1:  \def\Datum{October 10, 2006}

2: \magnification=\magstephalf

3: \hsize=16.2truecm

4: \vsize24truecm

5: \parskip4pt plus 1pt

6: \frenchspacing

7: \lineskiplimit=-2pt

8: \vglue6truemm

9: 

10: \def \Eins {1}

11: \def \Zwei {2}

12: \def \GP {2.1}

13: \def \LE {2.2}

14: \def \RM {2.3}

15: \def \Drei {3}

16: \def \EL {3.1}

17: \def \CR {3.2}

18: \def \HM {3.3}

19: \def \HN {3.4}

20: \def \UR {3.5}

21: \def \AF {3.6}

22: \def \US {3.7}

23: \def \PY {3.8}

24: \def \ZI {3.9}

25: \def \UQ {3.10}

26: \def \Vier {4}

27: \def \HJ {4.1}

28: \def \DC {4.2}

29: \def \ZU {4.3}

30: \def \ZB {4.4}

31: \def \ZA {4.5}

32: \def \ZV {4.6}

33: \def \PL {4.7}

34: \def \LO {4.8}

35: \def \PS {4.10}

36: \def \PQ {4.11}

37: \def \Fuenf {5}

38: \def \JP {5.1}

39: \def \AQ {5.2}

40: \def \PI {5.3}

41: \def \EB {5.4}

42: \def \NP {5.5}

43: \def \Sechs {6}

44: \def \CO {6.1}

45: \def \PR {6.3}

46: \def \CP {6.4}

47: \def \UH {6.5}

48: \def \UK {6.6}

49: \def \JG {6.7}

50: \def \UL {6.8}

51: \def \PT {6.9}

52: \def \JA {6.10}

53: \def \JC {6.11}

54: \def \KZ {6.12}

55: \def \KX {6.13}

56: \def \KI {6.14}

57: \def \KY {6.15}

58: \def \HA {6.16}

59: \def \HU {6.17}

60: \def \UF {6.18}

61: \def \VZ {6.19}

62: \def \KR {6.20}

63: \def \Sieben {7}

64: \def \IO {7.1}

65: \def \IP {7.2}

66: \def \ZR {7.3}

67: \def \AP {7.4}

68: \def \GO {7.5}

69: \def \KO {7.6}

70: \def \BP {7.7}

71: \def \MO {7.8}

72: \def \KW {7.9}

73: \def \KU {7.10}

74: \def \LP {7.11}

75: \def \HQ {7.12}

76: \def \RU {7.13}

77: \def \Acht {8}

78: \def \EI {8.1}

79: \def \EY {8.2}

80: \def \EZ {8.3}

81: \def \EX {8.4}

82: \def \EV {8.5}

83: \def \NM {8.6}

84: \def \LS {8.7}

85: \def \UA {8.8}

86: \def \UB {8.9}

87: \def \OQ {8.10}

88: \def \IZ {8.11}

89: \def \Neun {9}

90: \def \FJ {9.1}

91: \def \XA {9.2}

92: \def \FA {9.4}

93: \def \CA {9.5}

94: \def \MC {9.6}

95: \def \FC {9.7}

96: \def \FD {9.8}

97: \def \ST {9.9}

98: \def \PA {9.10}

99: \def \FG {9.11}

100: \def \Zehn {10}

101: \def \SS {10.1}

102: \def \LA {10.2}

103: \def \Elf {11}

104: \def \Zwoelf {12}

105: \def \LK {12.1}

106: \def \FM {12.2}

107: \def \EM {12.3}

108: \def \NO {12.4}

109: \def \FN {12.5}

110: \def \FO {12.6}

111: \def \FP {12.7}

112: \def \FQ {12.8}

113: \def \FR {12.9}

114: \def \FS {12.10}

115: \def \ZC {12.11}

116: \def \VB {12.12}

117: \def \Dreizehn {13}

118: \def \VD {13.1}

119: \def \Vierzehn {14}

120: \def \FT {14.1}

121: \def \FU {14.2}

122: \def \XY {14.3}

123: \def \FV {14.4}

124: \def \FW {14.5}

125: \def \Fuenfzehn {15}

126: \def \FI {15.1}

127: \def \FY {15.2}

128: \def \VJ {15.3}

129: \def \FZ {15.4}

130: \def \GQ {15.5}

131: \def \GR {15.6}

132: \def \VF {15.7}

133: \def \VG {15.8}

134: \def \VA {15.9}

135: \def \VK {15.12}

136: \def \VL {15.13}

137: \def \GS {15.14}

138: \def \VM {15.17}

139: \def \VN {15.18}

140: \def \VO {15.20}

141: \def \VP {15.21}

142: \def \Sechzehn {16}

143: \def \VQ {16.1}

144: \def \GT {16.2}

145: \def \GU {16.3}

146: \def \GV {16.5}

147: \def \GX {16.6}

148: \def \ANFR {1}

149: \def \ANHI {2}

150: \def \BELO {3}

151: \def \BHUR {4}

152: \def \BERO {5}

153: \def \BAHR {6}

154: \def \BARZ {7}

155: \def \BOGG {8}

156: \def \BOUR {9}

157: \def \BUSH {10}

158: \def \CART {11}

159: \def \CHMO {12}

160: \def \DAYA {13}

161: \def \DOKR {14}

162: \def \EAEZ {15}

163: \def \EBFT {16}

164: \def \EBEN {17}

165: \def \FELS {18}

166: \def \FEKA {19}

167: \def \GAME {20}

168: \def \HOCH {21}

169: \def \KNAP {22}

170: \def \HART {23}

171: \def \HORM {24}

172: \def \ISMI {25}

173: \def \KAZT {26}

174: \def \LOBO {27}

175: \def \LOBP {28}

176: \def \LOBQ {29}

177: \def \MENA {30}

178: \def \PALA {31}

179: \def \SAKA {32}

180: \def \SEVL {33}

181: \def \TANA {34}

182: 

183: 

184: \input cls.def\immediate\openout\aux=\jobname.aux

185: 

186: 

187: \arxtrue

188: 

189: \ifdraft

190: \footline={\hss\sevenrm \Datum\hss}

191: \else\nopagenumbers\fi

192: 

193: \headline={\ifnum\pageno>1\sevenrm\ifodd\pageno Homogeneous Levi

194: degenerate CR-manifolds

195: \hss{\tenbf\folio}\else{\tenbf\folio}\hss{Fels-Kaup}

196: \fi\else\hss\fi}

197: 

198: 

199: \centerline{\Gross Classification of  Levi degenerate homogeneous}

200: \bigskip

201: \centerline{\Gross CR-manifolds in dimension 5}

202: \bigskip\bigskip

203: \centerline{Gregor Fels\quad and\quad Wilhelm Kaup}

204: \bigskip

205: {\parindent0pt\footnote{}{\ninerm 2000 Mathematics

206: Subject Classification: 32M17, 32M25, 32V25.

207: }}

208: 

209: \def\Partskip{\vskip9mm}

210: 

211: \bigskip\bigskip

212: 

213: \ifarx

214: {\narrower\narrower\Klein{\bf\noindent Abstract:} In this paper we present new

215: examples of homogeneous $2$-nondegenerate CR-manifolds of dimension 5

216: and give, up to local CR-equivalence, a full classification of all

217: CR-manifolds of this type.\par}

218: 

219: \bigskip\bigskip

220: 

221: {\def \Einsa {1}

222: \def \Zweia {6}

223: \def \Dreia {7}

224: \def \Viera {11}

225: \def \Fuenfa {14}

226: \def \Sechsa {16}

227: \def \Siebena {22}

228: \def \Achta {26}

229: \def \Neuna {29}

230: \def \Zehna {33}

231: \def \Elfa {36}

232: \def \Zwoelfa {38}

233: \def \Dreizehna {41}

234: \def \Vierzehna {43}

235: \def \Fuenfzehna {44}

236: \def \Sechzehna {49}

237: \def \Refa {52}

238:  \SkIp=70pt

239: \advance\baselineskip8pt\Klein\advance\hsize-60pt

240: \line{\hbox to \SkIp{\hfill\Eins. }Introduction\dotfill\Einsa}

241: \vskip8pt

242: \line{\hbox to 90pt{\hfill}\elfrm PART 1: Levi degenerate CR-manifolds\hfill}

243: \vskip1pt

244: \line{\hbox to \SkIp{\hfill\Zwei. }Preliminaries\dotfill\Zweia}

245: \line{\hbox to \SkIp{\hfill\Drei. }Tube manifolds\dotfill\Dreia}

246: \line{\hbox to \SkIp{\hfill\Vier. }Tube manifolds over cones\dotfill\Viera}

247: \line{\hbox to \SkIp{\hfill\Fuenf. }Some examples\dotfill\Fuenfa}

248: \line{\hbox to \SkIp{\hfill\Sechs. }Levi degenerate CR-manifolds associated with an

249: endomorphism\dotfill\Sechsa}

250: \line{\hbox to \SkIp{\hfill\Sieben. }Homogeneous 2-nondegenerate manifolds of CR-dimension 2\dotfill\Siebena}

251: \line{\hbox to \SkIp{\hfill\Acht. }Homogeneous 2-nondegenerate CR-manifolds in dimension 5\dotfill\Achta}

252: \vskip8pt

253: \line{\hbox to 90pt{\hfill}\elfrm PART 2: The classification\hfill}

254: \vskip1pt

255: \line{\hbox to \SkIp{\hfill\Neun. }Lie-theoretic characterization of locally homogeneous

256: CR-manifolds\dotfill\Neuna}

257: \line{\hbox to \SkIp{\hfill\Zehn. }The Lie algebra $\7g$ has small semisimple part\dotfill\Zehna}

258: \line{\hbox to \SkIp{\hfill\Elf. }The cases $\7s\cong\so(2,3)$ and $\7s\cong\so(1,3)$\dotfill\Elfa}

259: \line{\hbox to \SkIp{\hfill\Zwoelf. }The case $\7s\cong\7{sl}(2,\RR)$\dotfill\Zwoelfa}

260: \line{\hbox to \SkIp{\hfill\Dreizehn. }The case $\7s\cong\su(2)$\dotfill\Dreizehna}

261: \line{\hbox to \SkIp{\hfill\Vierzehn. }Reduction to the case where $\7g$ is solvable and of dimension

262: 5\dotfill\Vierzehna}

263: \line{\hbox to \SkIp{\hfill\Fuenfzehn. }The existence of a 3-dimensional abelian ideal in

264: $\7g$ suffices\dotfill\Fuenfzehna}

265: \line{\hbox to \SkIp{\hfill\Sechzehn. }The final steps\dotfill\Sechzehna}

266: \vskip4pt

267: \line{\hbox to \SkIp{\hfill}References\dotfill\Refa}

268: }\fi

269: 

270: \newwrite\lst\immediate\openout\lst=\jobname.lst

271: 

272: 

273: \KAP{Eins}{Introduction}

274: 

275: The main topic of this paper is the study of real-analytic

276: CR-manifolds $M$ with everywhere degenerate Levi form. In particular,

277: for homogeneous manifolds of this type, we develop methods for the

278: computation of the Lie algebra s$\hol(M,a)$ of infinitesimal

279: CR-transformations at every $a\in M$. We also classify up to local

280: CR-equivalence all locally homogeneous degenerate

281: CR-manifolds in dimension 5.

282: 

283: In this context, a well studied example of a homogeneous Levi degenerate

284: CR-manifold is the quadratic hypersurface

285: $$

286: \5M:=\{z\in\CC^{3}: (\Re z_{1})^{2}+(\Re z_{2})^{2}=(\Re

287: z_{3})^{2},\;\Re z_{3} >0\} \;,

288: $$ compare

289: e.g. \Lit{EBFT}, \Lit{FEKA}, \Lit{KAZT} and \Lit{SEVL}.  This

290: 5-dimensional CR-manifold has several remarkable properties and

291: serves as motivation for various considerations in this paper.

292: Notice that $\5M$ can also be written as tube manifold

293: $$\5M=\5F+i\RR^{3}\subset\CC^{3},\steil{where}\5F:=\{x\in\RR^{3}:

294: x^{2}_{1}+x^{2}_{2}=x^{2}_{3},\,x_{3}>0\}$$ is the future light cone

295: in 3-dimensional space-time. A glance at this description shows that

296: $\5M$ is homogeneous under a group of complex-affine

297: transformations. It is less obvious that the Lie algebras of {\sl

298: global} and {\sl local} infinitesimal CR-transformations at $a\in\5M$,

299: $\hol(\5M)$ and $\hol(\5M,a)$, are both isomorphic to $\so(2,3)$, and

300: hence have dimension 10, compare \Lit{KAZT}. Also the following

301: `globalization' is known: The group $\SO(2,3)$ acts on the complex

302: quadric $\fett\3Q_3\subset\PP_{4}(\CC)$ by biholomorphic

303: transformations and has a hypersurface orbit that contains $\5M$ as a

304: dense domain.

305: 

306: \ifarx\medskip\fi

307: 

308: The cone $\5F$ clearly is a disjoint union of affine

309: half-lines. Therefore, $\5M$ is a disjoint union of complex

310: half-planes, actually $\5M$ is a fiber bundle with typical fiber

311: $\HH^{+}:=\{z\in\CC:\Re(z)>0\}$. However, the total space $\5M$ is not

312: even locally CR-equivalent to a product $\HH^+\Times M'$ with $M'$ a

313: CR-manifold.  Notice that the Levi form in both cases, i.e., for $\5M$

314: and for a product of $\HH^{+}$ with a Levi nondegenerate 3-dimensional

315: CR-manifold $M'$, has exactly one non-zero eigenvalue at every

316: point. Hence, one needs further invariants to distinguish those

317: CR-manifolds.  While every product $M'\times\CC$ is holomorphically

318: degenerate, the crucial fact here is that the light cone tube $\5M$ is

319: nondegenerate in a higher order sense: To be precise, $\5M$ is

320: 2-nondegenerate at every point, and we refer to \Lit{BAHR} and also to

321: \Lit{BERO}, 11.1, for the notion of $k$-nondegeneracy.

322: 

323: \ifarx\medskip\fi

324: 

325: In the non-homogeneous setting, for every $k\in \NN$ and fixed

326: manifold dimension it is not difficult to construct large classes of

327: CR-manifolds, even hypersurfaces, which are $k$-nondegenerate at some

328: points, but are Levi nondegenerate in a dense open subset.  It seems

329: to be much harder to construct CR-manifolds which are

330: $k$-nondegenerate everywhere for $k\ge 2$.  Note that the CR-dimension

331: of a {\sl homogeneous} $M$ is an upper bound for the degree $k$ of

332: nondegeneracy.  Hence, the lowest manifold dimension for which

333: everywhere $2$-nondegenerate CR-manifolds can exist is 5.  which

334: raises the intriguing question whether besides the light cone tube

335: there exist further 2-nondegenerate homogeneous CR-manifolds in

336: dimension 5. So far, all known examples in dimension finally turned

337: out to be locally CR-isomorphic to $\5M$, compare e.g. \Lit{EBEN},

338: \Lit{KAZT}, \Lit{GAME}, \Lit{FEKA}, \Lit{BELO}, and the belief arose

339: that there are no others.

340: 

341: \ifarx\medskip\fi

342: 

343: The main objective of this paper is to show that there are actually

344: infinitely many locally mutually non-equivalent examples and to

345: provide a full classification. Starting point is the following simple

346: observation: Suppose that $F\subset\RR^{n}$ is an affinely homogeneous

347: (connected) submanifold of dimension say $d<n$. Then the tube

348: $M:=F+i\RR^{n}$ is a generic CR-submanifold of $\CC^{n}$ of

349: CR-dimension $d$ and is homogeneous under a group of complex-affine

350: transformations. Indeed, every real-affine transformation leaving $F$

351: invariant extends to a complex-affine transformation leaving $M$

352: invariant and, furthermore, $M$ is invariant under all translations

353: $z\mapsto z+iv$ with $v\in\RR^{n}$. Clearly, the crucial question is,

354: when is $M$ $k$-nondegenerate and when are two tubes $M,M'$ of this

355: type locally CR-equivalent?

356: 

357: \ifarx\medskip\fi

358: 

359: The classification of all affinely homogeneous surfaces

360: $F\subset\RR^{3}$ can be found in \Lit{DOKR} and \Lit{EAEZ}. In

361: particular, a complete list (up to local affine equivalence and in a

362: slightly different formulation) of all degenerate types that are not a

363: cylinder, is given by the following examples (1) -- (3).

364: 

365: \item{(1)} $F=\5F$ the future light cone as above.

366: 

367: \item{(2a)} $F=\{\,r\,(\cos t,\sin t,e^{\omega

368: t})\in\RR^{3}:r\in\RRp,t\in\RR\}$ with $\omega>0$ arbitrary.

369: 

370: \item{(2b)} $F=\{\,r\,(1,t,e^{t})\in\RR^{3}:r\in\RRp,t\in\RR\}$.

371: 

372: \item{(2c)} $F=\{\,r\,(1,e^{t},e^{\theta

373: t})\in\RR^{3}:r\in\RRp,t\in\RR\}$ with $\theta>2$ arbitrary.

374: 

375: \item{(3)} $F=\{\,c(t)+r\re1c'(t)\in\RR^{3}: r\in\RRp,t\in\RR\}$,

376: where $c(t):=(t,t^{2},t^{3})$ parameterizes the {\sl twisted cubic}

377: $\{(t,t^{2},t^{3}):t\in\RR\}$ in $\RR^{3}$ and $c'(t)=(1,2t,3t^{2})$.

378: 

379: \noindent Notice that the limit case $\omega=0$ in (2a) gives the

380: future light cone $\5F$, while the limit case $\theta=2$ in (2c) gives

381: the linearly homogeneous surface

382: $\{x\in\RR^{3}:x_{1}x_{3}=x_{2}^{2},\,x_{1}>0,\,x_{2}>0\}$,

383: which is locally, but not globally linearly equivalent to $\5F$. In

384: fact, $\5F$ is linearly equivalent to the cone\nline

385: $\{x\in\RR^{3}:x_{1}x_{3}=x_{2}^{2},\,x_{1}+x_{3}>0\}$.

386: 

387: \ifarx\vfill\eject\else\medskip\fi

388: 

389: As our first main result we show, compare \ruf{UA} and \ruf{UB} for

390: details:

391: 

392: \noindent{\parskip0pt\bf Theorem I: \sl For every surface $F$ in {\rm(1) -- (3)}

393: the corresponding tube manifold $M:=F+i\RR^{3}$ is a homogeneous

394: $2$-nondegenerate CR-submanifold of $\CC^{3}$ and any two of them are

395: pairwise locally CR-inequivalent. Furthermore, for every $F$ in

396: {\rm(2a) -- (3)} and every $a\in M=F+i\RR^{3}$ the following holds: \0

397: The Lie algebra $\hol(M,a)$ is solvable and has dimension $5$. \1 The

398: stability group $\Aut(M,a)$ is trivial. \1 Every homogeneous

399: real-analytic CR-manifold $M'$, that is locally CR-equivalent to $M$,

400: is already globally CR-equivalent to $M$.\par}

401: 

402: \smallskip\noindent Notice that, a priori, there is no reason why the

403: $F$ in (1) -- (3), although known to be locally affinely inequivalent,

404: should have locally CR-inequivalent tubes (for nondegenerate affinely

405: homogeneous surfaces in $\RR^{3}$, for instance, there are

406: counterexamples).

407: 

408: \medskip We actually prove an analog of Theorem I in every dimension

409: $n\ge3$, where the same trichotomy occurs as above: Consider the

410: following surfaces $F\subset\RR^{n}$

411: 

412: \item{(1')} $F=\5F^{n}:=\{x\in\RR^{n}:x_{1}>0,\,x_{2}>0\steil{and}

413: x_{j}=x^{j-1}_{2}/x_{1}^{j-2} \steil{for}3\le j\le n\}$.

414: 

415: \item{(2')}

416: $F=\{r\re1e^{t\phi}(a)\in\RR^{n}:r\in\RRp,t\in\RR\}$, where

417: $\phi$ is an endomorphism of $\RR^{n}$ having $a\in\RR^{n}$ as cyclic

418: vector (i.e., the iterates $\phi^{k}(a)$, $k\ge0$ span $\RR^{n}$) and

419: the $n$ eigenvalues of $\phi$ do not form an arithmetic progression in

420: $\CC$.

421: 

422: 

423: \item{(3')} $F=\{c(t)+r\re1c'(t):r\in\RRp,t\in\RR\}$, where

424: $c(t):=(t,t^{2},\dots,t^{n})\in\RR^{n}$ parameterizes the twisted

425: $n$-ic in $\RR^{n}$.\par

426: 

427: \noindent In Sections \ruf{Sechs} and \ruf{Sieben} we show among other

428: statements: For every $F,F'$ from (1') -- (3') the tube manifolds

429: $M:=F+i\RR^{n}$, $M':=F'+i\RR^{n}$ are affinely homogeneous generic

430: $2$-nondegenerate submanifolds of $\CC^{n}$ with CR-dimension $2$.

431: Furthermore, $M$, $M'$ are locally CR-equivalent if and only if $F$,

432: $F'$ are globally affinely equivalent, and this holds if and only if

433: for given $a\in M$, $a'\in M'$ the Lie algebras $\hol(M,a)$,

434: $\hol(M',a')$ are isomorphic.  In case $F=\5F^{n}$ the Lie algebra

435: $\hol(M,a)$ contains a copy of $\gl(2,\RR)$ and hence is not

436: solvable.  In all other cases, i.e. $F$ comes from (2') or (3'), the

437: Lie algebra $\hol(M,a)$ is solvable of dimension $n+2$ and the

438: stability group $\Aut(M,a)$ has order at most $2$.

439: 

440: 

441: \medskip Let us briefly comment on the proof of Theorem I: Once

442: defining equations for an $F$ under consideration are explicitly known

443: (this is quite obvious for the types (1) -- (3), compare Section

444: \ruf{Acht}, but seems to be hard for the types (2'), (3')\hskip1pt),

445: one can compute by standard methods the order $k$ of

446: nondegeneracy. However, the amount of calculation necessary to

447: determine $k$ in such a way grows very fast with $k$ and with the

448: dimension or codimension of $M\subset \CC^n$. This is one of the

449: reasons, especially with an eye on possible generalizations, to choose

450: a different approach, which does not use explicit equations. For

451: instance, given an arbitrary submanifold $F\subset \RR^n$ which is

452: (locally) {\sl affinely homogeneous}, we present a simple criterion

453: (Proposition \ruf{US}) which allows to determine quickly the order $k$

454: of nondegeneracy for the corresponding tube manifold.

455: 

456: \noindent The hard part of the proof is to show that the various tubes

457: $F+i\RR^3$ are actually locally inequivalent as CR-manifolds.  Recall

458: that for real-analytic, not necessary homogeneous, {\sl hypersurfaces

459: with nondegenerate Levi form} there exist local invariants which

460: determine each $M$ up to local CR-equivalence due to the

461: fundamental work of Cartan, Tanaka and Chern-Moser.  However, an

462: analogues approach is not available for $M$ of higher codimensions or

463: when the Levi form is degenerate. To distinguish the various tubes

464: $F+i\RR^3,$ we develop a method (valid also in greater generality)

465: which enables us to determine explicitly the Lie algebras $\hol(M,a)$

466: of infinitesimal CR-transformations of the various CR-germs $(M,a)$

467: (see Section \ruf{Zwei} for basic definitions and Sections \ruf{Sechs}

468: for further details).

469: 

470: \medskip The CR-manifolds occurring in the previous theorem are quite

471: special as they all are tube manifolds.  Moreover, all but one (namely

472: the twisted cubic case in (3)\hskip1pt) are actually conical.  In the

473: Levi nondegenerate case, many (homogeneous) examples are known which

474: are not locally CR-equivalent to any tube manifold. For instance, the

475: unit sphere subbundle of $TS^3$ with its canonical CR-structure is

476: such an example. Therefore, our second main result came quite

477: unexpected to us:

478: 

479: \medskip\noindent {\bf Theorem II.} {\sl Every 5-dimensional locally

480: homogeneous 2-nondegenerate CR-manifold $M$ is locally

481: CR-equivalent to $F+i\RR^3$ with $F$ being one of the surfaces in {\rm

482: (1) -- (3)}}.

483: 

484: \medskip\noindent For the precise definition of local homogeneity we

485: refer to Section \ruf{Zwei}.  A priori, locally homogeneous

486: CR-manifolds might exist which are not locally CR-equivalent to any

487: globally homogeneous one.  As a by-product of the above 2 results we

488: can see that such a pathology does not happen in the case under

489: consideration. Theorem II gives a classification of all

490: (abstract) 2-nondegenerate locally homogeneous CR-manifolds in

491: dimension 5 up to local CR-equivalence.  In fact, using Cartan's

492: classification \Lit{CART} of the 3-dimensional Levi nondegenerate

493: homogeneous CR-manifolds, the following result holds.

494: 

495: \medskip\noindent {\bf Classification.}  {\sl Every 5-dimensional

496: locally homogeneous CR-manifold $M$ with degenerate Levi form is

497: locally CR-equivalent to one of the following: \0 $M=F+i\RR^3\subset

498: \CC^3$ and $F$ is one of the surfaces {\rm (1) -- (3)} from above.  \1

499: $M=\CC\times M'$, where $M'$ is one of the 3-dimensional Levi

500: nondegenerate CR-manifolds from Cartan's list in \Lit{CART}. \1

501: $M=\CC^2\times \RR$ or $M=\CC\times \RR^3$ with $\RR^3$ totally real.

502: 

503: \noindent The manifolds in {\rm (i)} are all $2$-nondegenerate and in

504: {\rm (ii) -- (iii)} are holomorphically degenerate. Also, the

505: manifolds in {\rm (iii)} are just the Levi flat ones.}

506: 

507: \smallskip With Theorem II the following question arises naturally for

508: higher codimension: {\sl Are there, up to local CR-equivalence,

509: further locally homogeneous $2$-nondegenerate CR-manifolds of

510: CR-dimension 2 besides those that are tubes over the surfaces $F$ in

511: {\rm (1') -- (3')} above?} Notice in this context that every locally

512: homogeneous $2$-nondegenerate CR-manifold of dimension $6$ necessarily

513: has CR-dimension $2$.

514: 

515: \smallskip For 5-dimensional CR-hypersurfaces with {\sl nondegenerate}

516: Levi form, that is, when Chern-Moser invariants are available, there

517: already exists a partial classification: Locally homogeneous

518: CR-hypersurfaces in $\CC^3$ with stability groups of positive

519: dimension have been classified by Loboda in terms of local equations

520: in normal form, see \Lit{LOBO}, \Lit{LOBP}, \Lit{LOBQ}.

521: 

522: \medskip The paper is organized as follows. After recalling some

523: necessary preliminaries in Section \ruf{Zwei} we discuss in Section

524: \ruf{Drei} tube manifolds $M=F\oplus i\RR^{n}\subset\CC^{n}$ over

525: real-analytic submanifolds $F\subset\RR^{n}$. It turns out that the

526: CR-structure of $M$ is closely related to the real-affine structure of

527: the base $F$. For instance, the Levi form of $M$ is essentially the

528: sesquilinear extension of the second fundamental form of the

529: submanifold $F\subset\RR^{n}$. Generalizing the notion of the second

530: fundamental form we define higher order invariants for $F$ (see

531: \ruf{HN}). In the uniform case these invariants precisely detect the

532: $k$-nondegeneracy of the corresponding CR-manifold $M=F\oplus i\RR^n$.

533: It is known that the (uniform) $k$-nondegeneracy of a real-analytic

534: CR-manifold $M$ together with minimality ensures that the Lie algebras

535: $\hol(M,a)$ are finite-dimensional, and is equivalent to this in the

536: special class of locally homogeneous CR-manifolds. For submanifolds

537: $F\subset\RR^{n}$ which in addition are homogeneous under a group of

538: affine transformations, a simple criterion for $k$-nondegeneracy of

539: the associated tube manifold $M$ is given in Proposition \ruf{US}.

540: 

541: 

542: In Section \ruf{Vier} these results are applied to the case where $F$

543: is conical in $\RR^n$, that is, locally invariant under dilations

544: $z\mapsto tz$ for $t$ near $1\in\RR$. In this case $M$ is always {\sl

545: Levi degenerate.} Assuming that $\hol(M,a)$ is finite dimensional

546: (which automatically is the case for minimal and finitely

547: nondegenerate CR-manifolds) we develop some basic techniques for the

548: explicit computation of $\hol(M,a)$.  The main results in Section

549: \ruf{Vier} are the following: We prove that under the finiteness

550: assumption $\hol(M,a)$ consists only of polynomial vector fields and

551: carries a natural graded structure, see Proposition \ruf{DC}. We prove

552: (under the same assumptions, see Proposition \ruf{ZB}.ii) that local

553: CR-equivalences between two such tube manifolds are always rational

554: maps (even if these manifolds are not real-algebraic). Furthermore,

555: jet determination estimates are provided (\ruf{ZB}.iv). In the special

556: situation where $\hol(M,a)$ consists of affine germs only, these

557: results are further strengthened.

558: 

559: In Section \ruf{Fuenf} we illustrate by examples how our methods can

560: be applied.  In Example \ruf{JP} we present to every $c\ge 1$ and

561: $k\in \{2,3,4\}$ a homogeneous submanifold of $\CC^{n}$ which is

562: $k$-nondegenerate, of codimension $c$ and of CR-dimension $k$.  We

563: close this section investigating the tubes $M^\alpha_{p,q}$ over the

564: cones $F^\alpha_{p,q}:=\{x\in \RR^{p{+}q}_+: \sum \epsilon_j

565: x_j^\alpha=0\}$ with $p$ positive and $q$ negative $\epsilon_j$'s and

566: $\alpha\in \RR^{*}$. Using our methods from the previous section we

567: explicitly determine the Lie algebras $\hol(M^\alpha_{p,q},a)$ for

568: arbitrary integers $\alpha\ge2$ and $p\ge q\ge 1$.

569: 

570: In Section \ruf{Sechs}, we construct homogeneous CR-submanifolds

571: $M=M^{\phi,d}\subset\CC^{n}$ of tube type, depending on the choice of

572: an endomorphism $\phi\in \End(\RR^n)$, an integer $1<d<n$ and a cyclic

573: vector $a\in\RR^{n}$ for $\phi$ in the following way: The powers

574: $\phi^0,\phi^1,\ldots,\phi^{d-1}$ span an abelian Lie algebra $\7h$

575: and, in turn, give a cone $F:=\exp(\7h)a\subset \RR^n$. For $\phi$

576: which admits a cyclic vector $a\in \RR^n$ the corresponding tube

577: manifold $M^{\phi,d}=F+i\RR^{n}$ is 2-nondegenerate, minimal and of

578: CR-dimension $d$.  The key result here is the explicit determination

579: of the CR-invariant $\hol(M^{\phi,d},a)$ for $\phi$ in 'general

580: position' (Propositions \ruf{UH} and \ruf{HU}). The precise meaning

581: for $\phi$ of being in 'general position' is stated in Lemma \ruf{UL}.

582: Further results are, again for $\phi$ in general position, that the

583: tube manifold $M$ is simply connected and has trivial stability group

584: at every point. As a consequence, the manifolds $M$ of this type have

585: the following remarkable property: {\sl Every homogeneous

586: (real-analytic) CR-manifold locally CR-equivalent to $M$ is globally

587: CR-equivalent to $M$}.

588: 

589: In Section \ruf{Sieben} the results from Section \ruf{Sechs} are

590: further refined in the case of homogeneous CR-manifolds

591: $M^\phi:=M^{\phi,2}:=F^\phi+i\RR^n\subset \CC^n$ of CR-dimension 2 but

592: without restrictions on the codimension.  In fact, every minimal and

593: locally homogeneous tube CR-manifold $M:=S+i\RR^n$ with a conical

594: 2-dimensional $S\subset \RR^n$ is locally CR-equivalent to

595: $M^{\phi,2}$ for some cyclic $\phi$.  In this section also the case

596: when $\phi$ is not in 'general position', i.e., the characteristic

597: roots do form an arithmetic progression (see Lemma \ruf{UL})

598: is treated.  The main results here are: Whether a cyclic $\phi$ is in

599: general position or not, the Lie algebras $\hol(M^\phi,a)$

600: (Proposition \ruf{ZR}) and the global automorphism groups

601: $\Aut(M^\phi)$ (Proposition \ruf{GO}) are determined.  Furthermore,

602: the problem of local and global CR-equivalence among the $M^\phi$'s is

603: solved (Propositions \ruf{KO} and \ruf{BP}) and a moduli space is

604: constructed (Subsection \ruf{MO}).

605: 

606: Part I of the paper is concluded with Section \ruf{Acht} where the

607: examples (1) -- (3) from Theorem I are presented in more detail.  The

608: results of the preceeding section are applied to this case of

609: 5-dimensional tube manifolds $M$. In particular Propositions \ruf{UA}

610: and \ruf{IZ} contain some additional information to that stated in

611: Theorem I and also complete the proof of Theorem I.

612: 

613: Part II of the paper is mainly devoted to prove Theorem II.  In the

614: preliminary section \ruf{Neun} we explain how the geometric properties

615: such as $k$-nondegeneracy, minimality or the CR-dimension of an

616: arbitrary locally homogeneous CR-germ $(M,o)$ can be encoded in a pure

617: Lie algebraic object, the associated CR-algebra $(\7g,\7q)$. This is

618: the key for our classification and is based on results taken from

619: \Lit{FELS}. Specified to 5-dimensional CR-germs, we formulate the

620: precise algebraic conditions on a CR-algebra $(\7g,\7q)$ ensuring that

621: the associated CR-germ $(M,o)$ is 2-nondegenerate.

622: 

623: Once the classification of 2-nondegenerate 5-dimensional locally

624: homogeneous CR-germs $(M,o)$ is reduced to a classification problem of

625: certain CR-algebras, we begin in Section \ruf{Zehn} to carry out the

626: details of the proof.  It is subdivided into several sections, lemmata

627: and claims and will only be completed in Section \ruf{Sechzehn}.  Our

628: proof relies on a quite subtle analysis of Lie algebraic properties of

629: CR-algebras and uses basic structure theory of Lie algebras and Lie

630: groups.  The methods are quite general and can be adapted to handle

631: also higher dimensional cases.

632: 

633: A more detailed outline of the proof can be found in the first part of

634: Section \ruf{Zehn}.

635: 

636: 

637: 

638: \ifarx\vfill\eject\else\Partskip\fi

639: 

640: \centerline{\twelverm PART 1: Levi degenerate CR-manifolds}

641: \bigskip

642: 

643: \KAP{Zwei}{Preliminaries}

644: 

645: In the following let $E$ always be a complex vector space of finite

646: dimension and $M\subset E$ an immersed connected real-analytic

647: submanifold. In most cases $M$ will be locally closed in $E$. Due to

648: the canonical identifications $T_aE=E,$ for every $a\in M$ we consider

649: the tangent space $T_{a}M$ as an $\RR$-linear subspace of $E$. Then

650: $H_{a}M:=T_{a}M\cap iT_{a}M$ is the largest complex linear subspace of

651: $E$ contained in $T_{a}M$. The manifold $M$ is called a {\sl

652: CR-submanifold} if $\dim_{\CC}\!H_{a}M$ does not depend on $a\in

653: M$. This dimension is called the {\sl CR-dimension} of $M$ and

654: $H_{a}M$ is called the {\sl holomorphic tangent space} at $a$, compare

655: \Lit{BERO} as general reference for CR-manifolds. Given a further

656: real-analytic CR-submanifold $M'$ of a complex vector space $E'$, a

657: smooth mapping $g:M\to M'$ is called {\sl CR} if for every $a\in M$

658: the differential $dg_{a}:T_{a}M\to T_{ga}M'$ maps the corresponding

659: holomorphic tangent spaces in a complex-linear way to each

660: other. Keeping in mind the identification $T_aE=E$, a vector field on

661: $M$ is a mapping $f:M\to E$ with $f(a)\in T_{a}M$ for all $a\in

662: M$. For better distinction we also write $\xi=f(z)\dd z$ instead of

663: $f$ and $\xi_{a}$ instead of $f(a)$, compare the convention (2.1) in

664: \Lit{FEKA}.

665: 

666: An {\sl infinitesimal CR-transformation} on $M$ is by definition a

667: real-analytic vector field $f(z)\dd z$ on $M$ such that the

668: corresponding local flow consists of CR-transformations. Let us denote

669: by $\hol(M)$ the space of all such vector fields, which is a real Lie

670: algebra with respect to the usual bracket. For every $f(z)\dd

671: z\in\hol(M)$ and every $a\in M$ there exist an open neighbourhood

672: $U\subset M$ of $a$ with respect to the manifold topology on $M$, an

673: open neighbourhood $W$ of $a$ with respect to $E$ and

674: a holomorphic mapping $h:W\to E$ with $f(z)=h(z)$

675: for all $z\in U\cap W$, compare \Lit{ANHI} or 12.4.22 in

676: \Lit{BERO}.

677: 

678: Further, for every $a\in M$ we denote by $\hol(M,a)$ the Lie algebra

679: of all germs of infinitesimal CR-transformations defined on arbitrary

680: open neighbourhoods of $a$ with respect to the manifold topology of

681: $M$. For simplicity and without essential loss of generality we always

682: assume that the CR-submanifold $M$ is {\sl generic} in $E$, that is,

683: $E=T_{a}M\oplus iT_{a}M$ for all $a\in M$. This assumption allows us to

684: consider $\hol(M,a)$ in a canonical way as a real Lie subalgebra of

685: the complex Lie algebra $\hol(E,a)$, what we always do in the

686: following.  The CR-manifold $M$ is called {\sl holomorphically

687: nondegenerate} at $a$ if $\hol(M,a)$ is totally real in $\hol(E,a)$,

688: that is, if $\hol(M,a)\cap i\,\hol(M,a)=0$ in $\hol(E,a)$. This

689: condition together with the minimality of $M$ at $a$ implies $\dim

690: \hol(M,a)<\infty$, see 12.5.3 in \Lit{BERO}. Here, the CR-submanifold

691: $M\subset E$ is called {\sl minimal} at $a\in M$ if $T_{a}R=T_{a}M$

692: for every submanifold $R\subset M$ with $a\in R$ and $H_{z}M\subset

693: T_{z}R$ for all $z\in R$. By Proposition 15.5.1 in \Lit{BERO}, $\dim

694: \hol(M,a)<\infty$ implies that $M$ is holomorphically nondegenerate at

695: $a$. The CR-manifold $M$ is called {\sl locally homogeneous} at the

696: point $a\in M$ if there exists a Lie subalgebra $\7g\subset\hol(M,a)$

697: of finite dimension such that the canonical evaluation map $\7g\to

698: T_{a}M$ is surjective. In case $M$ is locally homogeneous at $a$ the

699: condition $\dim \hol(M,a)<\infty$ is equivalent to $M$ being

700: holomorphically nondegenerate and minimal at $a$.

701: 

702: By $\aut(M,a):=\{\xi\in\hol(M,a):\xi_{a}=0\}$ we denote the {\sl

703: isotropy subalgebra} at $a\in M$. Clearly, $\aut(M,a)$ has finite

704: codimension in $\hol(M,a)$. Furthermore, we denote by $\Aut(M,a)$ the

705: group of all {\sl germs} of real-analytic CR-isomorphisms $ h:W\to

706: \tilde W$ with $ h(a)=a$, where $W,\tilde W$ are arbitrary open

707: neighbourhoods of $a$ in $M$. It is known that every germ in

708: $\Aut(M,a)$ can be represented by a holomorphic map $U\to E$, where

709: $U$ is an open neighbourhood of $a$ in $E$, compare e.g. 1.7.13 in

710: \Lit{BERO}. Furthermore, $\Aut(M)$ denotes the group of all global

711: real-analytic CR-automorphisms $h:M\to M$ and $\Aut(M)_a$ its isotropy

712: subgroup at $a$. There is a canonical group monomorphism

713: $\Aut(M)_a\hookrightarrow\Aut(M,a)$ as well as an exponential map

714: $\exp:\aut(M,a)\to\Aut(M,a)$ for every $a\in M$.

715: 

716: By $\aff(M)\subset\hol(M)$ we denote the Lie subalgebra of all

717: (complex) affine infinitesimal CR-transformations on $M$. For every

718: $a\in M$ furthermore $\aff(M,a)\subset\hol(M,a)$ is the Lie subalgebra

719: of all affine germs. The canonical embedding

720: $\aff(M)\hookrightarrow\aff(M,a)$ is an isomorphism for every $a\in

721: M$.

722: 

723: Suppose that $g\colon U\to U'$ is a biholomorphic mapping between open

724: subsets $U,U'\subset E$. Then

725: $$

726: g_{*}\big(f(z)\dd z\big)=g'(g^{-1}z)\big(f(g^{-1}z)\big)\,\dd

727: z\Leqno{GP}

728: $$

729: defines a complex Lie algebra isomorphism

730: $g_{*}\colon\hol(U)\to\hol(U')$, where $g'(u)\in\End(E)$ for every

731: $u\in U$ is the derivative of $g$ at $u$. For real-analytic

732: CR-submanifolds $M,M'\subset E$ every CR-isomorphism

733: $g\colon(M,a)\to(M',a')$ of manifold germs induces a Lie algebra

734: isomorphism $g_{*}\colon\7g\to\7g'$, where $\7g\!:=\hol(M,a)$ and

735: $\7g'\!:=\hol(M',a')$. From \Ruf{GP} it is clear that $g_{*}$ extends

736: to a complex Lie algebra isomorphism $\7l\to\7l'$, where the sums

737: $\7l:=\7g\oplus i\7g\subset\hol(E,a)$ and

738: $\7l':=\7g'\oplus i\7g'\subset\hol(E,a')$ are not necessarily direct. In

739: particular, $g\mapsto g_{*}$ defines a group homomorphism

740: $\Aut(M,a)\to\Aut(\7g)$.

741: 

742: \medskip A basic invariant of a CR-manifold is the (vector valued)

743: {\sl Levi form}. Its definitions found in the literature may differ by

744: a constant factor. Here we choose the following one: It is

745: well-known that for every point $a$ in the CR-manifold $M$ there is a

746: well defined alternating $\RR$-bilinear map

747: $$

748: \omega_{a}:H_{a}M\times H_{a}M\;\to\; E/H_{a}M

749: $$

750: satisfying $\omega_{a}(\xi_{a},\zeta_{a})\equiv[\xi,\zeta]_{a}

751: \steil{mod}H_{a}M,$ where $\xi,\zeta$ are arbitrary smooth vector

752: fields on $M$ with $\xi_{z},\zeta_{z}\in H_{z}M$ for all $z\in M$. We

753: define the Levi form

754: $$

755: \5L_{a}:H_{a}M\times H_{a}M\;\to\; E/H_{a}M\Leqno{LE}

756: $$

757: to be twice the sesquilinear part of $\omega_{a}$. By {\sl

758: sesquilinear} we always mean `conjugate linear in the first and

759: complex linear in the second variable', that is,

760: $$

761: \5L_{a}(v,w)=\omega_{a}(v,w)+i\omega_{a}(iv,w)\qquad \hbox{for all }

762: v,w\in H_{a}M\,.

763: $$

764: In particular, the vectors $\5L_{a}(v,v),$ $v\in H_aM,$ are contained

765: in $iT_{a}M/H_{a}M$, which can be identified in a canonical way with

766: the normal space $E/T_{a}M$ to $M\subset E$ at $a$.  The following

767: remark follows immediately from the way the Levi form is defined:

768: 

769: \Remark{RM} Suppose $Z$ is a complex manifold, $\phi:Z\to M$ is a smooth

770: CR-mapping and $a=\phi(c)$ for some $c\in Z$. Then every vector $v\in

771: d\phi_{c}(T_{c}Z)\subset H_{a}M$ satisfies $\5L_{a}(v,v)=0$. In

772: general, $v$ is not contained in the {\sl Levi kernel}

773: $$

774: K_{a}M:=\{v\in H_{a}M:\5L_{a}(v,w)=0\steil{for all}w\in H_aM\}\,.

775: $$

776: \Formend The CR-manifold $M$ is called {\sl Levi nondegenerate} at $a$

777: if $K_{a}M=0$. Generalizing that, the notion of {\sl

778: $k$-nondegeneracy} for $M$ at $a$ has been introduced for every

779: integer $k\ge1$ (see \Lit{BAHR}, \Lit{BERO}). As shown in 11.5.1 of

780: \Lit{BERO} a real-analytic and connected CR-manifold $M$ is

781: holomorphically nondegenerate at $a$ (equivalently: at every $z\in M$)

782: if and only if there exists a $k\ge1$ such that $M$ is

783: $k$-nondegenerate at some point $b\in M$. For $k=1$ this notion is

784: equivalent to $M$ being Levi nondegenerate at $b\in M$.

785: 

786: \bigskip In the second part of our paper we also need a more general

787: notion of a (real-analytic) CR-manifold. This is a connected

788: real-analytic manifold $M$ together with a subbundle $HM\subset TM$

789: (called the `holomorphic subbundle') and a bundle endomorphism $J$ of

790: $HM$ with $J^2= - \id$ such that $(HM,J)$ is involutive, compare 7.4

791: in \Lit{BOGG}. By a theorem in \Lit{ANFR} there exists an embedding

792: $M\into Z$ into a complex manifold $Z$, such that $H_zM$ corresponds

793: to $T_{z}M\cap\,iT_{z}M$, where $T_zZ\to T_zZ$ is simply the

794: multiplication with the imaginary unit $i$ (here the real-analyticity

795: is necessary).  The bundle homomorphism $J:HM\to HM$ is then the

796: restriction of that multiplication with $i$ to $H_{z}M$ for every

797: $z\in M$.  For local considerations one can always assume that $Z$ is

798: (an open subset of) $\CC^n$.

799: 

800: \KAP{Drei}{Tube manifolds}

801: 

802: Let $V$ be a real vector space of finite dimension and $E:=V\,\oplus\,

803: iV$ its complexification. Let furthermore $F\subset V$ be a connected

804: real-analytic submanifold and $M:=F+iV\subset E$ the

805: corresponding {\sl tube manifold.} $M$ is a generic CR-submanifold of

806: $E$, invariant under all translations $z\mapsto z+iv$, $v\in V$. In

807: case $V'$ is another real vector space of finite dimension, $E'$ its

808: complexification, $F'\subset V'$ a real-analytic submanifold and

809: $\phi:V\to V'$ an affine mapping with $\phi(F)\subset F'$, then

810: clearly $\phi$ extends in a unique way to a complex-affine mapping

811: $E\to E'$ with $\phi(M)\subset M'$. However, it should be noted that

812: higher order real-analytic maps $\psi:F\to F'$ also extend locally to

813: holomorphic maps $\psi:U\to E'$, $U$ open in $E$. But in contrast to

814: the affine case we have in general $\psi(M\cap U)\not\subset M'$. We

815: may therefore ask how the CR-structure of $M$ is related to the real

816: affine structure of the submanifold $F\subset V$.

817: 

818: For every $a\in F$ let $T_{a}F\subset V$ be the {\sl tangent space}

819: and $N_{a}F:=V/T_{a}F$ the {\sl normal space} to $F$ at $a$. Then

820: $T_{a}M=T_{a}F\oplus iV$ for the corresponding tube manifold $M$, and

821: $N_{a}F$ can be canonically identified with the normal space

822: $N_{a}M=E/T_{a}M$ of $M$ in $E$. Define the map $\ell_{a}:T_{a}F\times

823: T_{a}F\to N_{a}F$ in the following way: For $v,w\in T_{a}F$ choose a

824: smooth map $f:V\to V$ with $f(a)=v$ and $f(x)\in T_{x}F$ for all $x\in

825: F$ (actually it suffices to choose such an $f$ only in a small

826: neighborhood of $a\,$). Then put

827: $$

828: \ell_{a}(v,w):=f'(a)(w)\;\mod\;T_{a}F\,,\Leqno{EL}

829: $$

830: where the linear operator $f'(a)\in\End(V)$ is the derivative of $f$

831: at $a$. One shows that $\ell_a$ does not depend on the choice of $f$

832: and is a symmetric bilinear map. We mention that if $V$ is provided

833: with a flat Riemannian metric and $N_aF$ is identified with

834: $T_aF^\perp$ then $\ell$ is nothing but the second fundamental form of

835: $F$ (see the subsection II.3.3 in \Lit{SAKA}). The form $\ell_{a}$ can

836: also be read off from local equations for $F$, more precisely, suppose

837: that $U\subset V$ is an open subset, $W$ is a real vector space and

838: $h:U\to W$ is a real-analytic submersion with $F=h^{-1}(0)$. Then the

839: derivative $h'(a):V\to W$ induces a linear isomorphism $N_{a}F\cong W$

840: and modulo this identification $\ell_{a}$ is nothing but the second

841: derivative $h''(a):V\times V\to W$ at $a$, restricted to $T_{a}F\times

842: T_{a}F$. By

843: $$

844: K_{a}F:=\{w\in T_{a}F:\ell_{a}(v,w)=0\steil{for all}v\in T_{a}F\}

845: $$

846: we denote the {\sl kernel} of $\ell_{a}$. The manifold $F$ is called

847: (affinely) {\sl nondegenerate} at $a$ if $K_{a}F=0$ holds. The

848: following statement follows directly from the definition of $\ell_a$:

849: 

850: \Lemma{CR} Suppose that $\phi\in\End(V)$ satisfies $\phi(x)\in T_{x}F$

851: for all $x\in F$. Then $\phi(a)\in K_{a}F$ if and only if

852: $\phi\big(T_{a}F\big)\subset T_{a}F$.\Formend

853: 

854: \noindent Lemma \ruf{CR} applies in particular for $\phi=\id$ in case

855: $F$ is a {\sl cone,} that is, $rF=F$ for all real $r>0$. More

856: generally, we call the submanifold $F\subset V$ {\sl conical} if $x\in

857: T_{x}F$ for all $x\in F$. Then $\RR a\subset K_{a}F$ holds for all

858: $a\in F$.

859: 

860: \medskip

861: 

862: In the remaining part of this section we explain how the CR-structure

863: of the tube manifold $M$ is related to the real objects $\ell_a,\,

864: TF,\, KF$ and $ K^r\!F$, to be defined below, which depend only on the

865: affine geometry of $F$. In general it needs some effort to check

866: whether a given CR-manifold $M$ is $k$-nondegenerate at a point $a\in

867: M$ (in sense of \Lit{BHUR}).  For affinely homogeneous tube manifolds,

868: however, there are simple criteria, see Propositions \ruf{UR} and

869: \ruf{US}. We start with some preparations. For every $a\in F\subset M$

870: $$

871: H_{a}M=T_{a}F\oplus iT_{a}F\,\subset\, E

872: \Leqno{HM}

873: $$

874: is the holomorphic

875: tangent space at $a$, and $E/H_{a}M$ can be canonically identified

876: with $N_{a}F\oplus iN_{a}F$. It is easily seen that the Levi form

877: $\5L_{a}$ of $M$ at $a$, compare \Ruf{LE}, is nothing but the

878: sesquilinear extension of the form $\ell_{a}$ from $T_{a}F\times

879: T_{a}F$ to $H_{a}M\times H_{a}M$. In particular,

880: $$

881: K_{a}M=K_{a}F\oplus iK_{a}F

882: $$

883: is the {\sl Levi kernel} of $M$ at $a$. In case the dimension of

884: $K_{a}F$ does not depend on $a\in F$ these spaces form a subbundle

885: $KF\subset TF$. In that case to every $v\in K_{a}F$ there exists a

886: smooth function $f:V\to V$ with $f(a)=v$ and $f(x)\in K_{x}F$ for all

887: $x\in F$, i.e., the tangent vector $v$ extends to a smooth section in

888: $KF$. In any case, let us define inductively linear subspaces

889: $K_{a}^{r}F$ of $T_{a}F$ as follows:

890: 

891: \Definition{HN} For every real-analytic submanifold $F\subset V$,

892: every $a\in F$ and every $r\in\NN$ put \0 $K_{a}^{0}F:=T_{a}F$ and

893: define \1 $K_{a}^{r+1}\!F$ to be the space of all vectors $v\in

894: K_{a}^{r}F$ such that there is a smooth mapping $f:V\to V$ with

895: $f'(a)(T_{a}F)\subset K_{a}^{r}F$, $f(a)=v$ and $f(x)\in K_{x}^{r}F$

896: for all $x\in F$.

897: 

898: \medskip\noindent It is clear that $K_{a}^{1}F=K_{a}F$ holds. Let us

899: call $F$ of {\sl uniform degeneracy} or {\sl uniformly degenerate} if

900: for every $r\in\NN$ the dimension of $K_{a}^{r}F$ does not depend on

901: $a\in F$. In that case it can be shown that for every $v\in

902: K_{a}^{r}F$ the outcome of the condition $f'(a)(T_{a}F)\subset

903: K_{a}^{r}F$ in (ii) does not depend on the choice of the smooth

904: mapping $f:V\to V$ satisfying $f(a)=v$ and $f(x)\in K_{x}^{r}F$ for

905: all $x\in F$. For instance, $F$ is of uniform degeneracy if $F$ is

906: locally affinely homogeneous, that is, if there exists a Lie algebra

907: $\7a$ of affine vector fields on $V$ such that every $\xi\in\7a$ is

908: tangent to $F$ and such that the canonical evaluation map $\7a\to

909: T_{a}F$ is surjective for every $a\in F$. Clearly, if $F$ is locally

910: affinely homogeneous in the above sense then the corresponding tube

911: manifold $M=F+iV$ is locally homogeneous as CR-manifold.

912: 

913: \medskip Recall our convention that every smooth map $f:V\to V$ is

914: considered as the smooth vector field $\xi=f(x)\dd x$ on $V$. Our

915: computations below are considerably simplified by the obvious fact

916: that every smooth vector field $\xi$ on V has a unique smooth

917: extension to $E$ that is invariant under all translations $z\mapsto

918: z+iv$, $v\in V$. In case $\xi$ is tangent to $F\subset V$ the

919: extension satisfies $\xi_{z}\in H_{z}M$ for all $z\in M$.

920: 

921: In case the submanifold $F\subset V$ is uniformly degenerate in a

922: neighbourhood of $a\in F$ we call $F\,$ {\sl affinely

923: $k$-nondegenerate} at $a$ if $K_{a}^{k}F=0$ and $k$ is minimal with

924: respect to this property. It can be seen that `affine

925: $k$-nondegeneracy' is invariant under affine coordinate changes. As a

926: consequence of \Lit{KAZT}, compare the last 5 lines in the Appendix

927: therein, we state:

928: 

929: \Proposition{UR} Suppose that $F$ is uniformly degenerate in a

930: neighbourhood of $a\in F$. Then the corresponding tube manifold

931: $M=F+iV$ is $k$-nondegenerate as CR-manifold at $a\in M$ if and

932: only if $F$ is affinely $k$-nondegenerate at $a$.\Formend

933: 

934: \Corollary{AF} Suppose $\dim F \ge2$ and $K_{x}F=\RR x$ for all $x\in

935: F$. Then $F$ is affinely $2$-nondegenerate at every point.

936: 

937: \Proof The map $f=\id$ has the property $f(x)\in K_xF$ for every $x\in

938: F$. Hence, the relation $f'(x)(T_{x}F)=T_{x}F\not\subset K_{x}F$

939: implies $x\notin K_{x}^{2}F$ and thus $K_{x}^{2}F=0$ as well as

940: $x\ne0$. In particular, $F$ is uniformly degenerate.\qed

941: 

942: \medskip For locally affinely homogeneous submanifolds $F\subset V$

943: the spaces $K_{a}^{r}F$ can easily be characterized. For each affine

944: vector field $\xi=h(x)\dd x$ on $V$ denote by

945: $\xi^{\lin}:=h-h(0)\in\End(V)$ the {\sl linear part} of $\xi$.

946: 

947: \Proposition{US} Suppose that $\7A$ is a linear space of affine vector

948: fields on $V$ such that every $\xi\in\7A$ is tangent to $F$ and the

949: canonical evaluation mapping $\7A\to T_{a}F$ is a linear

950: isomorphism. Then, given any $r\in\NN$, the vector $v\in K_{a}^{r}F$

951: is in $K_{a}^{r+1}\!F$ if and only if $\xi^{\lin}(v)\in K_{a}^{r}F$

952: for every $\xi\in\7A$.

953: 

954: \Proof By the implicit function theorem, there exist open

955: neighbourhoods $Y$ of $0\in\7A$ and $X$ of $a\in M$ such that

956: $g(y):=\exp(y)a$ defines a diffeomorphism $g:Y\to X$. Define the

957: smooth map $f:X\to V$ by $f(x)=\mu_{y}(v)$, where $\mu_{y}$ for

958: $y:=g^{-1}(x)$ is the linear part of the affine transformation

959: $\exp(y)$. Then $f(a)=v$ and $f(x)\in K_{x}^{r}F$ for every $x\in

960: X$. A simple computation shows that\medskip

961: \centerline{$f'(a)\big(g'(0)\xi\big)=\xi^{\lin}(v)\steil{for every}

962: \xi\in\7A$.}  \nline In view of Definition \ruf{HN}.ii this identity

963: implies the claim. \qed

964: 

965: \medskip It is easily seen that a necessary condition for $M$ being

966: minimal as CR-manifold is that $F$ is not contained in an affine

967: hyperplane of $V$. For later use in Proposition \ruf{PS} we state the

968: following sufficient condition.

969: 

970: \Proposition{PY} Suppose that $\7A$ has the same properties as in

971: Proposition \ruf{US} and denote by $\Lambda\subset\End(V)$ the {\rm

972: associative} real subalgebra generated by

973: $\{\xi^{\lin}:\xi\in\7A\}$. Then the tube manifold $M=F+iV$ is

974: minimal at $a$ if $V$ is the linear span of all vectors $\lambda(v)$

975: with $v\in T_{a}F$ and $\lambda\in\Lambda$.

976: 

977: \Proof Without loss of generality we assume that the canonical

978: evaluation mapping $\7A\to T_{x}F$ is a linear isomorphism for every

979: $x\in F$. We also assume that $V$ is the linear span of all

980: $\lambda(v)$ as above. Define inductively for every integer $r\ge1$

981: the subbundle $H^{r}M\subset TM$ in the following way: $H^{1}M:=HM$

982: and every $H^{r+1}_{z}M$, $z\in M$, is the linear span of $H^{r}_{z}M$

983: together with all vectors $[\xi,\eta]_{z}$, where $\xi,\eta$ are

984: arbitrary smooth sections in $H^{r}M$ over $M$. For the proof it is

985: enough to show

986: $T_{a}M=H^{\infty}_{a}M:=\bigcup_{r\ge1}\!T_{a}M$.\nline From

987: $T_{a}F\oplus iT_{a}F\,\subset\, H_{a}^{r}M\,\subset\, T_{a}F\oplus

988: iV$ we see that for every $1\le r\le\infty$ there is a unique linear

989: subspace $H^{r}_{a}F\subset V$ with $H^{r}_{a}M=T_{a}F\oplus

990: iH^{r}_{a}F$ and $H^{1}_{a}F=T_{a}F$. Therefore it is enough to show

991: $H^{\infty}_{a}F=V$. We claim that $\xi^{\lin}(H^{r}_{a}F)\subset

992: H^{r+1}_{a}F$ holds for all $\xi\in\7A$. To see this fix an arbitrary

993: $w\in H^{r}_{a}F$ and an arbitrary vector field $\xi\in\7A$. Choose a

994: smooth section $\eta$ over $F$ in the bundle $iH^{r}_{a}F$ with

995: $\eta_{a}=iw$ and extend $\eta$ as well as $\xi$ in the unique way to

996: smooth vector fields on $M$ that are invariant under all translations

997: $z\mapsto z+iv$ with $v\in V$. Then $\xi,\eta$ are sections in

998: $H^{r}M$, and $[\xi,\eta]_{a}=i\xi^{\lin}(w)\in H^{r+1}_{a}M$ implies

999: $\xi^{\lin}(w)\in H^{r+1}_{a}F$ as required. Now define inductively

1000: the linear subspaces $W^{r}\subset V$ by $W^{1}:=H^{1}_{a}F=T_{a}F$

1001: and $W^{r+1}$ as the linear span of $W^{r}$ together with all

1002: $\xi^{\lin}(W^{r})$, $\xi\in\7A$. Then $V=\bigcup_{r\ge1}W^{r}$ by

1003: assumption and $W^{r}\subset H^{r}_{a}F$ by induction gives $V\subset

1004: H^{\infty}_{a}M\subset V$ as desired.\qed

1005: 

1006: \medskip

1007: 

1008: \Lemma{ZI} Suppose that $F\subset V$ is a submanifold such that for

1009: every $c\in V$ with $c\ne0$ there exists a linear transformation

1010: $g\in\GL(V)$ with $g(F)=F$ and $g(c)\ne c$ (this

1011: condition is automatically satisfied if $F$ is a cone). Then for

1012: $M=F+iV$ the CR-automorphism group $\Aut(M)$ has trivial center.

1013: 

1014: 

1015: \Proof Let an element in the center of $\Aut(M)$ be given and let

1016: $h:U\to E$ be its holomorphic extension to an appropriate connected

1017: open neighbourhood $U$ of $M$. Since $h$ commutes with every

1018: translation $z\mapsto z+iv$, $v\in V$, it is a translation itself:

1019: Indeed, for $a\in F$ fixed and $c:=h(a)-a$ the translation

1020: $\tau(z):=z+c$ coincides with $h$ on $a+iV$ and hence on $U$ by the

1021: identity principle. For every $g\in\GL(V)\cap\Aut(M)$ the identity

1022: $gh=hg$ implies $g(c)=c$. This forces $c=0$ by our

1023: assumption, that is, $h(z)\equiv z$.\qed

1024: 

1025: \Proposition{UQ} Suppose that the homogeneous CR-manifold $M$ is

1026: simply connected and that $\Aut(M)$ has trivial center. In case the

1027: stability group $\Aut(M,a)$ is trivial for some (and hence every)

1028: $a\in M$, the following properties hold: \0 Let $M'$ be an arbitrary

1029: homogeneous CR-manifold and $D\subset M$, $D'\subset M'$ non-empty

1030: domains. Then every real-analytic CR-isomorphism $h:D\to D'$ extends

1031: to a real-analytic CR-\-isomorphism $M\to M'$. \1 Let $M'$ be an

1032: arbitrary locally homogeneous CR-manifold and $D'\subset M'$ a domain

1033: that is CR-equivalent to $M$. Then $D'=M'$.

1034: 

1035: \Proof ad (i): Fix a point $a\in D$. To every $g\in G:=\Aut(M)$

1036: with $g(a)\in D$ there exists a transformation $g'\in G':=\Aut(M')$

1037: with $hg(a)=g'h(a)$. Because of $\Aut(M',a')=\{\id\}$ the

1038: transformation $g'$ is uniquely determined by $g$ and satisfies

1039: $hg=g'h$ in a neighbourhood of $a$. Since the Lie group $G$ is simply

1040: connected $g\mapsto g'$ extends to a group homomorphism $G\to G'$ and

1041: $h$ extends to a CR-covering map $h:M\to M'$. The deck transformation

1042: group $\Gamma:=\{g\in G:gh=h\}$ is in the center of $G$ and hence is

1043: trivial by assumption. Therefore, $h:M\to M'$ is a

1044: CR-isomorphism. \nline ad (ii): The proof is essentially the same

1045: as for Proposition 6.3 in \Lit{FEKA}.\qed

1046: 

1047: The condition `locally homogeneous' in Proposition \ruf{UQ}.ii cannot

1048: be omitted. A counterexample is given for every integer $k\ge3$

1049: by the tube $M'\subset\CC^{3}$ over the cone

1050: $$

1051: F':=\{x\in\RR^{3}:x_{2}^{k}=x_{1}^{k-1}x_{3},\;x_{1}^{2}+

1052: x_{2}^{2}>0\}\,.

1053: $$ Then with $\;\RRp:=e^{\RR}$ the tube $M$ over

1054: $F:=F'\cap(\RRp)^{3}$ is the Example \ruf{EX} below for $\theta=k$,

1055: and $M,M'$ satisfy for $D'=M$ the assumption of Proposition

1056: \ruf{UQ}.ii.

1057: 

1058: \ifarx\vfill\eject\fi

1059: 

1060: \KAP{Vier}{Tube manifolds over cones}

1061: 

1062: In this section we always assume that the submanifold $F\subset V$ is

1063: conical (that is, $x\in T_xF$ for every $x\in F$) and that

1064: $a\in F$ is a given point. Then, for $M:=F+iV$, the Lie algebra

1065: $\7g:=\hol(M,a)$ contains the Euler vector field $\delta:=z\dd z$.

1066: Denote by $\7P$ the complex Lie algebra of all polynomial holomorphic

1067: vector fields $f(z)\dd z$ on $E$, that is, $f:E\to E$ is a polynomial

1068: map. Then $\7P$ has the $\ZZ$-grading

1069: $$

1070: \7P=\bigoplus_{k\in\ZZ}\7P_{k}\,,\qquad[\7P_{k},\7P_{l}]\subset

1071: \7P_{k+l}\,,\Leqno{HJ}

1072: $$

1073: where $\7P_{k}$ is the $k$-eigenspace of $\ad(\delta)$ in $\7P$. Then

1074: $\7P_{k}$ is the subspace of all $({k+1})$-homogeneous vector fields

1075: in $\7P$ if $k\ge-1$ and is $0$ otherwise. Define

1076: $\7g_k:=\7g\cap\7P_k.$ Clearly, $\bigoplus_{k\in \ZZ}\7g_k$ is a

1077: graded, in general proper subalgebra of $\7g$, and

1078: $\aff(M,a)=\7g_{-1}\oplus\7g_{0}$.

1079: 

1080: \Proposition{DC} Retaining the above notation, suppose that

1081: $\7g=\hol(M,a)$ has finite dimension. Then \0 $\7g\subset\7P$, that

1082: is, every $f(z)\dd z\in\7g$ is a polynomial vector field on $E=V\oplus

1083: iV$. Furthermore, $f(iV)\subset iV$.\vskip3pt \1 $\displaystyle

1084: \7g=\bigoplus\limits_{k\ge-1}\7g_{k}\,,\quad[\7g_{k},\7g_{l}]\subset

1085: \7g_{k+l}\Steil{and}\7g_{-1}=\{iv\dd z: v\in V\}\;$.\vskip2pt \1 For

1086: every $z\in M$ the canonical map $\hol(M)\to\hol(M,z)$ is a Lie

1087: algebra isomorphism. \1 $\7g_{k}=0$ for some $k\in\NN$ implies

1088: $\7g_{j}=0$ for every $j\ge k$.

1089: 

1090: \Proof Consider $\7l:=\7g\oplus i\7g\subset\hol(E,a)$, which

1091: contains the vector field $\eta:=(z-a)\dd z$. We first show

1092: $\7l\subset \7P$: Fix an arbitrary $\xi:=f(z) \dd z\in\7l$. Then in a

1093: certain neighbourhood of $a\in E$ there exists a unique expansion

1094: $\xi=\sum_{k\in\NN}\xi_{k}$, where $\xi_{k}=p_{k}(z-a)\dd z$ for a

1095: $k$-homogeneous polynomial map $p_{k}:E\to E$. It is easily verified

1096: that the vector field $\ad(\eta)\xi\in\7l$ has the expansion

1097: $\ad(\eta)\xi=\sum_{k\in\NN}(k{-}1)\xi_{k}$. Now assume that for

1098: $d:=\dim \7l $ there exist indices $k_{0}<k_{1}<\dots<k_{d}$ such that

1099: $\xi_{k_{l}}\ne0$ for $0\le l\le d$. Since the Vandermonde matrix

1100: $\big((k_{l}-1)^{j}\big)$ in non-singular, we get that the vector

1101: fields $(\ad\eta)^{j}\xi=\sum_{k\in\NN}(k{-}1)^{j}\xi_{k}$, $0\le j\le

1102: d$, are linearly independent in $\7l$, a contradiction. This implies

1103: $\xi\in\7P$ as claimed.\nline Since $\7g\subset\7P$ has finite

1104: dimension, every $\eta\in\7g$ is a finite sum

1105: $\eta=\sum_{k=-1}^{m}\eta_{k}$ with $\eta_{k}\in\7P_k$ and $m\in\NN$

1106: not depending on $\eta$. For every polynomial $p\in\RR[X]$ then

1107: $p(\ad\delta)\eta=\sum_{k=-1}^{m}p(k)\eta_{k}$ shows

1108: $\eta_{k}\in\7g_{k}$ for all $k$, that is, $\7g=\bigoplus\7g_{k}$. The

1109: identity $\7g_{-1}=\{iv\dd z: v\in V\}$ follows from the fact that

1110: $\7g_{-1}$ is totally real in $\7P_{-1}$and this implies $f(iV)\subset

1111: iV$ for all $f(z)\dd z\in\7g_{k}$ by

1112: $[\7g_{-1},\7g_{k}]\subset\7g_{k-1}$ and induction on $k$. For the

1113: proof of the last claim assume $\7g_{k}=0$ for some $k\ge0$. Then

1114: $[\7P_{-1},\7g_{k+1}]\subset (\7g_{k}\oplus i\7g_{k})=0$ implies

1115: $\7g_{k+1}\subset\7g_{-1}$ and hence $\7g_{k+1}=0$. \qed

1116: 

1117: \Corollary{ZU} In case $\7g=\hol(M,a)$ has finite dimension the

1118: CR-manifold $M=F+iV$ is locally homogeneous at $a$ if and only

1119: if $F$ is locally linearly homogeneous at $a\in F$.

1120: 

1121: \Proof From Proposition \ruf{DC} follows that

1122: $W:=\{\xi_{a}:\xi\in\bigoplus_{k\in\NN}\7g_{2k}\}$ is a subspace of

1123: $V$ while $\{\xi_{a}:\xi\in\bigoplus_{k\in\NN}\7g_{2k-1}\}=iV.$ Hence,

1124: $M$ is locally homogeneous at $a$ if and only if $W=T_{a}F$. But for

1125: every $k\in\NN$ and every $\xi\in\7g_{2k}$ the vector field

1126: $\eta:=(\ad ia\dd z)^{2k}\xi$ is in $\7g_{0}$ and satisfies

1127: $\eta_{a}=(-1)^{k}(2k+1)!\,\xi_{a}$, that is,

1128: $W=\{\xi_{a}:\xi\in\7g_{0}\}$.\qed

1129: 

1130: Notice that the conclusion $\7g\subset\7P$ together with an eigenspace

1131: decomposition as in Proposition \ruf{DC} can for $V=\RR^{n}$ be

1132: obtained in the same way if instead of `$F$ conical' it is only

1133: assumed for $M=F+i\RR^{n}$ that $\7g=\hol(M,a)$ contains a

1134: vector field $\alpha_{1}z_{1}\dd{z_{1}}+

1135: \alpha_{2}z_{2}\dd{z_{2}}+\dots+\alpha_{n}z_{n}\dd{z_{n}}$ with

1136: $\alpha_{k}>0$ for all $k$, compare e.g. \Ruf{RU}.

1137: 

1138: \smallskip

1139: 

1140: 

1141: \Proposition{ZB} Assume that $\7g:=\hol(M,a)$ has finite dimension and

1142: that $F'\subset V$ is a further conical submanifold with tube manifold

1143: $M'=F'+iV$ and $\7g':=\hol(M',a')$ for some $a'\in F'$. Assume

1144: that the CR-germs $(M,a)$ and $(M',a')$ are isomorphic and let

1145: $g,\tilde g:(M,a)\to(M',a')$ be arbitrary CR-isomorphisms. Then \0

1146: $\dim\7g_{k}=\dim\7g_{k}'$ for all $k\in\ZZ$, where

1147: $\7g^{}_{k},\7g'_{k}$ are given by the decomposition \ruf{DC}.ii.  \1

1148: $g$ is represented by a rational transformation on $E$.  \1 In case

1149: $\7g_{1}=0$, $\,g$ is represented by a linear transformation in

1150: $\GL(V)\subset\GL(E)$ mapping every $K_{a}^{r}F$ onto

1151: $K_{a'}^{r}F'$. \1 $g=\tilde g$ if and only if $g,\tilde g$ have the

1152: same $d$-jet at $a$, where $d:=\min\{k\in\NN:\7g_k=0\}$.

1153: 

1154: \Proof Let $\7l:=\7g\oplus i\7g$ and $\7l_{k}:=\7l\cap\7P_{k}$ for all

1155: $k$. The Lie algebra automorphism $\Psi:=\exp(\ad a\dd z)$ of $\7l$

1156: maps every $f(z)\dd z$ to $f(z+a)\dd z$. For every $k$ denote by

1157: $\7l^{k}\subset\7l$ the subspace of all vector fields that vanish of

1158: order at least $k{+}1$ at $a$. Then $\Psi(\7l^{k})=\bigoplus_{j\ge

1159: k}\7l_{j}$ implies $\dim\7l_{k}=\dim\7l^{k}/\7l^{k+1}$. As a

1160: consequence, $\dim_{\RR}\7g_{k}=\dim_{\CC}\!\7l_{k}$ is a CR-invariant

1161: of the germ $(M,a)$ for every $k$.\nline For the proof of (ii), (iii)

1162: put $\7l':=\7g'+i\7g'$ and extend $g$ to a biholomorphic mapping

1163: $g:U\to U'$ with $g(a)=a'$ and $g(U\cap M)=U'\cap M'$ for suitable

1164: connected open neighbourhoods $U,U'$ of $a,a'\in E$. Consider

1165: the induced Lie algebra isomorphism $ g_{*}:\7l\to\7l'$, compare

1166: \Ruf{GP}. Its inverse $\Theta:=g_{*}^{-1}$ is given by

1167: $$

1168: \Theta\big(f(z)\dd z\re1\big)=g'(z)^{-1}\!f\big(g(z)\big)\re2\dd

1169: z\,.\Leqno{ZA}

1170: $$

1171: Since $\7l$ consists of polynomial vector fields (Proposition

1172: \ruf{DC}) there exist {\sl polynomial} maps $p:E\to E$ and $

1173: q:E\to\End(E)$ such that

1174: $$

1175: \Theta\big(z\dd z\big)=p(z)\dd z\Steil{and}\Theta\big(e\dd

1176: z\big)=\big(q(z)e\big)\dd z

1177: $$

1178: for all $e\in E$. Then \Ruf{ZA} implies $g'(z)^{-1}=q(z)$ and

1179: $g'(z)^{-1}g(z)=p(z)$, that is,

1180: $$

1181: g(z)=q(z)^{-1}p(z)\Leqno{ZV}

1182: $$

1183: in a neighbourhood of $a\in E$ and, in particular, $g$ is

1184: rational.\nline Now suppose $\7g_{1}=0$. Then also

1185: $\7g_{k}=\7g'_{k}=0$ for all $k\ge1$ by \ruf{DC}.iv and (i).  Clearly

1186: $\Theta(\7l'_{a'})=\7l_{a}$, where $\7l_{a}:=\{\xi\in\7l:\xi_{a}=0\}$

1187: and similarly $\7l'_{a'}\subset\7l'$ are the isotropy subalgebras at

1188: $a,a'$. Also, $\delta_{a}:=(z-a)\dd z$ is the unique element in

1189: $\7l_{a}$ such that $\ad(\delta_{a})$ induces the negative identity on

1190: the factor space $\7l/\7l_{a}$. Since $\delta_{a'}$ has the same

1191: uniqueness property for $\7l'_{a'}$ we must have

1192: $\Theta(\delta_{a'})=\delta_{a}$.  Since $\delta=z\dd z$ is the

1193: $\7g$-component of $\delta_{a}$ as well as of $\delta_{a'}$ in

1194: $\7g\oplus i\7g$ we actually get $\Theta(\delta)=\delta$, that is,

1195: $p(z)\equiv z$.  Also $\7l_{-1}$ is $\Theta$-invariant, implying that

1196: $q$ is constant. Therefore $g=q(a)^{-1}$.  \nline For the proof of

1197: (iv) we assume without loss of generality that $M=M'$, $a=a'$ and that

1198: $g,\id\in\Aut(M,a)$ have the same $d$-jet at $a$. This implies that

1199: $(g(z)-z)$ vanishes of order $>d$ and $(g'(z)-\id)$ vanishes of order

1200: $\ge d$ at $a$. Therefore also $(g'(z)^{-1}-\id)=(q(z)-\id)$ vanishes

1201: of order $\ge d$ at $a$. Since $q$ is a polynomial of degree $\le d$,

1202: there is a $d$-homogeneous polynomial $s:E\to\End(V)$ with

1203: $q(z)=\id+s(z-a)$. Consider the vector field $\eta:=(z-a)\dd z\in\7l$

1204: and define the holomorphic mappings $h,r:U\to E\,$ by

1205: $$

1206: h(z):=q(z)\big(g(z)-a\big)=(z-a)+r(z)\,.

1207: $$

1208: Then $\Theta\big(\eta)=h(z)\dd z\;\in\;\7g$ shows that $h$ and $r$ are

1209: polynomials of degree $\le d$. But

1210: $$

1211: r(z)=\big(g(z)-z\big)+s(z-a)\big(g(z)-a\big)

1212: $$

1213: vanishes of order $>d$ at $a$, that is, $r=0$ and

1214: $\Theta(\eta)=\eta$. This implies $\Theta(\7l_{-1})=\7l_{-1}$ since

1215: $\7l_{-1}$ is the $(-1)$-eigenspace of $\ad(\eta)$. Therefore $g$ is

1216: an affine transformation on $E$. From $g(a)=a$ and $g'(a)=\id$ we

1217: finally get $g=\id$.\qed

1218: 

1219: \Corollary{PL} Let $M:=F+iV$, $M':=F'+iV$ with conical

1220: submanifolds $F,F'\subset V$ and let $a\in F$, $a'\in F'$ be arbitrary

1221: points. Assume furthermore that $\hol(M,a)=\aff(M,a)$ holds. Then the

1222: following conditions are equivalent: \0 The manifold germs

1223: $(M,a),\,(M',a')$ are CR-equivalent. \1 The manifold germs

1224: $(M,a),\,(M',a')$ are affinely equivalent. \1 The manifold germs

1225: $(F,a),\,(F',a')$ are linearly equivalent.\Formend

1226: 

1227: 

1228: 

1229: \bigskip\noindent Recall that $\aut(M,a)\subset\7g=\hol(M,a)$ is

1230: defined as the isotropy subalgebra at $a$ and $\Aut(M,a)$ is the

1231: CR-automorphism group of the manifold germ $(M,a)$, also called the

1232: {\sl stability group} at $a\in M$.

1233: 

1234: \Proposition{LO} Let $F\subset V$ be conical and $M=F+iV$. In

1235: case $\7g=\hol(M,a)$ has finite dimension, the following conditions

1236: are equivalent: \0 $\7g_{1}=0$. \1 $\7g=\aff(M,a)$. \1 The tangential

1237: representation $\;g\mapsto g'(a)$ induces a group monomorphism

1238: $\Aut(M,a)\;\hookrightarrow\;\GL(V)$.  \smallskip\noindent Each of

1239: these conditions is satisfied if $\aut(M,a)=0$.

1240: 

1241: \Proof Let $\7l:=\7g\oplus i\7g\subset\hol(E,a)$ and

1242: $\7l_{k}:=\7g_{k}\oplus i\7g_{k}$ for all $k$. \nline\To12 Follows

1243: from the last claim in Proposition \ruf{DC}. \nline\To23 By

1244: Proposition \ruf{ZB}.iii every $g\in\Aut(M,a)$ is represented by a

1245: linear transformation on $V$. \nline\To31 Let $\xi\in\7g_{1}$ be an

1246: arbitrary vector field. Then there exists a unique symmetric bilinear

1247: map $b:E\times E\to E$ with $\xi=b(z,z)\dd z$. Now $(\ad ia\dd

1248: z)^{2}\xi=-2b(a,a)\dd z\in\7g$, that is, $\eta:=h(z)\dd z$ is in

1249: $\aut(M,a)$, where $h(z):=b(z,z)-b(a,a)$. For every $t\in\RR$

1250: therefore the transformation $\psi_{t}:=\exp(t\eta)\in\Aut(M,a)$ has

1251: derivative $\psi_{t}'(a)=\exp(th'(a))\in\GL(E)$ in $a$. But

1252: $\psi_{t}'(a)\in\GL(V)$ by \ruf{ZB}.iii and thus $2b(a,v)=h'(a)v\in V$

1253: for all $v\in V$. On the other hand $b(a,v)\in iV$ by Lemma \ruf{ZI},

1254: implying $\psi_{t}'(a)=\id$ for all $t\in\RR$. By the injectivity of

1255: the tangential representation therefore $\eta_{t}$ does not depend on

1256: $t$ and we get $\xi=0$. This proves (i) and thus the equivalence of

1257: (i) -- (iii).\nline Suppose $\aut(M,a)=0$ and that there exists a

1258: non-zero vector field $\xi\in\7g_{1}$. Then $\xi_{a}\in iV$ and there

1259: exists an $\eta\in\7g_{-1}$ with $\xi-\eta\in\aut(M,a)$, a

1260: contradiction.\qed

1261: 

1262: \Remark{} Notice that the condition (iii) in Proposition \ruf{LO}

1263: states that the tangential representation takes its values in the

1264: subgroup $\GL(V)\subset\GL(E)$. In general, the tangential

1265: representation is not injective and also takes values outside

1266: $\GL(V)$. The tube $\5M$ over the future light cone can serve as a

1267: counterexample for both of these phenomena.\Formend

1268: 

1269: 

1270: \Proposition{PS} Suppose that $M=F+iV$, $F\subset V$ a conical

1271: submanifold, is locally homogeneous and that

1272: $\7g=\7g_{-1}\oplus\7g_{0}$ for $\7g=\hol(M,a)$. Then the tangential

1273: representation at $a$ induces a group isomorphism

1274: $$

1275: \Aut(M,a)\;\cong\;\{g\in\GL(V):g\7g_{0}g^{-1}=\7g_{0} \steil{and}g(a)=

1276: a\}\,,

1277: $$

1278: where $\7g_{0}$ is considered in the canonical way as linear subspace

1279: of $\,\End(V)$.

1280: 

1281: \Proof The assumptions imply that $H(a)\cap F$ is a neighbourhood of

1282: $a$ in $F$ for $H:=\exp(\7g_{0})\subset\GL(V)$. Let $g\in\GL(V)$ be an

1283: arbitrary linear transformation with $g(a)=a$ and

1284: $g\7g_{0}g^{-1}=\7g_{0}$. Then $gHg^{-1}=H$ and hence

1285: $gH(a)=Hg(a)=H(a)$, that is $g\in\Aut(M,a)$. Conversely, by \ruf{ZB}.iii

1286: every element of $\Aut(M,a)$ can be represented by some

1287: $g\in\GL(V)$ with $g(a)=a$. The Lie algebra automorphism

1288: $\Theta=g_{*}$ of $\7g$ leaves $\delta$ and thus also $\7g_{0}$

1289: invariant. As a consequence, $\Theta(\phi)=g\phi g^{-1}$ for every

1290: $\phi\in\7g_{0}$, that is, $g\7g_{0}g^{-1}=\7g_{0}$.\qed

1291: 

1292: 

1293: \medskip The real Lie algebra structure of $\hol(M,a)$ is a

1294: CR-invariant for the manifold germ $(M,a)$. For certain classes of

1295: conical tube manifolds this gives a complete invariant:

1296: 

1297: \Proposition{PQ} Let $F,F'\subset V$ be conical submanifolds for which

1298: the corresponding tubes $M:=F+iV$, $M':=F'+iV$ are locally

1299: homogeneous CR-manifolds. Let $a\in F$, $a'\in F'$ be arbitrary points

1300: and assume that for $\7g:=\hol(M,a)$ the spaces $\7g_{k}$ occurring in

1301: the gradation \ruf{DC}.ii satisfy $\7g_{k}=[\7g_{0},\7g_{0}]=0$. Then

1302: the following conditions are equivalent: \0 The germs

1303: $(M,a),\,(M',a')$ are CR-equivalent. \1 $\hol(M,a)$, $\,\hol(M',a')$

1304: are isomorphic as real Lie algebras.

1305: 

1306: \Proof Only the implication \To21 is not obvious. Suppose that

1307: $\Theta:\7g\to\7g':=\hol(M',a')$ is a Lie algebra isomorphism. We use

1308: the same symbol for the complex linear extension $\Theta:\7l\to

1309: \7l'$. \nline Our first step is to show that $\Theta$ can be modified

1310: in such a way that it respects the gradations. To begin with,

1311: $[\7g',\7g']$ is abelian since our assumption implies that

1312: $[\7g,\7g]=\7g_{-1}$ has this property. With $\7g'=\bigoplus \7g_k'$

1313: being the gradation for $\7g'$ as in \ruf{DC}.ii assume that there

1314: exists a minimal integer $k\ge1$ with $\7g'_{k}\ne0$. Then

1315: $\7g'_{-1}=[\delta,\7g'_{-1}]$ and $\7c:=[\7g'_{-1},\7g'_{k}]\ne0$

1316: imply $\7g'_{-1},\7c\subset[\7g',\7g']$ together with

1317: $[\7g'_{-1},\7c]\ne0$, a contradiction. Therefore

1318: $\7g'=\7g'_{-1}\oplus\7g'_{0}$ with $[\7g',\7g']=\7g'_{-1}$, and as a

1319: consequence,

1320: $\Theta(\7g_{-1})=\Theta([\7g,\7g])=[\7g',\7g']=\7g_{-1}'$. Since

1321: $\ad:\7g'_{0}\to \gl(\7g'_{-1})$ is injective, $\delta+\7g'_{-1}$ is

1322: precisely the set of all $\xi\in\7g'$ such that $\ad\xi$ induces on

1323: $\7g'_{-1}$ the negative identity. Therefore, there exists

1324: $\eta\in\7g'_{-1}$ with $\Theta(\delta)=\delta-\eta$. Replacing

1325: $\Theta$ by $\exp(\ad \eta)\Theta=(\id+\ad \eta)\Theta,$ we get

1326: $\Theta(\delta)=\delta$ and finally $\Theta(\7g_{k})=\7g'_{k}$ for all

1327: $k$.\nline There exists a linear operator

1328: $\theta\in\GL(V)\subset\GL(E)$ with $\Theta(e\dd z)=\theta(e)\dd z$

1329: for all $e\in E=V\oplus iV$. We claim that $H'=\theta H\theta^{-1}$

1330: holds for the abelian subgroups $H:=\exp \7g_0$ and $H':=\exp\7g_0'$

1331: of $\GL(V)$. Indeed, application of $\Theta$ to $\big[e\dd

1332: z,\lambda(z)\dd z\big]=\lambda(e)\dd z\;$ yields

1333: $$

1334: \tilde\lambda \theta=\theta\lambda\Steil{ for all}\lambda(z)\dd

1335: z\in\7g_{0}\steil{and}\tilde\lambda(z)\dd z:=\Theta\big(\lambda(z)\dd

1336: z\big)\in\7g'_{0}\,.

1337: $$

1338: We may therefore assume (possibly after replacing $F$ by $\theta F$

1339: and $a$ by $\theta a$) that $H=H'$, $F=H(a)$ and $F'=H(a')$. Denote by

1340: $\Lambda\subset\End(V)$ the associative subalgebra generated by all

1341: $\lambda\in\End(V)$ with $\lambda(z)\dd z\in\7g_{0}$. Then $\Lambda$

1342: is abelian, contains the identity of $\End(V)$ and

1343: $H\subset\Lambda$. Since $\dim\7g<\infty$, the CR-manifold $F+

1344: iV$ is minimal and consequently $F$ cannot be contained in a

1345: hyperplane of $V$. This implies $\Lambda(a)=V$ and thus the existence

1346: of a $g\in\Lambda$ with $a'=g(a)$. From $gH=Hg$ we get

1347: $F'=g(F)$. Since $g(F)$ also cannot be contained in a hyperplane of

1348: $V$ finally $g\in\GL(V)$ follows.\qed

1349: 

1350: In Propositions \ruf{UH} an d \ruf{HU} a large class of linearly

1351: homogeneous conical submanifolds $F\subset V$ is given for which the

1352: corresponding tubes $M$ satisfy the condition

1353: $\7g_{1}=[\7g_{0},\7g_{0}]=0$ in Proposition \ruf{PQ}.

1354: 

1355: We note that we do not know a single example with $\dim\7g<\infty$ and

1356: $\dim\7g_{k}>\dim\7g_{-k}$ for some $k\in\NN$.  We also do not know

1357: any pair $M,M'$ of holomorphically nondegenerate conical tube

1358: manifolds, that are locally CR-equivalent but are not locally affinely

1359: equivalent. For Levi nondegenerate tube manifolds (which necessarily

1360: cannot be conical) such examples can be found in \Lit{DAYA}. In

1361: \Lit{ISMI} even two affinely homogeneous examples are contained which

1362: are locally affinely non-equivalent but whose associated tube

1363: manifolds are locally CR-equivalent.

1364: 

1365: 

1366: 

1367: \KAP{Fuenf}{Some examples}

1368: 

1369: In this section we present two classes of examples.  To our knowledge,

1370: the only known example of a homogeneous $k$-nondegenerate CR-manifold

1371: with $k\ge3$ occurs in \Lit{FELS} for the case $k=3$: That is a

1372: hypersurface $M$ in a $7$-dimensional compact complex manifold, on

1373: which the simple Lie group $\SO(3,4)^{0}$ acts by biholomorphic

1374: transformations with orbit $M$. In our first example we give for

1375: arbitrary CR-codimenion $c\ge1$ a minimal homogeneous

1376: $3$-nondegenerate as well as a minimal homogeneous $4$-nondegenerate

1377: CR-manifold.  The second class of examples deals with tubes $M$ over

1378: cones of the form $\{x\in(\RRp\!)^{n}:\sum_{j\le p}

1379: x_{j}^{\alpha}=\sum_{j>p} x_{j}^{\alpha}\}$ with $1\le p<n$ and

1380: $\alpha\ne0,1$. Using results from the preceding section, we

1381: explicitly determine all Lie algebras $\7g=\hol(M,a)$ for certain

1382: $\alpha$. Among these are all hyperquadrics (that is $\alpha=2$) of

1383: signature $(p,q)$ with $q:=n-p$, where $\7g$ turns out to be

1384: isomorphic to $\so(p{+}1,q{+}1)$.

1385: 

1386: The first example of CR-manifolds introduced below consists of tubes

1387: $M=\Gamma(a)+iV$ over certain group orbits $\Gamma(a)$, where the

1388: connected group $\Gamma:=\{g\in\GL(2,\RR):\det(g)>0\}$ acts linearly

1389: on a real vector space $V$. In that way we obtain homogeneous

1390: $k$-nondegenerate CR-manifolds for $k\in\{2,3,4\}$. For dimension

1391: reasons it is impossible to construct CR-manifolds of higher

1392: nondegeneracy, employing the group $\Gamma=\GL(2,\RR)^{0}$.  We do not

1393: know how to construct $k$-nondegenerate homogeneous tube manifolds with

1394: $k\ge5$ (should these exist) with suitable other groups, either.

1395: 

1396: \Example{JP} For fixed integers $k\in\{2,3,4\}$ and $c\ge1$ let

1397: $V\subset\RR[u_{1},u_{2}]$ be the subspace of all homogeneous

1398: polynomials of degree $m:=k+c-1$. Then the group $\Gamma$ (see above)

1399: acts irreducibly on $V$ by $p\mapsto p\circ g^{-1}$ for all

1400: $g\in\Gamma$ and has the subgroup $\{g\in\RR\id:g^{m}=\id\}$ as kernel

1401: of ineffectivity. For $a:=\sum^{k-2}_{j=0}u_{1}^{j}u_{2}^{m-j}\in V$

1402: the orbit $F=\5F^{k,c}:=\Gamma(a)$ is a connected conical submanifold

1403: of dimension $k$ in $V$.  With Proposition \ruf{US} it is easily seen

1404: that

1405: $$

1406: K^{r}_{a}F=\sum^{k-1-r}_{j=0}\RR\,u_{1}^{j}u_{2}^{m-j}\steil{for

1407: all}r\ge0\,.

1408: $$

1409: In particular, the tube manifold $\5M^{k,c}:=F+iV$ is a

1410: $k$-nondegenerate homogeneous CR-manifold of CR-dimension $k$ and

1411: CR-codimension $c$. The manifolds $\5M^{2,c}$ will be discussed in

1412: more detail in Section \ruf{Sieben}. In particular, $\5M^{2,1}$

1413: is linearly equivalent to the future light cone tube $\5M$.  For

1414: easier handling let us identify $\RR^{m+1}$ with $V$ via

1415: $(x_{0},x_{1},\dots,x_{m})

1416: \;\leftrightsquigarrow\;\sum^{m}_{j=0}x_{j}u_{1}^{j}u_{2}^{m-j}\;.$

1417: Since $\Gamma$ acts on $\5M^{k,c}$ by (linear) CR-automorphisms, the

1418: linear part $\7g_0$ of $\7g=\hol(\5M^{k,c},a)$ (compare \ruf{DC}.ii)

1419: contains a copy of $\gl(2,\RR)$. More explicitly, this subalgebra is

1420: spanned by the vector fields

1421: $$

1422: \eqalign{ \zeta_{1}:&=\sum^{m}_{j=0}jz_{j}\dd{z_{j}}\,,\mskip53.6mu

1423:     \xi^{-1,1}:=\sum^{m-1}_{j=0}(m-j)z_{j+1}\dd{z_{j}}

1424: \cr

1425: \xi^{1,-1}:&=\sum^{m}_{j=1}jz_{j-1}\dd{z_{j}} \,,\mskip59mu

1426:    \zeta_{2}:=\sum^{m}_{j=0}(m-j)z_{j}\dd{z_{j}}

1427: \;.\cr}\Leqno{AQ}

1428: $$

1429: In particular, the vector fields $\xi^{1,-1},\,\xi^{-1,1}$ and

1430: $\xi^{0,0}:=\zeta_{1}-\zeta_{2}=\sum^{m}_{j=0}(2j-m)z_{j}\dd{z_{j}}$

1431: span a copy of $\7{sl}(2,\RR)$ in $\7g_{0}$. In case $k=2$ a

1432: straightforward computation shows that

1433: $\Gamma_{a}=\Big\{\Matrix\alpha\beta0\epsilon\in\Gamma:

1434: \epsilon^{m}=1\Big\}$ is the isotropy subgroup at $a\in\5M^{2,c}$.

1435: Hence,

1436: $\Big\{\Matrix\alpha\beta{-\beta}\alpha\in\Gamma\Big\}\cong\CC^{*}$

1437: acts transitively\vadjust{\vskip4pt} on $F$, that is, $\5M^{2,c}$ is

1438: diffeomorphic to $\RR^{n}\times\CC^{*}$, where $n:=m+1=\dim V$.

1439: 

1440: \medskip\noindent It seems quite hard to find explicit global

1441: equations for $\5M^{k,c}$ in case $k>2$. For $k=2$ see \Ruf{IP}. Only

1442: for $k=3$ and $c=1$ we have the following: Consider in $\RR^{4}$ the

1443: algebraic hypersurface given by the homogeneous equation

1444: $$

1445: S:=\{x\in\RR^{3}:x^{2}_{0}x_{3}^2+4x_{0}x_{2}^{3}-6x_{0}x_{1}x_{2}x_{3}-

1446: 3x_{1}^{2}x_{2}^{2}+4x_{1}^{3}x_{3}=0\}\,.\Leqno{PI}

1447: $$ It is obvious that $S$ contains the point $a=(1,1,0,0)$. Since the

1448: application of every vector field in $\Ruf{AQ}$ to the defining

1449: function of $S$ gives a multiple of this function, $S$ is invariant

1450: under the group $\Gamma$. Therefore the $3$-nondegenerate homogeneous

1451: CR-submanifold $\5M^{3,1}=\Gamma(a)+i\RR^{4}$ of $\CC^{4}$ is an

1452: open piece of the tube $S+i\RR^{4}$.

1453: 

1454: 

1455: \bigskip\Example{EB} Fix integers $p\ge q\ge1$ with $n:=p+q\ge3$ and a

1456: real number $\alpha$ with $\alpha^2\ne\alpha$. Then

1457: $$

1458: F=F_{p,q}^{\alpha}:=\Big\{x\in(\RRp)^{n}:\sum_{j=1}^{p}

1459: x_{j}^{\alpha}=\sum_{j=p+1}^{n}

1460: x_{j}^{\alpha} \Big\}\,,

1461: $$

1462: is a hypersurface in $V:=\RR^{n}$. Furthermore, $F$ is a cone and

1463: therefore $\dim K_{a}F\ge1$ for every $a\in F$. On the other hand, the

1464: second derivative at $a$ of the defining equation for $F$ gives a

1465: non-degenerate symmetric bilinear form on $V\times V$, whose

1466: restriction to $T_{a}F\times T_{a}F$ then has a kernel of dimension

1467: $\le1$. Therefore $\dim K_{a}F=1$ for every $a\in F$ and by Corollary

1468: \ruf{AF} the CR-manifold $M=M_{p,q}^{\alpha}:=F_{p,q}^{\alpha}+

1469: iV$ is everywhere $2$-nondegenerate (compare Example 4.2.5 in

1470: \Lit{EBFT} for the special case $n=\alpha=3$). Since $M$ as

1471: hypersurface is also minimal, $\7g=\hol(M,a)$ has finite dimension.

1472: 

1473: For the special case $\alpha=2$ and $q=1$ the above cone

1474: $F=F_{n-1,1}^{2}$ is an open piece of the future light cone

1475: $$

1476: \{x\in\RR^{n}:x_{1}^{2}+\dots+x_{n-1}^{2}=x_{n}^{2},\;x_{n}>0\}

1477: $$

1478: in $n$-dimensional space-time, which is affinely homogeneous. In

1479: \Lit{KAZT} it has been shown that for the corresponding tube manifold

1480: $M$ the Lie algebra $\7g=\hol(M,a)$ is isomorphic to $\so(n,2)$ for

1481: every $a\in M$. In case $q>1$ the following result seems to be new:

1482: 

1483: \noindent{\fett Case $\alpha=2$:} Consider on $\CC^{n}$ the

1484: symmetric bilinear form $\langle

1485: z|w\rangle:=\sum\epsilon_{j}z_{j}w_{j}$. Then $F$ is an open piece of

1486: the hypersurface $\{x\in\RR^{n}:x\ne0\,,\;\langle

1487: x|x\rangle=0\}$, on which the reductive group $\RR^{*}\cd\O(p,q)$ acts

1488: transitively. Therefore, $\7s_{0}^{}:=\RR\delta\oplus\so(p,q)$ is

1489: contained in $\7g_{0}$. One checks that

1490: $$

1491: \7s:=\7g_{-1}\oplus\7s_{0}^{}\oplus\7s_{1}^{}\,,\qquad\7s_{1}^{}:=

1492: \big\{\big(2i\langle c|z\rangle z-i\langle z|z\rangle c\big)\dd

1493: z:c\in\RR^{n}\big\}

1494: $$

1495: is a Lie subalgebra of $\7g$. The radical $\7r$ of $\7s$ is

1496: $\ad(\delta)$-invariant and hence of the form

1497: $\7r=\7r_{-1}\oplus\7r_{0}\oplus\7r_{1}$ for

1498: $\7r_{k}^{}:=\7r\cap\7g_{k}$. From $\so(p,q)$ semisimple we

1499: conclude $\7r_{0}\subset\RR\delta$. But $\delta$ cannot be in $\7r$

1500: since otherwise $\7g_{-1}\subset\7r$ would give the false statement

1501: $[\7g_{-1},\7s_{1}^{}]\subset\RR\delta$. Therefore $\7r_{0}^{}=0$, and

1502: $[\7g_{-1},\7r_{1}^{}]=[\7r_{-1}^{},\7s_{1}^{}]=0$ implies

1503: $\7r=0$. Now Proposition 3.8 in \Lit{KAZT} implies $\7g=\7s$, and, in

1504: particular, that $\7g$ has dimension ${n+2\choose2}$. In fact, it can

1505: be seen that $\7g$ is isomorphic to $\so(p+1,q+1)$.

1506: 

1507: \medskip\noindent{\fett Case $\alpha$ an integer $\ge3$:} Then $F$

1508: is an open piece of the real-analytic submanifold

1509: $$

1510: S:=\Big\{x\in\RR^{n}:x\ne0\Steil{and}\sum_{j=1}^{n}\epsilon_{j}

1511: x_{j}^{\alpha}= 0\Big\}\Leqno{NP}

1512: $$

1513: which is connected in case $q>1$ and has two connected components

1514: otherwise. For every $x\in\RR^{n}$ let $d(x)\in\NN$ be the cardinality

1515: of the set $\{j:x_{j}=0\}$. It is easily seen that $\dim

1516: K_{x}S=1+d(x)$ holds for every $x\in S$. Now consider the group

1517: $$

1518: \GL(F):=\{g\in\GL(V):g(F)=F\}\,.

1519: $$

1520: Every $g\in\GL(F)$ leaves $S$ and hence also

1521: $H:=\{x\in\RR^{n}:d(x)>0\}$ invariant, that is, $g$ is the product of

1522: a diagonal with a permutation matrix. Inspecting the action of

1523: $\GL(F)$ on $\{c\in\overline F:d(c)=n-2\}$ we see that $\GL(F)$ as

1524: group is generated by $\RRp\!\cd\id$ and certain coordinate

1525: permutations. As a consequence, $\7g_{0}=\RR\delta$. Now suppose that

1526: there exists a non-zero vector field $\xi\in\7g_{1}$. Then

1527: $\xi=q(z,z)\dd z$ for some symmetric bilinear map

1528: $q:\CC^{n}\times\CC^{n}\to\CC^{n}$ with $q(c,z)\dd z\in\CC\delta$ for

1529: every $c\in\CC^{n}$. Because of $q\ne0$ symmetric therefore any two

1530: vectors in $\CC^{n}$ must be linearly dependent, which contradicts

1531: $n\ge3$.  Therefore $\7g_{1}=0$ and hence $\7g_{k}=0$ for all $k\ge1$

1532: by Proposition \ruf{ZB}.iv. In particular, $\dim\7g=n+1<\dim M$ for

1533: the tube manifold $M=F+iV$, that is, $M$ is not locally

1534: homogeneous. For $n=3$ this gives an alternative proof for Proposition

1535: 6.36 in \Lit{EBEN}.

1536: 

1537: \Proposition{} Let $a,a'\in F=F^{\alpha}_{p,q}$ be arbitrary

1538: points. Then in case $3\le\alpha\in\NN$ the CR-manifold germs $(M,a)$

1539: and $(M,a')$ are CR-equivalent if and only if $a'\in\GL(F)(a)$.

1540: 

1541: \Proof Suppose that $g:(M,a)\to(M,a')$ is an isomorphism of

1542: CR-manifold germs. From $\7g'=\7g_{-1}'\oplus\,\RR\delta$ for

1543: $\7g':=\hol(M',a')$ Proposition \ruf{ZB}.iii implies that $g$ is

1544: represented by a linear transformation in $\GL(V)$ that we also denote

1545: by $g$. But then $g(F)\subset S$ with $S$ defined in \Ruf{NP}. Because

1546: $g(F)$ has empty intersection with $H$ we actually have $g(F)\subset

1547: F$. Replacing $g$ by its inverse we get the opposite inclusion, that

1548: is $g\in\GL(F)$.\qed

1549: 

1550: 

1551: \KAP{Sechs}{Levi degenerate CR-manifolds associated with an

1552: endomorphism}

1553: 

1554: The lowest CR-dimension for which there exist homogeneous CR-manifolds

1555: that are Levi degenerate but not holomorphically degenerate is

1556: $2$. The construction recipe below will give, up to local affine

1557: equivalence, all affinely homogeneous conical tube submanifolds of

1558: $\CC^{n}$ with CR-dimension $2$. Indeed, it is based on the following

1559: simple observation: Suppose that $F\subset V:=\RR^{n}$ is a conical

1560: locally linearly homogeneous submanifold of dimension $2$. Denote by

1561: $\7a$ the Lie algebra of all linear vector fields on $V$ that are

1562: tangent to $F$. Then, fixing a point $a\in F$, there exists a

1563: $\phi\in\End(V)$ with $\phi(x)\dd x\in\7a$ and $T_{a}F=\RR

1564: a\oplus\RR\phi(a)$. Therefore, the orbit $H(a)$ under the subgroup

1565: $H:=\{\exp(r\id+t\,\phi):r,t\in\RR\}$ is an (immersed) surface in $V$

1566: having the same germ at $a$ as $F$.

1567: 

1568: 

1569: \Joker{CO}{Construction recipe} Throughout this section let $1<d<n$ be

1570: arbitrary integers and $V$ a real vector space of dimension $n$.  Let

1571: furthermore $\phi\in\End(V)$ be a fixed endomorphism and

1572: $\7h\subset\End(V)$ the linear span of all powers $\phi^{k}$ for

1573: $k=0,1,\dots,d-1$. Then $H:=\exp(\7h)\subset\GL(V)$ is an abelian

1574: subgroup, and for given $a\in V$ the orbit $F:=H(a)$ is a cone and an

1575: immersed submanifold of $V$ (not necessarily locally closed in case

1576: $n\ge4$). Furthermore, the tube $M=F+iV\subset E$ is an immersed

1577: CR-submanifold of $E$.

1578: 

1579: \Joker{}{Cyclic endomorphisms and vectors} A vector $a\in V$ is called

1580: {\sl cyclic} with respect to $\phi\in \End(V)$ if the $\phi^k(a)$,

1581: $k\ge 0$, span $V$. This is equivalent to

1582: $a,\phi(a),\ldots,\phi^{n-1}(a)$ being a basis of $V$. We call

1583: $\phi\in\End(V)$ {\sl cyclic} if it has a cyclic vector and denote

1584: by $\Cyc(V)\subset\End(V)$ the subset of all cyclic endomorphisms. If

1585: $a,b\in V$ both are cyclic vectors of $\phi$ then there exists a

1586: transformation $g\in\RR[\phi]\subset\End(V)$ with $b=g(a)$. But $g$

1587: commutes with every element of the group $H=\exp(\7h)$ and hence maps

1588: the orbit $H(a)$ onto $H(b)$. In particular, the CR-isomorphism type

1589: of $M=H(a)+iV$ only depends on $\phi$ and $d$, but not on the

1590: choice of the cyclic vector $a$. To emphasize this dependence we also

1591: write $M^{\phi,d}$ for $M$ and $F^{\phi,d}$ for $H(a)$, but only if

1592: $\phi$ is cyclic.  We tacitly assume that a choice for a cyclic

1593: vector $a$ has been made. In case $d=2$ we even write $M^{\phi}$ and

1594: $F^{\phi}$ instead of $M^{\phi,2}$ and $F^{\phi,2},$ respectively.

1595: The following proposition shows the relevance of these manifolds in

1596: our discussion.

1597: 

1598: \Proposition{PR} For $F=H(a)$ and $M=F+iV$ as in \ruf{CO} the

1599: following conditions are equivalent: \0 $\hol(M,a)$ has finite

1600: dimension. \1 $a$ is a cyclic vector of $\phi$.\par\noindent If these

1601: conditions are satisfied, $M$ is a minimal $2$-nondegenerate

1602: homogeneous CR-manifold with CR-dimension $d$ and Levi kernel

1603: $K_{a}M=\CC a$.

1604: 

1605: \Proof \To12 Condition (i) together with the homogeneity of $M$

1606: implies that $M$ is minimal. Let $W\subset V$ be the linear span of

1607: all vectors $\phi^{k}(a)$, $k\ge0$. Then $H\subset\RR[\phi]$ implies

1608: $H(a)\subset W$ and hence $W=V$ by the minimality of $M$. Therefore,

1609: $a$ is a cyclic vector, and the $\phi^{k}(a)$, $0\le k<d$, form a

1610: basis of the tangent space $T_{a}F$. In particular, $F$ has dimension

1611: $d$, which is also the CR-dimension of $M$.  \nline \To21 Suppose that

1612: $a$ is a cyclic vector of $\phi$.  Lemma \ruf{CR} gives $\RR a\subset

1613: K_{a}F$ since $F$ is a cone in $V$. For the proof of the opposite

1614: inclusion fix an arbitrary $w\in T_{a}F$ with $w\notin\RR a$. Then

1615: $w=\sum_0^m c_j \phi^j(a)$ with $c_m\ne0$ for some $1\le m<d$ and

1616: $\phi^{d-m}(w)\notin T_{a}F$ shows $w\notin K_{a}F$ by Proposition

1617: \ruf{US}. Therefore, $M$ is $2$-nondegenerate by Corollary \ruf{AF}

1618: and $K_{a}M=\CC a$. It remains to show that $M$ is minimal at $a$. But

1619: this immediately follows from Proposition \ruf{PY}.\qed

1620: 

1621: 

1622: For the manifolds $M=M^{\phi,d}$ it is possible to compute the Lie

1623: algebras $\7g=\hol(M,a)$ in `most cases'. Clearly, the Lie algebra

1624: $\7g_{0}$ contains the $d$-dimensional Lie algebra $\7h$ (as before,

1625: $\7g_{0}$ is canonically identified with a linear subspace of

1626: $\End(V)\,$). In the Propositions \ruf{UH} and \ruf{HU} we will show

1627: that the equality $\7g=\7g_{-1}\oplus\7h$ holds for all $\phi$ in

1628: `general position'.

1629: 

1630: \noindent Suppose that $\phi\in\End(V)$ has the cyclic vector $a\in V$

1631: and that the integer $d$ satisfies $1<d<n$. Let

1632: $$

1633: \chi^{}_{\phi}:=\prod_{j=1}^{s}(X-\lambda_{j})^{n_{j}}\;\in\;\RR[X]\,,\quad

1634: n_{j}\ge1\,,\Leqno{CP}

1635: $$ with mutually distinct eigenvalues

1636: $\lambda_{1},\dots,\lambda_{s}\in\CC$ be the characteristic polynomial

1637: of $\phi$.  Let furthermore $\alpha_{1},\dots,\alpha_{n}$ be the

1638: family of all roots of $\chi_{\phi}$, that is, each $\lambda_{j}$

1639: occurs $n_{j}$-times in this string.  As usual,

1640: $\alpha_{1},\dots,\alpha_{n}$ is called an {\sl arithmetic

1641: progression} in $\CC$ if there exists a $\beta\in\CC$ such that, after

1642: a suitable permutation, $\alpha_{j}=\alpha_{1}+(j-1)\beta$ for all

1643: $1\le j\le n$.

1644: 

1645: \Proposition{UH} Let $M=M^{\phi,d}$ and $\7g:=\hol(M,a)$ for a cyclic

1646: vector $a\in V\cap M$ of $\phi$. Then $\7g_{0}=\7h$ holds if one of the

1647: following conditions is satisfied. \0 $d=2$ and the characteristic

1648: roots $\alpha_{1},\dots,\alpha_{n}$ of $\phi$ do not form an {\sl

1649: arithmetic progression}. \1 $d\ge3$ and $s>d$.\Formend

1650: 

1651: \noindent For the proof we need several preparations.  To simplify the

1652: notation at various places let us introduce

1653: $$

1654: S:=\{1,\dots,s\},\quad m:=d-1\Steil{and}m_{k}:=n_{k}-1\steil{for

1655: all}k\in S\,.\Leqno{UK}

1656: $$

1657: For every $K\subset S$ define furthermore

1658: $$

1659: \Delta_{K}:= \big\{\beta_{jk}:j\in

1660: S,k\in K\big\}\steil{with}\beta_{jk}:=(\lambda_{j}^{}-\lambda_{k}^{},

1661: \lambda_{j}^{2}- \lambda_{k}^{2},

1662: \dots,\lambda_{j}^{m}-\lambda_{k}^{m})\in\CC^{m}\Leqno{JG}

1663: $$

1664: and denote by $\6K$ the set of all non-empty subsets $K\subset S$ such

1665: that there exist subsets $P\subset Q\subset (S\setminus K)$ with the

1666: following two properties, where the maximum over the empty set here is

1667: defined to be $0\,$: \0 $\sum_{k\in K}n_{k}\;+\;\sum_{q\in Q}n_q

1668: \;\;\ge\;\; d\;+\;\max\re2\{n_{p}-1:p\in P\}\,.$ \1 To every

1669: $\beta\in\Delta_{Q}\!\setminus\Delta_{K}$ there exist uniquely

1670: determined $j\in S$ and $q\in Q$ with $\beta=\beta_{jq}$ such that

1671: $n_{j}<n_{q}$ and $q\in P$.

1672: 

1673: \noindent Notice that $\6K$ contains every subset $K\subset S$ with

1674: $\;\sum_{k\in K}n_{k}\ge d\;$ (just take $Q=\emptyset$).  In the

1675: following Lemma we show that (i) or (ii) in Proposition \ruf{UH} will

1676: follow from a more general technical condition that will allow us to

1677: give a uniform proof of \ruf{UH} for all $d$.

1678: 

1679: \Lemma{UL} Suppose that $ s\ge d$ and $\bigcap_{K\in\6K}\!\Delta_{K}

1680: \,=\,\{0\}$ holds. Then one of the conditions (i) and (ii) in

1681: Proposition \ruf{UH} is satisfied.

1682: 

1683: \Proof {\fett $d=2$:} Assume that $\alpha_{1},\dots,\alpha_{n}$ is not

1684: an arithmetic progression and that there exists a non-zero

1685: $\beta\in\Cap_{K\in\6K}\Delta_{K}$. We claim that

1686: $\lambda_{1},\dots,\lambda_{s}$ is an arithmetic progression:

1687: Otherwise $s\ge3$ and there exists a $k$ with $1<k<s$ such that,

1688: without loss of generality, $\lambda_{j}=\lambda_1+(j-1)\beta$ for all

1689: $1\le j\le k$ and $\lambda_1-\beta\ne\lambda_{r}\ne \lambda_k+\beta$

1690: for all $r>k$. For every $j>k$ the set $\Delta_{\{k,j\}}$ contains the

1691: number $\beta$, that is, there is an $r\in S$ with

1692: $\lambda_{r}=\lambda_{k}+\beta$ or

1693: $\lambda_{r}=\lambda_{j}+\beta$. The first possibility violates our

1694: assumptions. In the second case necessarily $r>k$ must hold since

1695: $r\le k$ would imply $\lambda_{j}=\lambda_1-\beta$ and thus the second

1696: possibility cannot be true for {\it all}\/ $j$ with $k<j\le s$. This

1697: proves that $\lambda_1,...\,,\lambda_s$ is an arithmetic progression

1698: and also $s<n$ by assumption. In particular, $n_j>1$ for some $j\in

1699: S$. Next we claim that $\{1\}\in\6K$ and $\{s\}\in\6K$: To see that

1700: $\{1\}\in\6K,$ let $k\in \{1,...\,, s\}$ be minimal with

1701: $n_{k}>1$. With $P:=Q:=\{k\}\setminus\{1\}$ condition \Ruf{JG}.ii is

1702: fulfilled and the first part of the claim follows. A similar argument

1703: proves also $\{s\}\in\6K$.  But then

1704: $\Delta_{\{1\}}\cap\Delta_{\{s\}}=\{0\}$ gives a contradiction.\nline

1705: {\fett $d\ge3$:} Suppose that $0\ne \beta_{jk}\in

1706: \Cap_{K\in\6K}\Delta_{K}$ and that $s>d$. Then

1707: $L:=S\setminus\{k\}\in\6K$ (with $Q=\emptyset$) and there are $\ell\in

1708: L$, $r\in S$ with $\beta_{jk}=\beta_{r\ell}$. Since $d\ge 3,$ the

1709: equation $\beta_{jk}=\beta_{r\ell}$ implies $k=\ell$, a

1710: contradiction.\phantom{XXXX}\qed

1711: 

1712: \medskip\noindent{\fett Proof of

1713: Proposition \UH:} It is enough to assume that the assumption of Lemma

1714: \ruf{UL} is satisfied. Consider the decomposition

1715: $E=E_{1}\oplus\cdots\oplus E_{s}$ with $E_{k}$ the kernel of

1716: $(\phi-\lambda_{k})^{n}$ for every $k\in S$. Denote by

1717: $\pi_{k}:E\to E_{k}$ the canonical projection and by

1718: $\epsilon_{k}:E_{k}\to E$ the canonical injection. Then

1719: $a_{k}:=\pi_{k}(a)$ is a cyclic vector for

1720: $\phi_{k}:=\pi_{k}\phi\re1\epsilon_{k}\in\End(E_{k})$. Furthermore, if

1721: we put $a^{j}_{k}:=(\phi_{k}-\lambda_{k})^{j}(a_{k})$ for all $j\ge0$,

1722: then $a^{0}_{k},\dots,a^{m_{k}}_{k}$ is a basis of $E_{k}$. With

1723: $m=d-1$ as defined above let

1724: $\Phi:=(\phi^{1},\dots,\phi^{m})\in\End(V)^{m}$.

1725: 

1726: \noindent For every real (or complex) vector space $W$ and all tuples

1727: $t=(t_{1},\dots,t_{m})\in\RR^{m}$, $w=(w_{1},\dots,w_{m})\in W^{m}$

1728: let us write as shorthand $t\cd w:=\sum_{j}t_{j}w_{j}$. For every

1729: $t\in\RR^{m}$ the point $e^{t\cd\Phi}(a)$ is contained in $F=M\cap V$.

1730: 

1731: \noindent Now fix an arbitrary $\mu\in\7g_{0}\subset\End(V)$. Since

1732: the vector field $\mu$ is tangent to $F$, to every

1733: $t\in\RR^{m}$ there exist real coefficients

1734: $r_{0},r_{1},\dots,r_{m}$ with

1735: $$

1736: \mu e^{t\cd\Phi}(a)\;=\;Re^{t\cd\Phi}(a)\Steil{for}R:=

1737: \sum^{m}_{\ell=0}r^{}_{\li2\ell}\phi^{\ell}\,.\Leqno{PT}

1738: $$

1739: Actually, every $r_{\ell}$ has to be considered as a real valued function

1740: on $\RR^{m}$. Put

1741: $$

1742: \mu_{k}:=\pi_{k}\mu\Steil{and}N_{k}:=\big((\phi_{k}-\lambda_{k}),

1743: \dots,(\phi^{m}_{k}-\lambda_{k}^{m})\big)\;\in\;\End(E)^{m}

1744: $$

1745: for all $k\in S$. Applying $\pi_{k}$ to \Ruf{PT}

1746: gives

1747: $$

1748: \eqalign{ R\,e^{t\cd N_{k}}(a_{k})\;&=\;e^{t\cd N_{k}}R(a_{k})=\;

1749: \sum_{j=1}^{s}e^{t\cd\beta_{jk}} \mu_{k}e^{t\cd N_{j}}(a_{j})

1750: \,\quad\hbox{with}\cr

1751: R(a_{k})&=\;\sum^{m_{k}}_{j=0}\rho_{k,j}a^{j}_{k}\Steil{for}

1752: \rho_{k,j}(t):=\sum^{m}_{\ell=j}\textstyle{\ell\choose

1753: j}\lambda^{\ell-j}_{k}r^{}_{\ell}(t)\;\;.\cr }\Leqno{JA}

1754: $$

1755: For every subset $B\subset\CC^{m}$ denote by

1756: $\6F(B)\subset\5C(\RR^{m},\CC)$ the smallest linear subspace

1757: containing all functions $h(t)e^{t\cd\beta}$ with $\beta\in B$ and

1758: $h\in\CC[t]=\CC[t_{1},\dots,t_{m}]$ a polynomial function on

1759: $\RR^{m}$.  Then it is well known that every $f\in\6F(B)$ has a unique

1760: representation $f=\sum_{\beta\in B}f^{[\beta]}e^{t\cd\beta}$ with

1761: $f^{[\beta]}\in\CC[t]$ and $\{\beta\in B:f^{[\beta]}\ne0\}$ finite.

1762: Since $t\cd N_{j}$ is nilpotent, i.e., $e^{\pm t\cd N_j}$ are

1763: polynomial, as a consequence of the identities \Ruf{JA} we get for

1764: every $K\subset S$

1765: $$

1766: \rho_{k,j}\;\in\;\6F(\Delta_{K})\Steil{for every}k\in K\steil{and}0\le

1767: j\le m_{k}\,.\Leqno{JC}

1768: $$

1769: Denote by $\6B$ the set of all subsets $B\subset\CC^{m}$ with

1770: $r_{\ell}\in\6F(B)$ for all $\ell$.

1771: 

1772: \noindent{\bf Claim 1:} \quad {\sl $\bigcap_{K\in

1773: \6K}\Delta_K\in\6B$, that is, $r_{\ell}\in\CC[t]$ for all

1774: $\ell$.}

1775: 

1776: \ProofC Fix an arbitrary $K\in\6K$ and let $P\subset Q$ be as in the

1777: definition of $\6K$. Since $\6B$ is closed under intersections it is

1778: enough to show $\Delta_{K}\in\6B$. Assume to the contrary that this is

1779: not true.  Consider the linear system of equations for the $r_{\ell}$,

1780: compare also \ruf{HA},

1781: $$

1782: \rho_{k,j}=\sum^{m}_{\ell=j}\textstyle{\ell\choose

1783: j}\lambda^{\ell-j}_{k}r^{}_{\ell}\;\in\;\6F(\Delta_{K\cup

1784: Q})\Steil{for all}k\in K\cup Q\steil{and}0\le j\le

1785: m_{k}\,.\Leqno{KZ}

1786: $$

1787: The coefficient matrix is of generalized Vandermonde type and hence

1788: has rank $d=m+1$ since by the definition of $\6K$ the number of

1789: equations is at least $d$. This implies

1790: $r_{\ell}\in^{}\6F(\Delta_{K}\cup\Delta_{Q})$ for all $\ell$.  Since

1791: by assumption not all $r_{\ell}$ are in $\6F(\Delta_{K})$ there is a

1792: $\beta\in\Delta_{Q}\!\setminus\Delta_{K}$ such that the

1793: $\beta$-components $r^{[\beta]}_{\ell}\in\CC[t]$ do not vanish for all

1794: $\ell$ simultaneously. By the definition of $\6K$ there are uniquely

1795: determined $p\in S$ and $q\in P$ with $\beta=\beta_{pq}$.  Define

1796: $L:=K\cup Q\setminus \{q\}$.  Since $\beta\not\in\Delta_L$ we get from

1797: \Ruf{JA} the following linear system

1798: $$

1799: \rho^{[\beta]}_{k,j}=\sum^{m}_{\ell=j}\textstyle{\ell\choose

1800: j}\lambda^{\ell-j}_{k}r^{[\beta]}_{\ell}\;=0\,,\steil{where}k

1801: \steil{runs through}L\steil{and}0\le j\le m_{k}\,\,.\Leqno{KX}

1802: $$

1803: By the very definition of $K$, $P$ and $Q$ it follows that the above

1804: linear system consists of at least $d-1$ equations. Consequently we

1805: can write it in the form:

1806: $$

1807: \sum^{m-1}_{\ell=j}\textstyle{\ell\choose

1808: j}\lambda^{\ell-j}_{k}r^{[\beta]}_{\ell}\;\in\CC

1809: r^{[\beta]}_{m}\Steil{for all}k\in L \steil{and}\;0\le j\le m_{k}\;.

1810: $$

1811: Since its coefficient matrix is of generalized Vandermonde type, every

1812: $r^{[\beta]}_{\ell}$ is a complex multiple of $r^{[\beta]}_{m}$ and,

1813: in particular, $r^{[\beta]}_{m}\ne0$. We claim

1814: $\rho^{[\beta]}_{q,0}\ne0$. Indeed, otherwise we could add the

1815: equation $\rho^{[\beta]}_{q,0}=0$ to the linear system \Ruf{KX}, which

1816: then has a coefficient matrix of generalized Vandermonde type with

1817: rank $d$ contradicting $r^{[\beta]}_{m}\ne0$.  Denote by $D$ the

1818: degree of

1819: $\rho^{[\beta]}_{q,0}=\sum_{\ell}\lambda^{\ell}_{q}r^{[\beta]}_{\ell}$. Then

1820: $D=\deg \rho^{[\beta]}_{q,0}=\deg r_{m}^{[\beta]}\ge 0$ and all

1821: $r_{\ell}$ and $\rho^{[\beta]}_{q,j}$ have degree $\le D$. The

1822: equations \Ruf{JA} imply (replace $k$ by $q$, form the

1823: $\beta$-components for $\beta:=\beta_{pq}$ and carry out the

1824: multiplication with $e^{t\cd N_{q}}$)

1825: $$

1826: \mu_{q}e^{t\cd N_{p}}\re1(a_{p})\;=\;e^{t\cd

1827: N_{q}}\sum^{m_{q}}_{j=0}\rho^{[\beta]}_{q,j}a^{j}_{q}=\sum^{m_{q}}_{j=0}

1828: \big(f_{j}\rho^{[\beta]}_{q,0}+g_{j}\big)a^{j}_{q}\Leqno{KI}

1829: $$

1830: for certain polynomials $f_{j},g_{j}\in\CC[t]$ with

1831: $\deg(g_{j})<\deg(f_{j})=j$. Comparing degrees on both sides in \Ruf{KI}

1832: and keeping in mind $n_{p}<n_{q}$ (by the definition of $\6K$) we get

1833: $$

1834: D+m_{q}=\deg(f_{m_{q}}\rho^{[\beta]}_{q,m_{q}}+g_{m_{q}})\le

1835: m_{p}<m_{q}\,.

1836: $$

1837: This contradicts $r_m^{[\beta]}\ne 0$ and Claim 1 is proved.\qd

1838: 

1839: 

1840: \noindent{\bf Claim 2: \sl Every $r_{\ell}$ is a constant polynomial.}

1841: 

1842: \ProofC Fix a $k\in S$ with $D:=\deg(\rho_{k,0})=\max_{j\in

1843: S}\deg(\rho_{j,0})$. With $s\ge d$ a Vandermonde argument applied to

1844: the linear system $\rho_{j,0}=\sum_{\ell}\lambda^{\ell}_{j}r_{\ell}$,

1845: $\,j\in S$, gives that every $r_{\ell}$ has degree $\le D$. As in

1846: \Ruf{KI} we have

1847: $$

1848: \mu_{k}e^{t\cd N_{k}}\re1(a_{k})\;=\;e^{t\cd

1849: N_{k}}\sum^{m_{k}}_{j=0}\rho_{k,j}a^{j}_{k}=\sum^{m_{k}}_{j=0}

1850: \big(f_{j}\rho_{k,0}+g_{j}\big)a^{j}_{k}\Leqno{KY}

1851: $$ for polynomials $f_{j},g_{j}\in\CC[t]$ with

1852: $\deg(g_{j})<\deg(f_{j})=j$.  All coefficient polynomials in \Ruf{KY}

1853: in front of the $a^{j}_{k}$ have degree $\le m_{k}$, that is

1854: $D+m_{k}\le m_{k}$ and hence $D\le0$. This proves Claim 2.\qd

1855: \nline The proof of \ruf{UH} now is complete: Indeed, since

1856: $F$ contains a basis of $V$, the endomorphism $\mu$ is uniquely

1857: determined by the function tuple $(r_{\ell})$, that is, $\dim\7g_{0}\le

1858: d=\dim \7h $.\qed

1859: 

1860: \smallskip \Remark{HA} For given tuples

1861: $\lambda_{1},\dots,\lambda_{s}\in\CC$ and $\,n_{1},\dots,n_{s}\in\NN$

1862: with $n_{k}\ge1$ and $n:=\sum\!n_{k}$ let $L:=\{(k,j):1\le k\le

1863: s,\,0\le j<n_{k}\}$ endowed with the lexicographic order. Then it can

1864: be seen that the following $n\Times n$-matrix (every entry with $\ell<

1865: j$ is zero)

1866: $$

1867: \Big(\textstyle{\ell\choose

1868: j}\lambda^{\ell-j}_{k}\Big)_{0\le\ell<n,\re1(k,j)\in L}\Steil{has

1869: determinant}\displaystyle\prod\limits_{p<q}

1870: (\lambda_{q}-\lambda_{p})^{n_{p}n_{q}}\;.

1871: $$

1872: 

1873: 

1874: Notice from the proof of \ruf{UH} that the condition

1875: $\bigcap_{K\in\6K}\!\Delta_{K} \,=\,\{0\}$ guarantees that every

1876: $\mu\in\7g_{0}$ leaves every generalized eigenspace $E_{k}$ of $\phi$

1877: invariant, while $s\ge d$ guarantees that every such $\mu$ actually is

1878: in $\7h$. In the next section we will see that the condition (i) in

1879: Proposition \ruf{UH} for $d=2$ is optimal, compare Proposition \ruf{ZR}.

1880: 

1881: \Proposition{HU} Let $M=M^{\phi,d}$ and assume that $\7g_{0}=\7h$ for

1882: $\7g=\hol(M,a)$. Then $\7g=\aff(M,a)$ and $\aut(M,a)=0$.

1883: 

1884: \Proof For the proof of $\7g=\aff(M,a)$ it is enough to show

1885: $\7g_{1}=0$ by Proposition \ruf{DC}.iv. This is more easily done in

1886: the more general complex setting, compare the following Lemma

1887: \ruf{UF}. Finally, counting dimensions yields $\aut(M,a)=0$.\qed

1888: 

1889: \medskip\noindent It remains to show the next Lemma. As before,we

1890: identify the spaces $\End(E)$ and $\7P_{0}$, see \Ruf{HJ}.\vskip-30pt

1891: 

1892: \Lemma{UF} Let $\phi\in\Cyc(E)$ be an arbitrary cyclic

1893: endomorphisms. For given integer $d<n=\dim E$ let furthermore $\7h$ be

1894: the complex linear span of all powers $\phi^{j}$, $0\le j<d$. Then

1895: $$

1896: \big\{\xi\in\7P_{1}:[\7P_{-1},\xi]\subset\7h\big\}\;=0\;\,.

1897: $$

1898: \Proof We identify $E$ with $\CC^{n}$ in such a way that the

1899: matrix $\Phi$ of $\phi$ is in Jordan normal form, more precisely:

1900: $\Phi$ is a block diagonal matrix with Jordan blocks

1901: $J_{1},\dots,J_{s}$, where each block $J_{l}$ is lower triangular, has

1902: the eigenvalue $\lambda_{l}$ on its main diagonal and is of size

1903: $n_{l}\Times n_{l}$ for some $n_{l}\ge1$. We also introduce an

1904: equivalence relation on $\{1,\dots,n\}$ in the following way: Put

1905: $j\sim k$ if the $j^{\th}$ row and the $k^{\th}$ column in $\Phi$

1906: intersect in one of the Jordan blocks. \nline Now suppose that there

1907: exists a non-zero vector field $\xi\in \7P_{1}$ with

1908: $[\7P_{-1},\xi]\subset\7h$. This $\xi$ has a unique representation

1909: $$

1910: \xi=\sum_{j,k,p=1}^{n}c_{p}^{jk}z_{j}z_{k}\dd{z_{p}}\Steil{with}

1911: c_{p}^{jk}=c_{p}^{kj}\in\CC\,.

1912: $$

1913: For every $j\le n$ the vector field $\dd z_{j}$ is contained in

1914: $\7P_{-1}$. Therefore

1915: $$

1916: \xi_{j}:={1\over2}\big[\dd{z_{j}},\xi\big]=\sum_{k,p=1}^{n}

1917: c^{jk}_{p}z_{k} \dd{z_{p}}

1918: $$

1919: is contained in $\7h$ implying $c_{p}^{jk}=0$ if $k\not\sim p$ and, by

1920: symmetry, $c_{p}^{jk}=0$ if $j\not\sim p$. This implies

1921: $$

1922: \xi_{j}=\sum_{p\sim j}\sum_{k\sim j}c^{jk}_{p}z_{k}

1923: \dd{z_{p}}\,.\Leqno{VZ}

1924: $$ By assumption $\xi\ne0$ and therefore $\xi_{r}\ne0$ for some $r\le

1925: n$. Without loss of generality we may assume $r\sim1$. Replacing

1926: $\phi$ by $(\phi-\lambda_{1})$ we also may assume $\lambda_{1}=0$.

1927: Put $b:=n_{1}$, that is, the Jordan block $J_{1}$ has size $b\times b$

1928: and is nilpotent. Define for all $1\le p,k\le b$

1929: $$

1930: \eta_{p}:=z_{1}\dd{z_{p}}+z_{2}\dd{z_{p+1}}+\dots+z_{b+1-p}\dd{z_{b}}

1931: \,,

1932: $$

1933: $$

1934: \7D_{k}:=\sum_{j=k}^{b}\CC\re1\eta_{j}\Steil{and}\psi_{k}:=

1935: \phi^{k-1}\prod_{l=2}^{s}(\phi-\lambda_{l})^{n_{l}}\;

1936: \in\;\End(E)\,.

1937: $$

1938: From \Ruf{VZ} we know

1939: $\xi_{j}\in\7D_{1}\cap\7h$ for all $j\le b$. In particular,

1940: $$ \xi_{j}=\sum_{p=1}^{b}c^{j1}_{p}\eta_{p}\,,\qquad c_{p}^{jk}=

1941: \cases{c^{j1}_{p+1-k}&if $k\le p$\cr0&otherwise\cr}\qquad\hbox{and hence}

1942: $$

1943: $$

1944: \xi_{j}=c^{11}_{1}\eta^{}_{j}+c^{11}_{2} \eta^{}_{j+1}+

1945: \dots+c^{11}_{b+1-j}\eta^{}_{b}

1946: $$ for all $j$ due to the symmetry of $c^{jk}_{p}$ in the upper

1947: indices.  As a consequence there exists a minimal $q\le b$ with

1948: $q\ge1$ and $c^{11}_{q}\ne0$. But then

1949: $\xi_{b+1-q}=c^{11}_{q}\eta^{}_{b}$ implies

1950: $\eta^{}_{b}\in\7D_{b}\cap\7h$. We show that this cannot be

1951: true:\nline Since $\psi_{1}$ as polynomial in $\phi$ has constant term

1952: $\lambda_{2}\lambda_{3}\dots\lambda_{s}\ne0$, we get that $\psi_{k}$

1953: spans $\7D_{k}$ over $\7D_{k+1}$ for every $k\ge1$. This implies

1954: $\eta_{b}=\sum_{j=1}^{n}e_{j}\phi^{j-1}$ for suitable real

1955: coefficients $e_{j}$ with $e_{n}\ne0$ and thus

1956: $\eta_{b}\notin\7h$.\phantom{XXX}\qed

1957: 

1958: 

1959: \bigskip For the application of Proposition \ruf{UQ} to manifolds of

1960: the type $M=M^{\phi,d}$ it is necessary to know when $M$ is simply

1961: connected and when $\Aut(M,a)$ is the trivial group. \nline A

1962: sufficient condition for $M^{\phi,d}$ (and $F^{\phi,d}=H\Kl a$) to be

1963: simply connected is the following: {\sl There exist eigenvalues

1964: $\lambda_{1},\dots,\lambda_{d}$ of $\phi$ such that ${\det(A+\overline

1965: A)}\ne0$, where $A=(\lambda_{j}^{k-1})_{1\le j,k\le d}$ is the

1966: corresponding Vandermonde matrix}. \nline To get a partial answer to

1967: the second question consider to given $\epsilon\in\RR^{*}$ a $g\in\GL(V)$

1968: with $g(a)=a$ and $g\phi g^{-1}=\epsilon\phi$. From

1969: $g\phi^{k}g^{-1}=(\epsilon\phi)^{k}$ we get

1970: $g(\phi^{k}(a))=\epsilon^{k}\phi^{k}(a)$ for all $k\ge0$, that is, $g$

1971: is uniquely determined in $\GL(V)$ by the above assumptions since

1972: $\phi$ is cyclic. A further consequence is

1973: $g\re1\exp(t\phi^{k})\re2g^{-1}=\exp(t\epsilon^{k}\phi^{k})$ for all

1974: $t\in\RR$ and $k\ge0$. This means $gHg^{-1}=H$ for the group

1975: $H=\exp(\7h)$. But then $g(F)=F$ for $F=H(a)$ and consequently

1976: $g\in\Aut(M,a)$.

1977: 

1978: Notice that we always may assume without loss of generality that

1979: $\phi\in\Cyc(V)$ has trace $0$ (otherwise replace $\phi$ by

1980: $(\phi-c\,\id)\in\End(V)$ for $c:=n^{-1}\tr(\phi)$, since this

1981: procedure does not change the algebra $\7h)$.

1982: 

1983: \Lemma{KR} Suppose that $2d\le n+1$ for $n=\dim V $ and that $\phi$

1984: is trace-free. Suppose in addition that $\aut(M,a)=0$ holds for the

1985: cyclic vector $a\in V$ and $M:=M^{\phi,d}$. Then

1986: $$\Aut(M,a)=\big\{g\in\GL(V):g(a)=a\steil{and}g\phi

1987: g^{-1}=\pm\phi\big\}\,.$$ In particular, $\Aut(M,a)$ always has order

1988: $\le 2$ and is trivial, for instance, if the spectrum of $\phi$ in

1989: $\CC$ is not symmetric with respect to the origin of $\CC$.

1990: 

1991: 

1992: \Proof By Proposition \ruf{LO} the assumption $\aut(M,a)=0$ implies

1993: $\7g_{1}=0$ and $\7g_{0}=\7h$. As a consequence of Proposition

1994: \ruf{PS} therefore

1995: $\Aut(M,a)=\big\{g\in\GL(V):g(a)=a\steil{and}g\7hg^{-1}= \7h\big\}$

1996: holds. Let $g\in\Aut(M,a)$ be an arbitrary automorphism. Then $g\phi

1997: g^{-1}=\sum^{m}_{j=0}c_{j}\phi^{j}$ for real coefficients $c_{j}$ and

1998: $m:=d-1$. We show by induction on $k$ that $c_{j}=0$ holds for all

1999: $j>m/k$ and all $1\le k\le m$. For $k=1$ this is obvious. So fix a

2000: $k>1$ with $k\le m$. By induction hypothesis $g\phi^{k}

2001: g^{-1}=\big(\sum c_{j}\phi^{j}\big)^{k}

2002: =\sum^{2m}_{\ell=0}e_{\ell}\phi^{\ell}\in\7h$ with real coefficients

2003: $e_{\ell}$. Since $2m<n$ by assumption, we must have $e_{\ell}=0$ for

2004: all $\ell>m$, that is $c_{j}=0$ for all $j>m/k$. For $k=m$ this

2005: implies $c_{j}=0$ for all $j>1$. Taking traces finally gives $g\phi

2006: g^{-1}=\epsilon\phi$ for $\epsilon:=c_{1}$. By the above example

2007: $\phi$ cannot be nilpotent, that is, $\phi$ has non-zero spectrum in

2008: $\CC$. Since this spectrum is invariant under multiplication with

2009: $\epsilon$ necessarily $\epsilon=\pm1$ holds.\qed

2010: 

2011: 

2012: 

2013: \KAP{Sieben}{Homogeneous 2-nondegenerate manifolds of CR-dimension 2}

2014: 

2015: In this section we specialize to homogeneous tube manifolds $M=F+iV$

2016: in $E=V\oplus iV$ of CR-dimension $2$, that is, where $F\subset V$ is

2017: a surface of dimension $2$. We begin with manifolds of type

2018: $M^{\phi}=M^{\phi,2}$ that are obtained by the construction recipe

2019: \ruf{CO}. Since Propositions \ruf{UH} and \ruf{HU} of the preceding

2020: section do not cover the case where the characteristic roots

2021: $\alpha_{1},\dots,\alpha_{n}$ of $\phi\in\Cyc(V)$, $n=\dim V$, form an

2022: arithmetic progression, let us discuss this case first:\nline Possibly

2023: after replacing $\phi$ by $\phi-r\id$ with an appropriately chosen

2024: constant $r$ we may assume without loss of generality that $\phi$ is

2025: trace-free. Since multiplication of $\phi$ by any non-zero real number

2026: does not change the algebra $\7h=\RR\id\oplus\RR\phi$, there are

2027: essentially 3 different cases for $\phi$ with characteristic roots

2028: forming an arithmetic progression -- either $\phi$ is nilpotent or

2029: $\phi$ has pairwise different characteristic roots in $\RR$ or in

2030: $i\RR$.  Consider the manifold $M:=\5M^{2,n-2}=\5F^{2,n-2}+iV$ from

2031: Example \ruf{JP}. As already remarked in \ruf{JP} the conformal

2032: subgroup $H^{\i}:=\RRp\cd\SO(2)\subset\GL(2,\RR)$ acts transitively on

2033: $\5F^{2,n-2}$. The corresponding Lie algebra is

2034: $\7h_{\i}:=\RR\id\oplus\RR\phi^{}_{\i}$, where

2035: $\phi_{\i}:=\xi^{1,-1}-\xi^{-1,1}\in\7g_{0}$ is a trace-free

2036: semisimple endomorphism with eigenvalues in $i\RR$, see \Ruf{AQ} for

2037: the notation. Next consider the subgroup

2038: $H^{\r}:=\RRp\cd\SO(1,1)\subset\GL(2,\RR)$ with Lie algebra

2039: $\7h^{\r}:=\RR\id\oplus\RR\phi^{}_{\r}$, where

2040: $\phi_{\r}:=\xi^{1,-1}+\xi^{-1,1}\in\7g_{0}$ is a trace-free

2041: semisimple endomorphism with real eigenvalues. In particular,

2042: $H^{\r}(a)$ is open in $H^{\i}(a)$. Finally, consider the solvable

2043: subgroup

2044: $$

2045: H^{\z}:=\Big\{\Matrix\alpha0\beta\alpha\in\GL(2,\RR):\alpha>0\Big\}

2046: \steil{with Lie algebra}\7h^{\z}:=\RR\id\oplus\RR\phi^{}_{\z}\,,

2047: $$

2048: where $\phi_{\z}:=\xi^{1,-1}$ only has zero eigenvalues. Again, the

2049: orbit $H^{\z}(a)$ is open in $H^{\i}(a)$. This implies that the

2050: manifolds $M^{\phi_{\r}}$ and $M^{\phi_{\z}}$ are open subsets of

2051: $M^{\phi_{i}}=\5M^{2,n-2}$ and hence that all three of them are

2052: locally CR-equivalent. Since $\5M^{2,n-2}$ is minimal as CR-manifold

2053: the endomorphisms $\phi_{\i},\phi_{\r}$ and $\phi_{\z}$ have $a$ as

2054: cyclic vector by Proposition \ruf{PR} (what also easily can be

2055: verified). By construction, the characteristic roots of the

2056: endomorphisms $\phi_{\i}$, $\phi_{\r}$ and $\phi_{\z}$ form an

2057: arithmetic progression and represent the three types with imaginary,

2058: real and zero characteristic roots.  The estimate $\dim \7g_{0}\ge

2059: 4>\dim \7h^\psi$ for $\psi=\phi_{\i},\phi_r,\phi_n$ together with

2060: Propositions \ruf{UH} and \ruf{HU} implies that the characteristic

2061: roots of all 3 endomorphisms form an arithmetic progression.  Summing

2062: up we have proved  the following result.

2063: 

2064: \Proposition{IO} Let $\phi,\phi'\in\End(V)$ be endomorphisms with cyclic

2065: vectors $a,a'\in V$. Assume that for both endomorphisms the families

2066: of characteristic roots $\alpha_{1},\dots,\alpha_{n}$ and

2067: $\alpha'_{1},\dots,\alpha'_{n}$ form an arithmetic progression. Then

2068: the germs $(M^{\phi},a)$ and $(M^{\phi'},a')$ are

2069: CR-equivalent.\Formend

2070: 

2071: \medskip We can use Proposition \ruf{IO} to get explicit global

2072: equations for every $M^{\phi}$ where the characteristic roots of

2073: $\phi\in\Cyc(V)$ form an arithmetic progression: Indeed, we may take

2074: $\phi:=\xi^{1,-1}$ from \Ruf{AQ} on $\CC^{m+1}$ with coordinates

2075: $(z_{0},z_{1},\dots,z_{m})$ and $a:=(1,0,\dots,0)$. Then $\phi$ is

2076: nilpotent and

2077: $S:=^{}\exp(\RR\phi)(a)=\{(1,t,t^{2},\dots,t^{m}):t\in\RR\}$. Consequently,

2078: $M^{\phi}=F+iV$ is the tube over the cone $F$ generated by $S$ (that

2079: is $F=\RRp\!\cd S$). As a consequence, $F$ is an open piece of the

2080: algebraic surface given by the following explicit system of quadratic

2081: equations on $\RR^{m+1}$ with coordinates $(x_{0},x_{1},\dots,x_{m})$

2082: $$

2083: x_{0}x_{j+1}=x_{1}x_{j}\Steil{for}0<j<m\,.\Leqno{IP}

2084: $$

2085: This can be reformulated also in the following slightly different

2086: form: Let $C:=\{(t,t^{2},\dots,t^{n}):t\in\RR\}$ in $\RR^{n}$ be the

2087: {\sl twisted $n$-ic} (also called twisted cubic, quartic etc, see

2088: \Lit{HART} for interesting properties of these curves). Then the cone

2089: $\RRp\!\cd C$ generated by $C$ is a nonsingular surface outside the

2090: origin and the corresponding tube manifold is locally CR-equivalent to

2091: $M^{\phi}$ with $\phi$ as in \ruf{IO}. The twisted $n$-ic will also

2092: show up in another type of examples, compare \ruf{KU}.

2093: 

2094: \medskip Next we extend Propositions \ruf{UH} and \ruf{HU} to the case

2095: $d=2$ where the characteristic roots of $\phi$ do form an arithmetic

2096: progression.  Recall that for the light cone tube $\5M\cong\5M^{2,1}$

2097: the Lie algebra $\7g=\hol(\5M,a)$ is isomorphic to $\so(2,3)$, compare

2098: \Lit{KAZT}, \Lit{FEKA}. In particular, $\7g_{0}\cong\gl(2,\RR)$ and

2099: $\dim\7g_{1}=3$ in this case.

2100: 

2101: \Proposition{ZR} Assume that the characteristic roots of

2102: $\phi\in\Cyc(V)$ form an arithmetic progression. Then for

2103: $M=M^{\phi}$, $\,n=\dim V$ and $\7g=\hol(M,a)$ the following

2104: properties hold. \0 $\7g_{0}$ is isomorphic to $\gl(2,\RR)$ and hence

2105: has dimension $4$. \1 $\7g=\aff(M,a)$ in case $n\ge4$. In particular,

2106: $\dim\7g=n+4$ in this case.

2107: 

2108: \Proof ad (i): We may assume that $\lambda_{j}=(1-n)/2+(j-1)$ for

2109: all $j\in S$ using the notation in \Ruf{UK} and \Ruf{JG}. Then

2110: $\bigcap_{K\in\6K}\Delta_{K}=\{-1,0,1\}$. Solving \Ruf{PT} for

2111: $r_{0},r_{1}$ gives that all pairs

2112: $$

2113: \eqalign{r_{0}&=(n-1)(ue^{-t}+v_{0}-we^{t})\cr

2114: r_{1}&=2(ue^{-t}+v_{1}+we^{t})\,,\cr}\Leqno{AP}

2115: $$ $u,v_{0},v_{1},w\in\RR$ arbitrary, form the solution space. This

2116: implies $\dim\7g_{0}=4$. Since $M$ is locally CR-equivalent to

2117: $\5M^{2,n-2}$ the Lie algebra $\7g_{0}$ contains a copy of

2118: $\gl(2,\RR)$, that is $\7g_{0}\cong\gl(2,\RR)$.\nline ad (ii): Let

2119: $n\ge4$. We may assume that for $m=n-1$ the Lie algebra $\7g_{0}$ is

2120: the linear span of the vector fields \Ruf{AQ}. For every

2121: $\nu\in\ZZ^{2}$ let

2122: $\7g^{\nu}:=\{\xi\in\7g:[\zeta_{j},\xi]=\nu_{j}\xi\steil{for}j=1,2\}\,.$

2123: Then $\7g^{\nu}\subset\7g_{k}$ for $k=(\nu^{}_{1}+\nu^{}_{2})/m$ and

2124: $$

2125: \7g=\bigoplus_{\nu\in\ZZ^{2}}\7g^{\nu}\Steil{with}[\7g^{\nu},

2126: \7g^{\mu}]\subset\7g^{\nu+\mu}\,,

2127: $$ compare also (3.5) in \Lit{FEKA}.

2128: Clearly $i\dd{z_{k}}\in\7g^{-k,k-m}\steil{for all}0\le k\le m$.

2129: Because of Proposition \ruf{DC}.iv it is enough to show $\7g_{1}=0$.

2130: Assume to the contrary that there exists a non-zero $\xi\in\7g_{1}$.

2131: Then we may assume without loss of generality that $\xi\in\7g^{k,m-k}$

2132: for some $k\in\ZZ$. Let $c$ be the cardinality of $\{0\le j\le

2133: m:[\dd{z_{j}},\xi]\ne0\}$. From

2134: $\7g_{0}=\7g^{-1,1}\oplus\7g^{0,0}\oplus\7g^{1,-1}$ and

2135: $[i\dd{z_{j}},\xi]\in\7g^{k-j,j-k}$ we see $c\le3$. Assume $c=3$,

2136: which implies $0<k<m$. From $[\dd{z_{j}},\xi]\ne0$ for $j=k\pm1$ and

2137: the special form of $\xi^{-1,1}$, $\xi^{1,-1}$ in \Ruf{AQ} we see that

2138: $\xi$ must depend on all $n$ variables, a contradiction to $n>3$.  But

2139: $c\le2$ also gives a contradiction since in all spaces $\7g^{-1,1}$,

2140: $\7g^{0,0}$, $\7g^{1,-1}$ every non-zero vector field must depend on

2141: at least $n-2$ variables.\qed

2142: 

2143: \medskip Recall that $\Aut(M)$ is the group of all global

2144: CR-automorphisms and $\Aff(M)$ is the subgroup of all of affine

2145: transformations of $M$.

2146: 

2147: \Proposition{GO} Let $\phi\in\End(V)$ be a trace-free cyclic

2148: endomorphism. Then the groups $\Aut(M)$ and $\Aff(M)$

2149: coincide. Furthermore, with $n=\dim V$ the following dimension

2150: estimates hold.\0 $\dim\Aut(M)=n+4$ if the characteristic roots of

2151: $\phi$ are pairwise distinct and form an arithmetic progression in

2152: $i\RR$.\1 $\dim\Aut(M)=n+3$ if $\phi$ is nilpotent.\1

2153: $\dim\Aut(M)=n+2$ in all other cases.

2154: 

2155: \Proof Let $F:=M\cap V$ and denote by $\7a\subset\7g=\hol(M,a)$ the

2156: Lie algebra of $\Aut(M)$. Since $\7a$ contains the Euler vector field

2157: we have $\7a=\7a_{-1}\oplus\7a_{0}\oplus \7a_{1}$ for

2158: $\7a_{j}:=\7a\cap\7g_{j}$. We first determine $\dim\7a_{0}$. Because

2159: of $\7h:=\RR\id\oplus\RR\phi\subset\7a_{0}$ we have

2160: $\dim\7a_{0}\ge2$.\nline In case (i) $M$ is CR-equivalent to

2161: $\5M^{2,n-2}$, compare Example \ruf{JP}. Therefore $\GL(2,\RR)$ acts

2162: transitively on $F$ and $\dim\7a_{0}=4$ by Proposition \ruf{ZR}.\nline

2163: Next consider the case (ii), that is, $\phi$ is nilpotent. Then

2164: $\7a_{0}$ consists of all $\xi\in\7g_{0}\subset\End(V)$ with

2165: $\xi(c)\in\RR c$ for all $c\in\partial F:=\overline F\setminus F$,

2166: where $\overline F$ is the closure of $F$ in $V$. We may assume that

2167: $a=(1,0,\dots,0)\in\RR^{n}$ and $\phi=\xi^{1,-1}$ in the notation of

2168: \Ruf{AQ}. This implies

2169: $F=\{e^{s}(1,t,t^{2},\dots,t^{n-1})\in\RR^{n}:s,t\in\RR\}$ and hence

2170: $\partial F=\RR c$ for $c:=(0,\dots,0,1)$. Therefore, $\7a_{0}$ is the

2171: linear span of $\zeta_{1}$, $\zeta_{2}$ and $\xi^{1,-1}$ and

2172: $\dim\7a_{0}=3$ in this situation.\nline Next consider the case

2173: $V=V_{1}\oplus V_{2}\oplus\dots\oplus V_{n}$ where every $V_{j}$ is

2174: the $\big((1-n)/2+(j-1)\big)$ eigenspace of $\phi$, that is, the

2175: characteristic roots of $\phi$ form an arithmetic progression in

2176: $\RR$. Here $\partial F=V_{1}\cup V_{n}$ is easily verified.  The

2177: vector fields in $\7g_{0}$ are characterized by the function tuples

2178: $(r_{0},r_{1})$ in \Ruf{AP}. The condition $\xi(V_{j})\subset V_{j}$

2179: for $j=1,n$ implies that $r_{0},r_{1}$ are constant for every

2180: $\xi\in\7a_{0}$, that is $\7a_{0}=\7h$. On the other hand, if the

2181: characteristic roots of $\phi$ do not form an arithmetic progression,

2182: then also $\7g_{0}=\7h$ by Proposition \ruf{UH}.i, that is,

2183: $\dim\7a_{0}=2$ always holds in case (iii).\nline Next we show

2184: $\7a_{1}=0$ in all cases: For $n\ge4$ this follows from $\7g_{1}=0$,

2185: see Proposition \ruf{ZR}.ii. In case (iii) we have $\7a_{0}=\7h$ and

2186: the claim follows with Lemma \ruf{UF}. Therefore we only have to

2187: consider the cases (i) and (ii) for $n=3$. In case (i) $M$ is the

2188: future light cone tube $\5M$ and $\Aut(\5M)=\Aff(\5M)$ follows as a

2189: special case of Proposition 6.9 in \Lit{KAZT}. In case (ii) $M$ is a

2190: proper domain in $\5M$: We realize $\5M=\5F+i\RR^{3}$ in

2191: $\CC^{3}$ with coordinates $(z_{0},z_{1},z_{2})$ as

2192: $\5F=\{x\in\RR^{3}:x_{0}x_{2}=x^{2}_{1}\steil{and}x_{0}+x_{2}>0\}$.

2193: Then $\hol(\5M)$ is the linear span of the vector fields (3.5) and

2194: (3.7) in \Lit{FEKA}. We may assume without loss of generality that

2195: $\phi=\xi^{-1,1}=2z_{1}\dd{z_{0}}+z_{2}\dd{z_{1}}$. This implies

2196: $\5M\setminus M=\RRp\!\cd c+i\RR^{3}$ for $c:=(1,0,0)\in\5F$.

2197: We know already that $\7a$ is the linear span of the vector fields

2198: $\zeta_{1}$, $\zeta_{1}$ and $\xi^{-1,1}$. From Figure 1 and (3.7) in

2199: \Lit{FEKA} we therefore derive that either $\7a_{1}=0$ or

2200: $\7a_{1}=\RR\xi^{0,2}$ for

2201: $\xi^{0,2}:=iz^{2}_{1}\dd{z_{0}}+iz_{1}z_{2}

2202: \dd{z_{1}}+iz^{2}_{2}\dd{z_{2}}$. The latter possibility cannot occur

2203: since $\xi^{0,2}$ is not tangent to $\5M\setminus M$: Check, for

2204: instance, the point $(1,i,0)$. This proves $\7a=\7g_{-1}\oplus\7a_{0}$

2205: in all cases. As in the proof of Proposition \ruf{ZB}.iii it is shown

2206: that this implies $\Aut(M)=\Aff(M)$. The above dimension estimates for

2207: $\7a_{0}$ imply the dimension estimates in (i) -- (iii).\qed

2208: 

2209: \medskip Next we solve the local as well as the global CR-equivalence

2210: problem for all manifolds $M^{\phi}$. Because of Proposition \ruf{IO}

2211: in the local situation only the case has to be considered where the

2212: characteristic roots of $\phi$ do not form an arithmetic

2213: progression. Recall that without loss of generality we always may

2214: assume that $\phi$ is trace-free.

2215: 

2216: 

2217: \Proposition{KO} Let $\phi,\phi'\in\End(V)$ be trace-free cyclic

2218: endomorphisms with characteristic roots $\alpha_{1},\dots,\alpha_{n}$

2219: and $\alpha'_{1},\dots,\alpha'_{n}$ respectively. Suppose that the

2220: $\alpha_{1},\dots,\alpha_{n}$ do not form an arithmetic

2221: progression. Then for given cyclic vectors $a,a'\in V$ and

2222: corresponding $M=M^{\phi}$, $M'=M^{\phi'}$ the Lie algebras

2223: $\hol(M,a)$ and $\aff(M,a)$ coincide. Furthermore, the following

2224: conditions are equivalent. \0 The Lie algebras $\hol(M,a)$ and

2225: $\hol(M',a)$ are isomorphic . \1 The germs $(M,a)$ and $(M',a')$ are

2226: CR-equivalent. \1 $g\phi'g^{-1}=r\phi$ for suitable $g\in\GL(V)$ and

2227: $r\in\RR^{*}$. \1 There exists a permutation $\pi\in\7S_{n}$ and an

2228: $r\in\RR^{*}$ with $\alpha'_{j}=r\alpha_{\pi(j)}$ for all $j$.

2229: 

2230: \Proof $\7g:=\hol(M,a)=\aff(M,a)$ and $\dim\7g=n+2$ follows from

2231: Propositions \ruf{UH} and \ruf{HU}.\nline \To12 With $\7g$ also

2232: $\7g':=\hol(M',a')$ has dimension $n+2$. Therefore

2233: also $\alpha'_{1},\dots,\alpha'_{n}$ do not form an

2234: arithmetic progression, see Proposition \ruf{ZR}. This implies

2235: $\7g_{0}=\7h$, $\7g'_{0}=\7h'$ and (ii) follows with Proposition

2236: \ruf{PQ}. \nline\To23 Let $g$ be a CR-isomorphism

2237: $(M,a)\to(M',a')$. Then $g\in\GL(V)$ as a consequence of Proposition

2238: \ruf{ZB}.iii. and clearly $g\7h g^{-1}=\7h'$. Since

2239: $\RR\phi'\subset\7h'$ is precisely the subset of all trace-free

2240: endomorphisms (iv) follows. \nline The remaining implications are easy

2241: to check and left to the reader.\qed

2242: 

2243: \ifarx\eject\fi

2244: 

2245: \Proposition{BP} Let $\phi,\phi'\in\End(V)$ be trace-free cyclic

2246: endomorphisms and $M:=M^{\phi}$, $M':=M^{\phi'}$. Then the following

2247: conditions are equivalent. \0 The groups $\Aff(M)$ and $\Aff(M')$ are

2248: isomorphic. \1 $M$ and $M'$ are globally CR-equivalent. \1

2249: $g\phi'g^{-1}=r\phi$ for suitable $g\in\GL(V)$ and $r\in\RR^{*}$.

2250: \Proof \To13 Suppose that (i) holds. Because of Proposition \ruf{GO}

2251: we may assume $\dim\Aut(M)=n+2$ without loss of generality.  Then

2252: (iii) follows from Proposition \ruf{KO} if at least for one of the

2253: $\phi,\phi'$ the characteristic roots do not form an arithmetic

2254: progression. In the remaining cases the claim follows from Proposition

2255: \ruf{GO} since for both endomorphisms the characteristic roots form an

2256: arithmetic progression in $\RR$ and are pairwise distinct.\nline\To32

2257: $\Rightarrow$ (i) is obvious since $\Aut(M)=\Aff(M)$ and

2258: $\Aut(M')=\Aff(M')$ by Proposition \ruf{GO}.\qed

2259: 

2260: \ifarx\bigskip\else\medskip\fi

2261: \Joker{MO}{Some moduli spaces} For fixed $n=\dim V$ let $\6M$ be

2262: the space of all global CR-equivalence classes $[M^{\phi}]$ of

2263: manifolds $M^{\phi}$ with $\phi\in\Cyc(V)$, that is, every

2264: $[M^{\phi}]$ is the set of all $M^{\psi}$ that are globally

2265: CR-equivalent to $M^{\phi}$. Furthermore put

2266: $$\Phi:=\{\phi\in\Cyc(V):\tr(\phi)=0\}\Steil{and}\6M^{*}:=

2267: \big\{[M^{\phi}]\in\6M: \phi\in\Phi,\,\phi^{n}\ne0\big\}\,.$$ The

2268: reductive group $\RR^{*}\Times\GL(V)$ acts on $\End(V)$ by

2269: $\phi\mapsto rg\phi g^{-1}$ for every $(r,g)\in\RR^{*}\Times\GL(V)$

2270: and leaves the cone $\Phi$ invariant. By Proposition \ruf{BP} $\6M$

2271: can be identified as set with the quotient

2272: $\Phi/(\RR^{*}\Times\GL(V))$.  This quotient can be built in several

2273: steps: For every $j$ let $\sigma_{j}(\phi)\in\RR$ be the $j^{\th}$

2274: elementary symmetric function in $n$ variables evaluated on the

2275: characteristic roots of $\phi$, that is,

2276: $$X^{n}+\sum^{n}_{j=2}(-1)^{j}\sigma_{j}(\phi)X^{n-j}\;\in\;\RR[X]$$

2277: is the characteristic polynomial of $\phi\in\Phi$. Let $W:=\RR^{n-1}$

2278: with coordinates $(x_{2},\dots,x_{n})$ and denote by $\sigma:\Phi\to

2279: W$ the mapping given by

2280: $\phi\mapsto(\sigma_{2}(\phi),\dots,\sigma_{n}(\phi))$. Since every

2281: real polynomial factors in a product of linear and quadratic real

2282: polynomials, the map $\sigma$ is surjective and $\6M$ can be

2283: canonically identified as set with the quotient $W/\RR^{*}$, where

2284: $\RR^{*}$ acts on $W$ by

2285: $(x_{2},x_{3},\dots,x_{n})\mapsto(t^{2}x_{2},t^{3}x_{3},\dots,t^{n}x_{n})$

2286: for every $t\in\RR^{*}$. The subgroup $\{\pm1\}\subset\RR^{*}$ leaves

2287: the sphere $S^{n-2}=\{x\in W:\sum x_{j}^{2}=1\}$ invariant and

2288: $\6M^{*}$ can be identified with the quotient

2289: $Q^{n-2}:=S^{n-2}/\{\pm1\}$.  In general, $Q^{n-2}$ can be stratified

2290: into a finite number of manifolds. For instance, $Q^{1}$ is a compact

2291: line segment and $Q^{2}$ is homeomorphic to the sphere $S^{2}$. At

2292: this point a word of caution is necessary: We do not give a topology

2293: to $\6M$, the topology on $Q^{n-2}$ only serves for the readers

2294: imagination.

2295: 

2296: \ifarx\bigskip\fi

2297: Instead of $\6M$ we can also consider the space of {\sl local }

2298: CR-equivalence classes for manifolds of type $M^{\phi}$. By our

2299: results this space is of the form $\6M/\Sim\,$, where the equivalence

2300: relation $\Sim$ on $\6M$ just identifies the $3$ equivalence classes

2301: $[M^{\phi}]\in\6M$ such that the characteristic roots of $\phi$ form

2302: an arithmetic progression. Clearly, $\6M/\Sim=\6M^{*}/\Sim$ can be

2303: obtained by identifying two points in $Q^{n-2}$. In the spacial case

2304: $n=3$ the endpoints of the line segment $Q^{1}$ have to be identified,

2305: that is, $\6M$ can be thought of in this case as the circle

2306: $\overline{\RR}:=\RR\cup\infty$ (without topology) where the point

2307: $\infty$ corresponds to the class represented by the future light cone

2308: tube $\5M$. To be more specific, call in case $n=3$ for every

2309: $\phi\in\Phi$

2310: $$

2311: \mu\big(M^{\phi}\big):=-{\sigma_{2}(\phi)^{3}\over

2312: \sigma_{3}(\phi)^{2}} \;\;\in\;\;\overline{\RR}\Leqno{KW}

2313: $$

2314: the {\sl modulus} of the CR-manifold $M^{\phi}$ (with $t/0:=\infty$

2315: for all $t\in\RR$). It is clear that $M^{\phi}$ and $M^{\psi}$ in case

2316: $n=3$ are locally CR-equivalent if and only if they have the same

2317: modulus. A special meaning has the modulus $\mu_{0}:=27/4$:

2318: For real moduli $>\mu_{0}$ the endomorphism $\phi$ has $3$ distinct

2319: real eigenvalues while in case of real moduli $<\mu_{0}$ the

2320: endomorphism $\phi$ has one real and two purely imaginary eigenvalues.

2321: 

2322: \ifarx\vfill\eject\fi

2323: \bigskip\medskip\Joker{KU}{Another type of examples} For $k=3$ and

2324: $c\ge1$ let $V$, $\;\Gamma=\GL(2,\RR)^{0}$ and

2325: $a=u^{m}_{2}+u_{1}u^{m-1}_{2}\in V$ be as in Example

2326: \ruf{JP}. Consider the subgroup

2327: $$

2328: \Sigma:=\Big\{\Matrix\alpha0\beta1:\alpha\in\RRp,\beta\in\RR\Big\}

2329: \;\subset\;\Gamma\,.\Leqno{LP}

2330: $$

2331: Then $F:=\Sigma(a)$ is a homogeneous surface in

2332: $V$. For $o:=u_{2}^{m}$ the orbit $C:=\Sigma(o)$ is a homogeneous

2333: curve. Identify $V$ and $\RR^{m+1}$ with coordinates

2334: $(x_{0},x_{1},\dots,x_{m})$ as in Example \ruf{JP}. Then

2335: $a=(1,1,0,\dots,0),\,o=(1,0,\dots,0)\in\RR^{m+1}$ and the Lie algebra

2336: of $\Sigma$ corresponds to the linear span of the two vector fields

2337: $\zeta_{1}$ and $\xi^{1,-1}$ in \Ruf{AQ}. In particular

2338: $$

2339: C=\{(1,t,t^{2},\dots,t^{m}):t\in\RR\}

2340: $$

2341: and $T_{o}C=\RR\cd b$ with $b:=(0,1,0,\dots,0)$ for the tangent space

2342: at $o\in C$. On the other hand, the affine {half}line $o+\RRp\!\cd b$

2343: is contained in $F$. The geometric meaning of this is the following:

2344: The development $S:=\bigcup_{c\in C}(c+T_{c}C)$ of the curve $C$ is

2345: divided by $C$ in two $\Sigma$-orbits, one of which is $F$ (compare

2346: \Lit{EAEZ}, \p45 for the special case $m=3$). Now identify the

2347: $\Sigma$-invariant hyperplane $W:=\{x\in V:x_{0}=1\}$ with $\RR^{m}$

2348: by 'dropping the coordinate' $x_{0}$. Then $C$ becomes the twisted

2349: $m$-ic $\{(t,t^{2},\dots,t^{m}):t\in\RR\}$, $o$ becomes the origin and

2350: $a$ the first basis vector $(1,0,\dots,0)$ in $\RR^{m}$. In the

2351: coordinates of $\RR^{m}$ the vector fields $\zeta_{1}$ and

2352: $\xi^{1,-1}$ are affine and have the forms

2353: $$

2354: \zeta_{1}=\sum^{m}_{j=1}jz_{j}\dd{z_{j}}\Steil{and}

2355: \xi^{1,-1}=\dd{z_{1}}+\sum^{m}_{j=2}jz_{j-1}\dd{z_{j}}\,.\Leqno{HQ}

2356: $$

2357: With Proposition \ruf{US} it is easily verified that

2358: $K^{r}_{a}F=\{x\in\RR^{m}:x_{j}=0\steil{if}j+r>2\}$ for all

2359: $r\ge0$. Since the twisted $m$-ic is not contained in any hyperplane

2360: of $\RR^{m}$ we therefore get that the tube $M:=C+i\RR^{m}$ is a

2361: homogeneous minimal $2$-nondegenerate submanifold of $\CC^{m}$ with

2362: CR-dimension $2$ and CR-codimension $m-2$. Notice that for the cone

2363: $\RRp\!\cd F$ generated by $F$ in $V$ the tube $\RRp\!\cd

2364: F+iV$ is an open piece of $\5M^{3,n-3}$, $n:=m+1$, and hence is

2365: $3$-nondegenerate.\nline Denote for every integer $j$ by $\7g^ {(j)}$

2366: the $k$-eigenspace of $\ad(\zeta_{1})$ in $\7g:=\hol(M,a)$. Then every

2367: $\xi\in\7g^{(j)}$ is a complex linear combination of monomial vector

2368: fields $z_{1}^{\nu_{1}}z_{2}^{\nu_{2}}\cdots

2369: z_{m}^{\nu_{m}}\dd{z_{p}}$ with

2370: $\nu_{1}+2\nu_{2}+\dots+m\nu_{m}=j+p$. As in the proof of Proposition

2371: \ruf{DC}, it is shown that $\7g$ has the $\ZZ$-gradation

2372: $$

2373: \7g=\bigoplus_{j\ge-m}\7g^{(j)}\,.\Leqno{RU}

2374: $$ It can be seen that $\7g$ is the linear span of all $i\dd{z_{j}}$,

2375: $1\le j\le m$, as well as $\zeta_{1}$ and $\xi^{1,-1}$. In particular,

2376: $\7g$ is a solvable Lie algebra of dimension $m+2$, coincides with

2377: $\aff(M,a)$ and has commutator subgroup of dimension $m+1$. A proof

2378: will be sketched for the special case $m=3$ in Example \ruf{EV}.

2379: 

2380: 

2381: 

2382: \KAP{Acht}{Homogeneous 2-nondegenerate CR-manifolds in dimension 5}

2383: \medskip

2384: 

2385: In this section we specialize the examples of the previous section to

2386: the case $V=\RR^{3}$. We start with manifolds of type

2387: $M=M^{\phi}=F^{\phi}+i\RR^{3}$ in $\CC^{3}$, Then the local

2388: CR-equivalence classes of these manifolds are parameterized by the

2389: modules $\mu(M)\in\overline{\RR}$, compare \Ruf{KW}. The $\phi$

2390: occurring in the following examples not necessarily are trace-free but

2391: easily could be transformed to.  As defined in the previous section

2392: let $\mu_{0}:=27/4$.

2393: 

2394: 

2395: \Example{EI}{\fett ($\mu=\infty$) } Let

2396: $F:=\{x\in\RR^{3}:x_{1}^{2}+x_{2}^{2}=x^{2}_{3},\,x_{3}>0\}$ be the

2397: future light cone. This surface occurs as $F^{\phi}$ for

2398: $\phi:=x_{2}\dd{x_{1}}-x_{1}\dd{x_{2}}$ having spectrum $\{\pm i,0\}$.

2399: 

2400: 

2401: \Example{EY}{\fett ($\mu<\mu_{0}$) } For $\omega>0$ let

2402: $F\subset\RR^{3}$ be the orbit of $(1,0,1)$ under the group of all

2403: linear transformations $x\mapsto r(\cos t\,x_{1}-\sin t\,x_{2},\sin

2404: t\,x_{1}+\cos t\,x_{2},e^{\omega t}x_{3})$, $r\in\RRp,t\in\RR$.\nline

2405: With $r:=(x_{1}^{2}+x_{2}^{2})^{1/2}$, the manifold $F$ is given in

2406: $\{x\in\RR^{3}:r>0\}$ by the explicit equations

2407: $$

2408: x_{3}=r\exp\big(\omega\cos^{-1}(x_{1}/r)\big)=r\exp\big

2409: (\omega\sin^{-1}(x_{2}/r)\big)\,,

2410: $$

2411: where locally always one of these suffices. A suitable choice is

2412: $\phi=x_{1}\dd{x_{2}}-x_{2}\dd{x_{1}}+\omega x_{3}\dd{x_{3}}$ with

2413: spectrum $\{\pm i,\omega\}$.

2414: 

2415: \medskip

2416: 

2417: \Example{EZ}{\fett ($\mu=\mu_{0}$) } Let $F\subset\RR^{3}$ be the

2418: orbit of $(1,0,1)$ under the group of all linear transformations

2419: $x\mapsto r(x_{1},x_{2}+tx_{1},e^{t}x_{3})$ with

2420: $r\in\RRp,t\in\RR$, that is,

2421: $$

2422: F=\big\{x\in\RR^{3}:x_{1}>0,\;x_{3}^{}= x_{1}^{}\re1e^{ x_{2}/x_{1}}

2423: \big\}\,.

2424: $$

2425: Here $\phi=x_{1}\dd{x_{2}}+x_{3}\dd{x_{3}}$ has characteristic roots

2426: $0,0,1$.

2427: 

2428: \medskip

2429: 

2430: \Example{EX}{\fett ($\mu>\mu_{0})$ } For $\theta>2$ let

2431: $$

2432: F:=\{x\in(\RRp)^{3}:x_{3}=x_{1}(x_{2}/x_{1})^{\theta}\}\,.

2433: $$

2434: Here $\phi=x_{2}\dd{x_{2}}+\omega x_{3}\dd{x_{3}}$ has eigenvalues

2435: $\{0,1,\omega\}$.

2436: 

2437: \bigskip

2438: 

2439: The following is Example \ruf{KU} specialized to $m=3$.

2440: 

2441: \Example{EV} Let $\Sigma$ be the group generated by the following two

2442: one-parameter groups

2443: $$

2444: x\mapsto(e^{t}x_{1},e^{2t}x_{2},e^{3t}x_{3})\,,\qquad

2445: x\mapsto(x_{1}+t,x_{2}+2tx_{1}+t^{2},x_{3}+3tx_{2}+3t^{2}x_{1}+

2446: t^{3})\Leqno{NM}

2447: $$

2448: of affine transformations on $\RR^{3}$. Then $\Sigma$ is isomorphic to

2449: the group defined in \Ruf{LP}. For $a:=(1,0,0)$ the orbit

2450: $F:=\Gamma(a)$ is

2451: $$

2452: F\;=\;\{(t,t^{2},t^{3})+r(1,2t,3t^{2})\in\RR^{3}:r\in\RRp,t\in\RR\}\,.

2453: $$

2454: The tube $M:=F+i\RR^{3}$ is an affinely homogeneous

2455: $2$-nondegenerate CR-manifold. The Lie algebra of $\Sigma$ is spanned

2456: by the affine vector fields

2457: $$

2458: \zeta_{1}:=z_{1}\dd{z_{1}}+2z_{2}\dd{z_{2}}+3z_{3}\dd{z_{3}}

2459: \Steil{and}\xi^{1,-1}:=\dd{z_{1}}+z_{1}\dd{z_{2}}+z_{2}\dd{z_{3}}

2460: \,,

2461: $$

2462: compare also \Ruf{HQ}. The Lie algebra $\7g:=\hol(M,a)$ is of finite

2463: dimension and has the gradation \Ruf{RU} for $m=3$, where $\7g^ {(k)}$

2464: is the $k$-eigenspace of $\ad(\zeta_{1})$.  We claim that $\7g$ has

2465: dimension $5$ and coincides with $\aff(M,a)$. The proof consists of

2466: several elementary steps which we only sketch here. To begin with,

2467: define $\7a^{(k)}\subset\7g^{(k)}$ by

2468: $$

2469: \7a^{(-3)}:=\RR\, i\dd{z_{3}}\,,\; \7a^{(-2)}:=\RR

2470: i\dd{z_{2}}\,,\;\7a^{(-1)}:=\RR

2471: i\dd{z_{1}}\oplus\RR\phi\,,\;\7a^{(0)}:=\RR\zeta

2472: $$

2473: and $\7a^{(k)}:=0$ for all other $k$. By induction on $k$ it is seen

2474: that $\7g^{(k)}=\7a^{(k)}$ holds for all $k$: For $k\le-3$ this is

2475: obvious. For $k=-2$ suppose there exists a

2476: $\xi\in\7g^{(-2)}\setminus\7a^{(-2)}$. Then

2477: $\xi=\alpha\dd{z_{2}}+\beta z_{1}\dd{z_{3}}$ for some

2478: $\alpha,\beta\in\CC$. From $[\xi,\7a^{(-1)}]\subset\7a^{(-3)}$ we get

2479: $\beta\in\RR$ and then $\xi_{a}\in T_{a}M$ implies $\beta=0$. But then

2480: $\xi\in\7a^{(-2)}$ since $\7a^{(-2)}\!\oplus

2481: i\7a^{(-2)}\not\subset\7g^{(-2)}$, a contradiction. For $k\ge-1$ the

2482: procedure is as follows. Suppose there exists a

2483: $\xi\in\7g^{(k)}\setminus\7a^{(k)}$. Then write $\xi$ as complex

2484: linear combination of monomial vector fields as mentioned above and

2485: subtract from $\xi$ a suitable element of $\7a^{(k)}$ thus killing as

2486: many coefficients in front of monomial terms of the form

2487: $f(z)\dd{z_{1}}$ as possible. By induction hypothesis

2488: $[\7g^{(k)},\7a^{(j)}]\subset\7a^{(k+j)}$ holds for all $j<0$ and

2489: gives $\xi=0$, a contradiction.  This proves the claim and also the

2490: first part of the following statement.

2491: 

2492: 

2493: \Lemma{LS} Let $M:=F+i\RR^{3}$ for $F=H(a)\subset\RR^{3}$ as in

2494: Example \ruf{EV}. Then $\7g=\hol(M,a)$ is a solvable Lie algebra of

2495: dimension $5$ with commutator algebra $[\7g,\7g]$ of dimension

2496: $4$. Furthermore, $\Aut(M,a)=\{\id\}$.

2497: 

2498: \Proof Fix $a:=(1,0,0)\in F$ and write $\xi_j:=i\ddz{j}$ for

2499: $j=1,2,3$. Let $h\in \Aut(M,a)$ and $\Theta$ the induced Lie algebra

2500: automorphism of $\;\7l:=\7g\oplus i\7g$. With $\7g$ the subspaces

2501: $$

2502: \7n:=[\7g,\7g]=\bigoplus_{k<0}\7g^{(k)}\;,\qquad

2503: [\7n,\7n]=\RR\xi_{2}\oplus

2504: \RR\xi_{3}\Steil{and}[\7n,[\7n,\7n]]=\RR\xi_{3}

2505: $$

2506: are stable under $\Theta$ and hence also

2507: $$

2508: \7n\cap\;K_aM = \RR\xi_1\Steil{and}[\7n,\7n]\cap H_aM= \RR \xi_2\,,

2509: $$

2510: where $\7g$ and the tangent space $T_aM$ are identified via the

2511: evaluation map. In particular, $\Theta(\7P_{-1})=\7P_{-1}$ and

2512: $h(z)=g(z)+c$ for a diagonal matrix $g\in\GL(3,\RR)$ with $c=g(a)-a$,

2513: compare \Ruf{GP}. Taking commutators of $h$ with all elements in the

2514: second $1$-parameter group of \Ruf{NM} and then taking the derivative

2515: by $t$ at $t=0$ gives $c\dd z\in\7g$. From $\7g\cap\;i\7g=0$ we

2516: conclude $c=0$ and thus $g_{11}=1$ for the diagonal matrix $g$. Now

2517: $\psi$ is the unique vector field in $\7g$ that has the value $\dd z$

2518: at both points $0,a$, implying $\Theta(\psi)=\psi$. Therefore $g$ is

2519: the unit matrix and $h$ is the identity in $\Aut(M,a)$.\qed

2520: 

2521: 

2522: \smallskip For every $M=F+i\RR^{3}$ with $F\subset\RR^{3}$ a

2523: cone from Example \ruf{EI} -- \ruf{EX} the commutator of $\hol(M,a)$

2524: has either dimension $10$ (Example \ruf{EI}) or dimension $3$ (all the

2525: others). As a consequence of Proposition \ruf{KO} and Lemma \ruf{LS}

2526: we therefore get:

2527: 

2528: \Proposition{UA} The CR-manifolds $M=F+i\RR^{3}$ with

2529: $F\subset\RR^{3}$ occurring in the examples \ruf{EI} -- \ruf{EV}, are

2530: all homogeneous and $2$-nondegenerate. Furthermore, they are mutually

2531: locally CR-inequivalent.\Formend

2532: 

2533: 

2534: \medskip With an argument from \Lit{FEKA} together with 2.5.10 in

2535: \Lit{HORM} a holomorphic extension property for {\sl global}

2536: continuous CR-functions $f$ on $M=F\oplus i\RR^{3}\subset\CC^{3}$, $F$

2537: one of the cones from examples \ruf{EI} -- \ruf{EV}, can be obtained:

2538: Every such $f$ has a unique continuous extension to the convex hull

2539: $\hat{M}$ of $M$ in $\CC^{3}$ that is holomorphic on the interior of

2540: $\hat{M}$ with respect $\CC^{3}$. Since $M$ is completely contained in

2541: the interior of $\hat{M}$ in case $F$ belongs to Example \ruf{EY},

2542: every global continuous CR-function on such an $M$ is real-analytic.

2543: 

2544: \bigskip For the tube $\5M$ over the future light cone (that is

2545: Example \ruf{EI}) there exist many (even simply-connected) homogeneous

2546: CR-manifolds that are all locally CR-equivalent to $\5M$ but are

2547: mutually non-diffeomorphic, compare \Lit{KAZT}. In contrast to this,

2548: using already Theorem II from Section \ruf{Neun}, we can state the

2549: following global result:

2550: 

2551: 

2552: \Proposition{UB} Let $M$ be a homogeneous $2$-nondegenerate

2553: CR-manifold that is not locally CR-equiva\-lent to the tube $\5M$ over

2554: the future light cone. Then $M$ is simply connected and $\Aut(M)$ is a

2555: solvable Lie group of dimension 5 acting transitively and freely on

2556: $M$. For every $a\in M$ the stability group $\Aut(M,a)$ is trivial and

2557: every homogeneous real-analytic CR-manifold $M'$, that is locally

2558: CR-equivalent to $M$, is already globally CR-equivalent to $M$.

2559: 

2560: \Proof By Theorem II $M=F+iV$ for  $F$ as in one of the

2561: examples \ruf{EY} -- \ruf{EV}. It is easily checked that $F$ and hence

2562: $M$ is simply connected. In case $F$ is a cone, the claim follows with

2563: Lemma \ruf{KR}. Therefore we may assume that $F$ is the

2564: submanifold of Example \ruf{EV}. But then $\Aut(M,a)$ is the trivial

2565: group by Lemma \ruf{LS} and $\Aut(M)$ has trivial center by

2566: Proposition \ruf{ZI}. But then the claim follows from Proposition

2567: \ruf{UQ}.~ \qed

2568: 

2569: \bigskip The affinely homogeneous surfaces $F\subset\RR^{3}$ of

2570: examples \ruf{EI} -- \ruf{EV} occur already in \Lit{EAEZ} \p43. There

2571: the surfaces are presented in their affine normal forms: Our Example

2572: \ruf{EI} ($\mu=\infty$) corresponds to P4, the Examples \ruf{EY} --

2573: \ruf{EX} ($\mu\in\RR$) to P3 and Example \ruf{EV} to P1. The remaining

2574: degenerate types in \Lit{EAEZ}, types P2$^{\pm}$ and P5, do not show

2575: up among our examples since the associated tube manifolds are

2576: holomorphically degenerate.

2577: 

2578: Let us consider the type P3 in \Lit{EAEZ} \p43 a little bit closer.

2579: This is the family of local surfaces in $\RR^{3}$ given by the local

2580: equations in affine normal form

2581: $$x_{3}=x_{1}^{2}+x_{1}^{2}x_{2}+x_{1}^{2}x_{2}^{2}+x_{1}^{5}+

2582: x_{1}^{2}x_{2}^{3}+4x_{1}^{5}x_{2}+x_{1}^{2}x_{2}^{4}+ax_{1}^{7}

2583: +10x_{1}^{5}x_{2}^{2}+x_{1}^{2}x_{2}^{5}+{\rm O}(8),\Leqno{OQ}$$ where

2584: $a\in\RR$ is an arbitrary parameter and ${\rm O}(8)$ for every fixed

2585: $a$ is a convergent power series in $x=(x_{1},x_{2},x_{3})$ vanishing

2586: of order $\ge8$ at the origin and uniquely determined by the

2587: requirement, that \Ruf{OQ} defines near the origin of $\RR^{3}$ a

2588: locally affinely homogeneous surface. Different values of $a\in\RR$

2589: give locally affinely inequivalent surfaces and there tubes in

2590: $\CC^{3}$ correspond in a 1-1-way to our Examples \ruf{EY} --

2591: \ruf{EX}. It is not difficult to see that the modulus $\mu$ of every

2592: such surface is related to the parameter $a$ in \Ruf{OQ} by the

2593: formula $100\mu=(28a)^{3}$.

2594: 

2595: \medskip In \Lit{DOKR}, \Lit{EAEZ} all locally affinely homogeneous

2596: surfaces in $\RR^{3}$ have been classified up to local affine

2597: equivalence. Inspecting the degenerate surfaces in these classifications

2598: gives together with our results the following

2599: 

2600: \Proposition{IZ} Let $F$ be a locally affinely homogeneous surface in

2601: $\RR^{3}$ and assume that the corresponding tube $M:=F+i\RR^{3}$

2602: in $\CC^{3}$ is $2$-nondegenerate. Then \0 $M$ is locally

2603: CR-equivalent to a manifold occurring in the Examples \ruf{EI} --

2604: \ruf{EV}, and \1 for every further locally affinely homogeneous

2605: surface $F'$ in $\RR^{3}$ the corresponding tubes $M$ and

2606: $M':=F'+i\RR^{3}$ are locally CR-equivalent if and only if $F$

2607: and $F'$ are locally affinely equivalent.\Formend

2608: 

2609: 

2610: \Partskip

2611: 

2612: \ifarx\bigskip\fi

2613: 

2614: \centerline{\twelverm PART 2: The classification}

2615: 

2616: 

2617: \KAP{Neun}{Lie-theoretic characterization of locally homogeneous

2618: CR-manifolds}

2619: 

2620: In this part of the paper we  classify

2621: all homogeneous 5-dimensional 2-nondegenerate CR-manifolds up to

2622: local CR-equivalence, that is, we

2623: carry out the proof of

2624: 

2625: 

2626: \medskip\noindent

2627: {\bf Theorem II.} {\sl

2628:    Let $M$ be a locally homogeneous $2$-nondegenerate

2629: real-analytic CR-manifold of dimension $5$. Then $M$ is locally

2630: CR-equivalent to a tube $F+ i\RR^{3}\;\subset\;\CC^{3}$, where

2631: $F\subset\RR^{3}$ is one of the affinely homogeneous surfaces

2632: occurring in the Examples \ruf{EI} -- \ruf{EV}.}

2633: 

2634: \smallskip\noindent We call (in accordance with Section \ruf{Zwei}) a

2635: real-analytic CR-manifold $M$ {\sl locally homogeneous} at a point

2636: $o\in M$ if there exists a Lie subalgebra $\7g\subset\hol(M,o)$ of

2637: finite dimension such that the canonical evaluation map $\7g\to

2638: T_{o}M$ is surjective, that is, such that the tangent vectors

2639: $\xi_{o}$, $\xi\in\7g$, span the tangent space $T_{o}M$. If this is

2640: the case we also call the corresponding CR-germ $(M,o)$ {\sl locally

2641: homogeneous}. If a particular locally transitive $\7g\subset\hol(M,o)$

2642: has been fixed we also say that the germ $(M,o)$ is $\7g$-homogeneous.

2643: 

2644: The proof of Theorem II relies on a natural equivalence (see

2645: \Lit{FELS}, Prop. 4.1) between the category of CR-manifold germs with

2646: a locally transitive Lie algebra action and a certain purely

2647: algebraically defined category. Before we briefly outline the main

2648: steps of our proof, we recall the notion of a CR-algebra, taken from

2649: \Lit{MENA}, and introduce some notation.

2650: 

2651: \Definition{FJ} A CR-algebra is a pair $(\7g,\7q)$, where $\7g$ is a

2652: real Lie algebra of {\sl finite} dimension and $\7q$ is a complex Lie

2653: subalgebra of the complexification $\7l:=\7g\oplus i\7g$. The

2654: CR-algebra $(\7g,\7q)$ is called {\sl effective} if $0$ is the only

2655: ideal of $\7g$ contained in $\7g\cap\7q$. \Formend

2656: 

2657: \Remarks{XA} \0 In \Lit{MENA} also the case is allowed where $\7g$ has

2658: infinite dimension, but $\7q$ has to have finite codimension in $\7l$.

2659: In this part of the paper however, only finite-dimensional Lie

2660: algebras occur. \1 The CR-algebras form in an obvious way a category:

2661: A morphism $(\7g,\7q)\to(\7g',\7q')$ of CR-algebras is a Lie algebra

2662: homomorphism $\7g\to\7g'$ in the usual sense whose complex linear

2663: extension $\7l\to\7l'$ maps $\7q$ to $\7q'$. Unfortunately, the

2664: resulting notion of isomorphism between CR-algebras is too strict for

2665: our purposes. We therefore mainly work with the coarser notion of

2666: geometric equivalence between CR-algebras to be introduced later. \1

2667: The geometric situation behind the notion of a CR-algebra is the

2668: following: Let $Z$ be a complex manifold homogeneous under a complex

2669: Lie group $L$, let $o\in Z$ be a point with isotropy subgroup

2670: $Q\subset L$ at $o$ and $G\subset L$ a connected real form of $L$,

2671: that is, the connected identity component of the fixed point set

2672: $L^\sigma$ for an involutive antiholomorphic automorphism $\sigma$ of

2673: $L$. Then each $G$--orbit $M$ in $Z$ is a generic (immersed)

2674: CR-submanifold. Let $\7g,\7q$ be the Lie algebras of $G$ and $Q$,

2675: respectively. Then $(\7g,\7q)$ is a CR-algebra that completely

2676: describes the CR-germ $(M,o)$ together with the local action of $G$

2677: near $o$. \1 In general, however, not every CR-algebra $(\7g,\7q)$ can

2678: be obtained from a global situation as described in (iii), only from a

2679: more general local setting.  Nevertheless, the set \CRG of all

2680: CR-equivalence classes of locally homogeneous CR-germs and the set

2681: \CRA of all geometric equivalence classes of CR-algebras stand in a

2682: canonical 1-1-correspondence, compare also Section 4 in

2683: \Lit{FELS}. For convenience of the reader, we briefly describe below

2684: this correspondence in both directions. As a reference for a general

2685: discussion of `local' and `infinitesimal' actions we refer to the

2686: original paper of Palais, \Lit{PALA}.\Formend

2687: 

2688: \medskip\noindent In the following let $M$ always be a locally

2689: homogeneous real-analytic CR-manifold with base point $o\in M$ that

2690: {\sl locally} can be embedded in some $\CC^{n}$ (equivalently, $M$

2691: is an (abstract) real-analytic {\sl involutive} CR-manifold as defined

2692: at the end of Section \ruf{Zwei}). Each such CR-manifold

2693: can globally and generically be embedded into a complex manifold $Z$

2694: \Lit{ANFR}. Since we are only interested in the local structure of $M$

2695: at $o$ and therefore mostly deal with CR-germs $(M,o)$, we may assume

2696: without loss of generality that $M$ is embedded in a complex vector

2697: space $E\cong \CC^n$ as a locally closed generic CR-submanifold.

2698: 

2699: Next we describe the interplay between locally homogeneous CR-germs

2700: and CR-algebras more closely. In particular, we give two canonical

2701: constructions that induce the 1-1-correspondence between the sets \CRG

2702: and \CRA mentioned in \ruf{XA}.iv and also allow the precise

2703: definition of `geometric equivalence' for CR-algebras:

2704: 

2705: Let $(M,o)$ be a CR-germ and $\7g\subset\hol(M,o)$ a locally

2706: transitive Lie subalgebra of finite dimension. Then an effective

2707: CR-algebra $(\7g,\7q)$ can be associated in the following way: To

2708: begin with, realize $\hol(M,o)$ in the canonical way as real Lie

2709: subalgebra of the complex Lie algebra $\hol(E,o)$. This is possible,

2710: since we assumed $M$ to be generic in $E$ and we can use Proposition

2711: 12.4.22 in \Lit{BERO}. As in Definition \ruf{FJ} we always denote by

2712: $\7l=\7g^{\CC}=\7g\oplus i\7g$ the formal complexification of

2713: $\7g$. Now $\Xi(\xi+ i\eta):=\xi+J\eta$ defines a Lie algebra

2714: homomorphism $\Xi:\7l\to\hol(E,o)$, where $J$ denotes the complex

2715: structure tensor $J:TE\to TE$. We will not make a notational

2716: distinction between the complex structures in $\7l$ or $TZ$, and write

2717: `$i$' for it. The homomorphism $\Xi$ is in general not injective (it

2718: is, if $M$ is holomorphically nondegenerate). Let $\7q\subset\7l$ be

2719: the $\Xi$-preimage of the isotropy subalgebra

2720: $\{\xi\in\hol(E,o):\xi_{o}=0\}$. Then the CR-algebra $(\7g,\7q)$ is

2721: called a {\sl CR-algebra associated} to the locally homogeneous

2722: CR-germ $(M,o)$. It is obvious that $\7g\cap\7q$ is nothing but the

2723: isotropy subalgebra $\7g_{o}=\{\xi\in\7g:\xi_{o}=0\}$ and the tangent

2724: space $T_{o}M$ can be canonically identified with $\7g/\7g_{o}$. Also

2725: the holomorphic tangent space $H_{o}M$ and the partial complex

2726: structure $J:H_oM\to H_oM$ (equivalently: the decomposition

2727: $H^{1,0}_oM\oplus H^{0,1}_oM$ of the complexification

2728: $H^{\CC}_oM=H_oM\otimes \CC$) can be read off the CR-algebra

2729: $(\7g,\7q)$: Let $\sigma$ be the conjugate linear involution of $\7l$

2730: with $\7l^{\sigma}:=\Fix(\sigma)=\7g$. Then it is easily verified that

2731: $\7H:=(\7q+\sigma\7q)^{\sigma}$ coincides with $\{\xi\in\7g:\xi_{o}\in

2732: iT_{o}M\}$, that is, $\7H/\7g_{o}$ is canonically isomorphic to

2733: $H_{o}M$ (the capital letter for $\7H$ is chosen to indicate that

2734: $\7H$ in general is not a Lie algebra, only a linear

2735: subspace). Further, $H^{0,1}_oM=\qu{\7q}{\7q\!\cap\!\sigma\7q}$ and

2736: $H^{1,0}_oM=\qu{\sigma\7q}{\7q\!\cap\! \sigma\7q}$. In \Lit{FELS} it

2737: has been shown that the geometric properties of the CR-structure of

2738: the CR-germ $(M,o)$ like minimality, $k$-nondegeneracy, holomorphic

2739: degeneracy can be read off every CR-algebra $(\7g,\7q)$ associated

2740: with $(M,o)$. The facts relevant for our classification will be

2741: discussed below.

2742: 

2743: There is also a canonical way to associate a locally homogeneous

2744: CR-manifold germ $(M,o)$ to a given CR-algebra $(\7g,\7q)$ (not

2745: necessarily effective): Choose a complex Lie group $L$ with Lie

2746: algebra $\7l=\7g\oplus i\7g$ and a complex linear subspace

2747: $E\subset\7l$ with $\7l=\7q\oplus E$. Then there exist open

2748: neighbourhoods $U$ of $0\in E$, $V$ of $0\in\7q$ and $R$ of $\id\in L$

2749: such that $(u,v)\mapsto\exp(u)\exp(v)$ defines a biholomorphic mapping

2750: $\phi:U\times V\to R$. Choose an open neighbourhood $P$ of $0\in\7g$

2751: with $\exp(P)\subset R$. Then, in our particular situation, with

2752: $\pi:U\times V\to U$ being the canonical projection, the mapping

2753: $\psi:=\pi\circ\phi^{-1}\circ\exp:P\to U$ has constant rank. Without

2754: loss of generality we therefore may assume that $M:=\psi(P)$ is a

2755: connected real-analytic submanifold of $E$ containing the origin $0\in

2756: E$. The Lie algebra $\7l$ can be identified with the Lie algebra of

2757: all right-invariant vector fields on $L$ and every $\xi\in\7l$ can be

2758: projected along $\pi\circ\phi^{-1}:R\to U$ to a holomorphic vector

2759: field $\tilde\xi\in\hol(U)$. Thus the real subalgebra

2760: $\tilde\7g:=\{\tilde\xi:\xi\in\7g\}\subset\hol(U)$ is a homomorphic

2761: image of $\7g$ and spans at every $x\in M$ the tangent space

2762: $T_{x}M$. In particular, $M$ is a generic CR-submanifold of $E$ and

2763: $\tilde\7g$ is a locally transitive subalgebra. It is not difficult so

2764: see that $\tilde\7g$ is obtained from $\7g$ by factoring out the

2765: kernel of ineffectivity, more precisely, let $\7j$ be the largest

2766: ideal in $\7g$ with $\7j\subset \7g\cap\7q $. Then $\tilde\7g$ is

2767: isomorphic to $\7g/\7j$. If there exists a Lie group $L$ with Lie

2768: algebra $\7l$ such that the subalgebra $\7q$ corresponds to a {\sl

2769: closed} complex subgroup $Q\subset L$ then we may take $L/Q$ for $U$

2770: and the $G$--orbit through $[Q]\in L/Q$ for $M.$ We call $(M,o)$ the

2771: CR-{\sl germ associated} to the CR-algebra $(\7g,\7q)$.

2772: 

2773: \Definition{} The CR-algebras $(\7g,\7q)$ and $(\7g',\7q')$ are called

2774: {\sl geometrically equivalent} if the associated CR-germs are

2775: CR-equivalent.

2776: 

2777: Notice that CR-algebras are always geometrically equivalent if they

2778: are isomorphic in the categorical sense of \ruf{XA}.ii, but not

2779: conversely in general. Notice also that every CR-algebra is

2780: geometrically equivalent to an effective one. In the following

2781: paragraphs \ruf{FA}--\ruf{FD}, we fix

2782: 

2783: \vskip5pt\centerline{\sl for the rest of the paper}

2784: 

2785: \noindent the basic setup and notation, which are mainly taken from

2786: \Lit{FELS}:

2787: 

2788: \Notation{FA} Given a CR-algebra $(\7g,\7q)$, let $(M,o)$ be the

2789: associated CR-germ. Write $\7l:=\7g^{\CC}=\7g\oplus i\7g$ for the

2790: complexification and $\sigma:\7l\to \7l$ for the complex conjugation

2791: with $\7l^\sigma=\7g$. Then \0 $\7g_o : =\7g\cap \7q$ is called the

2792: {\sl real isotropy subalgebra}. Define

2793: $\7l_o:=\7q^{(\infty)}:=\7q\cap\sigma\7q$ and note that $\7l_o$ is

2794: the complexification $(\7g_o)^{\CC}$ of $\7g_{o}$.  \1 $\7g/\7g_{o}$

2795: and $\7H/\7g_{o}\subset\7g/\7g_{o}$ for

2796: $\7H:=(\7q+\sigma\7q)^\sigma\subset\7g$ are called the {\sl real} and

2797: the {\sl holomorphic tangent space} respectively.  \1 The descending

2798: chain $\7q^{(0)}\supset\7q^{(1)}\supset\7q^{(2)}\supset\cdots\supset

2799: \7q^{(\infty)}$ of complex subalgebras is inductively defined by

2800: $\7q^{(0)}:=\7q$, $\7q^{(\infty)}:=\7q\cap \sigma\7q$ and

2801: $\7q^{(k+1)}:=\{w\in\7q^{(k)}:[w,\sigma\7q]\subset\7q^{(k)}\li1+

2802: \sigma\7q\}$, for $k\in \NN$.

2803: 

2804: \noindent If $(M,o)$ is the manifold germ associated to the CR-algebra

2805: in \ruf{FA} then the holomorphic tangent space

2806: $\7H/\7g_{o}\subset\7g/\7g_{o}$ in the sense of (ii) can be

2807: canonically identified with the holomorphic tangent space $H_{o}M$ in

2808: the geometric sense. As shown in \Lit{FELS}, the mapping $\7q\to\7H$,

2809: $w\mapsto w+\sigma w$, induces a complex linear isomorphism

2810: $\7q/\7q^{(\infty)}\cong\7H/\7g_{o}\,$. The former quotient is

2811: canonically isomorphic to $H^{0,1}_oM$ (similarly, $H^{1,0}_oM\cong

2812: \sigma\7q/\7q^{(\infty)}$). The Levi kernel $K_{o}M$ and its higher

2813: order analogues $K^{r}_{o}M$ can be considered as complex linear

2814: subspaces of $\7q/\7q^{(\infty)}$. We will make extensive use of some

2815: of the main results of \Lit{FELS}, such as Theorem 5.10:

2816: 

2817: \Joker{CA}{Algebraic characterization of $k$-nondegeneracy} For every

2818: $r\ge0$ the space $\7q^{(r)}$ is a Lie subalgebra of $\7q$ and the

2819: $r^{{\th}}$ Levi kernel $K^{r}_{o}M$ is isomorphic to

2820: $\7q^{(r)}/\7q^{(\infty)}$. In particular, for every $k\ge1$ the

2821: locally homogeneous CR-manifold $M$ is $k$-nondegenerate if and only

2822: if

2823: 

2824: \vskip5pt\centerline{$\7q^{(k-1)}\ne\;\7q^{(k)}=\;\7q^{(\infty)}$.}\Formend

2825: 

2826: \medskip\noindent To handle the $5$-dimensional case we also introduce

2827: the following abbreviations \1 $\7f:=\7q^{(1)}=\{v\in

2828: \7q:[v,\sigma\7q]\subset \7q+\sigma\7q\}$ and

2829: $\7F:=(\7f+\sigma\7f)^\sigma$.

2830: 

2831: \noindent Then $\7g_o\subset\7F\subset \7H\subset\7g$ are real

2832: subspaces stable under $\ad(\7g_o)$ and $\7l_o\subset\7f\subset\7q$

2833: are complex subalgebras. In Lemma 5.9. of \Lit{FELS} it has been shown

2834: that actually $\7F$ is a Lie algebra and coincides with

2835: $N_{\7g}(\7H)\cap \7H$ (here, given $W\subset \7g$, $N_\7g(W):=\{v\in

2836: \7g:[v,W]\subset W\}$).

2837: 

2838: \smallskip Summarizing the above discussion, the following result is

2839: the key for our classification.

2840: 

2841: \Proposition{MC} Let $(M,o)$ be an $\7g$-homogeneous CR-germ and let

2842: $(\7g,\7q)$ be the corresponding CR-algebra.  Then $M$ is

2843: 5-dimensional, minimal and 2-nondegenerate if and only if

2844: $$

2845: \codim_{\7g}(\7g_{0})=5\Steil{and}

2846: \7q\ne\7f\ne\7q^{(2)}=\7q^{(\infty)}:=\7q\cap\sigma\7q\;.

2847: $$

2848: 

2849: \ifarx\eject\fi

2850: 

2851: {\openup2pt\Lemma{FC} The Lie algebraic terms in \Ruf{MC} are

2852: equivalent to the following set of conditions: \o

2853: $\dim\qu{\7F}{\7g_o}=\dim \qu{\7H}{\7F}=2,\quad

2854: \dim_{\RR}\qu{\7g}{\7H}=1=

2855: \dim_{\CC}\qu{\7f}{\7l_o}=\dim_{\CC}\qu{\7q}{\7f}$

2856: 

2857: \l $[\7g_o,\7F]\subset\7F,\quad[\7F,\7F]\subset\7F,\quad

2858: [\7F,\7H]\subset\7H$

2859: 

2860: \l $[\7q,\sigma\7q]\not\subset\7q\,{+}\kern1pt\sigma\7q$\hfill\hbox

2861: to 7truecm{\Klein\sl $M$ is not Levi flat\hss}

2862: 

2863: \l $[\7f,\sigma\7q]\,\subset \7q\,{+}\kern1pt\sigma\7q,\quad

2864: \7f\ne\7l_o$\hfill\hbox to 7truecm{\Klein\sl $M$ is Levi

2865: degenerate\hss}

2866: 

2867: \l $[\7f,\sigma\7q]\,\not\subset\,\7f\,{+}\,\sigma\7q\,$\hfill\hbox to

2868: 7truecm{\Klein\sl $M$ is 2-nondegenerate.\hss}

2869: }

2870: \Formend

2871: 

2872: \smallskip\noindent We will frequently use the fact that the condition

2873: `$\,[\7F,\7H]\subset \7F\,$' (instead of $\subset \7H$) violates

2874: condition \25.

2875: 

2876: \medskip By the canonical bijection between the classes \CRG and \CRA

2877: mentioned in \ruf{XA}.iv and made precise above our classification

2878: problem is transferred to the classification of certain effective

2879: CR-algebras up to geometric equivalence. Unfortunately, to a given

2880: locally homogeneous CR-germ $(M,o)$ there may be associated many

2881: CR-algebras $(\7g,\7q)$ for which the $\7g$'s are non-isomorphic. For

2882: instance, if $M=\5M$ is the tube over the future light cone, then to

2883: $(\5M,o)$ there are associated CR-algebras $(\7g,\7q)$ with

2884: $\7g\cong\so(2,3)$, $\so(1,3)$, $\so(2,2)$ together with a bunch of

2885: other Lie algebras that are not semisimple, see \Lit{FEKA} for

2886: explicit realizations. Therefore, the best we can do in the following

2887: is to consider only CR-algebras $(\7g,\7q)$ such that $\dim\7g$ is

2888: minimal in the geometric equivalence class of $(\7g,\7q)$.  But, also

2889: then, we are still left in the example $(\5M,o)$ with several

2890: non-isomorphic solvable Lie algebras of dimension 5 as well as one

2891: non-solvable Lie algebra of dimension $5$ with $2$-dimensional

2892: non-abelian radical and Levi part $\cong\7{sl}(2,\RR)$.

2893: 

2894: 

2895: \smallskip \Joker{FD}{Fundamental Assumption} In the following, every

2896: CR-algebra $(\7g,\7q)$ under consideration is assumed to satisfy the

2897: condition \Ruf{MC} (equivalently \21 -- \25\hskip1pt) as well as the

2898: following additional condition. \l For every CR-algebra $(\7g',\7q')$,

2899: which is geometrically CR-equivalent to $(\7g,\7q)$, the dimension

2900: estimate $\dim\7g'\ge\dim\7g$ holds.

2901: 

2902: \smallskip\noindent Condition \26 implies, in particular, the

2903: following two conditions.

2904: 

2905: \item{\26$_{1}$} $(\7g,\7q)$ is effective.

2906: 

2907: \item{\26$_{2}$} There is no proper subalgebra $\7g'\subset \7g$ with

2908: $\7g'+\7g_o=\7g$ for $\7g_o=\7g\cap\7q$.

2909: 

2910: \noindent Indeed, $(\7g',\7q')$ with $\7q':=(\7g'+i\7g')\cap\7q$ is a

2911: CR-algebra that is geometrically equivalent to $(\7g,\7q)$ in case

2912: $\7g'+\7g_o=\7g$.

2913: 

2914: \bigskip\Joker{ST}{Basic structure theory} In the following we

2915: frequently use standard facts concerning reductive Lie algebras and

2916: parabolic subalgebras, see \Lit{KNAP}, Chap.VI-VII, as a general

2917: reference. To fix our notation, let $\7s$ be a complex reductive Lie

2918: algebra and $\7t$ a Cartan subalgebra of $\7s$. Denote by

2919: $\Phi=\Phi(\7s,\7t)\subset \7t^*$ the corresponding root system and by

2920: $\Pi\subset\Phi$ the subset of simple roots. A subalgebra

2921: $\7r\subset\7s$ is called {\sl parabolic} if it contains a maximal

2922: solvable subalgebra (also called a {\sl Borel subalgebra}) $\7b$ of

2923: $\7s$. Conjugacy classes of parabolic subalgebras in $\7s$ are

2924: parameterized by the subsets of $\Pi$: For every $\6P\subset \Pi$ and

2925: $\langle\kern-2pt\langle \6P \rangle\kern-2pt\rangle:=\Phi\cap

2926: \bigoplus_{\alpha\in \6P} \ZZ \alpha$ the corresponding parabolic

2927: subalgebra is defined by

2928: $$

2929: \7r=\7r_\6P:=\7r^{\red}\oplus

2930: \7r^{\nil}\Steil{with}\7r^{\red}:=\7t\;\oplus\!

2931: {\textstyle\bigoplus\limits_{\alpha\in\langle\!\langle

2932: \6P\rangle\!\rangle}}\7s_\alpha\Steil{and}\7r^{\nil}:=\!

2933: {\textstyle\bigoplus\limits_{\alpha\notin\langle\!\langle

2934: \6P\rangle\!\rangle}}\7s_\alpha \;\;.\Leqno{PA}

2935: $$

2936: The case of a reductive {\sl real} Lie algebra $\7s$ is a little bit

2937: more sophisticated. In contrast to the complex situation there may

2938: exist several conjugacy classes of Cartan subalgebras. Among these the

2939: class most suitable for our purposes consists of the so-called {\sl

2940: maximally split Cartan subalgebras} $\7t\subset\7s$, defined as

2941: follows: Select a Cartan decomposition $\7s=\7k\oplus\7p$ and let

2942: $\7a$ be a maximal abelian subalgebra of $\7p$. Consider the

2943: centralizer $\7m:=C_\7k(\7a)$ and put $\7t:=\7a\oplus \7t_\7m$, where

2944: $\7t_\7m\subset \7m$ is a maximal abelian subalgebra. The conjugacy

2945: classes of real parabolic subalgebras in $\7s$ are parameterized by

2946: the subsets of the simple roots $\Pi\subset \Phi(\7s,\7a)$ in the

2947: restricted root system $\Phi(\7s,\7a)\subset \7a^*$. Each parabolic

2948: subalgebra $\7r$ in $\7s$ has the decomposition

2949: $\7r=\7r^{{\red}}\ltimes \7r^{\nil}$ into the reductive and nilpotent

2950: parts (once a maximally split Cartan subalgebra $\7t\subset\7r$ is

2951: selected, the reductive factor with $\7r^{\red}\supset \7t$ is

2952: unique). Following a common convention the roots $\lambda$ occurring

2953: in the root space decomposition of the nilpotent ideal

2954: $\7r^{\nil}=\bigoplus_{\lambda\in\Lambda} \7s_\lambda$ are negative

2955: and we write $\7r^{-\n}:=\7r^{\nil}$, $\;\7r^{\r}:=\7r^{\red}$ and

2956: $\Phi^{-\n}:=\Phi(\7r^{\nil},\7a)=\Lambda$. Each parabolic subalgebra

2957: $\7r$ containing $\7t$ determines the decomposition

2958: $$

2959: \7s=\7r^{\n}\oplus \7r^{\r}\oplus \7r^{-\n}\Steil{with}

2960: \7r^{\n}:=\!\!{\textstyle\bigoplus\limits_{\lambda\in\Phi(\7r^

2961: {\nil},\7a)}}\li{12}\7s_{-\lambda}\;,\Steil{and we

2962: put}\7r^{\opp}:=\7r^{\r}\ltimes \7r^{\n}\;.\Leqno{FG}

2963: $$

2964: \vskip-6pt\noindent We call $\rk(\7s):=\dim \7t$ the {\sl rank} of

2965: $\7s$ and $\rk_{\RR}(\7s):=\dim \7a$ with $\7a\subset \7p$ as above

2966: the {\sl real rank} of $\7s.$

2967: 

2968: \KAP{Zehn}{The Lie algebra $\7g$ has small semisimple part}

2969: 

2970: 

2971: In the preceding section we have explained how 2-nondegeneracy can be

2972: expressed in pure Lie algebraic terms. In this section we start with

2973: the actual proof of Theorem II. Since this proof will be quite

2974: involved, we subdivide it into several sections, lemmata and claims

2975: until the final step is completed in Section \ruf{Sechzehn}. For the

2976: convenience of the reader we briefly outline the main steps.

2977: 

2978: As explained above the classification can be reduced to the

2979: determination of all CR-algebras satisfying the fundamental assumption

2980: \ruf{FD}.  Once all possible CR-algebras are known, we have to

2981: identify the underlying CR-germs. In general, it may happen that

2982: algebraically non-equivalent CR-algebras give rise to equivalent

2983: CR-germs. For this last part of the proof we use results from Section

2984: \ruf{Acht}.

2985: 

2986: Our proceeding will be to show that the assumption \ruf{FD} severely

2987: restricts the possibilities for $(\7g,\7q)$.  This will be achieved by

2988: a detailed structural study of the Lie algebras $\7g$ occurring in

2989: $(\7g,\7q)$. Recall that every Lie algebra $\7h$ has a Levi-\Malcev

2990: decomposition $\7h=\7h^{\s}\ltimes\rad(\7h)$, where $\7h^{\s}$ is

2991: semisimple and is uniquely determined up to an inner automorphism of

2992: $\7h$. Furthermore, $\rad(\7h)$ is the radical of $\7h$, i.e., the

2993: unique maximal solvable ideal in $\7h$. In the first part of the proof

2994: we investigate the various possibilities for $\7g^{\s}$ where

2995: $(\7g,\7q)$ satisfies certain conditions stated in the previous

2996: section. To be precise: {\sl For the rest of the paper the fundamental

2997: assumption \ruf{FD} remains in force for {\rm all} CR-algebras

2998: $(\7g,\7q)$ under consideration}. In particular, every $(\7g,\7q)$ is

2999: effective (condition \26$_{1}$, and therefore $\7g$ can be considered

3000: as a transitive Lie subalgebra of $\hol(M,o)$). Furthermore, condition

3001: \26$_{2}$ states that there is no proper Lie subalgebra of $\7g$ that

3002: also is transitive on $(M,o)$.

3003: 

3004: 

3005: Let an arbitrary CR-algebra $(\7g,\7q)$ subject to \ruf{FD} be given

3006: and let $\7g^{\s}\ltimes \rad(\7g)$ be a Levi-\Malcev decomposition

3007: of $\7g$.  In this and in the following few sections we assume that

3008: $\7g^{\s}\ne 0$ and investigate which simple factors can occur in

3009: $\7g^{\s}$. Thereby we use the following notation: We fix a simple

3010: ideal $\7s$ in $\7g^{\s}$ and denote by $\7s'$ the corresponding

3011: complementary ideal, i.e.,

3012: $$

3013: \7g=\7g^{\s}\ltimes \rad{\7g}\Steil{and} \7g^{\s}=\7s\times \7s'\;.

3014: \Leqno{SS}

3015: $$

3016: 

3017: In this section we show that $\7g^{\s}$ only can contain simple

3018: factors isomorphic to one of the Lie algebras $\so(2,3),\;

3019: \so(1,3),\;\su(2)$ and $\7{sl}(2,\RR)$.  This result is obtained by

3020: analyzing which simple real Lie algebras can contain proper

3021: subalgebras of very low codimensions.  In the next sections we exclude

3022: further possibilities for $\7s$: In Section \ruf{Elf} we show that the

3023: factors $\so(2,3)$ and $\so(1,3)$ cannot occur in $\7g^{\s}$ as it

3024: will turn out that their existence in the Levi factor would violate

3025: the minimality assumption \26. Nevertheless notice that there exist

3026: CR-algebras $(\so(2,3),\7q)$ and $(\so(1,3),\7q)$ satisfying \21 --

3027: \25 and \26$_1$. All the underlying CR-germs of such CR-algebras are

3028: locally CR-equivalent to the light cone tube $\5M$. In Section

3029: \ruf{Zwoelf} we find first examples of CR-algebras which satisfy

3030: \ruf{FD}. In these cases $\7g^{\s}$ contains a simple factor $\7s\cong

3031: \7{sl}(2,\RR)$, and then necessarily $\7g\cong \7{sl}(2,\RR)\times

3032: \7r$, where $\7r$ is a 2-dimensional non-abelian Lie algebra. Also all

3033: such CR-algebras give rise only to CR-germ locally CR-equivalent to

3034: the tube over the light cone.  In Section \ruf{Dreizehn} we finally

3035: eliminate the possibility $\7s\cong \su(2)$ for the simple factor

3036: $\7s$ in $\7g^{\s}$.  At that stage of the proof we will have proved

3037: the following dichotomy: Let $(\7g,\7q)$ be an arbitrary CR-algebra

3038: which obeys \ruf{FD}. If $\7g^{\s}\ne 0$ then $\7g^{\s}\cong

3039: \7{sl}(2,\RR)$ and furthermore the underlying CR-germ is locally

3040: CR-equivalent to the light cone tube $\5M$.  Or, $\7g^{\s}=0$, i.e.,

3041: $\7g$ is solvable.

3042: 

3043: For that reason, from Section \ruf{Vierzehn} on we only consider

3044: CR-algebras $(\7g,\7q)$ with $\7g$ solvable and show that then

3045: necessarily $\dim \7g=5=\dim M$. In Section \ruf{Fuenfzehn} we look

3046: closer at the nilcenter $\7z$ of $\7g$ and find out that $\dim

3047: \7z\in \{1,3\}$. The Main Lemma \ruf{GS}, which might be of interest

3048: for itself, gives a sufficient condition for a CR-algebra to be

3049: associated to a tube $F+i\RR^{3}\subset\CC^{3}$ over an affinely

3050: homogeneous surface $F\subset\RR^{3}$. In the last section we show by

3051: ad-hoc methods that indeed every 5-dimensional solvable Lie algebra

3052: $\7g$ occurring in the CR-algebra $(\7g,\7q)$ under consideration

3053: fulfills the assumption of the Main Lemma.  Hence, due to the

3054: aforementioned assertions and since (up to local affine equivalence)

3055: all affinely homogeneous surfaces in $\RR^{3}$ occur among the

3056: examples \ruf{EI} -- \ruf{EV}, this completes the proof of the

3057: classification theorem.

3058: 

3059: \medskip

3060: 

3061: We now begin with the proof of Theorem II:

3062: 

3063: 

3064: 

3065: \Lemma{LA} Let $(\7g,\7q)$ be a CR-algebra subject to \Ruf{FD}.

3066: Then the simple Lie subalgebra $\7s\subset \7g^{\s}$ can only

3067: be isomorphic to

3068: $\;\so(2,3)\;,\;\so(1,3)\,,\,\7{sl}(2,\RR)\;$ or $\;\su(2)$.

3069: 

3070: \Proof The proof is carried out in several reduction steps. To begin

3071: with, we write as shorthand

3072: $$

3073: \7h_o:=\7g_o\cap \7h\,,\quad\7h_\7F:=\7F\cap

3074: \7h\steil{and}\7h_\7H:=\7H\cap\7h

3075: $$

3076: for every subalgebra $\7h\subset\7g$. Notice that $\7h_o\subset

3077: \7h_{\7F}$ are subalgebras and $\7h_{\7H}$ is a linear

3078: $\ad(\7h_{\7F})$--stable subspace of $\7h.$ \nline Consider the

3079: subalgebras $\7s_o\subset \7s_\7F$ of $\7s$. The case

3080: $\7s_o=\7s_\7F=\7s$, that is $\7s\subset\7g_o$, can be ruled out since

3081: then $\7s'\oplus\rad(\7g)$ would be a proper locally transitive

3082: subalgebra of $\7g$, contradicting assumption \26$_{2}$. Therefore, at

3083: least one of the inclusions $\7s_{o}\subset \7s_\7F\subset\7s$ is

3084: proper. Consequently, there is always a proper subalgebra of

3085: codimension $\le3$ in $\7s$. Indeed, in case $\7s_\7F\ne\7s$ the

3086: subalgebra $\7s_\7F$ has this property and in case $\7s_\7F=\7s$ the

3087: proper subalgebra $\7s_{o}$ has codimension $\le2$. Hence, there

3088: exists a maximal proper subalgebra $\7h$ of $\7s$ with either

3089: $\7s_{\7F}\subset\7h$ or $\7s_o\subset \7h.$ Such a maximal subalgebra

3090: $\7h$ has codimension $\le3$ in $\7s$. Due to \Lit{BOUR}, Chap.~VIII,

3091: \S 10, Cor.~1, every maximal proper subalgebra of $\7s$ is either

3092: reductive or parabolic. In the following claims we list all simple Lie

3093: algebras $\7s$ which admit proper maximal subalgebras of such low

3094: codimensions. We discuss the reductive and parabolic case separately.

3095: 

3096: \noindent{\fett Claim 1: } \sl Let $\7k$ be a simple real algebra and

3097: $\7h\subset \7k$ reductive with

3098: $0<\codim_\7s\7h\le 3$. Then $\7k$ can only be isomorphic to

3099: $\7{sl}(2,\RR),\;\su(2)$ or $\so(1,3)$.

3100: 

3101: \ProofC Let $\7h=\7h_1\times \cdots \times \7h_p\times \7z$ be the

3102: decomposition of the reductive subalgebra $\7h$ into the simple

3103: factors $\7h_j$ and the center $\7z.$ Weyl's Theorem implies the

3104: existence of an $\ad(\7h)$-stable complement $\7v\subset \7s.$ Let

3105: $\rho:\7h\to \gl(\7v)$ be the induced adjoint representation. Every

3106: restriction $\rho_j:\7h_j\to \gl(\7v)$ must be faithful since

3107: otherwise $\7h_j$ would be an ideal in $\7s$. The crucial condition

3108: here is $\dim \7v\le 3$ which, in turn, implies that each $\7h_j$ is

3109: isomorphic either to $\7{sl}(2,\RR)$, $\so(3)$ or $\7{sl}(3,\RR)$. As

3110: a consequence

3111: $$

3112: \dim \7h\le \cases{8k&$r=2k$\cr8k+3&$r=2k+1$\cr}

3113: \Steil{for}r:={\rk}(\7h)\le{\rk}(\7s)\,.

3114: $$

3115: On the other hand, a

3116: glance at the classification of simple Lie algebras shows

3117: $$

3118: \dim \7s \ge \cases{2k^2+4k=\dim_{\RR} \7{sl}(k+1,\CC)& $r=2k$\cr

3119: 4k^2+8k+3=\dim\7{sl}(2k+2,\RR) &$r=2k+1\,.$}

3120: $$

3121: Putting both inequalities together and bearing in mind $\dim\7s-\dim

3122: \7h\le 3$ shows that the rank of $\7s$ can only be one of the numbers

3123: $1,2,4$. Since $\7{sl}(2,\RR)$ and $\su(2)$ are the only simple real

3124: Lie algebras of rank $1$ we may assume that $\7s$ has rank $2$ or $4$.

3125: The case $\rk(\7s)=4$ can be ruled out in the following way: Consider

3126: first the situation where $\7s$ is {\sl of complex type,} that is,

3127: $\7s$ is the underlying real Lie algebra of a complex simple Lie

3128: algebra $\7c$ of (complex) rank $2$. Then $\7c$ is either

3129: $\7{sl}(3,\CC)$ or $\so(5,\CC)$. But in both cases a proper reductive

3130: real subalgebra has at least (real) codimension 4 (in fact, 8). If

3131: $\7s$ is {\sl of real type} then the above estimates give

3132: $\dim\7h\le16$ and $\dim\7s\ge24=\dim \7{sl}(5,\RR)$. For the remaining

3133: case $\rk(\7s)=2$ either $\7s\cong \so(1,3)$, which is in the list of

3134: the claim, or $\7s$ is isomorphic to a real form of $\7{sl}(3,\CC)$ or

3135: $\so(5,\CC)$. In both cases every proper reductive complex subalgebra

3136: has at least codimension 4. This proves Claim 1.\qd

3137: 

3138: \noindent{\fett Claim 2:} \sl Let $\7k$ is a simple real Lie algebra

3139: and $\,\7h\subset \7k$ parabolic with $0<\codim_\7s\7h\le 3$. Then

3140: $\7s$ is isomorphic to $\so(1,3)$, $\7{sl}(4,\RR)$, $\su(2)$ or to a

3141: non-compact real form of $\7{sl}(3,\CC) $ or $\so(5,\CC)$.

3142: 

3143: \ProofC Since every parabolic subalgebra of a compact Lie algebra is

3144: trivial, apart from $\7s\cong \su(2)$ we only have to consider the

3145: case where $\7s$ is non-compact. The estimate

3146: $\rk(\7s)\le\dim\7s-\dim\7h$, compare \Ruf{ST}, implies

3147: $\rk(\7s)\le3$. We work out the various cases separately.

3148: 

3149: \noindent {\fett $\rk\7s=3\,$:} The complexification $\7s^{\CC}$ of

3150: $\7s$ is one of the Lie algebras $\7{sl}(4,\CC)$, $\7{so}(7,\CC)$ or

3151: $\7{sp}(3,\CC)$. The latter two can be immediately ruled out since a

3152: glance at the corresponding Satake diagrams shows that every proper

3153: parabolic subalgebra of them is at least of codimension 4. In the

3154: remaining case $\7s^{\CC}\cong \7{sl}(4,\CC)$ only the normal real

3155: form $\7{sl}(4,\RR)$ has a parabolic subalgebra of codimension 3. (A

3156: glance at the Satake diagrams for the remaining non-compact real forms

3157: of $\7{sl}(4,\CC)$ excludes further possibilities).

3158: 

3159: \noindent {\fett $\rk\7s=2\,$ or $1$:} The rank conditions imply that

3160: $\7s$ is the underlying real Lie algebra of $\7{sl}(2,\CC)$ or a real

3161: form of $\7{sl}(3,\CC)$ or $\so(5,\CC).$ The Lie algebra $\so(1,3)$ as

3162: well as every non-compact real form of $\7{sl}(3,\CC)$ or $\so(5,\CC)$

3163: contains a parabolic subalgebra of codimension less or equal 3. Since

3164: $\dim \7{sl}(2,\RR)=\dim\su(2)=3$, all simple Lie algebras of rank one

3165: also contain parabolic subalgebras of the required codimension.

3166: 

3167: \noindent In order to further reduce the list of possibilities for

3168: $\7s$ we have to look at the particular real forms obtained in Claim 1

3169: and 2 but not in the list of Lemma \ruf{LA} more closely:

3170: 

3171: \noindent {\fett Elimination of $\7s\cong \7{sl}(4,\RR)$:} Up

3172: to an automorphism of $\7{sl}(4,\RR)$ there is only one such parabolic

3173: subalgebra $\7h$ of codimension $3$. Let $\7t=\7a\subset \7h$ be the

3174: split Cartan subalgebra and $\7h=\7h^{\r}\ltimes \7h^{-\n}$ the

3175: decomposition as in \Ruf{FG}. Note that here the reductive factor

3176: $\7h^{\r}\cong \gl(3,\RR)$ acts irreducibly on $\7h^{\pm\n}\cong

3177: \RR^3$. The only possibility for the flag $\7s_o\subset \7s_\7F\subset

3178: \7s_\7H\subset \7s$, which cannot be trivially excluded, is when

3179: $\7s_o\subset \7s_\7F=\7h\subsetneq \7s_\7H\subsetneq \7s$ and

3180: $\codim_\7h(\7s_o)\le 2.$ This cannot be true since

3181: $\dim\qqu{\7s_\7H}{ \7s_\7F}=2$, $[\7s_\7F:\7s_\7H]\subset \7s_\7H$

3182: (condition \22) but $\7h=\7s_\7F$ acts irreducibly on the

3183: 3-dimensional space $\qqu{\7s}{\7s_\7F}\cong \7h^{\n}.$

3184: 

3185: 

3186: \noindent{\fett Elimination of $\7s\cong\7{su}(1,2)\,$:} In this case

3187: of real rank 1, there exists up to conjugacy only one parabolic

3188: subalgebra $\7h$: This is the minimal one $\7h=\7h_\emptyset

3189: =\7m\oplus \7a\oplus\7n$ which is solvable with $\codim_{\7s}\7h

3190: =3$. The reductive part $\7h^{\r}=\7m\oplus \7a=:\7t$ of $\7h$ is a

3191: maximally split Cartan subalgebra. As before, the only situation which

3192: cannot be trivially disposed of is when $\7s_o\subset \7s_\7F=\7h$ and

3193: $c:=\codim_\7h(\7s_o)\in \{0,1,2\}.$ Recall that $\7h$ yields the

3194: decomposition $\7s=\7h^{\n}\oplus \7t\oplus \7h^{-\n}$. If $c=0,$ that

3195: is $\7s_o=\7h$, then $\7h^{\n}\oplus\rad(\7g)$ would be a proper

3196: locally transitive subalgebra of $\7g$ in contradiction to assumption

3197: \26$_{2}$. If $c=1$ then either $\7t\cap \7s_o\ne \7t$, and then

3198: $\7h^{{\opp}}\oplus\rad(\7g)$ would be a proper locally transitive

3199: subalgebra of $\7g$ (again contradicting \26$_{2}$) or $\7t\subset

3200: \7s_o.$ But taking into account the particular structure of

3201: $\7h^{-\n}=\7s_{-\lambda}\oplus \7s_{-2\lambda}$ this also does not

3202: occur since otherwise $\7s_o\cap \7h^{-\n}$ would be a $\7t$-stable

3203: 2-dimensional subalgebra, which is impossible since $\ad(\7t)$ acts

3204: irreducibly on the 2-dimensional root spaces $\7s_{\pm\lambda}$ and

3205: $[\7s_{\lambda},\7s_{\lambda}]=\7s_{2\lambda}.$ \nline It remains the

3206: case $c=2,$ that is, $\7s=\7g\cong \su(1,2)$ is locally

3207: transitive. Then $\7l=\7s^{\CC}\cong\7{sl}(3,\CC)$ and $\7q$ is a

3208: subalgebra of complex dimension $5$. Consequently $\7q$ is contained

3209: in a maximal (6-dimensional) parabolic subalgebra

3210: $\7h\cong\7{gl}(2,\CC)\ltimes\CC^2$ of $\7l$, that is, either $\7q$

3211: coincides with a Borel subalgebra $\7b$ or is conjugate to a

3212: subalgebra $\7j\cong \7{sl}(2,\CC)\ltimes \CC^2$. In both cases the

3213: subgroups $Q$ of $L=\SL(3,\CC)$ corresponding to $\7b$ or $\7j$ are

3214: closed and the underlying CR-germ $(M,o)$ is locally CR-equivalent to

3215: an $\SU(1,2)$-orbit either in $L/B\cong\FF(\CC^3)$, the complex

3216: manifold of full flags in $\CC^3$, or in the $\CC^*$-principal bundle

3217: $L/J$ over $\PP_2(\CC)$. A direct check shows that in both cases there

3218: do not exist 2-nondegenerate $\SU(1,2)$-orbits.

3219: 

3220: \noindent{\fett Elimination of $ \7s\cong\7{sl}(3,\RR)$.} Then

3221: $3\le\dim\7s_o\le6$ holds except for the trivial case $\7s_o=\7s$.  If

3222: $\dim\7s_o=3$ and hence $\7s$ is locally transitive, then as in the

3223: previous situation each CR-germ $(M,o)$ associated with a CR-algebra

3224: $(\7{sl}(3,\RR),\7q)$ is locally CR-equivalent to an

3225: $\SL(3,\RR)$-orbit either in $\SL(3,\CC)/B$ or $\SL(3,\CC)/J$ (as

3226: discussed above, with $\SL(3,\RR)$ in place of $\SU(1,2)$). Again,

3227: none of these orbits is 2-nondegenerate. It remains the case $\dim

3228: \7s_o\ge 4$. But then there always exists a parabolic (proper)

3229: subalgebra $\7h\subset\7s$ with $\7s_o+\7h=\7s$, that is,

3230: $\7h\oplus\rad(\7g)$ is a proper locally transitive subalgebra of

3231: $\7g$ excluding the case $\7s^{\CC}\cong\7{sl}(3,\CC)$.

3232: 

3233: \noindent{\fett Elimination of $ \7s\cong\so(1,4)$.} There exists up

3234: to conjugacy a unique parabolic subalgebra

3235: $\7h=\7h^{\r}\ltimes\7h^{-\n}\subset\7s$ of codimension $3$, and we

3236: have to investigate the cases $\7s\supsetneq \7s_\7H\supsetneq

3237: \7s_\7F=\7h\supset \7s_o$ only. A close look at the minimal (and

3238: maximal proper) parabolic subalgebra $\7h=\7m\oplus \7a\oplus

3239: \7s_{\lambda}$ shows that $\7m\cong \so(3)$ and $\7m$ acts irreducibly

3240: on the 3-dimensional nilpotent ideal $\7h^{-\n}=\7s_{-\lambda}$.

3241: Consequently $\7h$ acts irreducibly on $\7s/\7h =\7s/\7s_\7F\cong

3242: \RR^3.$ This leads to a contradiction as

3243: $[\7s_{\scriptscriptstyle\7F},\7s_{\scriptscriptstyle\7H}]\subset

3244: \7s_{\scriptscriptstyle\7H}$, i.e.,

3245: $\qqu{\7s_{\scriptscriptstyle\7H}}{\7s_{\scriptscriptstyle\7F}}$ would

3246: be a 2-dimensional stable subspace. The proof of Lemma \ruf{LA} is now

3247: complete.\qed

3248: 

3249: \medskip

3250: In Theorem 6 of \Lit{BELO} it is claimed that for every

3251: $2$-nondegenerate real-analytic hypersurface $M\subset\CC^{3}$ the Lie

3252: algebra $\hol(M,a)$ has dimension $\le11$ at every point. Using this

3253: result would save few arguments in the proof of Lemma \ruf{LA}.

3254: Instead, we preferred to present a self-contained proof of the

3255: proposition.

3256: 

3257: 

3258: \KAP{Elf}{The cases $\7s\cong\so(2,3)$ and $\7s\cong\so(1,3)$}

3259: 

3260: 

3261: We continue the proof of Theorem II.  So far we have proved that for

3262: the CR-algebra $(\7g,\7q)$ under consideration the simple factor $\s$

3263: of $\7g^{\s}$, see \Ruf{SS}, can only be isomorphic to one of the

3264: simple Lie algebras listed in Lemma \ruf{LA}. In this section we show

3265: that from these the possibilities $\7s\cong \so(2,3)$ and $\7s\cong

3266: \so(1,3)$ cannot occur. Here and in the following upper case roman

3267: numerals refer to the conditions around \ruf{FD}.

3268: 

3269: 

3270: We like to mention that for the tube $\5M$ over the future light cone

3271: $\hol(\5M,o)\cong\so(2,3)$ holds and that there exists a copy of

3272: $\so(1,3)$ in $\hol(\5M,a)$ that is also locally transitive. Since

3273: these Lie algebras have dimensions 10 and 6 and since, on the other

3274: hand, there are transitive subalgebras of $\hol(\5M,o)$ of dimension 5

3275: these two Lie algebras do not satisfy the minimality condition \26.

3276: In the following we consider both cases separately:

3277: 

3278: \noindent{\fett $\7s\cong\so(2,3)$.} The only Lie subalgebras $\7h$ in

3279: the normal real form $\so(2,3)$ with $\codim_\7s\7h\le 3$ are the

3280: 3-codimensional maximal parabolic subalgebras. We need to take a

3281: closer look at the structure of these subalgebras. There are up to

3282: isomorphisms of $\7s$ only two such parabolic subalgebras: If

3283: $\Pi(\7a)=\{\alpha,\beta\}$ is a basis of the root system $\Phi(\7a)$

3284: ($\7a$ split Cartan subalgebra, $\alpha$ long, $\beta$ short) then

3285: 

3286: 

3287: \smallskip $\displaystyle \eqalign{\7h_1:&=(\7a\oplus \7s_\alpha\oplus

3288: \7s_{-\alpha}) \;\;\oplus \;\; \7s_{-\beta}\oplus

3289: \7s_{-\alpha-\beta}\oplus \7s_{-\alpha-2\beta}=\7h_1^{\r}\ltimes

3290: \7h_1^{-\n} \cr \7h_2:&=(\7a\oplus \7s_\beta\oplus

3291: \7s_{-\beta})\;\;\oplus\;\; \7s_{-\alpha}\oplus

3292: \7s_{-\alpha-\beta}\oplus \7s_{-\alpha-2\beta}=\7h_2^{\r}\ltimes

3293: \7h_2^{-\n}}$

3294: 

3295: \smallskip\noindent are representatives of the corresponding isomorphy

3296: classes. The only instance where the filtration $\7s_o\subset

3297: \7s_{\scriptscriptstyle\7F}\subset \7s_{\scriptscriptstyle\7H}\subset

3298: \7s$ could be nontrivial (recall that $\7s_o,\;

3299: \7s_{\scriptscriptstyle\7F}$ are subalgebras,

3300: $\7s_{\scriptscriptstyle\7H}$ is a

3301: $\ad(\7s_{\scriptscriptstyle\7F})$-stable subspace) arises when

3302: $\7s_o\subset \7s_{\scriptscriptstyle\7F}=\7h_j\subset

3303: \7s_{\scriptscriptstyle\7H}\subset \7s$ for $j=1,2.$ The possibility

3304: `$\,\7h_2=\7s_{\scriptscriptstyle \7F}$' cannot occur since the

3305: adjoint representation of $\7h_2^{\r}$ on $\7h^{\n}_2\cong

3306: \qu{\7s}{\7s_{\scriptscriptstyle\7F}}$ is irreducible, contradicting

3307: the existence of a 2-dimensional

3308: $\ad(\7s_{\scriptscriptstyle\7F})$--stable subspace

3309: $\qqu{\7s_{\scriptscriptstyle\7H}}{\7s_{\scriptscriptstyle\7F}}.$ It

3310: remains the possibility ${\7s_{\scriptscriptstyle\7F}}=\7h_1\supset

3311: \7s_o$. From now on $\7h:=\7h_1$ and we analyze the various

3312: possibilities for $\dim {\7s_{\scriptscriptstyle\7F}}-\dim

3313: \7s_o\in\{0,1,2\}.$ The equation ${\7s_{\scriptscriptstyle\7F}}=\7s_o$

3314: contradicts condition \26$_{2}$ since in that case the proper

3315: subalgebra $(\7h^{\n}\oplus \7s')\ltimes \rad(\7g)$ would be

3316: transitive on $(M,o).$ The case when $\7s_o$ is of codimension 1 in

3317: ${\7s_{\scriptscriptstyle\7F}}=\7h$ can also be ruled out: Either

3318: $\7h^{-\n}\subset \7s_o$ and then $(\7h^{\opp}\oplus \7s')\ltimes

3319: \rad(\7g)$ would be a transitive proper subalgebra of $\7g$, or the

3320: intersection of $\7s_o$ with $\7h^{-\n}=(\7s_{-\beta}\oplus

3321: \7s_{-\beta-\alpha}) \oplus \7s_{-\alpha-2\beta}$ is a 2-dimensional

3322: subalgebra. In such a case the image $\pi(\7s_o)$ of the projection

3323: $\pi:\7h=\7h^{-n}\rtimes \7h^{\r}\to \7h^{\r}$ coincides with

3324: $\7h^{\r}.$ But this also leads to a contradiction: Neither the

3325: intersection $\7s_o\cap \7h^{-n}$ can coincide with

3326: $\7s_{-\beta}\oplus \7s_{-\beta-\alpha}$ (since it is not a

3327: subalgebra), nor $(\7s_{-\beta}\oplus \7s_{-\beta-\alpha})\cap \7s_o$

3328: can be 1-dimensional (since $\7h^{\r}$ acts irreducibly on

3329: $(\7s_{-\beta}\oplus \7s_{-\beta-\alpha})$). Finally we are left with

3330: the case $\dim {\7s_{\scriptscriptstyle\7F}}-\dim \7s_o=2, $ i.e.,

3331: $\7s$ is transitive on $(M,o).$ Thus $\7g=\7s$ by assumption \26.  But

3332: then $\dim\7g_{o}=5$ and there always exist proper subalgebras

3333: $\7g'\subset \7g$ with $\7g'+\7g_{0}=\7g$, a contradiction to

3334: condition \26$_{2}$.

3335: 

3336: \medskip \noindent{\fett $\7s\cong\so(1,3)\cong\7{sl}(2,\CC) $.} We

3337: work out this case by investigating various possibilities for $\dim

3338: \7s_o.$ Assume first that \nline $\Bullet\;\dim \7s_o\ge 3.$ We claim

3339: that then there exists a solvable subalgebra $\7r\subset \7s$ such

3340: that $\7s_o+\7r=\7s:$ The case $\dim \7s_o\ge 4$ is easily settled as

3341: all 4-dimensional subalgebras in $\7s$ are maximal, i.e., they are

3342: Borel subalgebras of $\7{sl}(2,\CC)$ and consequently have nilpotent

3343: complementary subalgebras. There do not exist (real) subalgebras of

3344: $\7{sl}(2,\CC)$ of dimension 5. If $\dim \7s_o=3$ then $\7s$ is either

3345: semisimple or solvable. In the semisimple case we work with an

3346: explicit matrix realization $\7s\subset \CC^{2\times 2}$: Either

3347: $\7s_o \cong \su(2)$ or $\cong\su(1,1),$ i.e.,

3348: $$

3349: \7s_o\cong \Big\{\!\!\pmatrixe{it & \epsilon\overline z \cr z &

3350: -it}\!\!:t\in \RR,z\in \CC

3351: \Big\}\Steil{for}\epsilon=1\steil{or}\epsilon=-1\;.

3352: $$ In both cases the upper triangular Borel subalgebra $\7b^+\subset

3353: \7{sl}(2,\CC)\subset \CC^{2\times 2}$ forms a linear complement of

3354: $\7s_o.$ This cannot happen, since then condition \26$_{2}$ would be

3355: violated. \nline If $\7s_o$ is solvable then $\7s_o$ is contained as

3356: certain 1-codimensional (real) subalgebra in a (complex) Borel

3357: subalgebra $\7b=\7t\ltimes \7b^{\nil}$ of $\7s\cong\7{sl}(2,\CC)$. We

3358: claim that $\7s_o\supset \7b^{\nil}:$ Otherwise (i.e., if

3359: $\dim^{}_{\RR }\7s_o\cap \7b^{\nil}=1$) we would have

3360: $\pi(\7s_o)=\7t$, where $\pi:\7b\to \7t$ is the projection

3361: homomorphism, and consequently $\7s_o$ would contain a certain complex

3362: Cartan subalgebra $\7t'$ of $\7b$ which acts $\RR$-irreducibly on

3363: $\7b^{\nil}$. \nline However, from our claim it follows that the

3364: opposite Borel subalgebra is a complementary subspace of $\7s_o$ in

3365: $\7s$. The case `$\dim \7s_o=3$' now is completely ruled out. Next, we

3366: investigate the case \nline {$\Bullet\;\dim \7s_o=2.$} It follows that

3367: $\7s_o$ is solvable and either complex or totally real. Since it is

3368: immediate that every complex subalgebra in $\7{sl}(2,\CC)$ has a

3369: complementary subalgebra (simple check), it remains only to deal with

3370: the totally real case, i.e., with $\7s_o$ being a real form of a Borel

3371: subalgebra $\7b\subset \7{sl}(2,\CC).$ Let $\tau:\7b\to \7b$ be the

3372: conjugation with $\7b^\tau=\7s_o.$ It is well-known that there exists

3373: a $\tau$-stable Cartan subalgebra $\7t\subset \7b$ and consequently we

3374: have the $\tau$-stable decomposition $\7b=\7t\ltimes

3375: [\7b,\7b]=:\7t\ltimes \7n.$ Select $\4h\in\7t,\;\4e\in \7n$ such that

3376: $[\4h,\4e]=2\4e.$ By construction $\tau(\4h)=a\cd \4h$ and

3377: $\tau(\4e)=b\cd \4e$ for suitable $a,b\in\CC$. Since $\tau$ is an

3378: automorphism, $a=1.$ Since $\tau$ is an involution, $|b|=1.$ This

3379: shows that up to a conjugation we may assume that $\7b$ is the

3380: upper-triangular Borel subalgebra of $\7{sl}(2,\CC)$ and the real

3381: forms $\7s_o\subset\7b^+$ have the following realizations:

3382: $$

3383: \7s_o\cong \Big\{\!\!\pmatrixn{t & sc\cr 0& -t}:t,s\in

3384: \RR\Big\}\steil{for some}c\in\CC^{*}\;.

3385: $$ For every $c\in\CC^{*}$ at least one of the Borel subalgebras

3386: $\CC\big({0\kern5pt 1 \atop 1\kern5pt 0}\big)\oplus \CC\big({1\kern2pt

3387: -1 \atop 1\kern2pt -1}\big)$ or $\CC\big({0\kern2pt -i \atop i\kern7pt

3388: 0}\big)\oplus \CC\big({1\kern7pt i \atop i\kern2pt -1} \big)$ then is

3389: complementary to $\7s_o$ in $\7{sl}(2,\CC).$ Finally, we need to deal

3390: with the case \nline $\Bullet\;\dim \7s_o= 1,$ i.e., $\7s$ itself is

3391: locally transitive and thus $\7g=\7s$ by the minimality condition

3392: \26$_{2}$. Consider the following matrix realization:\vskip-20pt

3393: $$

3394: \displaylines{\7l=\7{sl}(2,\CC)\times\7{sl}(2,\CC) \;\subset

3395: \;\CC^{2\times 2}\!\times \CC^{2\times 2} \cr

3396: \sigma(\4x,\4y)=(\overline{\4y},\overline{\4x})\,, \qquad

3397: \7g=\{(\4x,\overline{\4x}):\4x\in \7{sl}(2,\CC)\}\subset \7l\;.}

3398: $$

3399: The 3-dimensional complex subalgebra $\7q\subset\7l$ is either simple

3400: or solvable: \nline {\sl $\7q$ simple:} Then $\7q\cong \7{sl}(2,\CC)$

3401: and $\7q$ is either one of the two factors of $\7{sl}(2,\CC)\times

3402: \7{sl}(2,\CC)$ or $\7q$ is conjugate in $\7l$ to $\7g$. The first case

3403: can be ruled out immediately since then $\7q\cap \7g=0=\7g_o$ which is

3404: absurd.  In the second case, all simple subalgebras $\7q\subset \7l$

3405: which are not ideals are conjugate to each other. We may select the

3406: particular subalgebra $\7q=\{(\4x,-\4x^t):\4x\in

3407: \7{sl}(2,\CC)\}\subset \7l$, where $\4x^t$ is the transpose of

3408: $\4x$. The corresponding Lie subgroup $Q=\{(\4q,(\4q^{-1})^t):\4q\in

3409: \SL(2,\CC)\}\subset L:=\SL(2,\CC)\times\SL(2,\CC)$ is closed and the

3410: map $(\4x,\4y)\mapsto \4x\cd \4y^t$ identifies the quotient with the

3411: affine quadric $\SL(2,\CC)\subset\CC^{2\Times2}$ on which

3412: $S:=\SL(2,\CC)$ (considered as real Lie group) acts by the holomorphic

3413: transformations $(\4s,\4z)\mapsto \4s\4z\overline{\4s}{}^t$. Hence, in

3414: this situation every CR-germ associated with a CR-algebra $(\7s,\7q)$

3415: is globalizable. It is well-known (see, for instance \Lit{FEKA}) that

3416: the hypersurface $S$-orbits in $L/Q$ are either Levi-nondegenerate or

3417: locally CR-equivalent to the tube $\5M$ over the light cone. However,

3418: $\dim \7g=6$ and in this case $(\7g,\7q)$ satisfies \21 -- \25,

3419: \26$_{1}$ and \26$_{2}$, but not \26.\nline {\sl $\7q$ solvable:} We

3420: will show that also this case contradicts the fundamental assumption

3421: \ruf{FD}: \nline If $\7l_o=\CC(\4t,\overline{\4t})$ is ad-semisimple

3422: (and without loss of generality $\4t=\big({t \kern1em \atop \kern.7em

3423: -t}\big )$ for some $t\in \CC^{*}$) then $\7l_o$ is a regular torus in

3424: $\7l.$ Consequently, the centralizer $C_{\7l}(\7l_o)=:\7t_1\times

3425: \7t_2$ is a $\sigma$-stable Cartan subalgebra. Either

3426: $C_{\7l}(\7l_o)\subset \7q$ (but then $\7q+\sigma\7q$ is of

3427: codimension 2 in $\7l$) or $C_{\7l}(\7l_o)\cap \7q=\7l_o.$ In the

3428: latter case denote by $\7l^{\pm\alpha_1},$ $\7l^{\pm\alpha_2}$ the

3429: root spaces with respect to $\7t_1\times \7t_2.$ A direct check shows

3430: that $\7q$ is the direct sum of $\7l_o$ with two further root spaces

3431: (also if $\overline{\4t}=\4t$). Since $\7q$ is solvable, there are 4

3432: possibilities of choosing such pairs of root spaces. In all 4 cases

3433: either $\7q+\sigma\7q$ is too small or $\7f=\7l_o,$ i.e., the

3434: corresponding CR-germ is Levi nondegenerate. \nline It remains to

3435: discuss the case when $\7l_o=\CC(\4n,\overline{\4n})$ is

3436: ad-nilpotent. Since $\7q$ is solvable, it is contained in a Borel

3437: subalgebra $\7b$ of $ \7l.$ Consequently, $\7l_o\subset

3438: \7b^{\nil}=C_\7l(\7l_o)\cong \CC\4n \times \CC\overline{\4n}.$ Let

3439: $\pi:\7b\to \7b/\7b^{\nil}$ be the canonical projection. The image

3440: $\pi(\7q)$ cannot be surjective since there is no 3-dimensional

3441: subalgebra $\CC(\4n,\overline{\4n})\subset \7q \subset \7b$ with this

3442: property. It remains only the possibility $\7b^{\nil}\subset \7q$ but

3443: then $\7q\cap\sigma\7q \supset \7b^{\nil},$ which is too big.

3444: 

3445: Summarizing, we have for the CR-algebra $(\7g,\7q)$ satisfying

3446: \ruf{FD} that $\7g^{\s}$ must be a finite direct sum of copies of

3447: $\7{sl}(2,\RR)$ and $\su(2)$. In the next section we direct our

3448: attention to factors of type $\7{sl}(2,\RR)$.

3449: 

3450: 

3451: \KAP{Zwoelf}{The case $\7s\cong\7{sl}(2,\RR)$}

3452: 

3453: In this section we continue the proof of Theorem II and consider only

3454: CR-algebras $(\7g,\7q)$ subject to \ruf{FD}, for which the simple

3455: factor $\7s$ of $\7g^{\s}$ is isomorphic to $\7{sl}(2,\RR)$, compare

3456: \Ruf{SS}. As already proved, the remaining simple factors in $\7s'$

3457: (if there are any) are isomorphic to $\su(2)$ or $\7{sl}(2,\RR)$.

3458: 

3459: From $\hol(\5M,o)\cong\so(2,3)$ for the light cone tube $\5M$ it is

3460: clear that $\hol(\5M,o)$ contains copies of

3461: $\so(2,2)\cong\7{sl}(2,\RR)\Times\7{sl}(2,\RR)$. Fixing a subalgebra

3462: $\7r\subset\7{sl}(2,\RR)$ (which is necessarily non-abelian) the

3463: 5-dimensional Lie algebra $\7{sl}(2,\RR)\times\7r$ can be embedded

3464: into $\hol(\5M,o)$, and this can be done in such a way that the image

3465: is a transitive subalgebra. Therefore, the simple factor

3466: $\7{sl}(2,\RR)$ of $\7g^{\s}$ cannot avoided in the classification

3467: proof. However, we will show that this factor only occurs in the

3468: instance described above, that is, in connection with $\5M$.

3469: 

3470: Our first result is that in $\7g^{\s}$ the simple factor

3471: $\7{sl}(2,\RR)$ can occur at most once. Here and in the following we

3472: repeatedly use the basic fact that each proper subalgebra of

3473: $\7{sl}(2,\RR)$ has a solvable complementary subalgebra.  Furthermore,

3474: through this subsection we denote by $\pi:\7g=(\7s\times \7s')\ltimes

3475: \rad(\7g) \to \7s$ the canonical projection. We analyze various

3476: possibilities for the image of the isotropy subalgebra $\7g_{o}$ under

3477: $\pi$. \nline If $\pi(\7g_o)\ne 0$ then there exists a solvable

3478: complement $\7r\subset \7s$ to $\pi(\7g_o)$ and $(\7r\times

3479: \7s')\ltimes\rad(\7g)$ is a proper transitive subalgebra of $\7g$

3480: violating \26$_{2}$.\nline If $\pi(\7g_o)= 0$ then $\7g_o\subset

3481: \7s'\ltimes \rad(\7g)$ and is there of codimension 2. It follows that

3482: there is no factor $\7h\cong\7{sl}(2,\RR)$ in $\7s'$: Otherwise,

3483: counting dimensions we would have $\dim \7h \cap\7s_o\ge 1$ which

3484: implies that there is a proper transitive subalgebra of $\7g$,

3485: violating \26$_{2}$. This proves that $\7s'$ is a product of factors

3486: which all are isomorphic to $\su(2)$. Finally, we show that $\7s'$

3487: consists of at most {\sl one} such simple factor: Indeed, suppose that

3488: $\7s'=\7h\times\7s''$ for some ideal $\7h\cong\su(2)$ of $\7s'$.

3489: Denote by $\pi_{\7h}:=(\7s\times\7h\times\7s'')\ltimes

3490: \rad(\7g)\to\7h$ the canonical projection. Then $\pi_{\7h}(\7g_{o})$

3491: has codimension $\le2$ in $\7h$. Since $\su(2)$ does not have a

3492: subalgebra of dimension $2$, the dimension of $\pi_{\7h}(\7g_{o})$ can

3493: only be $1$ or $3$. But dimension $3$ violates \26$_{2}$ since then

3494: $\7g=\7g_{o}+\ker(\pi_{\7h})$. It remains the case

3495: $\dim\pi_{\7h}(\7g_{o})=1$. Then $\pi_{\7h}(\7g_o)$ is a torus in

3496: $\7h$ and $\pi(\7g_o)=\7h\cap \7g_o$. Since $\7g':=\7s\Times \7h$ is a

3497: transitive subalgebra of $\7g$ we must have $\7g'=\7g$ by

3498: \26$_{2}$. But this is not possible:

3499: 

3500: \Lemma{LK} $\7s\cong\7{sl}(2,\RR)$ implies

3501: $\7g\;=\;\7s\!\ltimes\!\rad(\7g)$ and $\7g_o\subset\rad(\7g)\,.$

3502: 

3503: \Proof By the above discussion we only have to rule out the case

3504: $\7g=\7s\times\7s'$ with $\7g_{o}\subset\7s'=\su(2)$. The key point

3505: here is that the isotropy subalgebra $\7g_o$ is toral in $\su(2)$ and

3506: that there exists a unique $\ad(\7g_o)$--stable subspace

3507: $\7p^\star\subset\su(2)$ on which the adjoint representation of

3508: $\7g_o$ is irreducible. Let $\pi_{\su}:\7s\oplus \su(2)\to \su(2)$ be

3509: the projection onto the second factor. Either $\pi_{\su}(\7F)=\su(2)$

3510: or $\pi_{\su}(\7F)=\7g_o.$ In the first case, we actually have

3511: $\7F=\su(2)$ as $[\7g_o,\7F]\subset \7F$ and $[\7s,\7g_o]=0$. But this

3512: cannot be true: Independently of what exactly $\7H=W\oplus \su(2)$

3513: would be, we always would have $[\7H,\7F]\subset \7F$. But this

3514: violates the nondegeneracy condition \25.  \nline It remains the case

3515: $\pi_{\su}(\7F)=\7g_o$, i.e., $\7F=\7b\oplus \7g_o$, where

3516: $\7b=\7F\cap \7s\subset\7s$ is a 2-dimensional real subalgebra. By a

3517: dimension argument $\dim \7H\cap \su(2) \ge 2.$ But this intersection

3518: must be $\ad(\7g_o)$-stable, which implies $\7H=\7b\oplus \su(2)$,

3519: i.e., $\7H$ is a subalgebra of $\7g=\7s\oplus \su(2)$.  Clearly, this

3520: violates condition \23.\qed

3521: 

3522: \smallskip In the next Lemma we show that the semidirect product

3523: $\7s\ltimes\rad(\7g)$ in \ruf{LK} actually is a direct product

3524: $\7s\Times\rad(\7g)$ and that the radical has dimension 2, i.e.  $\7g$

3525: has dimension 5. Recall the obvious fact that, up to isomorphy, there

3526: exist precisely two Lie algebras of dimension 2, the abelian Lie algebra

3527: $\RR^{2}=\RR\Times\RR$ and a non-abelian one.

3528: 

3529: \Lemma{FM} $\7s\cong\7{sl}(2,\RR)$ implies $\7g\;=\;\7s\Times\7r$ with

3530: $\7r:=\rad(\7g)$ being non-abelian and of dimension 2.

3531: 

3532: \Proof Let $\pi:\7g\to \7s$ be the canonical projection. We use the

3533: same symbol also for the complex extension $\7l\to

3534: \7s^{\CC}$. Because of \21 and $\7g_o\subset\7r$ the image

3535: $\pi(\7F)$ has dimension $\le2$.

3536: 

3537: \noindent $\Bullet\dim \pi(\7F)=0$ implies $\7F=\7r,$ i.e., $\7F$

3538: is an ideal in $\7g.$ But this violates \25.

3539: 

3540: \noindent $\Bullet\dim \pi(\7F)=1$ implies

3541: $\pi(\7f)=\pi(\sigma\7f)$ and $\dim\pi(\7f)=1$. Consequently

3542: $\pi(\7q),\pi(\sigma\7q)$ are 2-dimensional (Borel) subalgebras

3543: which generate $\7{sl}(2,\CC)$ as a linear space (otherwise $(M,o)$

3544: would be Levi flat). It follows $\pi^{-1}(\pi(\sigma\7q))\cap

3545: \7q=\7f.$ But this implies $[\7f,\sigma\7q]\subset \7f+\sigma\7q,$

3546: a contradiction to \25.

3547: 

3548: \noindent $\Bullet\dim \pi(\7F)=2.$ Note that $\pi(\7f),\;

3549: \pi(\sigma\7f)$ are 1-dimensional since $\7l_o\subset \7r^{\CC}$.

3550: Since $\pi(\7f)+\pi(\sigma\7f)$ is a 2-dimensional subalgebra,

3551: $\7t:=\pi(\7f),$ $\7t':=\pi(\sigma\7f)=\sigma\7t$ are tori as follows

3552: with the elementary structure theory of $\7{sl}(2,\CC)$. The case

3553: `$\pi(\7f)=\pi(\7q)$' can be excluded since otherwise we would have

3554: $\pi(\7q+\sigma\7q)=\pi(\7f+\sigma\7f)$ which is a subalgebra, a

3555: contradiction to \23. Hence, the only possibility remaining is

3556: $\pi(\7f)\ne \pi(\7q)$. This implies $\7q\cap \rad= \sigma\7q\cap

3557: \rad=\7l_o$. Furthermore, $\7l_o$ is an ideal in $\7q,\sigma\7q,$ and,

3558: since $[\7q,\sigma\7q]=\7l,$ even an ideal of $\7l.$ By the

3559: effectivity assumption \26$_{1}$, this implies $\7g_o=0,$ i.e., $\7g$

3560: has dimension $5$ and thus

3561: $$\7g\cong\7{sl}(2,\RR)\ltimes_\rho\RR^2\Steil{ or }\7g\cong

3562: \7{sl}(2,\RR)\times \7r\steil{with}\dim\7r=2\,,\Leqno{EM}$$ where

3563: $\rho:\7{sl}(2,\RR)\to \End(\RR^2)$ is the canonical inclusion.

3564: 

3565: \Claim{NO} The first case in \Ruf{EM}, that is

3566: $\7g\cong\7{sl}(2,\RR)\ltimes_\rho\RR^2$, cannot occur.

3567: 

3568: \ProofC Let $\pi:\7l\to \7{sl}(2,\CC)$ be the canonical projection and

3569: $\sigma:\7l\to \7l$ the complex conjugation defining the real form

3570: $\7g$ of $\7l$. The possibility \nline $\Bullet\;\pi(\7q)=0\,$

3571: ($=\pi(\sigma\7q)\,$) can be excluded since then

3572: $\7q=\rad(\7l)=\sigma\7q$, violating \23. Also

3573: 

3574: \noindent $\Bullet\dim \pi(\7q)=1$ can be ruled out: Then

3575: $\pi(\7q)\ne\pi(\7f)$ since otherwise

3576: $\pi(\7q+\sigma\7q)=\pi(\7f+\sigma\7f)$ would be a subalgebra,

3577: violating \23. Hence, $\7f=\7q\cap \rad(\7l)$. But this contradicts

3578: \25 since then $[\7f,\sigma\7q]\subset\rad(\7l)=\7f\oplus

3579: \sigma\7f$.

3580: 

3581: \noindent $\Bullet\dim \pi(\7q)=2$ (i.e., $\7q\cap \rad(\7l)=0$) is

3582: the most involved case. Then $\pi(\7q)=\7b$ is a Borel subalgebra of

3583: $\7{sl}(2,\CC)$. To rule out also this case we need some

3584: preparations. First realize $\7l$ as \vskip-20pt

3585: $$

3586: \7{sl}(2,\CC)\ltimes_{\rho} \CC^2=\{(X,w): X\in \7{sl}(\CC^2), \;w\in

3587: \CC^2 \}\steil{with}[(X,w),(Y,u)]=([X,Y]\;, \;Xu-Yw)

3588: $$ and complex conjugation $\sigma$ given by $(X,w)\mapsto(\overline

3589: X,\overline w)$. Since $(M,o)$ is not Levi flat (compare \23), we

3590: necessarily have $\7b+\sigma\7b=\7{sl}(2,\CC),$ and the 1-dimensional

3591: intersection $\7b\cap \sigma\7b$ is a toral subalgebra

3592: $\7t\subset\7{sl}(2,\CC)$ (we use here the well-known fact that the

3593: [$\sigma$--stable] intersection of any two Borel subalgebras contains

3594: a [$\sigma$--stable] Cartan subalgebra). In the next two paragraphs we

3595: recall some elementary facts from the representation theory of

3596: $\7{sl}(2,\CC)$ which we need to complete our proof.

3597: 

3598: \Joker{FN}{$\sigma$-adapted $\7{sl}_2$-triples} Let $\4h\in \7t\subset

3599: \7{sl}(2,\CC)$ be the element for which $[\4h,E]=2E$ and $[\4h,F]=-2F$

3600: for every $E\in \7b^{\nil},$ $F\in

3601: \sigma(\7b^{\nil})=(\sigma\7b)^{\nil}.$ Since $\sigma$ interchanges

3602: the eigenspaces of $\ad(\4h),$ we have $\sigma(\4h)=-\4h$. There are

3603: crucial technical points here: We claim that there exists $\4e\in

3604: \7b^{\nil}$ such that $[\4e,\sigma(\4e)]=\4h,$ i.e., $(\4e,\4h,\4f)$

3605: with $\4f:=\sigma(\4e)$ being an $\7{sl}_{2}$-triple in

3606: $\7{sl}(2,\CC)$ (i.e.,

3607: $[\4h,\4e]=2\4e,\;[\4h,\4f]=-2\4f,\;[\4e,\4f]=\4h$). \nline{\bf

3608: Construction:} We have to be careful here about signs: Since $\sigma$

3609: defines a {\sl non-compact} real form of $\7{sl}(2,\CC),$ the

3610: nondegenerate Hermitian 2-form $\kappa(\cdot\,,\sigma(\,\cdot\, ))$

3611: has precisely one negative eigenvalue, where $\kappa$ denotes the

3612: Killing form. Since $\kappa(\4h,\sigma(\4h))=\kappa(\4h,-\4h)<0$, it

3613: follows $\kappa(E,\sigma(E))>0$ and consequently $[\lambda

3614: E,\sigma(\lambda E)]=\4h$ for an appropriately chosen $\lambda\in

3615: \CC^*$ (keeping in mind the general formula $[E,F]=\kappa(E,F)H$ where

3616: $H$, the coroot, is a positive multiple of $\4h$). Define then

3617: $\4e:=\lambda E$ and $\4f:=\sigma(\4e)$. Each $\7{sl}_{2}$-triple

3618: $\4e',\4h',\4f'\in \7s^{\CC}\cong \7{sl}(2,\CC)$ with

3619: $\sigma(\4e')=\4f'$ is called $\sigma$--{\sl adapted} in the

3620: following.

3621: 

3622: \Joker{FO}{\hskip-3pt} Complexifying $\7{sl}(2,\RR)\ltimes_\rho \RR^2$, we

3623: briefly discuss how the 2-dimensional abelian radical $\CC^2,$

3624: considered as an $\7{sl}(2,\CC)$-module, is related to the real

3625: structure. Let a $\sigma$-adapted $\7{sl}_2$-triple $\4e,\4h,\4f$ be

3626: given.  Let $\CC^2=V_+\oplus V_-$ be the decomposition into $(\pm 1)$

3627: $\4h$-eigenspaces. They are interchanged by $\sigma$. We claim that

3628: there exists a $v_+\in V_+$ with

3629: $v_-:=\sigma(\4e)v_+=\sigma(v_+)\ne0$: Indeed, choose $w_+\in V_+\sm

3630: \{0\}$ arbitrarily. Then there is a $c\in \CC^*$ with $\4f

3631: w_+=c\sigma(w_+)$, and a direct check shows $|c|=1.$ Choose a

3632: $b\in\CC$ with $b^{2}=\overline c$ and put $v_+:=b\cd w_+$.  \nline We

3633: use the linear basis $v_+,v_-$ of the radical $\rad(\7l)\cong\CC^2$ in

3634: the following computations.

3635: 

3636: \noindent We now resume the proof of Claim \ruf{NO}. For short,

3637: write $\7r^{\CC}  :=\rad(\7l)$

3638: for the abelian radical.

3639: Since  $\7q\cap \7r^{\CC}=0$, we can write

3640: $\7q=\CC\cd(\4e,w_1)\oplus \CC\cd(\4h,w_2)$ for suitable $w_1,w_2\in

3641: \CC^2.$ Clearly, the $w_j$'s are not arbitrary: Write

3642: $w_1=\lambda_1v_+ +\mu_1v_-$ and $w_2=\lambda_2v_+ +\mu_2v_-.$ Note

3643: that $w_2\ne -\overline w_2$ (otherwise $\7q\cap\sigma\7q\ne 0$),

3644: i.e., $\lambda_2\ne -\overline \mu_2.$ The fact that $\7q$ is a

3645: subalgebra imposes a further condition on the coefficients

3646: $\lambda_j,\mu_j:$ A simple computation shows that $\mu_1=0$ and

3647: $\mu_2=-\lambda_1,$ i.e., for each $\lambda,\mu \in \CC$ with

3648: $\lambda\ne \overline \mu$ we have the two 2-dimensional complex

3649: subalgebras $\7q=\7q_{\lambda,\mu}$ and $\sigma\7q:$

3650: 

3651: \noindent \hfil $\7q=\CC\cd(\4e,\lambda v_+)\oplus

3652: \CC\cd(\4h,\mu v_+-\lambda v_-),\quad ~ \sigma\7q=\CC\cd(\4f,\overline

3653: \lambda v_-)\oplus \CC\cd(-\4h,-\overline \lambda v_++\overline\mu

3654: v_-)\,. $

3655: 

3656: \noindent Recall the definition $\7f=\{u\in

3657: \7q:[u,\sigma\7q]\subset\7q+\sigma\7q\}$ from \ruf{CA}.iv. A

3658: straightforward calculation shows that in all cases $\7f=0$

3659: holds. This contradicts \24 and proves \ruf{NO}. \qd

3660: 

3661: \noindent We proceed the proof of Lemma \ruf{FM} by restricting  the

3662: radical of $\7g\cong\7{sl}(2,\RR)\times\7r$ further:

3663: 

3664: \Claim{FP} The Lie algebra $\7g$ cannot be isomorphic to

3665: $\7{sl}(2,\RR)\times \RR^2$.

3666: 

3667: \ProofC Assume to the contrary that $\7g=\7{sl}(2,\RR)\times\RR^{2}$

3668: and denote by $\pi_{1}:\7l\to\7{sl}(2,\CC)$, $\pi_{2}:\7l\to\CC^{2}$

3669: the canonical projections.  As in the proof of the previous claim, the

3670: cases $\dim \7q\cap \rad(\7l)\ge 1$ can be ruled out immediately. We

3671: therefore only have to exclude the case $\7q\cap\CC^{2}=0$: In that

3672: case, let $\7b$ be the 2-dimensional $\pi_{1}$-image in

3673: $\7{sl}(2,\CC)$. It follows $\sigma(\7b)\ne \7b$ since otherwise

3674: $\pi_\7s(\7q+\sigma\7q)\subset \7b$ would contradict \23. Consequently

3675: there exist $\4e\in \7b,$ $\4h\in \7b\cap \sigma\7b$ and

3676: $\4f=\sigma(\4e)\in \sigma(\7b)=\pi(\sigma\7q)$ with the properties

3677: described in \ruf{FN}. The fact that $\7q,\sigma\7q$ are subalgebras

3678: and that $\pi(\7q)=\CC\4h \oplus \CC \4e$ determine $\7q,\sigma\7q$ as

3679: follows: $\7q=\CC\cd(\4h,w)\oplus\CC\cd (\4e,0)\steil{and}

3680: \sigma\7q=\CC\cd (\4h,-\overline w) \oplus \CC\cd (\4f,0)\steil{for a}

3681: w\in \CC^2\,.$\nline We may further assume that $w,\overline w$ are

3682: linearly independent, otherwise $\pi^{}_{2}(\7q+\sigma\7q)\ne\CC^{2}$

3683: would contradict \23. As a consequence, $(\4h,0)\not\in

3684: \7q+\sigma\7q$. The definition of $\7f$ then shows $

3685: \7f=\CC\cd(\4h,w)\;$.  On the other hand, for this $\7f$ we have

3686: $[\7f,\sigma\7q]\subset \7f\oplus \sigma\7q$ in contradiction to

3687: \25. This proves the claim and Lemma \ruf{FM}.\qed

3688: 

3689: \medskip So far we know that under the assumption $\7s\cong

3690: \7{sl}(2,\RR)$

3691: necessarily $\7g\cong\7{sl}(2,\RR)\times\7r$, where $\7r$ is the

3692: 2-dimensional non-abelian Lie algebra. To determine the full CR-algebra

3693: $(\7g,\7q)$ we still have to find out how the complex subalgebra

3694: $\7q\subset\7l$ sits inside $\7{sl}(2,\CC)\times\7r^{\CC}$.

3695: 

3696: \Claim{FQ} Up to a CR-algebra automorphism of $(\7g,\7q)$ the

3697: subalgebra $\7q\subset\7l$ is obtained in the following way: Fix a

3698: linear basis $\4x,\4z$ of $\,\7r^{\CC}$ with $[\4x,\4z]=\4z$ and a

3699: $\sigma$--adapted $\7{sl}_{2}$-triple $\4e,\4h,\4f\in\7{sl}(2,\CC)$

3700: (see \ruf{FN} for the definition). Then

3701: $\7q=\CC(\4h\,,2\4x+\mu\4z)\;\oplus\;\CC(\4e\;, \nu\4z)\;$ for

3702: suitable $\mu,\nu\in\CC$ with $\Im(\mu)=\pm2$ and $|\nu|=1$.

3703: 

3704: \ProofC The case $\7q=\7r^{\CC}$ can clearly be

3705: excluded. If the intersection $\7q\cap \7r^{\CC}$ would be

3706: 1-dimensional then $\7q\cap \7r^{\CC}\;\oplus\;\sigma\7q\cap

3707: \7r^{\CC}=\7r^{\CC}$ (as the sum $\7q+\sigma\7q$ must be

3708: direct). Since $\7f$ must also be 1-dimensional, we have

3709: $\7f=\7r^{\CC}\cap \7q$. But then $[\7f,\sigma\7q]\subset

3710: \7r^{\CC}=\7f\oplus \sigma\7f$, violating condition \25. \nline It

3711: remains the case $\7q\cap \7r^{\CC}=0,$ i.e., $\pi_\7s(\7q)$ is a

3712: Borel subalgebra $\CC \4e \oplus \CC \4h,$ and

3713: $\pi(\sigma\7q)=\CC\4h\oplus \CC \4f$ where $\4h,\4e,\4f\in

3714: \7{sl}(2,\CC)$ are chosen as explained in \ruf{FN}. Since

3715: $\7q\subset \7s^{\CC}\times \7r^{\CC}$ is a Lie sub{\sl algebra}, it

3716: necessarily has the following form:

3717: $$ \diagram{\7q=\CC\cd (\4h,\lambda \4x+\mu \4z)\oplus \CC\cd (\4e,

3718: \nu \4z)\Steil{with}\sigma\7q= \CC\cd (-\4h,\overline\lambda

3719: \4x+\overline \mu \4z)\oplus \CC\cd (\4f, \overline \nu \4z) \cr

3720: \hbox{and} \quad \lambda,\mu,\nu\in \CC\Steil{satisfying}

3721: \lambda=2\steil{if}\nu\ne 0\,.}\Leqno{FR}

3722: $$ {\fett$\nu =0$:} Then $\lambda \4x+\mu \4z$ and $\overline \lambda

3723: \4x+\overline \mu \4z$ must be linearly independent in $\7r^{\CC}$,

3724: i.e., $\lambda\overline\mu\ne \overline\lambda\mu$ (otherwise

3725: $\pi_\7r(\7q{+}\sigma\7q)\ne \7r^{\CC}$). Observe that

3726: $(\4h,0)\not\in\7q+\sigma\7q$ and $\lambda\ne0$. A direct verification

3727: shows $\7f=\CC\cd(\4h,\lambda \4x+\mu \4z)$ if $\lambda\in i\RR$ and

3728: $\7f=0$ otherwise. But then $\7f\oplus \sigma\7q$ is a subalgebra in

3729: contradiction to \25.\nline {\fett $\nu\ne0$:} Possibly after

3730: replacing $\4z$ by $|\nu|\4z$ we may assume $|\nu|=1$ in

3731: \Ruf{FR}. Employing the definition of $\7f$ in \ruf{CA}.iv, a

3732: simple calculation shows

3733: $$

3734: \7f=\cases{\CC(\4h\mp2i\overline\nu\,\4e\,,\,2\4x +\Re(\mu)\4z)\quad

3735: & $\Im(\mu) =\pm 2$\cr 0 & otherwise,}\Leqno{FS}

3736: $$

3737: that is, $\7f$ is 1-dimensional only if $\Im\mu =\pm 2$. This proves

3738: the claim.\qd

3739: 

3740: \Lemma{ZC} For the CR-algebra $(\7g,\7q)$ in \ruf{FQ} the associated

3741: CR-germ $(M,o)$ is CR-equivalent to $(\5M,a)$, where $\5M$ is the tube

3742: over the future light cone.

3743: 

3744: \Proof We start by giving a particular representing manifold for the

3745: associated germ, compare Example 6.6 in \Lit{FEKA}. Let $\mu,\nu$ be

3746: the constants occurring in \ruf{FQ} and consider the affine quadric

3747: $Z:=\SL(2,\CC)\subset \CC^{2\times 2}$, on which the group $\widehat

3748: L:=\SL(2,\CC)\times \SL(2,\CC)$ acts holomorphically by $z\mapsto

3749: gzh^{-1}$ for all $(g,h)\in\widehat L$. Then $\widehat

3750: G:=\SL(2,\RR)\times \SL(2,\RR)$ is a real form of $\widehat L$. Via

3751: $$

3752: \eqalign{ \4x:=&{1\over 2}\pmatrix{ 0 &1 \cr 1 & 0},\quad \4z:={1\over

3753: 2}\pmatrix{1 &\mkern-10mu -1 \cr 1 &\mkern-10mu -1}\quad

3754: \in\;\7r\qquad\cr \4e:=&{\nu\over 2}\pmatrix{1 & i\cr i &

3755: \mkern-12mu -1},\quad \4h:=\pmatrix{0 & \mkern-10mu -i \cr i &

3756: 0},\quad \4f:=\sigma(\4e)\;\;\;\in\;\;\7s^{\CC} }\Leqno{VB}

3757: $$ we consider $\7l=\7s^{\CC}\times\7r^{\CC}$ as a complex subalgebra

3758: of $\widehat\7l=\7{sl}(2,\CC)\times\7{sl}(2,\CC)$ and $\7g$ as a

3759: subalgebra of $\widehat{\7g}$. \nline We recall some basic facts,

3760: compare \Lit{FEKA}: The polynomial function $\psi(z):=\det(z+\overline

3761: z)-2$ on $Z$ is invariant under the action of the group $\widehat

3762: G$. Furthermore, the nonsingular part $M$ of the algebraic subset

3763: $S:=\psi^{-1}\big(\{\pm2\}\big)$ is a (non-connected) hypersurface in

3764: $Z$, locally CR-equivalent to $\5M$. On the other hand, the singular

3765: part of $S$ is a totally real submanifold of $Z$. Consider the point

3766: $$

3767: o:={e^{-\pi i/4}\over4}\pmatrix{4+\mu & -\mu \cr i\mu

3768: &i(4-\mu)}\;\in\;Z

3769: $$

3770: which actually is in $S$ because of $\psi(o)=\Im(\mu)$. A direct

3771: computation shows that the isotropy subalgebra $\7l_{o}$ of

3772: $\7l\subset\widehat\7l$ at $o\in Z$ is $\CC(\4h,2\4x+\mu\4z)\oplus \CC

3773: (\4e,\nu\4z)$ and that the isotropy subalgebra of

3774: $\7g\subset\widehat\7g$ at $o$ is trivial. This implies $a\in M$ and

3775: also that the CR-algebra $(\7g,\7q)$ is associated to the germ

3776: $(M,o)$. Since $M$ is locally CR-equivalent to the tube $\5M$ over the

3777: future light cone the proof for \ruf{ZC} is complete.\qed

3778: 

3779: 

3780: 

3781: 

3782: \KAP{Dreizehn}{The case $\7s\cong\su(2)$}

3783: 

3784: 

3785: The status of our proof so far can be summarized as follows: For the

3786: CR-algebra $(\7g,\7q)$ satisfying \ruf{FD} and $\7g^{\s}\ne 0$ the

3787: simple ideal $\7s$ of $\7g^{\s}$ can only by isomorphic to

3788: $\7{sl}(2,\RR)$ or $\su(2)$, Furthermore, the case

3789: $\7s\cong\7{sl}(2,\RR)$ only occurs if the associated CR-germ is

3790: equivalent to $(\5M,o)$ with $\5M$ being the light cone tube, and then

3791: $\7g\cong \7{sl}(2,\RR)\times \7r$ with nonabelian 2-dimensional $\7r$.

3792: Consequently, only the situation needs to be investigated when

3793: $\7g^{\s}$ only contains simple ideals isomorphic to $\su(2)$. We

3794: assume this throughout this section and state as main lemma:

3795: 

3796: \Lemma{VD} There is no simple factor of $\7g^{\s}$ isomorphic to

3797: $\su(2)$.\Formend

3798: 

3799: \noindent The {\sl proof} will be subdivided into several claims.

3800: Note that contrary to $\7{sl}(2,\RR),$ the only non-trivial proper

3801: subalgebras of $\su(2)$ are 1-dimensional tori.  As before let $\7s$

3802: be a fixed simple factor of $\7g^{\s}$ and denote by $\pi:\7g\to\7s$

3803: the canonical projection. Then the image $\pi(\7g_o)$ must be a proper

3804: subalgebra since otherwise $\7g'=\pi^{-1}(0)$ would violate

3805: \26$_{2}$. Our first observation is

3806: \Claim{} For $\7g$ and $\7r:=\rad(\7g)$ only the following cases may

3807: occur\vskip-23pt

3808: $$

3809: \displaylines{ \7g\cong\su(2)\ltimes\7r,\qquad \7g\cong\su(2)\times

3810: \su(2) \qquad\hbox{\rm or} \cr \7g\cong\big(\su(2)\times

3811: \su(2)\big)\ltimes\7r \Steil{ with }\7r\ne

3812: 0\steil{and}\7g_o=(\RR\kern1pt \4t_1 \times \RR\kern1pt \4t_2)

3813: \ltimes(\7g_o\!\cap\7r)\steil{for}\4t_{j}\in\su(2)\;.}

3814: $$

3815: 

3816: \ProofC Write $\7g=(\7s_1\oplus \7s_2\oplus \7s'')\ltimes \7r$ where

3817: $\7s_1,\7s_2\cong\su(2)$, $\7s''$ is the complementary ideal in

3818: $\7g^{\s}$ and $\pi_j$ denotes the projection onto $\7s_1$ or $\7s_2$,

3819: respectively.  Only one of the following possibilities could occur:

3820: \nline $\Bullet\pi_1(\7g_o)=0$. Then $\7s_1\oplus \7s_2$ is already a

3821: transitive subalgebra of $\7g$ and $\7s''=\7r=0$. \nline

3822: $\Bullet\pi_1(\7g_o)\ne 0\ne \pi_2(\7g_o)$. Then either

3823: $\7g=\7s_1\oplus\7s_2$ or $\7g_o\cap (\7s_1\oplus \7s_2)=

3824: (\7g_o\cap\7s_1)\oplus (\7g_o\cap \7s_2)= \RR \4t_1 \oplus \RR \4t_2$

3825: for suitable nonzero $\4t_1,\7t_2\in \su(2)$. In the latter case

3826: $\7g_o\cap (\7s''\ltimes \7r)$ is of codimension 1 in $\7s''\ltimes

3827: \7r$. But then we conclude that $\7s''=0$ as there is no

3828: 1-codimensional subalgebras in $\su(2)$. \qd

3829: 

3830: \noindent For the proof of \ruf{VD} we only need to investigate the

3831: above three types of $\7g$. We repeatedly use the fact that each 1- or

3832: 2-dimensional representation of $\su(2)\cong\so(3)$ is trivial and

3833: that each toral subalgebra $\7t\subset \su(2)$ acts irreducibly on

3834: $\qu{\su(2)}{\7t}.$

3835: 

3836: 

3837: \Claim{} $\7g\cong(\su(2)\times \su(2))\ltimes\7r$ implies

3838: $\7r=0$.

3839: 

3840: \ProofC Let $\pi_1,\pi_2$ be the projection onto the first and the

3841: second simple factor, respectively. As observed in the preceeding

3842: Claim, $\7g_o=(\7t_1\times \7t_2)\ltimes \7r_o$ with $\7r_o:=\7g_o\cap

3843: \7r$ and 1-dimensional toral subalgebras $\7t_j.$ Recall that we have

3844: the $\7g_o$-stable filtration $\7g\supset \7H\supset \7F\supset

3845: \7g_o.$ At least one of the images $\pi_j(\7F)$ coincides with

3846: $\7s_j,$ say, for $j=2.$ Since $\7t_j$ acts irreducibly on

3847: $\qu{\7s_j}{\7t_j},$ it follows $\7F=\7s_2\ltimes \7r_o$. Since then

3848: $\7H\cap \7s$ is at least 2-dimensional, the irreducibility of the

3849: action of $\7t_1$ implies $\7H=\7s \times \7s_2\ltimes \7r_o.$ But

3850: this would imply $[\7H,\7H]\subset \7H,$ contradicting \23.\qd

3851: 

3852: \Claim{} $\7g\not\cong\su(2)\times\su(2)$.

3853: 

3854: \noindent\rm {\sl Proof of the Claim.} Suppose to the contrary that

3855: $\7g=\su(2)\times\su(2)$. Since $\7F\subset\7g$ is a subalgebra of

3856: dimension 3 and there is no solvable subalgebra of this dimension,

3857: necessarily $\7F\cong \su(2)$. Consequently, either $\7F$ is one of

3858: the simple factors of $\7g$ or $\7F$ is the graph of an automorphism

3859: of $\su(2)$. Both possibilities lead to a contradiction: In the first

3860: case $\7F$ is an ideal in $\7g,$ contradicting \25, and in the second

3861: case $\7F$ acts irreducibly on $\qu{\7g}{\7F}$, violating the

3862: existence of the $\ad(\7F)$-stable proper subspace $

3863: \qu{\7H}{\7F}\subset\qu{\7g}{\7F}$.\qd

3864: 

3865: \noindent The remaining case $\7g=\su(2)\ltimes\7r$ with

3866: $\7r:=\rad(\7g)$ is the most involved. In this situation the

3867: projection of $\7g_o$ under the canonical projection

3868: $\pi:\7g\to\su(2)$ is of dimension $\le1$. As usual we denote the

3869: canonical projection $\7l\to\7{sl}(2,\CC)$ by the same symbol

3870: $\pi$. We investigate the various possibilities for the $\pi$-images

3871: of the Lie subalgebras $\7l_o\subset\7f\subset\7q$ defined in

3872: \Ruf{FA}.\nline $\Bullet\pi(\7l_o)=\pi(\7f)=\pi(\7q)$.  Since

3873: $\pi(\7l_o)$ is a $\sigma$-stable torus, it would follow

3874: $\pi(\7q+\sigma\7q)=\pi(\7l_o).$ Counting dimensions, this cannot

3875: happen. \nline $\Bullet\pi(\7l_o)\subsetneq\pi(\7f)=\pi(\7q)$. Here,

3876: we have to rule out the 2 possibilities $\dim \pi(\7l_o)\le1$. \nline

3877: If $\dim\pi(\7l_o)=0$ then $\pi(\7F)$ is a 1-dimensional toral

3878: subalgebra in $\su(2),$ and consequently

3879: $\pi(\7f)=\pi(\sigma\7f)=\pi(\7q)=\pi(\sigma\7q)$ is

3880: 1-dimensional. This contradicts the fact that $\7q+\sigma\7q$ is a

3881: hyperplane in $\7l.$ It remains the case when $\pi(\7l_o)$ is

3882: 1-dimensional. Then $\pi(\7F)=\7s,$ and possibly after replacing

3883: $\7s\cong \su(2)$ by the Levi factor contained in $\7F$ we may assume

3884: that $\7s\subset \7F.$ By dimension reasons $\7r_o:=\7g_o\cap

3885: \7r=\7F\cap \7r$. Further, $\7r_o$ is an ideal in $\7F$ and we have

3886: $\7F=\7s\ltimes \7r_o$ as well as $\7H=\7s\ltimes \7r_\7H$ with

3887: $\7r_\7H:=\7H\cap \7r.$ Since $\dim \7r_\7H-\dim \7r_o=2,$ the

3888: ad-representation of $\7s$ on $\qu{\7r_\7H}{\7r_o}$ is trivial, i.e.,

3889: $[\7s,\7r_\7H]\subset \7r_o.$ Since $\7f=\7l_o\oplus \CC \4y$ for some

3890: $\4y\in \7s^{\CC}\cong\7{sl}(2,\CC)$ and $\sigma\7q=\sigma\7f\oplus

3891: \CC\overline{\4x}$ for some $\overline{\4x}\in \7r^{\CC}_\7H,$ we

3892: would have $[\7f,\sigma\7q]=[\7l_o\oplus \CC \4y,\sigma\7f\oplus \CC

3893: \overline{\4x}]\subset \7f+\sigma\7f,$ in contradiction to \25. \nline

3894: $\Bullet\pi(\7l_o)=\pi(\7f)\subsetneq\pi(\7q).$ The possibility

3895: $0=\pi(\7f)$ can be excluded since otherwise $\7F=\7r$ and

3896: $\7f+\sigma\7f$ would be an ideal in $\7l,$ contradicting condition

3897: \25. Next we deal with the case when

3898: $\pi(\7l_o)=\pi(\7f)=\pi(\sigma\7f)$ is a (1-dimensional) toral

3899: subalgebra. By assumption $\7b:=\pi(\7q)$ is then 2-dimensional, i.e.,

3900: a Borel subalgebra, which implies that $\7b+\sigma{\7b}=\7s^{\CC}\cong

3901: \7{sl}(2,\CC),$ or equivalently $\pi(\7H)=\7s.$ Counting dimensions,

3902: $\7r_\7F:=\7F\cap \7r=\7H\cap \7r$. Further, $[\7r_\7F,\7H]\subset

3903: \7r\cap \7H=\7r_\7F$ and since, due to condition \23, $\7H$ generates

3904: $\7g$ as a Lie algebra, we deduce that $\7r_\7F$ is an ideal in $\7g.$

3905: From $\pi(\7l_o)=\pi(\7f)$ follows that $\7f=\7l_o\oplus \CC\4r$ with

3906: $\4r\in \7r^{\CC}_\7F.$ But this implies $[\7f,\sigma\7q]=

3907: [\7l_0\oplus \CC\4r,\sigma \7q]\subset \sigma\7q+\7r^{\CC}_\7F\subset

3908: \sigma\7q\oplus \7f,$ violating \25. It remains one last possibility

3909: for the flag $\7l_o\subset \7f\subset \7q$ in $\7l$:

3910: 

3911: \Claim{} If $\pi(\7l_o)\subsetneq\pi(\7f)\subsetneq\pi(\7q)$ then

3912: $\7l_o=0$ and $\7g\cong \su(2)\times\7r$ with $\7r:=\rad(\7g)$ of

3913: dimension 2.

3914: 

3915: \ProofC The properness of the inclusions implies $\7l_o\cap

3916: \7r^{\CC}=\7f\cap\7r^{\CC}=\7q\cap \7r^{\CC}=\sigma\7q\cap \7r^{\CC}$.

3917: Since $\7q$ and $\sigma \7q$ generate $\7l$ as a Lie algebra, it

3918: follows that $\7l_o\cap \7r^{\CC}$ is an ideal in $\7l,$ or

3919: equivalently $\7g_o\cap \7r\vartriangleleft \7g.$ By the effectivity

3920: assumption \26$_{1}$, we have $\7g_o\cap \7r=0.$ Consequently, since

3921: $\dim\pi(\7l_o)\le1$ the same estimate holds for $\dim\7l_o$. Next we

3922: show that the case $1=\dim_{\RR}\pi(\7g_o)\,$ ($\,=\dim

3923: \pi(\7l_o)=\dim \7l_o$) cannot happen: Assume on the contrary that

3924: $\pi(\7g_o)=1$. Then $\pi(\7F)=\7s$ and, possibly after replacing the

3925: Levi factor $\7s\subset \7g$ by a conjugate one, we may assume

3926: $\7s\subset \7F.$ Counting dimensions yields then $\7F=\7s\cong

3927: \su(2).$ Since $[\7F,\7H]\subset \7H$ by \22, we have a representation

3928: of $\7s$ on $\qu{\7H}{\7F}$. But this yields the contradiction as

3929: $\dim \qu{\7H}{\7F}=2$, compare \21, implies that this representation

3930: is trivial, i.e. $[\7s,\7H]=[\7F,\7H]\subset \7F$, violating

3931: \25. \nline We have proved that the only possibility for $\7g$ is

3932: $\su(2)\ltimes\7r$ with a 2-dimensional radical $\7r.$ Since $\su(2)$

3933: can act only trivially on such an $\7r$, the above semidirect product

3934: is in fact direct. This proves the claim.\qd

3935: 

3936: \Claim{} The case $\7g\cong\su(2)\times \7r$, $\dim \7r=2$, also

3937: cannot occur.

3938: 

3939: \ProofC We write $\pi=\pi_\7s$ and $\pi_\7r$ for the projections onto

3940: $\7s^{\CC}\cong \7{sl}(2,\CC)$ and $\7r^{\CC},$ respectively. The

3941: proof analyses various possibilities for $\7f\subset \7q.$ \nline If

3942: $\pi_\7s(\7F)=0$ then $\7F=\7r$ would be an ideal in $\7g$

3943: contradicting \25. It remains the case when $\pi_\7s(\7F)$ as well as

3944: $\pi_\7s(\7f)$ is 1-dimensional. In that situation necessarily

3945: $\pi_\7s(\7q)\ne \pi_\7s (\7f),$ and $\pi(\7q)$ is a Borel subalgebra

3946: in $\7{sl}(2,\CC).$ Further,

3947: $\pi_\7s(\7q)+\sigma(\pi_\7s(\7b))=\7{sl}(2,\CC)$ and

3948: $\pi_\7s(\7q)\cap\pi_\7s(\sigma\7b)$ is a Cartan subalgebra of

3949: $\7{sl}(2,\CC).$ Similar to the situation considered in \ruf{FN} we

3950: also can choose here a {\sl standard triple} $\4e,\4f,\4h\in

3951: \7s^{\CC}$ with $\pi_\7s(\7q)=\CC \4h \oplus \CC \4e $ and

3952: $\sigma(\4h)=-\4h$, but now $\sigma(\4e)=-\4f$ for the Cartan

3953: involution $\sigma$. The radical $\7r$ cannot be abelian: Otherwise,

3954: exactly as in the proof of Claim \ruf{FP}, we obtain a

3955: contradiction. In the remaining non-abelian case we proceed as in the

3956: proof of Claim \ruf{FQ} by investigating the various positions of

3957: $\7q$ in $\7l$: The cases $\dim \7r^{\CC}\cap \7q>0$ are easily ruled

3958: out (same argument as in the proof of \ruf{FQ}, following

3959: \Ruf{FR}). Hence, we may assume that $\7q$ and $\sigma\7q$ are given

3960: by the formula \Ruf{FR}, except that now $\sigma\7q =\CC\cd

3961: (-\4h,\overline\lambda \4x+\overline \mu \4z)\oplus \CC\cd (-\4f,

3962: \overline \nu \4z)$, i.e., the sign in front of $\4f$ has

3963: changed. This slight difference is precisely the reason why in case

3964: $\nu\ne 0,$ contrary to \Ruf{FS}, the CR-germ $(M,o)$ associated to

3965: $(\7g,\7q)$ would be Levi nondegenerate, as shown by a simple

3966: computation. This contradicts our fundamental assumption and concludes

3967: the proof of the claim as well as the proof of Proposition

3968: \ruf{VD}.\qed

3969: 

3970: 

3971: \KAP{Vierzehn}{Reduction to the case where $\7g$ is solvable and of

3972: dimension 5}

3973: 

3974: Striking the balance for the proof of Theorem II obtained so far,

3975: we have shown the following:

3976: 

3977: \noindent{\sl Let $(M,o)$ be an arbitrary

3978: locally homogeneous $2$-nondegenerate CR-germ of dimension $5$ and let

3979: $(\7g,\7q)$ be an associated CR-algebra. If the Lie algebra $\7g$ is

3980: not solvable then $M$ is locally

3981: CR-equivalent at $o\in M$ to the tube $\5M$ over the future light cone.}

3982: 

3983: For the rest of the proof we therefore assume that every CR-algebra

3984: $(\7g,\7q)$ under consideration satisfies the fundamental assumption

3985: \ruf{FD} and that $\7g$ is solvable. As main result of this section we

3986: show

3987: 

3988: 

3989: \Lemma{FT} The solvable Lie algebra $\7g$ has dimension $5$, i.e.,

3990: $\7g_o=0$.\Formend

3991: 

3992: \noindent The {\sl proof} of this Lemma will be subdivided into several

3993: steps. Recall that by definition the {\sl nilradical}

3994: $\7g^{\nil}$ of $\7g$ is the maximal nilpotent ideal in $\7g$. It is

3995: well-known that $\7g^{\nil}$ contains the commutator subalgebra

3996: $[\7g,\7g]$, and each element $\xi\in\7g^{\nil}$ is ad-nilpotent in

3997: $\7g.$ Similarly, we denote the (complex) nilradical of $\7l$ by

3998: $\7l^{\nil}.$ We retain the notation from \ruf{FA} and \ruf{CA}.

3999: 

4000: 

4001: \Claim{FU} ${\7g}_o\subset \7g^{\nil}$.

4002: 

4003: \ProofC Since $[\7g,\7g]\subset \7g^{\nil},$ the quotient

4004: $\qu{\7g}{\7g^{\nil}}$ is abelian. Let $\pi:\7g\to

4005: \qu{\7g}{\7g^{\nil}}$ be the projection. Assume that $\7g_o\not\subset

4006: \7g^{\nil}$, i.e., $\pi(\7g_o)\ne 0.$ Select a subalgebra $\7r\subset

4007: \qu{\7g}{\7g^{\nil}}$ (possibly 0) which is a complement of

4008: $\pi(\7g_o)$. But then $\7g':=\pi^{-1}(\7r)$ would violate \26$_{2}$.

4009: Consequently we necessarily have $\pi(\7g_o)=0$, i.e., $\7g_o\subset

4010: \7g^{\nil}$ as claimed. \qd

4011: \Parag{XY} Recall that for the CR-algebra $(\7g,\7q)$ under

4012: consideration there exist flags $\7l_o\subset\7f\subset\7q$ and

4013: $\7l_o\subset\sigma\7f\subset\sigma\7q$ of subalgebras in $\7l$, as

4014: defined in \ruf{FA}. For the subsequent considerations we select the

4015: following elements in $\7l$: $\4y\in\7f\setminus \7l_o$ and $\4x\in

4016: \7q\setminus \7f,$ and write $\overline{\4x}:=\sigma(\4x)$ and

4017: $\overline{\4y}:=\sigma(\4y),$ for short. Then

4018: 

4019: \noindent \hfil$\7f= \CC\4y \oplus \7l_o,\quad \7q=\CC\4x\oplus

4020: \CC\4y\oplus \7l_o, \quad \sigma\7f= \CC\overline{\4y} \oplus

4021: \7l_o\Steil{and} \sigma\7q=\CC\overline{\4x}\oplus

4022: \CC\overline{\4y}\oplus \7l_o\,.$

4023: 

4024: \noindent The inclusion $\7l_o\subset \7l^{\nil}$ is guaranteed by

4025: \ruf{FU}. Hence, $\7l_o$ acts by ad-nilpotent endomorphisms on

4026: $\7l.$ In particular, $[\7l_o,\7f]\subset\7l_o$ and

4027: $[\7l_o,\7q]\subset \7f.$ The condition \23 means

4028: that $[\4x,\overline{\4x}]\not\in \7q +\sigma\7q$, and

4029: \25 is equivalent to $[\overline{\4x},\4y]\not\in

4030: \7f+\sigma\7q.$ \nline Let $\7z\subset \7l^{\nil}$ be the center of

4031: the nilradical (which is nontrivial if $\7l \ne 0$), to which we refer

4032: as to the {\sl nilcenter} of $\7l$. As for every characteristic

4033: ideal, we have $\sigma(\7l^{\nil})=\7l^{\nil}$ and $\sigma(\7z)=\7z.$

4034: 

4035: 

4036: \Claim{FV} {\rm (i)} The nilcenter $\7z$ of $\7l$ is not contained in

4037: $\7q+\sigma\7q.$ \ite=2 \1 $\7l_o=0$ if $\dim \7z\ge 2$.

4038: 

4039: \Proof {\fett ad (i):} Assume that (i) is not true and let

4040: $\4x,\4y,\overline{\4x},\overline{\4y}\in \7l$ be as in \ruf{XY}.  Let

4041: $\zeta:=a\4x+\overline a\overline{\4x}+b\4y+\overline

4042: b\overline{\4y}+\gamma_o,$ $\gamma_o\in\7g_o,$ $a,b\in \CC$, be an

4043: arbitrary element in $\7z^\sigma.$ Then

4044: $[\overline{\4x},\zeta]\in\7z\subset \7q+\sigma\7q$ since $\7z$ is an

4045: ideal in $\7l$.  On the other hand

4046: $$

4047: [\overline{\4x},a\4x+\overline a\overline{\4x}+b\4y+\overline

4048: b\overline{\4y}+\gamma_o

4049: ]=a[\overline{\4x},\4x] + [\overline{\4x},b\4y+\overline

4050: b\overline{\4y}+\gamma_o

4051: ]\equiv a[\overline{\4x},\4x]\equiv0\;\mod\; \7q{+}\sigma\7q\,,

4052: $$

4053: which implies $a=0.$ This shows that $\7z\subset \FFF.$ Given then

4054: $\zeta=b\4y+\overline b\overline{\4y}+\gamma_o\in \7z^\sigma$, the

4055: inclusion $[\overline{\4x},\zeta]\subset\7z$ holds since $\7z$ is an

4056: ideal. On the other hand

4057: $$

4058: [\overline{\4x},b\4y+\overline

4059: b\overline{\4y}+\gamma_o\;

4060: ]=b[\overline{\4x},\4y]+[\overline{\4x},\overline

4061: b\overline{\4y}+\gamma_o]\equiv b[\overline{\4x},\4y]\equiv0\;\mod \;

4062: \7f{+}\sigma\7q\;.

4063: $$ Hence, $b=0$ as the consequence of the above equation and \25. But

4064: this cannot be true since then $\7z\subset \7l_o$ would be a

4065: nontrivial ideal of $\7l$, violating \26$_{1}$. \nline {\fett ad

4066: (ii):} Assume $\dim \7z\ge 2,$ i.e. $\7z\cap \QQQ \ne 0$. Recall that

4067: $\7l_o\subset \7l^{\nil}$ by \ruf{FU} and consequently

4068: $[\4y,\7l_o]\subset\7l_o,$ $[\overline{\4y},\7l_o]\subset\7l_o.$ Let

4069: an arbitrary $\zeta :=a\4x+\overline a\overline{\4x}+b\4y+\overline

4070: b\overline{\4y}+\gamma_o \in \7z^\sigma\cap \QQQ$ be given. Computing

4071: its bracket with $\7l_o$ we obtain:

4072: $$

4073: [\zeta,\7l_o]=a[\4x,\7l_o]+\overline a[\overline{\4x},\7l_o]\equiv0

4074: \;\mod\;\7l_o \;.

4075: $$ Either $a=0$ for all such $\zeta,$ and then $\7z\cap \QQQ=\7z\cap

4076: \FFF$, or $a\ne 0,$ and then

4077: $[\4x,\7l_o]\subset\7l_o\supset[\overline{\4x},\7l_o].$ In the latter

4078: case it follows that $\7l_o$ is an ideal in $\7l$ (due to

4079: \ruf{FC}.{\ninecmss III}, $\4x,\4y,\overline{\4x},\overline{\4y}$ and

4080: $\7l_o$ generate $\7l$ as a Lie algebra), hence, $\7l_o=0$ as claimed.

4081: The other possibility would be $\7z\cap (\7q+\sigma\7q)=\7z\cap

4082: (\7f+\sigma\7f)$ and we show that this cannot happen: Given an

4083: arbitrary $\zeta=b\4y+\overline b\overline{\4y}+\gamma_o\in

4084: \7z^\sigma\cap (\7f+\sigma\7f)$, note that

4085: $$

4086: [\overline{\4x},\zeta]\in \7z\cap \QQQ= \7z\cap\FFF.

4087: $$

4088: More explicitly, $[\overline{\4x}\,,\,b\4y+\overline

4089: b\overline{\4y}+\gamma_o ]=b[\overline{\4x},\4y]+[\overline{\4x}\,,\,

4090: \overline b\overline{\4y}+\gamma_o]\in

4091: b[\overline{\4x},\4y]+\sigma\7q$. But then the above equation together

4092: with \25 would imply $b=0,$ i.e., $\7z\cap \QQQ=\7z\cap\7l_o\ne 0.$

4093: This is absurd, as

4094: $$

4095: [\7q+\sigma\7q,\7z\cap\7l_o]\subset \7z\cap

4096: (\7q{+}\sigma\7q)=\7z\cap\7l_o \ne 0,

4097: $$

4098: i.e., since $\7q+\sigma\7q$ generates $\7l$ as a Lie algebra, $\7z\cap

4099: \7l_o$ would be a nontrivial ideal in $\7l$. \qd

4100: 

4101: \noindent

4102: In remains the case when the nilcenter is 1-dimensional.

4103: 

4104: \Claim{FW}Suppose $\dim\7z=1.$ Then

4105: $\7l=\7z\oplus \QQQ\,$ and $\,\7g_o=\7l_o=0.$

4106: 

4107: \Proof Since $\7z\not\subset \7q+\sigma\7q$ by \ruf{FV}, the sum $\7z

4108: + \QQQ $ is direct. Recall that the recursively defined subspaces

4109: $C_0(\7l^{\nil}):=0,$ $C_k:=C_k(\7l^{ \nil}):=\{u\in \7l^{

4110: \nil}:[u,\7l^{ \nil}]\subset C_{k-1}(\7l^{ \nil})\}$ for every $k>0$

4111: form the ascending central series of $\7l^{\nil}.$ Clearly, $\7z=C_1$

4112: and $\sigma(C_k)=C_k$ for all $k.$ Either $\7z=C_1=C_2=\7l^{\nil}$ and

4113: consequently $\7l_o=0$ (due to \26$_{1}$ $\7l_o$ must be a proper

4114: subalgebra of the 1-dimensional algebra $\7l^{ \nil}$) or $C_1\ne

4115: C_2.$ In the latter case $C_2\cap \QQQ\ne 0.$ Let $\eta=a\4x+\overline

4116: a \overline{\4x}+b\4y+\overline b\overline{\4y}+\gamma_o\in

4117: C_2^\sigma\cap \QQQ$ be arbitrary. Since $[\eta,\7l_o]\subset \7z\cap

4118: \QQQ=0,$ we have

4119: $$

4120: [\eta,\7l_o]\equiv a[\4x,\7l_o]+\overline

4121: a[\overline{\4x},\7l_o]\equiv0\;\mod\;\7l_{o}\,.

4122: $$

4123: If $a\ne 0$ then $[\4x,\7l_o],[\overline{\4x},\7l_o] \subset \7l_o$

4124: and $\7l_o$ is an ideal in $\7l,$ i.e., $\7l_o=0$ by \26$_{1}$. If

4125: $a=0$ for every choice of $\eta\in C_2^\sigma\cap \QQQ$ as above, then

4126: $C_2\cap \QQQ=C_2\cap \FFF.$ This possibility can be ruled out as

4127: follows: For a nonzero $\eta=b\4y+\overline

4128: b\overline{\4y}+\gamma_o\in C_2^\sigma\cap\FFF$ we have $[\eta,\4x]\in

4129: \QQQ \cap C_2=\FFF\cap C_2,$ i.e.,

4130: $$

4131: [b\4y+\overline b\overline{\4y}+\gamma_o,\4x]=\overline

4132: b[\overline{\4y},\4x] \;+

4133: ([b\4y+\gamma_o,\4x])\equiv0\;\mod\;\FFF\,,

4134: $$

4135: which is only possible if $\overline b=0.$ This would imply $\QQQ\cap

4136: C_2=\7l_o\cap C_2.$ On the other hand, the identity $\QQQ\cap\

4137: C_2=\7l_o\cap C_2$ implies that $\7l_o\cap C_2$ is an ideal of $\7l.$

4138: The effectivity of $(\7g,\7q)$ forces then $\7l_o\cap C_2=0,$

4139: contradicting $\QQQ\cap C_2\ne 0.$ This completes the proof of \ruf{FW}

4140: and, together with \ruf{FU}, \ruf{FV} also the proof

4141: of Lemma \ruf{FT}. \qed

4142: 

4143: 

4144: 

4145: \KAP{Fuenfzehn}{The existence of a 3-dimensional abelian ideal in

4146: $\7g$ suffices}

4147: 

4148: The proof of Theorem II has brought us to the point where we may and

4149: do henceforth assume that the Lie algebra $\7g$ in the CR-algebra

4150: $(\7g,\7q)$ satisfying \ruf{FD} is solvable. The

4151: subalgebra $\7q\subset\7l=\7g\oplus i\7g$ then necessarily is of

4152: complex dimension $2$.

4153: 

4154: The main result \ruf{GS} of this section states that the CR-germ

4155: associated to $(\7g,\7q)$ is represented by the tube $F+i\RR^3$ over

4156: an affinely homogeneous surface $F\subset \RR^3$ if $\7g$ is

4157: isomorphic to a semidirect product $\7h\ltimes \7r$ with $\7h$ being a

4158: 2-dimensional Lie subalgebra and $\7r\cong \RR^3$ being an abelian

4159: ideal.  Once this Main Lemma is proved, our proof of the

4160: classification theorem will be complete as soon as we can show that

4161: every 5-dimensional solvable Lie algebra $\7g$ occurring in $(\7g,\7q)$

4162: indeed is isomorphic to a semidirect product as above. This will be

4163: achieved in the final section \ruf{Sechzehn}.  In this section we only

4164: prove the partial result that if $\7q$ is abelian, then also the

4165: commutator $[\7g,\7g]$ is abelian and 3-dimensional. Moreover, there

4166: exists an abelian subalgebra $\7h\subset \7g$ such that

4167: $\7g=\7h\ltimes [\7g,\7g]$.

4168: 

4169: 

4170: 

4171: Since there is no general structure theory for solvable Lie algebras,

4172: we develop ad hoc methods and describe the structure constants in

4173: $\7l=\7g^{\CC}$ with respect to a particularly chosen basis. Every

4174: CR-algebra $(\7g,\7q)$ under consideration gives rise to the

4175: 1-dimensional subalgebras $\7f$ and $\sigma\7f$ of the 2-dimensional

4176: subalgebras $\7q$ and $\sigma\7q,$ respectively. We construct a basis

4177: of $\7l$ which reflects the conditions \21-\25 and investigate the

4178: various possibilities for the values of the corresponding structure

4179: constants. Select a non-zero $\4z\in \7z^{-\sigma}\setminus \QQQ$

4180: (this is possible due to Claim \ruf{FW}) and an $\4x\in \7q\setminus

4181: \7f$ such that for $\overline{\4x}:=\sigma\4x$ the congruence

4182: $[\4x,\overline{\4x}]\equiv \4z\; \mod\;\7q +\sigma\7q$ holds (this is

4183: possible due to \23). By \25 it is further possible to select $\4y\in

4184: \7f\setminus \7l_o$ such that for $\overline{\4y}:=\sigma\4y$ the

4185: structure equations of $\7l$ are of the following form (in particular,

4186: the coefficient in front of $\overline{\4x}$ in the second equation is

4187: $b_2=1$):

4188: $$

4189: \diagram{[\4x,\overline\4x]&\;=\;&\4z+\;& a_1\4x &\;- \;&\overline

4190: a_1\overline \4x&\;+\;& a_2\4y &\;-\;&\overline a_2\overline \4y \cr

4191: [\4x,\overline{\4y}]&\;=\;& & b_1\4x&+& \phantom{\;1\,}\overline\4x

4192: &+&b_3\4y&+&\hfill b_4 \overline{\4y} \cr [\4y,\overline{\4y}]&\;=\;&

4193: & & & & &\hfill c\,\4y & - &\hfill\overline c\,\overline{\4y} \cr

4194: [\4y,\4x]&\;=\;& & d_1\4x & & &+ & d_2\4y & & \cr

4195: [\4z,\eta]&\in&\7z(\7l^{\nil}) \rlap{ \qquad for every $\eta\in \7l\;.$}

4196: }\Leqno{FI}

4197: $$

4198: The brackets $ [\4y,\overline{\4x}]$ and

4199: $[\overline{\4y},\overline{\4x}]$ are completely determined by the

4200: above 5 equations due to the fact that $\sigma:\7l\to \7l$ is an

4201: antilinear Lie algebra automorphism. Of course, not for all values of

4202: the constants the above identities give rise to a Lie {\sl

4203: algebra}. In fact the above structure constants $a_1,...\,,d_2,$ are

4204: subject to further constraints, imposed by (1) the Jacobi identity,

4205: (2) our assumption that $\7l$ is solvable and, (3) our assumption

4206: $\4z\in \7z(\7l^{\nil}).$ \nline Conversely, let a 5-dimensional

4207: solvable complex Lie algebra $\7l=\CC\4z\oplus \CC \4x\oplus\CC

4208: \overline{\4x}\oplus \CC\4y\oplus \CC\overline{\4y}$ be given with

4209: structure equations as in \Ruf{FI} (together with

4210: $[\overline{\4x},\4y]= \4x +\overline b_1\overline\4x + \overline

4211: b_4\4y +\overline b_3\overline{\4y} $ and

4212: $[\overline{\4y},\overline{\4x}]=\overline d_1\overline\4x + \overline

4213: d_2\overline{\4y}\,$) for certain $a_1,..\,,d_2\in\CC$ and

4214: $\4z\in \7z^{-\sigma},$ where $\7z$ is the nilcenter of $\7l$. Define

4215: $$

4216: \7g:=\RR \,i\4z \oplus \RR(\4x+\overline{\4x})\oplus

4217: \RR(i\4x-i\overline{\4x})\oplus \RR(\4y+\overline{\4y})\oplus

4218: \RR(i\4y-i\overline{\4y})\Steil{ and }\7q:=\CC\4x\oplus \CC\4y \;\;.

4219: $$

4220: Then the CR-algebra $(\7g,\7q)$ satisfies the fundamental assumption

4221: \ruf{FD}.

4222: 

4223: 

4224: 

4225: \medskip As already mentioned, besides the geometrically motivated

4226: conditions \ruf{FC}, which already are incorporated in \Ruf{FI},

4227: further conditions affect the particular values of the structure

4228: constants, for example those given by the Jacobi identity. We use

4229: $\Jac{\xi_1}{\xi_2}{\xi_3}:=[[\xi_1,\xi_2],\xi_3]+

4230: [[\xi_2,\xi_3],\xi_1]+[[\xi_3,\xi_1],\xi_2]$ as shorthand. The

4231: identity $\Jac{\4x}{\4y}{\overline{\4y}}=0$ implies

4232: $$

4233: |c|=1, \qquad d_1=\overline c -\overline b_1 \qquad \hbox{and}\quad

4234: \overline c\re2b_1\in\RR\,.\Leqno{FY}

4235: $$

4236: 

4237: 

4238: \Remark{VJ} From $c\ne 0$, i.e., $[\4y,\overline{\4y}]\ne0$,

4239: follows that the solvable subalgebra $\7f+\sigma\7f,$ and in turn

4240: $\7l,$ cannot be nilpotent. \Formend

4241: 

4242: \medskip\noindent Keeping in mind the identities \Ruf{FY} it is

4243: possible to readjust the basis $\4x,...\,,\4z$ of $\7l$ as follows:

4244: Write $c=e^{2}$ for the coefficient $c$ in the third equation of

4245: \Ruf{FI} and replace $\4x$ by $e\4x$ as well as $\4y$ by $c\4y.$ After

4246: this replacement the structure equations \Ruf{FI} keep their form,

4247: only $c$ changes to $c=1$ and $b_{1}$ becomes real.

4248: 

4249: 

4250: \noindent {\sl Notational agreement:} For the rest of this section we

4251: use $\alpha,\beta,\gamma$ and $\delta$ to denote real numbers, while

4252: $a_j, b_j,c_j$ and $d_j$ stand for complex numbers. In particular, we

4253: write $\beta_1:=b_1$ to underline that the structure

4254: constant $b_1$ is real.

4255: 

4256: \noindent The Jacobi identity $\Jac{\4x}{\4y}{\overline{\4y}{}}=0$

4257: implies further relations between the coefficients in the structure

4258: equations \Ruf{FI}:\vskip-25pt

4259: $$

4260: \diagram{c=1,\qquad\qquad \beta_1:=b_1\in \RR \qquad\cr

4261: d_1=1-\beta_1\;,\qquad \overline b_3-d_2=b_4(\beta_1-1)\;,\qquad

4262: -\beta_1(b_3+d_2)=b_4-\overline b_4 \;.}\Leqno{FZ}

4263: $$ \Joker{GQ}{Perfect basis.} Summarizing, there exists a basis

4264: $\4z,\4x,\overline{\4x},\4y,\overline{\4y}$

4265: ($\overline{\4x}:=\sigma(\4x), \;\overline{\4y}:=\sigma(\4y)$) of

4266: $\7l=\7g^{\CC}$ with $\4z\in \7z^{-\sigma}\setminus

4267: (\7q{\oplus}\sigma\7q)$ and $\7q=\CC\4x\oplus \CC\4y$, $\7f=\CC \4y$

4268: such that the corresponding structure constants in \Ruf{FI} satisfy

4269: \Ruf{FZ}. We call each such basis {\sl perfect}.

4270: 

4271: \def\FFF{(\7f\mkern-1mu\oplus\mkern-1mu\sigma\7f)}

4272: \def\QQQ{(\7q\mkern-1mu\oplus\mkern-1mu\sigma\7q)}

4273: 

4274: There are still more constraints for the structure

4275: constants given by the Jacobi identity for further triples of elements

4276: and also by the fact that $\7z$ is in the center of

4277: $\7l^{\nil}$. Later on, we characterize these additional conditions

4278: more explicitly. For now, we elaborate the particular structure of the

4279: nilcenter:

4280: 

4281: \Lemma{GR} The nilcenter $\7z\subset \7g$ has dimension 1 or 3.

4282: 

4283: \Proof We closely analyze the conditions in \ruf{FC} in order to get

4284: the dimension estimates. It is clear that $1\le \dim \7z\le 4$, see

4285: Remark \ruf{VJ}. Assume $\dim \7z=4$. Then $\dim\7z\cap \QQQ=3$

4286: follows by Lemma \ruf{FV}. But since $\7f+\sigma\7f$ is not abelian,

4287: i.e., $\dim \7z\cap \FFF=1,$ there exist elements in $\7z$ of the form

4288: $\4x+\eta_1,\overline{\4x}+\eta_2$ with $\eta_j\in \7f+\sigma\7f.$

4289: This leads to a contradiction: Indeed, on the one side

4290: $[\4x+\eta_1,\overline{\4x}+\eta_2]=0$, and on the other side \25

4291: implies $[\4x+\eta_1,\overline{\4x}+\eta_2]

4292: \equiv[\4x,\overline{\4x}]\not\equiv0\;\mod\;{\7q +\sigma\7q}$. Hence,

4293: we have proved that $1\le \dim \7z\le 3.$

4294: 

4295: \noindent

4296: The main difficulty is to rule out the possibility $\dim\7z=2$.

4297: Select a perfect basis in $\7l$ as described in

4298: \ruf{GQ} keeping in mind \Ruf{FI} and \Ruf{FZ}. We have to deal with 2

4299: subcases.

4300: 

4301: \Claim{VF} If $[\4x,\4y]\ne 0$, i.e., $\7q$ is not abelian, then

4302: $\dim \7z=1$.

4303: 

4304: \ProofC We first show that $\7z\cap \QQQ=\7z\cap \FFF$. Select

4305: an arbitrary

4306: $\4z':=\lambda\4x+\overline\lambda\overline{\4x}+\mu\4y+\overline\mu

4307: \overline{\4y} \in \7z^\sigma\cap \QQQ.$ We have to investigate the 2

4308: possibilities `$\beta_1\ne 1$' and `$\beta_1=1$': In the first

4309: case we get

4310: 

4311: \noindent \hfil

4312: $[\4z',(1-\beta_1)\4x+ d_2\4y]\equiv\overline\lambda(1-\beta_1)

4313: [\overline{\4x},\4x]\equiv0\;\mod\;\7q +\sigma\7q$

4314: 

4315: \noindent since $[\4y,\4x]\in \7l^{\nil}$. Therefore$\lambda=0$ by the

4316: condition \23, i.e., $\4z'\in \7f +\sigma\7f.$ If $\beta_1=1,$ i.e,

4317: $d_1=0$, then by our assumption $d_2\ne 0,$ and in turn $\4y\in

4318: \7l^{\nil}.$ Hence,

4319: $[\4z',\4y]\equiv\overline\lambda[\overline{\4x},\4y]\equiv0\;\mod\;

4320: \7f +\sigma\7f,$ which, together with \25 also forces

4321: $\lambda=0$. This proves $\7z\cap \QQQ=\7z\cap \FFF$. \nline We claim

4322: that this identity can only h old if both sides vanish, i.e., $\dim

4323: \7z=1$ by Lemma \ruf{FV}.i: Assuming to the contrary that $\7z\cap

4324: (\7q\oplus\sigma\7q)\ne 0$, then on the one hand there exists

4325: $\4z'=\mu\4y+\overline\mu\overline{\4y} \in \7z\cap \FFF$ with $\mu\ne

4326: 0.$ On the other hand,

4327: $[\4x,\4z']=[\4x,\mu\4y+\overline\mu\overline{\4y}]\in

4328: \7z\cap\QQQ=\7z\cap \FFF$ which in view of \25 is only possible if

4329: $\mu=0$. This shows $\dim\7z=1$.

4330: 

4331: \noindent It remains to rule out the second subcase:

4332: 

4333: \Claim{VG} If $[\4y,\4x]=0$ , i.e., $\7q$ is abelian, then the

4334: commutator $[\7l,\7l]$ is a 3-dimensional abelian ideal. Further, if

4335: $\dim \7z\ge 2$ then $[\7l,\7l]=\7l^{\nil}$ and consequently

4336: $\7z=[\7l,\7l]$ is 3-dimensional.

4337: 

4338: 

4339: \ProofL This is the most involved case. The assumption $[\4y,\4x]=0$,

4340: i.e, $d_1=d_2=0$ implies $\beta_1=1,$ $b_3=0$ and $b_4:=\beta_4\in

4341: \RR$, see the table below:

4342: $$

4343: \diagram{[\4x,\overline\4x]&\;=\;&\4z \;+\;\;\;& a_1\4x &\;- \;

4344: &\overline a_1\overline \4x&\;+\;& a_2\4y &\;-\;&\overline

4345: a_2\overline \4y\cr [\4x,\overline{\4y}]&\;=\;& &\hfill \4x&+&

4346: \hfill\overline\4x && &+ &\beta_4\overline{\4y} \cr

4347: [\4y,\overline{\4y}]&\;=\;& & & & & &\hfill\4y & -

4348: &\hfill\overline{\4y} \cr [\4y,\4x]&\;=\;&0\hfill \cr [\4z,\eta]&\;=

4349: \;&\rlap{$ c_\eta \4z+ \4q_\eta$\quad $\eta\in \7l$, $\;\4q_\eta\in

4350: \7z\cap \QQQ\;.$ }\hfill }\Leqno{VA}

4351: $$

4352: We need to analyze the relations between the structure constants in

4353: more detail. Let\nline

4354: \centerline{$\4q_{\4x}=z_1\4x+z_2\overline{\4x}+z_3\4y+z_4\overline{\4y}

4355: \Steil{and}\4q_{\4y}=w_1\4x+w_2\overline{\4x}+w_3\4y+w_4\overline{\4y}$.}

4356: \nline The fact that the $\4q_\eta$ commute with all elements of

4357: $[\7l,\7l]\subset \7l^{\nil}$ (and in particular with

4358: $\4y-\overline{\4y}$) implies $z_1=z_2$, $w_1=w_2$ and

4359: $z_4=z_1\beta_4-z_3$, $ w_4=w_1\beta_4-w_3$.  The Jacobi identity

4360: $\Jac{\4x}{\overline{\4x}}{\4y}=0$ yields

4361: $$

4362: w_j=0,\; c_{\4y}=1,

4363: \Steil{i.e.,}[\4z,\4y]=[\4z,\overline{\4y}]=\4z,\qquad

4364: a_1=i\alpha_1,\qquad a_2=\alpha_2+{i\over 2}\alpha_1\beta_4\;,

4365: \Leqno{}

4366: $$

4367: for some $\alpha_1,\alpha_2\in \RR$, and

4368: $\Jac{\overline{\4x}}{\4y}{\4z}=0$ implies

4369: $$

4370: \Im z_1=0, \qquad \Re z_3=\qu{z_1\beta_4}2 \Steil{ and }\Re

4371: c_{\4x}=-\qu{\beta_4}2\;.\Leqno{}

4372: $$

4373: Summarizing,

4374: $$

4375: [\7l,\7l]=\CC

4376: \4z\;\oplus\;\CC(\4x+\overline{\4x}+\hbox{\Klein$\displaystyle

4377: {\beta_4\over 2}$} \4y+\hbox{\Klein$\displaystyle {\beta_4\over

4378: 2}$}\overline{\4y})\;\oplus\; \CC(\4y-\overline{\4y})\;, \Leqno{VK}

4379: $$

4380: and, using the above relations between the structure constants, a

4381: simple check shows that this ideal is abelian. Moreover, the subset

4382: $$

4383: \7h:=\RR \; i\big(\4x-\overline{\4x}+\hbox{\Klein$\displaystyle

4384: {\beta_4\over 2}$} \4y-\hbox{\Klein$\!\displaystyle {\beta_4\over

4385: 2}$}\overline{\4y}\big) \;\oplus \;\RR (\4y+\overline{\4y}) \Leqno{VL}

4386: $$

4387: is an abelian subalgebra of $\7g$

4388: (and $\7g=\7h\oplus [\7g,\7g]$; in fact $\7g=\7h\ltimes [\7g,\7g]$).

4389: \nline

4390: It should be noted that the nilcenter $\7z$ may be 1-dimensional,

4391: and then $\7l^{\nil}$ properly contains

4392: $[\7l,\7l]$.

4393: 

4394: \noindent We next show that the situation `$\dim \7z=2$' does not

4395: occur: Assume to the contrary that $\dim \7z= 2$. Then $\7z\cap \QQQ$

4396: is nonzero. Let $\4z':=\lambda\4x + \overline \lambda\overline{\4x} +

4397: \mu\4y +\overline \mu\overline{\4y}\in \7z^\sigma\cap \QQQ$ be

4398: arbitrary. Either, for all such $\4z'$ the coefficient $\lambda$ is 0,

4399: i.e., $\7z\cap \QQQ=\7z\cap \FFF$, and then this case can be ruled out

4400: by a similar argument as in the proof of the above claim. Or, there

4401: exists $\4z'$ with $\lambda\ne 0.$ In such a situation the identity

4402: $[\4z',\4y-\overline{\4y}]=0$ implies $\lambda\in \RR^*$ and without

4403: loss of generality we may suppose $\4z'=\4x + \overline{\4x} + z_3\4y

4404: +\overline z_3\overline{\4y}$ for some $z_3\in \CC$. The condition

4405: $[\4z,\4z']=0$ gives $\Re z_3=\qu{\beta_4}2$. But then

4406: $\4z'=(\4x+\overline{\4x}+\qu{\beta_4}2\cd

4407: (\4y+\overline{\4y}))+i\,\Im z_3 \cd (\4y-\overline{\4y})\in

4408: [\7l,\7l]$, compare \Ruf{VK}. We claim, that in the situation under

4409: consideration, $\7l^{\nil}=[\7l,\7l]$. To prove this, we simply

4410: compute the ad-action of $\4y+\overline{\4y}$ and $\4x-\overline{\4x}$

4411: on $[\7l,\7l]$. Since $[\4x-\overline{\4x},\4z' ]\in \CC \4z\oplus \CC

4412: \4z'$, the relation $\beta_4^2+4(\Im z_3)^2=4\alpha_1 \Im

4413: z_3+4\alpha_2$ must also be fulfilled. Once again, a simple

4414: computation yields

4415: $$

4416: \ad(\4y+\overline{\4y})\rest {[\7l,\7l]}=-2\cd\id, \qquad

4417: [\4x-\overline{\4x},\4z']=2\4z+2i(\alpha_1-\Im\,z_3)\4z'\;.

4418: $$

4419: The above identities show that for every $\4v:=u_1\cd

4420: (\4x-\overline{\4x})+u_2(\4y+\overline{\4y})+u_3\eta,$ $\,u_j\in \CC$,

4421: $\eta\in [\7l,\7l]$, the condition $[\4v,\4z']=0$ implies

4422: $u_1=u_2=0$, i.e., the centralizer $C_{\7l}(\4z')$ coincides with

4423: $[\7l,\7l]$. This proves $[\7l,\7l]=\7l^{\nil}=C_{\7l}(\4z')$. But

4424: this is absurd, since then the nilcenter $\7z$ would coincide with

4425: the 3-dimensional abelian ideal $[\7l,\7l]$, contrary to our

4426: assumption `$\dim \7z=2$'.

4427: 

4428: \noindent Finally, we need to investigate the case $\dim \7z=3.$ We

4429: claim that $\7z=[\7l,\7l]=\7l^{\nil}$: To see this, it enough to show

4430: that $\7l^{\nil}$ is 3-dimensional, as, due to \Ruf{VK}, $[\7l,\7l]$

4431: is 3-dimensional, too. As already mentioned (see the sentence

4432: following \Ruf{FY}) $\7l^{\nil}$ can be at most 4-dimensional. But the

4433: 4-dimensional case can be excluded, otherwise $\7l^{\nil}=\7z\oplus

4434: \CC \4n$ for $\4n\in \7l^{\nil}\setminus [\7l,\7l]$ which would imply

4435: that $\7l^{\nil}$ is abelian. Hence the nilradical is

4436: 3-dimensional. This proves \ruf{VG} and Lemma \ruf{GR}.\qed

4437: 

4438: 

4439: \medskip The next statement is one of the key points in our

4440: classification of 5-dimensional 2-nondegenerate homogeneous

4441: CR-germs. Before stating it we first fix some notation. Given a vector

4442: space $V$, write $\aff(V)$ for the Lie algebra consisting of affine

4443: maps of $V$. This Lie algebra (as well as the corresponding Lie group

4444: $\Aff(V)$) has the natural semidirect product structure:

4445: $\aff(V)=V\rtimes \gl(V)$ (with $\gl(V)=\{X\in \aff(V):X(0)=0\}$). Let

4446: $\pi:\aff(V)\to \gl(V)$ be the projection homomorphism. We use similar

4447: notation on the Lie group level and write, for instance,

4448: $\pi:\Aff(V)=V\rtimes \GL(V)\to \GL(V)$ for the corresponding group

4449: homomorphism. Sometimes we simply write $\psi^{\lin}:=\pi(\psi)$ for

4450: the linear part of an element in $\aff(V)$ or $\Aff(V).$

4451: 

4452: \ifarx\eject\fi

4453: 

4454: \MLemma{GS} Let $(\7g,\7q)$ be a CR-algebra satisfying the fundamental

4455: assumption \ruf{FD} and let $\7g$ be solvable and of dimension 5.

4456: Suppose that there exists a 3-dimensional abelian ideal $\7v\subset

4457: \7g$ and a 2-dimensional subalgebra $\7h\subset \7g$ with $\7h\cap

4458: \7v=\{0\}$. Then the associated CR-germ $(M,o)$ is locally

4459: CR-equivalent to a tube $F\times i\RR^3 \subset \CC^3$, where

4460: $F\subset \RR^3$ is an affinely homogeneous surface.\Formend

4461: 

4462: \noindent The {\sl proof} is divided in several steps which give more

4463: precise (but also more technical) information concerning the structure

4464: of the Lie groups corresponding to $\7g$, $\7l$ and $\7q$ and a

4465: realization of the CR-germ $(M,o)$:

4466: 

4467: 

4468: \Claim{} The adjoint representation $\ad:\7h\to \gl(\7v)$ is

4469: faithful. Consequently, identifying $\7h$ with the subalgebra

4470: $\ad(\7h)\subset\gl(\7v)$, the Lie algebras $\7g$ and $\7l$ can be

4471: realized as Lie subalgebras of affine transformations:

4472: $\;\;\7g=\7v\rtimes \7h=\7v\rtimes \ad(\7h)\subset \aff(\7v)\cong

4473: \aff(\RR^3)\;$ {and} $\;\7l=\7v^{\CC}\!\rtimes \7h^{\CC}\subset

4474: \aff(\7v^{\CC})\cong \aff(\CC^3).$

4475: 

4476: \ProofC Let $\7n\subset \7h$ be the kernel of the adjoint

4477: representation $\ad:\7h\to \gl(\7v).$ The case $\dim \7n=1$ can be

4478: excluded, otherwise $\7n\oplus\7v=\7g^{\nil}=\7z$ would be

4479: 4-dimensional, contradicting Lemma \ruf{GR}. The case $\dim

4480: \7n=2$ can be also excluded: Otherwise $\7g=\7v\times \7h$ would be

4481: abelian or contain a 4-dimensional abelian nilradical which in both

4482: cases would contradict \23 -- \25.\qd

4483: 

4484: 

4485: Write $V\cong \RR^3$ for a vector group with Lie algebra $\7v$ and

4486: $E:=V^{\CC}$ for its complexification. Let $H_{\GL}\subset \GL(V)$ and

4487: $H_{\GL}^{\CC}\subset \GL(E)$ be the Lie subgroups, corresponding to

4488: the Lie algebras $\ad(\7h)$ and $\ad(\7h^{\CC})$, respectively. Since

4489: $\GL(V)\cong \GL(3,\RR)$ contains no compact torus of dimension

4490: $\ge2$, each subgroup, in particular $H_{\GL}$, is closed. This is in

4491: general not true for the complex subgroup $H_{\GL}^{\CC}$. Let

4492: $H^{\CC}$ be the simply connected Lie group with Lie algebra

4493: $\7h^{\CC}$, $\pr:H^{\CC}\to H_{\GL}^{\CC}\subset\GL(\7v^{\CC})$ the

4494: homomorphism induced by $\ad:\7h^{\CC}\to

4495: \ad(\7h^{\CC})\subset\gl(\7v^{\CC})$ and $L:=V^{\CC}\rtimes

4496: H^{\CC}$. For simplicity, for each $h\in H^{\CC}$ we also write

4497: $h^{\lin}\subset \GL(E)$ instead of $\pi(h)$. Let $G=V\rtimes H\subset

4498: L$ be the real form. Since every 2-dimensional Lie algebra is

4499: solvable, we deduce that also $\7l=\7v^{\CC}\!\rtimes \7h^{\CC}$ (as

4500: well as $\7g$, $L$ and $G$) is solvable.

4501: 

4502: 

4503: \Claim{} Let $Q\subset L$ be the subgroup corresponding to the Lie

4504: subalgebra $\7q\subset \7l$. Then $Q$ is closed and $Q\cap

4505: V^{\CC}=\{e\}$. Hence $L=V^{\CC}\!\rtimes Q$ is a semidirect product.

4506: 

4507: \ProofC Let $\pi:\7l\to \7h^{\CC}$ be the projection homomorphism. Our

4508: first observation is that $\pi(\7q)=\7h^{\CC}$: The case

4509: `$\pi(\7q)=0$' can clearly be excluded as in such a situation

4510: $\7q\subset \7v^{\CC}$, and in turn $\7q+\sigma\7q\subset \7v^{\CC}$,

4511: which is absurd. \nline The possibility $\dim \pi(\7q)=1$ can be

4512: ruled out as follows: We may assume that $\pi(\7q\oplus

4513: \sigma\7q)=\7h^{\CC}$ (otherwise $\7q\oplus \sigma\7q$ would be a

4514: subalgebra). Either $\pi(\7f)=0,$ i.e., $\7q\cap \7v^{\CC}=\7f$, and

4515: in turn $(\7q\oplus\sigma\7q)\cap \7v^{\CC}=\7f\oplus \sigma\7f$: This

4516: leads to a contradiction since then $[\7F,\7H]\subset \7H\cap

4517: \7v=\7F$, violating \25. Or, $\pi(\7f)=\pi(\7q)$. But then

4518: $\7q=\7f\oplus \CC \4x$ with a nonzero $\4x\in \7q\cap \7v^{\CC}$, and

4519: in turn $[\7q,\sigma\7q]\subset \7q\oplus\sigma\7q$, contradicting

4520: \23. Summarizing, $\pi(\7q)=\7h^{\CC}$. \nline On the group level,

4521: since $L$ is simply connected and solvable, every connected subgroup

4522: is closed. The restriction of $\pi$ to $Q$ induces a surjective

4523: homomorphism $Q\to H^{\CC}$. Since both groups are 2-dimensional,

4524: this homomorphism is a covering. Our assumption that $H^{\CC}$ is

4525: simply connected finally implies that

4526: $\pi\lower1pt\hbox{\mathsurround=0pt$|$}_{Q}:Q\to H^{\CC}$ is an

4527: isomorphism. In particular $Q\cap V^{\CC}=Q\cap \ker\pi=\{e\}$.\qd

4528: 

4529: 

4530: \Claim{VM}With respect to the identification $Z:=L/Q=V^{\CC}\cong

4531: \CC^3$ the real form $G$ acts on $L/Q$ by affine transformations and

4532: $V\subset G$ by translations.

4533: 

4534: \ProofC The existence of the decomposition $L=V^{\CC}\!\rtimes Q$

4535: implies that there are well-defined functions $v\colon L\to V^{\CC}$,

4536: $q\colon L\to Q$ such that $\ell=v(\ell)\cd q(\ell)$ for every

4537: $\ell\in L.$ Let $g=w\cd h\in V\rtimes H=G$ (with $w\in V, h\in H$) be

4538: arbitrary. Then, for any $z\in V^{\CC}$ we have

4539: $$

4540: g\cd zQ=w\cd h \cd zQ =w \cd v(h)\cd q(h)\cd zQ = w \cd v(h)\cd

4541: (q(h)\cd z \cd q(h)^{-1})Q

4542: $$

4543: and $ q(h)\cd z \cd q(h)^{-1}= v(h)^{-1}h \cd z \cd h^{-1}v(h)=

4544: h^{\lin}(z)$. Hence, with respect to the identification $V^{\CC}=L/Q$

4545: (induced by the inclusion $V^{\CC}\into L$ such that $0$ corresponds

4546: to the point $eQ\in L/Q$), the action of $G$ can written as follows:

4547: $$

4548: g\cdot z= h^{\lin}(z) + v(h) +w \qquad g=w\cd h\in L,\;\; z\in V^{\CC}\,.

4549: \Leqno{VN}

4550: $$

4551: In particular, the subgroup $V\subset G$ acts by translations

4552: $z\mapsto z+w$. \qd

4553: 

4554: \noindent

4555: Consequently $M:= G\cdot 0=VH\cdot 0= V +

4556: F\subset V\oplus iV$ with $F:=M\cap iV$.

4557: It should be noted, however, that in general $F :=(G\cdot

4558: 0)\cap iV \ne H\cdot 0$. Nevertheless, as we shortly will see, $F$ is

4559: affinely homogeneous under a slightly different subgroup of

4560: $\Aff(iV)$. Clearly, multiplying a tube manifold $F+ iV\subset

4561: V\oplus iV$, $F\subset V$, by the imaginary unit $i$ we get the

4562: CR-equivalent realization $V+ iF=V+ F'$ with $F'=iF\subset

4563: iV$. The latter form of a tube manifold is more suitable in our

4564: particular setup, and we keep this notation until the end of the

4565: proof of the Main Lemma.

4566: 

4567: \Claim{} Retaining the previous notation, there exists a subgroup

4568: $B\subset iV\!\rtimes \GL(iV)=\Aff(iV)$ such that $F:= (G\cdot 0)\cap

4569: iV =B\cdot 0$.

4570: 

4571: \ProofC Let $\pr^{i}:V\oplus iV\to iV$ be the linear projection. A

4572: glance at \Ruf{VN} shows that $F=\pr^{i}\{v(h): h\in H\}.$ In order to

4573: determine $v(h)$ more explicitly we need to analyze the position of

4574: $Q$ in $V^{\CC}\!\rtimes H^{\CC}$ in greater detail: Since $\7h$ is

4575: 2-dimensional, there exists a basis $\4s,\4n\in \7h$ such that

4576: $[\4s,\4n]=\epsilon \4n$ with $\epsilon\in \{0,1\}$. Recall that the

4577: projection map $\pi:\7q\to \7h^{\CC}$ is an isomorphism. Consequently

4578: there exist $\4w_{\4s},\4w_{\4n}\in \7v^{\CC}=V^{\CC}$ such that the

4579: elements $\4w_{\4s}+\4s$ and $\4w_{\4n}+\4n$ in

4580: $\7l=V^{\CC}\oplus\7h^{\CC}$ generate $\7q$. Since then $\sigma\7q=\CC

4581: (\overline {\4w}_{\4s}+\4s)\oplus \CC (\overline{\4w}_{\4n}+\4n)$, we

4582: must have $\4w_{\4s}\ne \overline {\4w}_{\4s}$ and $\4w_{\4n}\ne

4583: \overline {\4w}_{\4n}$ (otherwise $\7q\cap\sigma\7q \ne 0$). Let

4584: $\exp:\7l\to L$ and $\Exp:\ad(\7h^{\CC})\to \GL(V^{\CC})$ be the

4585: exponential maps (i.e., $\Exp({\ad \4v})=\pi(\exp \4v)$ with $\pi$ as

4586: in the paragraph preceeding \ruf{GS}). Furthermore, let $\Psi$ be the

4587: entire function defined by $$\Psi(z)={e^{z}-1\over

4588: z}=\sum^{\infty}_{k=0}{z^{k}\over(k+1)!}z^{k}\,.$$

4589: Then for $\4S^t:=\Psi(t\ad(\4s))$ and $\4N^u:=\Psi(u\ad(\4n))$

4590: a simple computation shows:

4591: $$

4592: \eqalign{ H&=\{\exp(t\4s)\cd \exp(u\4n):t,u\in \RR\} \cr Q&=\{\exp

4593: t(\4w_{\4s}+\4s)\cdot\exp u(\4w_{\4n}+\4n): t,u\in \CC \}\cr &=\{

4594: \4S^t(t\4w_{\4s})\cd\exp(t\4s)\cdot

4595: \4N^u(u\4w_{\4n})\cd\exp(u\4n):t,u\in \CC \} \cr

4596: &=\{\big(\4S^t(t\4w_{\4s})\cd \Exp(t\ad(\4s))\,(\4N^u(u\4w_{\4n}))\big)

4597: \cdot \exp (t\4s)\cd\exp(u\4n):t,u\in \CC \} \;\;\subset\;\; V^{\CC}

4598: \cdot H^{\CC}=L\,. } \Leqno{VO}

4599: $$

4600: The explicit form of $v(h)$ (compare the proof of Claim \ruf{VM}) can

4601: be read off the last line in \Ruf{VO}:

4602: $$

4603: v(h)=v(\exp (t\4s)\exp(u\4n))

4604: =\big(\4S^t(t\4w_{\4s})\big)^{-1}\cdot\big(\Exp(t\ad(\4s))\,

4605: (\4N^u(u\4w_{\4n}))\big)^{-1}\,.

4606: $$

4607: Since $\ad(\4s), \4N^u$ and $\4S^t$ are real operators, it follows for

4608: $h=\exp(t\4s)\cdot\exp( u\4n)$ as before:

4609: $$

4610: \eqalign{ \pr^{i}(v(h))\;&=\;\big(\4S^t(t\4w_{\4s})\big)^{-1}\!\cdot

4611: \big(\Exp(t\ad(\4s))\, (\4N^u(u\4w_{\4n}))\big)^{-1}\cr&=\;\; \exp

4612: t(-\4w^{i}_{\4s}+\4s)\cd \exp u(-\4w^{i}_{\4n}+\4n)\cdot 0 \;\subset

4613: iV\;.\cr }\Leqno{VP}

4614: $$

4615: (Using additive notation, $\pr^{i}(v(h))=-\4S^t(t\4w^{i}_{\4s})-

4616: \Exp(t\ad(\4s))\,(\4N^u(u\4w^{i}_{\4n}))\;\subset iV\,$). Define

4617: 

4618: \hfil $\7b:=\RR(-\4w^{i}_{\4s}+\4s)\oplus

4619: \RR(-\4w^{i}_{\4n}+\4n)\;\;\subset \; \7l=V^{\CC}\!\rtimes

4620: \7h^{\CC}$\hfil

4621: 

4622: \noindent and check that this is a Lie algebra. Then $B:=\exp

4623: \RR(-\4w^{i}_{\4s}+\4s)\cd \exp \RR(-\4w^{i}_{\4n}+\4n)$ is the

4624: subgroup of $L$ with Lie algebra $\7b$, and \Ruf{VP} shows that

4625: $F=\pr^{i}\{v(h):h\in H\}=B\cdot 0$. This finishes the proof of the

4626: claim and of the Main Lemma.\qed

4627: 

4628: \KAP{Sechzehn}{The final steps}

4629: 

4630: 

4631: Our final step toward the complete classification of all

4632: 5-dimensional 2-nondegenerate and homogeneous CR-germs is to deduce

4633: that each 5-dimensional solvable Lie algebra $\7g$ which occurs in a

4634: CR-algebra $(\7g,\7q)$ subject to \ruf{FD}, also satisfies the

4635: assumptions of the preceeding Main Lemma \ruf{GS}:

4636: 

4637: 

4638: 

4639: \Lemma{VQ} Let $(\7g,\7q)$ be a CR-algebra satisfying \ruf{FD} and

4640: suppose that $\7g$ is solvable and of dimension 5.  Then there exists a

4641: semidirect product decomposition $\7g=\7v\rtimes \7h$ with a

4642: 3-dimensional abelian ideal $\7v\subset \7g$ and a 2-dimensional

4643: subalgebra $\7h$.  \Formend

4644: 

4645: \Proof We have already observed in Lemma \ruf{GR} that the nilcenter

4646: $\7z$ has dimension 1 or 3. If $\dim \7z=3$ then, due to \Ruf{VF}, we

4647: can apply \ruf{VG}: Consequently, we can choose $\7v=[\7g,\7g]$

4648: (compare \Ruf{VK}) and $\7h$ as defined in \Ruf{VL}.

4649: 

4650: \noindent The situation $\dim \7z=1$ requires some more elaborated

4651: work. We classify all CR-algebras $(\7g,\7q)$ under consideration in

4652: terms of the corresponding structure equations with respect to some

4653: perfect basis $\4x,...\,,\4z$ of $\7l$ (see \ruf{GQ}): Given

4654: $(\7g, \7q)$, let the corresponding structure equations be as in

4655: \Ruf{FI}, taking into account \Ruf{FZ}. To handle the various sets of

4656: relations between the structure constants, we divide the class of

4657: CR-algebras under consideration into the subclasses A, B and C, see

4658: below.

4659: 

4660: \noindent {\fett Case A: $\beta_1\ne \pm 1$}. In this situation it is

4661: possible to assume $a_1=0$ (simply replace $\4x$ by

4662: $\4x+\lambda\4y$ with $\lambda=u+iv$ defined by $u:=\qu{\Re

4663: a_1}{(1-\beta_1)}$ and $v:=\qu{\Im a_1}{(1+\beta_1)}$). The structure

4664: equations then read

4665: $$

4666: \diagram{[\4x,\overline\4x]&\;=\;&\4z+\;& & & &\;+\;& a_2\4y &\;-\;

4667: &\overline a_2\overline \4y \cr [\4x,\overline{\4y}]&\;=\;& &\kern1em

4668: \beta_1\4x&+& \;\overline\4x &+&b_3\4y&+&b_4\overline{\4y} \cr

4669: [\4y,\overline{\4y}]&\;=\;& & & & & & \;\;\4y & - & \;\;\overline{\4y}

4670: \cr [\4y,\4x]&\;=\;& & \kern-10pt(1-\beta_1)\4x & & &+ &d_2\4y & &

4671: \rlap{\qquad\qquad $d_2=\overline b_3+(1-\beta_1)b_4$} \cr

4672: [\4z,\eta]&\;= \;&c_\eta\4z \rlap{ \qquad for every $\eta\in \7l\;$

4673: ,}}\qquad\qquad\qquad\qquad \Leqno{GT}

4674: $$

4675: and we now work out further constraints imposed on the constants: An

4676: explicit evaluation of the Jacobi identity

4677: $\Jac{\4x}{\overline{\4x}}{\4y}=0$ yields $c_\4y=c_{\overline{\4y}}

4678: =2\beta_1-1.$ Furthermore, we obtain the equations $\overline

4679: b_3(1+\beta_1)=b_4(\beta_1^2-\beta^{}_2)$ and $\overline

4680: b_3(1+\beta_1)=(b_4-\overline b_4)(\beta_1-1)$ which imply

4681: $b_4(1-\beta_1)=\overline b_4.$ \nline In order to investigate most

4682: conveniently further relations between the structure constants, we

4683: deal separately with the following 3 subcases:

4684: 

4685: \medskip\centerline

4686: { {\bf AI:} $b_4\in i\RR^{*}$ and $\beta_1=2,$\hfil

4687: {\bf AII:} $b_4\in \RR^{*}$ and $\beta_1=0,$\hfil

4688: {\bf AIII:} $b_4=0.$}

4689: 

4690: \noindent {\bf Ad AI:} Put $\beta_{4}:=-ib_{4}$. In this particular

4691: situation the identity $\Jac{\4x}{\overline{\4x}}{\4y}=0$ implies

4692: $b_3=-{2\over 3} i\beta_4,$ $a_2=-{2\over 9}\beta_4^2$ and

4693: $\Jac{\4x}[\4y]{\4z}=0$ implies $c_\4x=-i\beta_4.$ There are no more

4694: conditions imposed by the Jacobi identity and the structure

4695: equations of $\7l$ are now

4696: $$

4697: \diagram{[\4x,\overline\4x]&\;=\;&\4z+\;& & & &\;-\;& 2 \gamma^2\4y

4698: &\;+\;&2 \gamma^2\overline{\4y} \cr [\4x,\overline{\4y}]&\;=\;&

4699: &\kern1em 2\4x&+& \; \overline\4x &-&2i \gamma\4y&+&3i

4700: \gamma\overline{\4y} \cr [\4y, \overline{\4y}]&\;=\;& & & & & &

4701: \;\;\4y & - & \;\;\overline{\4y} \cr [\4y,\4x]&\;=\;& & \kern-3pt-\4x

4702: & & &- &i \gamma \4y & & \cr [\4z,\4x]&\;= \;&\rlap{$-3i \gamma\,\4z$

4703: \qquad\qquad $[\4z,\4y]=-3\4z\;,$}\hfill }\qquad\qquad\qquad\qquad

4704: \Leqno{GU}

4705: $$ where $ \gamma:=\beta_4/3\in \RR^{*}$. Keeping in mind Lemma

4706: \ruf{GS}, the structure of $\7l$, $\7g$ and the position of $\7q$ is

4707: determined by \Ruf{GU}. This can be seen more clearly by decomposing

4708: $\7g$ and $\7l$ into the eigenspaces of $\ad(\4s),$ where

4709: $\4s:=-(\4y+\overline{\4y}/2:$ Define the following elements

4710: $\4n,\4v_1,\4v_2,\4v_3$ from $\7g:$

4711: $$

4712: \diagram{\4n&\;:= \;& i\4x-i\overline{\4x}-\;\gamma\4y-\;

4713: \gamma\overline{\4y} \cr

4714: \4v_1&:=& \qquad\qquad \hbox{\klein$\displaystyle{i\over 2}$}

4715: \4y-\hbox{\klein$\displaystyle{i\over 2}$}\overline{\4y} \cr

4716: \4v_2&:=& \;\4x+\;\overline{\4x}-2i\gamma\4y+2i \gamma\overline{\4y}\cr

4717: \4v_3&:=& 2i\4z\;.\vphantom{{}^{\textstyle|}}\hfill}

4718: $$

4719: It is clear that $\4s,\4n,\4v_1,\4v_2,\4v_3$ form a basis of $\7g.$

4720: The bracket relations are:

4721: $$

4722: [\4s,\4v_k]=k\4v_k,\;[\4s,\4n]=\4n,\quad [\4n,\4v_1]=\4v_2,\;

4723: [\4n,\4v_2]=\4v_3,\;[\4n,\4v_3]=0.

4724: $$ Further, $\7v:=\RR\4v_1\oplus \RR\4v_2\oplus \RR\4v_3\cong \RR^3$

4725: is an abelian ideal in $\7g$ with $[\7g,\7g]=\7g^{\nil}=\RR\4n\oplus

4726: \7v$. Hence, $\7g$ has the structure of the semidirect product

4727: $\7v\rtimes\7h$ with $\7h=\RR\4s\oplus\RR\4n$ and the Main Lemma

4728: applies.

4729: 

4730: \noindent {\bf Remark.} A direct verification shows that for every

4731: $\gamma\in\RR^*$ $(\7g,\7q)$ in \Ruf{GU} is associated to Example

4732: \ruf{EV}.

4733: 

4734: \noindent We show that in the next case the Lie algebra cannot be

4735: solvable, hence, this case can be discarded.

4736: 

4737: \noindent {\bf Ad AII:} Write $\beta_4:=b_4$.  The Jacobi identity

4738: implies $a_2=0=b_3$ and $d_2=\beta_4$, and \Ruf{VA} reads \vskip-18pt

4739: $$

4740: \diagram{[\4x,\overline\4x]&\;=\;&\4z\;\;\cr

4741: [\4x,\overline{\4y}]&\;=\;& \;\;& \kern1em & \kern1em \;&

4742: \;\overline\4x & \;& \;&+&\beta_4\overline{\4y} \cr

4743: [\4y,\overline{\4y}]&\;=\;& & & & & & \kern.7em \;\4y & - & \kern.7em

4744: \; \overline{\4y} \cr [\4y,\4x]&\;=\;& & \;\4x & & &+ &\beta_4 \4y & &

4745: \cr [\4z,\4x]&\;= \; &\rlap{$\beta_4\4z$ \qquad\qquad

4746: $[\4z,\4y]=-\4z\;.$}\hfill }\qquad\qquad\qquad\qquad \Leqno{}

4747: $$

4748: But then the linear span of the vectors

4749: $$

4750: \displaylines{ \4e^+:=\4y-\overline{\4y}, \qquad\qquad

4751: \4h:=-\frac1{\beta_4}\big (\4x+\overline{\4x}+(\beta_4-1)\4y \;+

4752: \;(\beta_4+1) \overline{\4y}\big),\kern1em \cr \4e^-:=

4753: \frac1{4\beta_4^2}\big(2\4z+(2-2\beta_4)\4x+(2+2\beta_4)

4754: \overline{\4x}-(1-\beta_4)^2\4y+(1+\beta_4)^2\overline{\4y}\big) }

4755: $$

4756: is a copy of $\7{sl}(2,\CC)$ in $\7l$, that is, $\7l$ is not solvable.

4757: 

4758: \noindent{\bf Ad AIII:}. The condition $b_4=0$ together with the Jacobi

4759: identity $\Jac{\4x}{\overline{\4x}}{\4y}=0$ implies $a_2=b_3=0$ and

4760: $c_\4y=2\beta_1-1$. Since $(1-\beta_1)\ne 0$, the identity

4761: $[\4z,[\4y,\4x]]=0$ implies $[\4z,\4x]=0$, see the table below.

4762: $$

4763: \diagram{[\4x,\overline\4x]&\;=\;&\4z\;& \cr

4764: [\4x,\overline{\4y}]&\;=\;& &\kern1em \beta_1\4x&+& \;\overline\4x &

4765: \cr

4766: [\4y,\overline{\4y}]&\;=\;& & & & & & \;\;\4y & - & \;\;\overline{\4y}

4767: \cr

4768: [\4y,\4x]&\;=\;& & \kern-6pt(1-\beta_1)\4x & & & & & &

4769: \cr

4770: [\4z,\4x]&\;= \;&0\hfill

4771: \rlap{\qquad,\qquad $[\4z,\4y] \;=\; (2\beta_1-1)\4z\;.$} \hfill

4772: }\qquad\qquad\qquad\qquad

4773: \Leqno{GV}

4774: $$

4775: Select the following basis of $\7g$:

4776: $$

4777: \diagram{\4n\,&\;:= \;& \hbox{\klein$\displaystyle{i\over

4778: 2}$}(\4x-\overline{\4x}) \qquad,\qquad\quad \4s\,:=

4779: \hbox{\Klein$\displaystyle{1\over 2-2\beta_1}$}(\4y+\overline{\4y})\cr

4780: \4v_1&:=& \qquad\qquad i\4y\;- i\overline{\4y} \hfill\cr \4v_2&:=&

4781: \;\;\4x \; + \;\overline{\4x} \hfill\cr \4v_3&:=&

4782: i\4z\;.\vphantom{\big|}\hfill}

4783: $$

4784: One checks immediately that $\7v:=\RR \4v_1\oplus\RR \4v_2\oplus\RR

4785: \4v_3$ is an abelian ideal and $\7h:= \RR \4s\oplus\RR \4n$ is a

4786: subalgebra with $[\4s,\4n]=\4n$. Further,

4787: $$

4788: [\4s,\4v_j]= \hbox{\klein$\displaystyle{2-j+(j-1)\beta_1\over

4789: \beta_1-1}$}\4v_j\steil{for}j=1,2,3\Steil{and}[\4n,\4v_1]=\4v_2,\;

4790: [\4n,\4v_2]=\4v_3,\;[\4n,\4v_3]=0.

4791: $$

4792: Hence, $\7g=\7v\rtimes \7h$ as claimed, and the Main Lemma applies.

4793: Also in this case for all $\beta_1\ne \pm 1$ the CR-algebra

4794: $(\7g,\7q)$ is associated to Example \ruf{EV}.

4795: 

4796: 

4797: \medskip It remains to discuss the cases $\beta_1=\pm1.$

4798: 

4799: \noindent {\fett Case B: $\beta_1=1$.} Plugging $\beta_1$ into

4800: \Ruf{FZ} gives $d_2=\overline b_3$ and $d_1=0.$ A direct check shows

4801: that $\Jac{\4x}{\overline{\4x}}{\4y}=0$ implies $0=b_3=d_2,$ i.e.,

4802: $[\4x,\4y]=0$. Lemma \ruf{VG} then gives that $\7g$

4803: is isomorphic to the semidirect product $\7v\rtimes\7h$ with

4804: $\7v=[\7g,\7g]$ and the abelian subalgebra $\7h$ as defined in

4805: \Ruf{VL}.

4806: 

4807: \medskip\noindent {\fett Case C: $\beta_1=-1$.} We proceed as in the

4808: preceeding cases. Starting from the structure equations \Ruf{FI} with

4809: respect to some perfect basis

4810: $\4x,\overline{\4x},\4y,\overline{\4y},\4z\,$, we first evaluate

4811: \Ruf{FZ} for this particular value of $\beta_1$. We get $d_1=2$ and

4812: $d_2=\overline b_3+2b_4.$ Next, the Jacobi identity

4813: $\Jac{\4x}{\overline{\4x}}{\4y}=0$ implies $b_4:=\beta_4\in \RR$ and

4814: $a_1=\beta_4.$ Further, $\Jac{\4x}{\4y}{\overline{\4y}}=0$ implies

4815: $b_3=-\beta_4 +i\gamma,$ $\gamma\in \RR.$ Next,

4816: $\Jac{\4x}{\overline{\4x}}{\4y}=0$ determines the value of $a_2:$

4817: $a_2=(3\beta_4^2+\gamma^2)/4-i\beta_4\gamma/2.$ Finally

4818: $\Jac{\4x}{\4y}{\4z}=0$ implies

4819: $[\4z,\4x]=(\quu{3\beta_4}2-\quu{3i\gamma}2)\4z,$ see diagram below:

4820: $$

4821: \diagram{[\4x,\overline\4x]&\;=\;&\4z+\;& \beta_4\4x & -

4822: &\beta_4\overline{\4x} &\;+\;& a_2\4y &\;-\;&\overline a_2\overline

4823: \4y \rlap{\qquad

4824: $a_2=\quu{(3\beta_4^2+\gamma^2)}4-\quu{i\gamma\beta_4}2 $} \cr

4825: [\4x,\overline{\4y}]&\;=\;& &\kern1em -\4x&+& \;\overline\4x &

4826: &+b_3\4y&+&\beta_4\overline{\4y} \rlap{\qquad $b_3=-\beta_4+i\gamma$}

4827: \cr [\4y,\overline{\4y}]&\;=\;& & & & & & \;\;\4y & - &

4828: \;\;\overline{\4y} \cr [\4y,\4x]&\;=\; & & \kern-4pt 2\4x & & &-

4829: &b_4\4y & & \cr [\4z,\4x]&\;= \;\rlap{$

4830: (\quu{3\beta_4}2-\quu{3i\gamma}2) \4z$ \qquad $[\4z,\4y]=-3\4z\;\;$.}

4831: }\qquad\qquad\qquad\qquad\qquad\qquad\Leqno{GX}

4832: $$

4833: Select the following basis of $\7g$:

4834: $$

4835: \diagram{\4n\;&\;:= \;& 2i\4x-2i\overline{\4x} \;-ib_3\4y

4836: \;+\;i\overline b_3\overline{\4y} \qquad\qquad \4s\;:=

4837: -\hbox{\klein$\displaystyle{1\over 4}$}(\4y+\overline{\4y})\cr

4838: \4v_1&:=& \qquad\qquad\kern1.3em \hbox{\Klein$\displaystyle{i\over

4839: 2}$}\4y\;\;- \hbox{\Klein$\displaystyle{i\over 2}$}\overline{\4y}

4840: \hfill\cr \4v_2&:=& \;2\4x + 2\overline{\4x} -\;b_3\4y \;- \;\;\;

4841: \overline b_3 \overline{\4y} \hfill\cr \4v_3&:=&

4842: 4i\4z\;.\vphantom{\big|}\hfill}

4843: $$

4844: A direct computation (using \Ruf{GX}) shows that $\7v:=\RR\4v_1\oplus

4845: \RR\4v_2\oplus \RR\4v_3$ is an abelian ideal and

4846: $$

4847: [\4s,\4n]=\4n, \qquad \diagram{[\4s,\4v_1]=

4848: -\hbox{\klein$\displaystyle{1\over 2}$}\4v_1 , \quad[\4s,\4v_2]=

4849: \hbox{\klein$\displaystyle{1\over 2}$}\4v_2 , \quad[\4s,\4v_3]=

4850: \hbox{\klein$\displaystyle{3\over 2}$}\4v_3 , \cr \kern.6em

4851: [\4n,\4v_1]=\;\4v_2 , \kern2.4em [\4n,\4v_2]= \;\4v_3 , \kern1.6em

4852: [\4n,\4v_3]= \;0\;.\qquad }

4853: $$ Again, this shows that $\7g$ is isomorphic to the semidirect

4854: product $\7v\rtimes \7h$ with $\7h=\RR \4n\oplus \RR \4s$. Actually,

4855: an explicit realization of the corresponding CR-manifold $M$ along the

4856: lines of proof of the Main Lemma shows that $(\7g,\7q)$ is associated

4857: to the tube over the future light cone. \qed

4858: 

4859: \medskip We close by stating that Lemma \ruf{GS} together with Lemma

4860: \ruf{VQ} finishes the proof of the classification theorem.

4861: 

4862: 

4863: \Partskip

4864: {\gross\noindent References\write\lst{\def\csname Refa\endcsname{\folio}}}

4865: \bigskip

4866: 

4867: {\Klein

4868: \parindent 15pt\advance\parskip-1pt

4869: \def\Masca#1{Math. Scand. {\bf #1}}

4870: \def\Springer{Ber\-lin-Hei\-del\-berg-New York: Sprin\-ger~}

4871: \def\LecNotes#1{Lecture Notes in Mathematics Vol. #1. \Springer}

4872: \def\Manus#1{ma\-nus\-crip\-ta math. {\bf #1}}

4873: \def\PAMS#1{Proc. Amer. Math. Soc. {\bf #1}} \def\TAMS#1{Transactions

4874: Amer. Math. Soc. {\bf #1}} \def\Quaox#1{Quart. J. Math. Ox\-ford {\bf

4875: #1}}

4876: 

4877: 

4878: \font\klCC=txmia scaled 900

4879: 

4880: 

4881: \Ref{ANFR}Andreotti, A., Fredricks, G.A.: Embeddability of real analytic Cauchy-Riemann manifolds. Ann. Scuola Norm. Sup. Pisa Cl. Sci. {\bf 6} (1979), 285--304.

4882: \Ref{ANHI}Andreotti, A., Hill, C.D.: Complex characteristic coordinates and tangential Cauchy-Riemann equations. Ann. Scuola Norm. Sup. Pisa Sci. Fis. {\bf 26} (1972), 299-324.

4883: \Ref{BELO}Beloshapka, V.K.: Symmetries of Real Hypersurfaces in Complex 3-Space. Math. Notes {\bf 78} (2005), 156--163.

4884: \Ref{BHUR}Baouendi, M.S., Huang, X., Rothschild, L.P.: Regularity of CR mappings between algebraic hypersurfaces. Invent. Math. {\bf 125} (1996), 13-36.

4885: \Ref{BERO}Baouendi, M.S., Ebenfelt, P., Rothschild, L.P.: {\sl Real Submanifolds in Complex Spaces and Their Mappings}. Princeton Math. Series {\bf 47}, Princeton Univ. Press, 1998.

4886: \Ref{BAHR}Baouendi, M.S., Huang, X., Rothschild, L.P.: Regularity of CR mappings between algebraic hypersurfaces. Invent. Math. {\bf 125} (1996), 13-36.

4887: \Ref{BARZ} Baouendi, M.S., Rothschild, L.P., Zaitsev, D.: Equivalences of real submanifolds in complex space. J. Differential Geom. 59 (2001), 301--351.

4888: \Ref{BOGG}Boggess, A.: {\sl CR manifolds and the tangential Cauchy-Riemann complex.} Studies in Advanced Mathematics. CRC Press. Boca Raton~ Ann Arbor~ Boston~ London 1991.

4889: \Ref{BOUR}Bourbaki, N.: {\sl \'El\'ements de math\'ematique: groupes et alg\`ebres de Lie.} Chapitre I-IX. Herman, Paris 1960.

4890: \Ref{BUSH}Burns, D., Shnider, S.: Spherical hypersurfaces in complex manifolds. Invent. Math. {\bf 33} (1976), 223--246.

4891: \Ref{CART}Cartan, \'E.: Sur la g\'eom\'etrie pseudo-conforme des hypersurfaces de l'espace de deux variables complexes. Annali di Matematica Pura ed Applicata {\bf 11/1} (1933) 17--90.

4892: \Ref{CHMO}Chern, S.S., Moser, J.K.: Real hypersurfaces in complex manifolds. Acta. Math. {\bf 133} (1974), 219-271.

4893: \Ref{DAYA}Dadok, J.,Yang, P.: Automorphisms of tube domains and spherical hypersurfaces. Amer. J. Math. {\bf 107} (1985), 999-1013.

4894: \Ref{DOKR}Doubrov, B., Komrakov, B., Rabinovich, M.: Homogeneous surfaces in the three-dimenaional affine geometry. In: {\sl Geometry and Topology of Submanifolds, VIII,} World Scientific, Singapore 1996, pp. 168-178

4895: \Ref{EAEZ}Eastwood, M., Ezhov, V.: On Affine Normal Forms and a Classification of Homogeneous Surfaces in Affine Three-Space. Geometria Dedicata {\bf 77} (1999), 11-69.

4896: \Ref{EBFT}Ebenfelt, P.: Normal Forms and Biholomorphic Equivalence of Real Hypersurfaces in \hbox{\klCC\char131}$^3$. Indiana J. Math. {\bf 47} (1998), 311-366.

4897: \Ref{EBEN}Ebenfelt, P.: Uniformly Levi degenerate CR manifolds: the 5-dimensional case. Duke Math. J. {\bf 110} (2001), 37--80.

4898: \Ref{FELS}Fels, G.: Locally homogeneous finitely nondegenerate CR-manifolds. To appear. {\tt arXiv:math.CV/0606032}

4899: \Ref{FEKA}Fels, G., Kaup, W.: CR-manifolds of dimension 5: A Lie algebra approach. J. Reine Angew. Math, to appear. {\ninett http://arxiv.org/pdf/math.DS/0508011}

4900: \Ref{GAME}Gaussier, H., Merker, J.: A new example of a uniformly Levi degenerate hypersurface in \hbox{\klCC\char131}$^3$. Ark. Mat. {\bf 41} (2003), 85--94.

4901: \Ref{HOCH}Hochschild, G.: {\sl The structure of Lie groups.} Holden-Day, Inc., San Francisco-London-Amsterdam 1965

4902: \Ref{KNAP}Knapp, A.W.: {\it Lie groups beyond an introduction}. Second edition. Progress in Mathematics {\bf 140}. Birkh\"auser Boston, Inc., Boston, MA, 2002.

4903: \Ref{HART}Hartshorne, R.: {\sl Algebraic Geometry}. Graduate Texts in Mathematics {\bf 52}. \Springer 1977

4904: \Ref{HORM}H\"ormander, L.: {\sl An Introduction to Complex Analysis in Several Variables}. Third edition. North-Holland Publishing Co. Amsterdam - New York, 1990.

4905: \Ref{ISMI}Isaev, A.V., Mishchenko, M.A.: Classification of spherical tube hypersurfaces that have one minus in the Levi signature form. Math. USSR-Izv. {\bf 33} (1989), 441-472.

4906: \Ref{KAZT}Kaup, W., Zaitsev, D.: On local CR-transformations of Levi degenerate group orbits in compact Hermitian symmetric spaces. J. Eur. Math. Soc. {\bf 8} (2006), 465-490.

4907: \Ref{LOBO}Loboda, A.V.: Homogeneous real hypersurfaces in \hbox{\klCC\char131}$^3$ with $2$-dimensional isotropy groups. Tr. Mat. Inst. Steklova {\bf 235} (2001), 114-142 and Proc. Steklov Inst. Math {\bf 235} (2001), 107-135.

4908: \Ref{LOBP}Loboda, A.V.: Homogeneous nondegenerate surfaces in \hbox{\klCC\char131}$^{3}$ with two-dimensional isotropy groups. (Russian) translation in  Funct. Anal. Appl. {\bf 36}  (2002), 151--153.

4909: \Ref{LOBQ}Loboda, A.V.: On the determination of a homogeneous strictly pseudoconvex hypersurface from the coefficients of its normal equation. (Russian)  translation in  Math. Notes  {\bf 73}  (2003), 419--423.

4910: \Ref{MENA}Medori, C., Nacinovich, M.: Algebras of infinitesimal CR automorphisms.  J. Algebra  {\bf287}  (2005), 234--274.

4911: \Ref{PALA}{Palais, R.~S.:} {\sl A global formulation of the Lie theory of transformation groups.} Mem. AMS 1957.

4912: \Ref{SAKA}Sakai, T.: {\sl Riemannian geometry.} Translations of Mathematical Monographs, 149. American Mathematical Society, Providence, RI, 1996.

4913: \Ref{SEVL}Sergeev, A.G., Vladimirov, V.S.: Complex analysis in the future tube. In: {\sl Several Complex Variables II,} Encyclopedia Math. Sci {\bf 8}, \Springer 1994.

4914: \Ref{TANA}Tanaka, N.: On the pseudo-conformal geometry of hypersurfaces of the space of $n$ complex variables. J. Math. Soc. Japan {\bf 14} (1962), 397-429.

4915: \bigskip

4916: }

4917: 

4918: \medskip\centerline{Mathematisches Institut, Universit\"at

4919: T\"ubingen, Auf der Morgenstelle 10, 72076 T\"ubingen, Germany}

4920: \smallskip

4921: \centerline{e-mail: {\ninett gfels@uni-tuebingen.de, kaup@uni-tuebingen.de}}

4922: 

4923: 

4924: \closeout\aux\closeout\lst\bye

4925: 

4926: 

4927: 

4928: %%% Local Variables:

4929: %%% mode: plain-tex

4930: %%% TeX-master: t

4931: %%% End:

4932: 

4933: 

4934: