},\bbC)^\bS$ is $\bbQ$-defined as well. 

4004: This also implies that the representation of $\Aut_{a}(\bH_{\Gamma})$

4005: factors via a $\bbQ$-defined representation $\eta$ 

4006: of $\Out_a(\bH_\Gamma)$. 

4007: In particular, the induced representation 

4008: of $\Aut(\Gamma)$ on $\H^*(\cK_{\lu},\bbC)^\bS$ factors through     

4009: $\Out(\Gamma)$, and,  taking a basis with respect to the above 

4010: constructed $\BQ$-structure, 

4011: every element of  

4012: $\Out(\Gamma)$ acts by a matrix with rational entries on  

4013: $\H^*(\cK_{\lu},\bbC)^\bS$.

4014: 

4015: 

4016: \begin{proposition}   \label{prop:cequivariance}

4017: The natural 

4018: representation of\/ $\Out(\Gamma)$ on $\H^*(\Gamma,\bbC)$

4019: is arithmetic. 

4020: \end{proposition} 

4021: \begin{proof} 

4022: By construction, $\Out(\Gamma)$ is contained in 

4023: the $\bbQ$-defined linear algebraic group $\cO_\Gamma$

4024: as a Zariski-dense arithmetic subgroup. Via the 

4025: homomorphism $\pi_{\cO_\Gamma}: \cO_\Gamma \ra \Out_a(\bH_\Gamma)$

4026: the representation of $\Out(\Gamma)$

4027: on $\H^*(\cK_{\lu},\bbC)^\bS$ is induced by a $\bbQ$-defined

4028: representation of $\cO_\Gamma$. Let $\rho_{\cO_\Gamma}$ denote the

4029: representation of  $\cO_\Gamma$ on $\H^*(\Gamma,\bbC)$

4030: which is obtained by conjugating with the isomorphism 

4031: $\mathsf{n}_{\Gamma}$ defined in \eqref{epili102}. Let $\rho$

4032: denote the natural representation of $\Out(\Gamma)$ on 

4033: $\H^*(\Gamma, \bbC)$. Since $\mathsf{n}_{\Gamma}$ is $\Aut(\Gamma)$-equivariant,

4034: $\rho_{\cO_\Gamma}\big(\Out(\Gamma)\big) = \rho\big(\Out(\Gamma)\big)$ consists of 

4035: integral matrices in $\GLHC$ (with respect to a basis of 

4036: $\H^*(\Gamma,\bbC)$ taken in the image

4037: of $\H^*(\Gamma,\BZ)$).

4038: 

4039: We show that $\rho_{\cO_\Gamma}$ is a $\bbQ$-defined representation 

4040: for the natural $\bbQ$-structure on $\H^*(\Gamma,\bbC)$.

4041: As $\Out(\Gamma)$ is Zariski-dense in $\cO_\Gamma$ and consists of

4042: rational points, it follows that the homomorphism $\rho_{\cO_\Gamma}$  maps

4043: a Zariski-dense subset of rational points of  $\cO_\Gamma$

4044: to rational points of  $\GLHC$. 

4045: The Galois-criterion for rationality applied to the case of the extension 

4046: $\bbC$ over $\bbQ$ implies that $\rho_{\cO_\Gamma}$ is defined over $\bbQ$.

4047: By AR1, we infer that the image $\rho\big(\Out(\Gamma)\big)$ in $\GLHC$

4048: is arithmetic in $\rho_{\cO_\Gamma}(\cO_\Gamma)$. 

4049: \end{proof} 

4050: 

4051: Proposition  \ref{prop:cequivariance} proves 

4052: Theorem \ref{topolo1} of the introduction. Remark that 

4053: the kernel of $\rho$ is a finitely

4054: generated group since it is an arithmetic subgroup of the kernel of

4055: $\rho_{\cO_\Gamma}$.  

4056: 

4057: \begin{remark}

4058: The proof of Proposition \ref{prop:cequivariance} gives no information

4059:  about the rationality of the isomorphism ${\mathsf n}_{\Gamma}$.

4060: In fact, the representation of $\cO_\Gamma$ might be trivial.

4061: \end{remark}

4062: 

4063: For $\Gamma=\Theta$ a nilpotent group, some considerations on

4064: rationality questions for the isomorphism ${\mathsf n}_\Theta$ 

4065: may be found  in \cite{LP}. We state here:

4066: 

4067: \begin{proposition}   The isomorphism ${\mathsf n}_\Gamma$ is defined over $\bbQ$.

4068: \end{proposition}

4069: \begin{prf}{Sketch of proof.} We  show that the map 

4070: ${\mathsf  n}:   \H^*(\cK_{\lu},\bbC)^\bS   \ra  \H^*_{\rm DR}(M_{\Gamma},\BC)$ is defined over $\bbQ$. This can be seen as follows. 

4071: In exponential coordinates for $\bU(\bbR)$, the forms in $\cK_{\lu}(\bbQ)$ define polynomial differential 

4072: forms with rational coefficients. The de Rham isomorphism  ${\mathsf I}^*$ 

4073: factorizes over the

4074: cohomology of the piecewise differential forms, see \cite[VIII]{GM}.

4075: The natural map from differential forms to piecewise-forms, maps the forms 

4076: corresponding to elements of $\cK_{\lu}(\bbQ)$ into piecewise linear forms with rational coefficients.  

4077: By Sullivan's  \emph{P.L.\ de Rham 

4078: theorem} (see \cite{Sullivan, GM}) these are mapped into $\H^*(M_\Gamma, \bbQ)$. 

4079: \end{prf} 

4080: 

4081: 

4082: 

4083: \begin{thebibliography}{99}

4084: 

4085: \bibitem{Ausl1} L.\ Auslander,

4086: \emph{The automorphism group of a polycyclic group},

4087: Ann.\ of Math.\ (2) {\bf 89} (1969), 314-322.

4088: 

4089: \bibitem{Ausl2}  \bysame, 

4090: \emph{On a problem of Philip Hall},

4091: Ann.\ of Math.\ (2) {\bf 86} (1967), 112-116.

4092: 

4093: \bibitem{AJ} L.\ Auslander, F.E.A.\ Johnson,

4094: {\em On a conjecture of C. T. C. Wall\/},

4095: J.\ London Math.\ Soc.\ (2) {\bf 14} (1976), no.\ 2, 331-332.

4096: 

4097: \bibitem{AusBaumslag} L.\ Auslander, G.\ Baumslag,

4098: \emph{Automorphism groups of finitely generated nilpotent groups}, 

4099: Bull.\ Amer.\ Math.\ Soc.\ {\bf 73} (1967) 716-717.

4100: 

4101: \bibitem{Baumslag} G.\ Baumslag, \emph{

4102: Automorphism groups of nilpotent groups},

4103: Amer.\ J.\ Math.\  {\bf 91} (1969),  1003-1011.

4104: 

4105: \bibitem{Baumslag1} \bysame, \emph{

4106: Lectures on Nilpotent Groups},

4107: C.B.M.S. Regional Conference Series {\bf 2}, A.M.S. Providence, R.I.

4108: (1971).

4109: 

4110: \bibitem{Baues1} O.\ Baues, {\it 

4111: Finite extensions and unipotent shadows of affine crystallographic groups}, 

4112: C.\ R.\ Acad.\ Sci.\ Paris, Ser.\ I {\bf 335} (2002), 785-788.

4113: 

4114: \bibitem{Baues2}  \bysame , {\em Infrasolvmanifolds and 

4115: rigidity of subgroups in

4116: solvable linear algebraic groups}, Topology, {\bf  43} (2004),  903-924.

4117: 

4118: \bibitem{Borel1} A.\ Borel, {\it Arithmetic properties of 

4119: linear algebraic groups.}, Proc.\

4120: I.C.M.\ Stockholm (1962), 10-22.

4121: 

4122: \bibitem{Borel2}  A.\ Borel, 

4123: {\it Density and maximality of arithmetic subgroups.}, J.\ Reine

4124: Angew.\ Math.\ {\bf 224}, (1966), 78-89.

4125: 

4126: \bibitem{Borel}  \bysame, {\em Linear algebraic groups},  

4127: Second edition,  Graduate Texts in Mathematics {\bf 126},  

4128: Springer-Verlag, (1991).

4129: 

4130: \bibitem{BS} 

4131: A.\ Borel,  J.-P.\  Serre, 

4132: {\em Th\' eor\` emes de finitude en cohomologie galoisienne},

4133: Comment.\ Math.\ Helv.\ {\bf  39} (1964),  111--164. 

4134: 

4135: \bibitem{BT} 

4136: A.\ Borel,  J.\  Tits, 

4137: {\em Groupes re\'ductive},

4138: Publ. Math. IHES {\bf 27} (1965), 55--150. 

4139: 

4140: \bibitem{Brown} K.S.\ Brown, {\em Cohomology of groups},

4141: Graduate Texts in Mathematics, {\bf 87}. Springer-Verlag, 

4142: New York-Berlin, (1982).

4143: 

4144: \bibitem{BG}  R.M.\ Bryant,  J.R.J.\ Groves, 

4145: \emph{Algebraic groups of automorphisms of nilpotent groups and Lie algebras},

4146: J.\ London Math.\ Soc.\ (2) {\bf 33} (1986), no. 3, 453-466.

4147: 

4148: \bibitem{Deligne} P.\ Deligne,

4149: \emph{Extensions centrales non 

4150: r\'esiduellement finies de groupes arithm\'etiques},

4151: C.\ R.\ Acad.\ Sci.\ Paris S\'er.\ A-B {\bf 287} (1978), no. 4, A203--A208.

4152: 

4153: \bibitem{duSautoy} M.\ du Sautoy, 

4154: \emph{Polycyclic groups, analytic groups and algebraic groups},

4155: Proc.\ London Math. Soc.\ (3) {\bf 85} (2002), no. 1, 62-92.

4156: 

4157: \bibitem{GM}  P. A.\  Griffiths, J.W.\ Morgan, \emph{ 

4158: Rational homotopy theory and differential forms},

4159:  Progress in Mathematics {\bf 16}, 

4160:  Birkh\"auser, Boston, Mass., (1981)

4161: 

4162: \bibitem{GP1}  F.\ Grunewald, V.\ Platonov,

4163: {\em Solvable arithmetic groups and arithmeticity problems},  

4164: Internat.\ J.\ Math.\  {\bf 10}   (1999),  no.\ 3, 327-366.

4165: 

4166: \bibitem{GP2}  \bysame, {\it On finite extensions of arithmetic

4167: groups}, C.\ R.\ Acad.\ Sci.\ Paris {\bf 325} (1997), 1153-1158.

4168: 

4169: \bibitem{GP3}  \bysame, {\it Rigidity Results for Groups with

4170: Radical, Cohomology of Finite Groups and Arithmeticity Problems},

4171: Duke Math.\ J.\  {\bf 100} (1999), 321-358. 

4172: 

4173: \bibitem{GP4}   \bysame, {\it Non-arithmetic polycyclic groups},

4174: C.\ R.\ Acad.\ Sci.\ Paris, {\bf 326} (1998), 1359-1364.

4175: 

4176: \bibitem{GP5}   \bysame, {\it Rigidity and Automorphism Groups of

4177: Solvable Arithmetic Groups.} C.\ R.\ Acad.\ Sci.\ Paris, {\bf 327} (1998), 

4178: 427-432. 

4179: 

4180: \bibitem{GO}  F.\ Grunewald, J.\ O'Halloran, \emph{

4181: Nilpotent groups and unipotent algebraic groups},  

4182: J. Pure Appl. Algebra {\bf 37} (1985), no. 3, 299-313.

4183: 

4184: \bibitem{GS}  F.\ Grunewald, D.\ Segal, 

4185: {\em On affine crystallographic groups},

4186: J.\  Differential Geom.\ {\bf  40} (1994), no.\ 3, 563-594.

4187: 

4188: \bibitem{Koszul} J.-L.\ Koszul, 

4189: \emph{Homologie et cohomologie des alg\'ebres de Lie},

4190: Bull.\ Soc.\ Math.\ France {\bf 78} (1950), 65-127. 

4191: 

4192: \bibitem{LP}  L.A.\ Lambe, S.B.\  Priddy, \emph{ 

4193: Cohomology of nilmanifolds and torsion-free, nilpotent groups},

4194: Trans.\ Amer.\ Math.\ Soc.\ {\bf  273}  (1982), no. 1, 39-55.

4195: 

4196: \bibitem{Macl} S.\  Maclane, {\em Homology},

4197: Springer-Verlag Berlin, G\"ottingen, Heidelberg, Grundlehren der

4198: Mathematischen Wissenschaften {\bf 114}, (1963).

4199: 

4200: \bibitem{Malcev} A.I.\ Malcev, {\em On a class of homogeneous spaces},

4201: Amer.\ Math.\ Soc.\ Translation {\bf 39} (1951).

4202: 

4203: \bibitem{Merzljakov} Ju.\ I.\, Merzljakov,

4204: {\em Integer representation of the holomorphs of polycyclic groups}, (Russian)

4205: Algebra i Logika {\bf 9} (1970), 539-558.

4206: 

4207: \bibitem{Mostow1}  G.D.\ Mostow, {\em  Representative functions on 

4208: discrete groups and solvable arithmetic subgroups}, 

4209: Amer.\ J.\ Math.\  {\bf 92} (1970), 1-32.

4210: 

4211: \bibitem{Mostow2}  \bysame,  {\em

4212: Some applications of representative functions to solvmanifolds},

4213: Amer.\  J.\ Math.\  {\bf 93}  (1971), 11-32.

4214: 

4215: \bibitem{Nomizu}  K.\  Nomizu, {\em  

4216: On the cohomology of compact homogeneous spaces of nilpotent Lie groups}, 

4217: Ann.\  of Math. (2)  {\bf 59} (1954),  531-538.

4218: 

4219: \bibitem{PickelRoit} P.F.\ Pickel, J.\ Roitberg, 

4220: \emph{Automorphism groups of nilpotent groups and spaces},

4221: J.\ Pure Appl.\ Algebra {\bf 150} (2000), no.\ 3, 307-319.

4222: 

4223: \bibitem{PR} V.\ Platonov,  A.\ Rapinchuk, {\it Algebraic

4224: Groups and Number Theory}, Academic Press (1994)

4225: 

4226: \bibitem[22]{Raghunathan} M.S.\ Raghunathan, {\em 

4227: Discrete subgroups of Lie groups},  

4228: Ergebnisse der Mathematik und ihrer Grenzgebiete {\bf 68},

4229: Springer-Verlag,  (1972) 

4230: 

4231: \bibitem{Roit} J.\ Roitberg,

4232: \emph{Genus and symmetry in homotopy theory},

4233: Math.\ Ann.\ {\bf 305} (1996), no.\ 2, 381-386.

4234: 

4235: \bibitem{Rutt} J. W.\ Rutter,

4236: \emph{Spaces of homotopy self-equivalences},

4237: LNM {\bf 1662}, Springer Verlag, (1997).

4238: 

4239: \bibitem{Segal} D.\ Segal, {\em Polyclic groups}, Cambridge Univ.\ Press,

4240: London, 1983

4241: 

4242: \bibitem{Segal2} \bysame, {\em On the outer automorphism group 

4243: of a polycyclic group}, 

4244: Proceedings of the Second International Group Theory 

4245: Conference (Bressanone, 1989),

4246: Rend.\ Circ.\ Mat.\ Palermo (2) Suppl.\ No.\ 23, (1990), 265-278. 

4247: 

4248: \bibitem{Serre}  J.-P.\  Serre, {\it Arithmetic groups},  

4249: London Math. Soc. Lecture Notes

4250: {\bf 36}, (1979), 105-135.

4251: 

4252: \bibitem{Serre2} \bysame, \emph{Cohomologie des groupes discrets},

4253: Prospects in mathematics 

4254: (Proc. Sympos., Princeton Univ., Princeton, N.J., 1970), 

4255: pp.\ 77-169, Ann.\ of Math. Studies, No.\ 70,

4256: Princeton Univ. Press, Princeton, N.J., 1971. 

4257: 

4258: \bibitem{Sullivan} D.\ Sullivan, \emph{Infinitesimal computations in topology},

4259:   Publ.\ I.H.E.S.\  {\bf 47} (1977), 269-331. 

4260:   

4261: \bibitem{Sullivan2} \bysame, \emph{Genetics of homotopy theory and the Adams conjecture}

4262: Ann.\ of Math.\ (2) {\bf 100}  (1974), 1-79.

4263: 

4264: \bibitem{Wehrfritz}  B.A.F.\ Wehrfritz, 

4265: {\it Two Remarks on Polycyclic Groups},  

4266: Bull.\ London Math.\ Soc.\

4267: {\bf 26}, (1994), 543-548.

4268: 

4269: \bibitem{Wehrfritz2} \bysame, 

4270: \emph{On the holomorphs of soluble groups of finite rank},

4271: J.\ Pure Appl.\ Algebra {\bf 4} (1974), 55-69.

4272: 

4273: \end{thebibliography}

4274: 

4275: 

4276: 

4277: 

4278: 

4279: 

4280: 

4281: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4282: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4284: 

4285: \end{document}

4286: 

4287: 

4288: 

4289: 

4290: 

4291: 

4292: 

4293: