1: \documentclass[11pt]{article}

2: \usepackage{amsmath}

3: \usepackage{amsthm}

4: \input{epsf.tex}

5: 

6: \newtheorem{theorem}{Theorem}[section]

7: \newtheorem{proposition}[theorem]{Proposition}

8: \newtheorem{lemma}[theorem]{Lemma}

9: \newtheorem{corollary}[theorem]{Corollary}

10: \newtheorem{define}[theorem]{Definition}

11: 

12: 

13: 

14: \def\Empty{}

15: 

16: %

17: % Definitions that use @ : 

18: \catcode`\@=11

19: % Redefine section heading to get smaller letters (this

20: % is copied from the original in rep12.sty)

21: 

22: \def\section{\@startsection {section}{1}{\z@}{-3.5ex plus -1ex minus 

23: -.2ex}{2.3ex plus .2ex}{\large\bf}}

24: 

25: 

26: % Our own caption style --

27: 

28: \def\fnum@figure{{\small Figure \thefigure}}

29: \def\fakefigure{\def\@captype{figure}}

30: 

31: \long\def\@makecaption#1#2{

32:     \vskip 10pt 

33:     \def\FCap{#2} \def\NoCap{\ignorespaces}

34:     \ifx \FCap\NoCap

35:        \setbox\@tempboxa\hbox{#1}  % This is to avoid the damn colon.

36:       \else

37:        \setbox\@tempboxa\hbox{#1: \small \it #2}

38:     \fi

39:     \ifdim \wd\@tempboxa >\hsize   % IF longer than one line:

40:         \unhbox\@tempboxa\par      %   THEN set as ordinary paragraph.

41:       \else                        %   ELSE  center.

42:         \hbox to\hsize{\hfil\box\@tempboxa\hfil}  

43:     \fi}

44: 

45: 

46: % Our page heading style:

47: \pagestyle{headings}

48: \oddsidemargin 0.00in

49: \evensidemargin 0.00in

50: \textwidth 6.6in

51: \textheight 9.0in

52: \headsep 16pt

53: \topmargin -16pt

54: %\def\@evenhead{\rm \thepage\hfil \leftmark}%        Left heading.

55: \def\@oddhead{\hbox{}\rightmark \hfil \rm\thepage}% Right heading.

56: %\def\chaptermark#1{\markboth {\sc {\ifnum \c@secnumdepth >\m@ne

57: %      \@chapapp\ \thechapter. \ \fi #1}}{}}%

58: \def\sectionmark#1{\markright {\sc{\ifnum \c@secnumdepth >\z@

59:       \S\thesection.\hskip 1em\relax \fi #1}}}

60: %\def\thebibliography#1{\chapter*{References\@mkboth

61: %{\sc References}{\sc References}}\list

62: % {[\arabic{enumi}]}{\settowidth\labelwidth{[#1]}\leftmargin\labelwidth

63: %    \advance\leftmargin\labelsep

64: %    \usecounter{enumi}}

65: %    \def\newblock{\hskip .11em plus .33em minus -.07em}

66: %    \sloppy

67: %    \sfcode`\.=1000\relax}

68: 

69: \catcode`\@=12

70: %

71: 

72: 

73: \def\oplabel#1{

74:   \def\OpArg{#1} \ifx \OpArg\Empty {} \else

75:   	\label{#1}

76:   \fi}

77: 

78: 

79: 	

80: \def\lref#1{\ref{#1} (#1)}% Prints label as well as number

81: %

82: %

83: \def\defineterm#1{{\it #1}}

84: %

85: % Make a blank figure of given size with label in center.

86: %	\blankfig{label}{# of inches}{caption}

87: %

88: \newlength{\saveu}

89: \def\blankspace#1#2{

90:   \setlength{\saveu}{#2in}

91:   \divide\saveu by 2

92:   \vskip\saveu

93:   \centerline {\small \mbox{}#1}

94:   \vskip\saveu

95: }

96: 

97: \long\def\placefig#1#2#3#4{

98:   \begin{figure}[#4]

99:     \blankspace{#1}{#2}

100:     \caption[#1]{#3}

101:     \oplabel{#1}

102:   \end{figure}}

103: \long\def\blankfig#1#2#3{

104:   \placefig{#1}{#2}{#3}{htbp}}

105: \long\def\herefig#1#2#3{

106:   \par\noindent\vbox{

107:   \blankspace{#1}{#2}

108:   \fakefigure\caption[#1]{#3}

109:   \oplabel{#1}

110:   \vspace{\intextsep}

111:   }}

112: \long\def\botfig#1#2#3{

113:   \placefig{#1}{#2}{#3}{b}}

114: 

115: 

116: 

117: % Definitions that use @ : 

118: \catcode`\@=11

119: 

120: % page heading on even pages gives chapter, but on same side of page

121: % as for odd pages. 

122: 

123: 

124: \newcommand{\zz}{\mbox{${\bf Z} \oplus {\bf Z}$}}

125: \newcommand{\pa}{\mbox{$<_{\alpha}$}}

126: \newcommand{\of}{\mbox{$\overline f$}}

127: \newcommand{\cd}{\mbox{${\cal D}$}}

128: \newcommand{\pb}{\mbox{$<_{\beta}$}}

129: \newcommand{\pu}{\mbox{$<_u$}}

130: \newcommand{\poa}{\mbox{$>_{\alpha}$}}

131: \newcommand{\pou}{\mbox{$>_u$}}

132: \newcommand{\oh}{\mbox{$\overline h$}}

133: \newcommand{\og}{\mbox{$\overline g$}}

134: \newcommand{\spr}{\mbox{${\bf S}^2 \times {\bf S}^1$}}

135: \newcommand{\cs}{\mbox{${\bf S}^1$}}

136: \newcommand{\ch}{\mbox{$\widehat{\bf C}$}}

137: \newcommand{\pp}{\mbox{$\Phi_{\theta}$}}

138: \newcommand{\ii}{\mbox{${\cal I}$}}

139: \newcommand{\dd}{\mbox{${\cal D}$}}

140: \newcommand{\ccl}{\mbox{${\cal L}$}}

141: \newcommand{\kk}{\mbox{${\cal K}$}}

142: \newcommand{\ca}{\mbox{${\cal A}$}}

143: \newcommand{\is}{\mbox{${\cal I}^s$}}

144: \newcommand{\iu}{\mbox{${\cal I}^u$}}

145: \newcommand{\res}{\mbox{${\cal R}^s$}}

146: \newcommand{\reu}{\mbox{${\cal R}^u$}}

147: \newcommand{\bp}{\mbox{$B^+$}}

148: \newcommand{\bn}{\mbox{$B^-$}}

149: \newcommand{\ba}{\mbox{${\bf A}$}}

150: \newcommand{\aaa}{\mbox{${\cal A}$}}

151: \newcommand{\bb}{\mbox{${\cal B}$}}

152: \newcommand{\ak}{\mbox{${\cal A}_{\kappa}$}}

153: \newcommand{\at}{\mbox{${\cal A}_{\tau}$}}

154: \newcommand{\otsf}{\mbox{${\overline \tau^s_F}$}}

155: \newcommand{\otuf}{\mbox{${\overline \tau^u_F}$}}

156: \newcommand{\ots}{\mbox{${\overline \tau^s}$}}

157: \newcommand{\otu}{\mbox{${\overline \tau^u}$}}

158: \newcommand{\ap}{\mbox{${\cal A}_{\alpha}$}}

159: \newcommand{\ab}{\mbox{${\cal A}_{\beta}$}}

160: \newcommand{\ag}{\mbox{${\cal A}_g$}}

161: \newcommand{\lat}{\mbox{${\cal L A}_{\tau}$}}

162: \newcommand{\lak}{\mbox{${\cal L A}_{\kappa}$}}

163: \newcommand{\lal}{\mbox{${\cal L A}_{\alpha}$}}

164: \newcommand{\lag}{\mbox{${\cal L A}_{\gamma}$}}

165: \newcommand{\lab}{\mbox{${\cal L A}_{\beta}$}}

166: \newcommand{\la}{\mbox{${\cal L A}$}}

167: \newcommand{\af}{\mbox{${\cal A}_f$}}

168: \newcommand{\uu}{\mbox{${\cal U}$}}

169: \newcommand{\ro}{\mbox{${\cal R}$}}

170: \newcommand{\vv}{\mbox{${\cal V}$}}

171: \newcommand{\we}{\mbox{${\cal W}$}}

172: \newcommand{\re}{\mbox{${\cal R}$}}

173: \newcommand{\wwx}{\mbox{${\cal WX}$}}

174: \newcommand{\et}{\mbox{$\eta _-$}}

175: \newcommand{\tsv}{\mbox{$\sigma ^s_e$}}

176: \newcommand{\hhu}{\mbox{${\cal H} ^u$}}

177: \newcommand{\hhs}{\mbox{${\cal H} ^s$}}

178: \newcommand{\hu}{\mbox{${\cal H}(\widetilde {\cal F} ^u)$}}

179: \newcommand{\hs}{\mbox{${\cal H}(\widetilde {\cal F} ^s)$}}

180: \newcommand{\hp}{\mbox{${\cal H}$}}

181: \newcommand{\tup}{\mbox{$\Theta ^u_+$}}

182: \newcommand{\tun}{\mbox{$\Theta ^u_-$}}

183: \newcommand{\tsp}{\mbox{$\Theta ^s_+$}}

184: \newcommand{\tsn}{\mbox{$\Theta ^s_-$}}

185: \newcommand{\vs}{\mbox{$\varphi_s$}}

186: \newcommand{\vu}{\mbox{$\varphi_u$}}

187: \newcommand{\vg}{\mbox{$\varphi_G$}}

188: \newcommand{\vf}{\mbox{$\varphi_F$}}

189: \newcommand{\vif}{\mbox{$\varphi^{\infty}_F$}}

190: \newcommand{\vig}{\mbox{$\varphi^{\infty}_G$}}

191: \newcommand{\vil}{\mbox{$\varphi^{\infty}_L$}}

192: \newcommand{\vi}{\mbox{$\varphi^{\infty}$}}

193: \newcommand{\vr}{\mbox{$\varphi _s$}}

194: \newcommand{\vvv}{\mbox{${\cal V}$}}

195: \newcommand{\hlf}{\mbox{$X^L_F$}}

196: \newcommand{\hfl}{\mbox{$X^F_L$}}

197: \newcommand{\zf}{\mbox{$\zeta_F$}}

198: \newcommand{\zl}{\mbox{$\zeta_L$}}

199: \newcommand{\pin}{\mbox{$\partial _{\infty}$}}

200: \newcommand{\cc}{\mbox{$\cal C$}}

201: \newcommand{\lsf}{\mbox{${\cal L}^s_F$}}

202: \newcommand{\luf}{\mbox{${\cal L}^u_F$}}

203: \newcommand{\oo}{\mbox{$\cal O$}}

204: \newcommand{\oos}{\mbox{${\cal O}^s$}}

205: \newcommand{\oou}{\mbox{${\cal O}^u$}}

206: \newcommand{\ooo}{\mbox{$\cal J$}}

207: \newcommand{\uo}{\mbox{${\cal J}^u$}}

208: \newcommand{\so}{\mbox{${\cal J}^s$}}

209: \newcommand{\ou}{\mbox{${\cal J}^u_+$}}

210: \newcommand{\rrr}{\mbox{$\times {\bf R}$}}

211: \newcommand{\rrrr}{\mbox{${\bf R}$}}

212: \newcommand{\rp}{\mbox{$[0,+\infty)$}}

213: \newcommand{\rr}{\mbox{$\times \{ 0 \}$}}

214: \newcommand{\pn}{\mbox{$\partial N$}}

215: \newcommand{\mi}{\mbox{$\widetilde M$}}

216: \newcommand{\wl}{\mbox{$\widetilde \lambda$}}

217: \newcommand{\ls}{\mbox{$\Lambda^s$}}

218: \newcommand{\lu}{\mbox{$\Lambda^u$}}

219: \newcommand{\wgs}{\mbox{$\widetilde \Lambda^s$}}

220: \newcommand{\wls}{\mbox{$\widetilde \Lambda^s$}}

221: \newcommand{\wlu}{\mbox{$\widetilde \Lambda^u$}}

222: \newcommand{\wcl}{\mbox{$\widetilde {\cal L}$}}

223: \newcommand{\wlsf}{\mbox{$\widetilde \Lambda^s_F$}}

224: \newcommand{\wluf}{\mbox{$\widetilde \Lambda^u_F$}}

225: \newcommand{\wlse}{\mbox{$\widetilde \Lambda^s_E$}}

226: \newcommand{\wlsg}{\mbox{$\widetilde \Lambda^s_G$}}

227: \newcommand{\wlsfi}{\mbox{$\widetilde \Lambda^s_{F_i}$}}

228: \newcommand{\wlsl}{\mbox{$\widetilde \Lambda^s_L$}}

229: \newcommand{\ww}{\mbox{$\widetilde W$}}

230: \newcommand{\wws}{\mbox{$\widetilde W^s$}}

231: \newcommand{\ti}{\mbox{$\partial \widetilde T$}}

232: \newcommand{\ttt}{\mbox{$\widetilde T$}}

233: \newcommand{\ppn}{\mbox{$\partial _+ N$}}

234: \newcommand{\lll}{\mbox{$l _{\epsilon}$}}

235: \newcommand{\iT}{\mbox{$i _{\widetilde T}$}}

236: \newcommand{\wwp}{\mbox{$\widetilde \Phi$}}

237: \newcommand{\wwr}{\mbox{$\widetilde \Phi_{\rrrr}$}}

238: \newcommand{\w}{\mbox{$\widetilde {\partial _+ C}$}}

239: \newcommand{\pw}{\mbox{$\varphi _{\widetilde T}$}}

240: \newcommand{\lmun}{\mbox{$\Gamma ^s \cup \Gamma ^u$}}

241: \newcommand{\gc}{\mbox{${\cal G}$}}

242: \newcommand{\gsgg}{\mbox{${\cal G}^s_*$}}

243: \newcommand{\gl}{\mbox{${\cal G}$}}

244: \newcommand{\gugg}{\mbox{${\cal G}^u_*$}}

245: \newcommand{\gsg}{\mbox{${\cal G}^s$}}

246: \newcommand{\gug}{\mbox{${\cal G}^u$}}

247: \newcommand{\ga}{\mbox{$\Gamma$}}

248: \newcommand{\gs}{\mbox{$\Gamma ^s$}}

249: \newcommand{\gu}{\mbox{$\Gamma ^u$}}

250: \newcommand{\ggs}{\mbox{${\cal F}^s_T$}}

251: \newcommand{\ggu}{\mbox{${\cal F}^u_T$}}

252: \newcommand{\gsr}{\mbox{$\Gamma ^s \times {\bf R}$}}

253: \newcommand{\gur}{\mbox{$\Gamma ^u \times {\bf R}$}}

254: \newcommand{\se}{\mbox{$S _{\epsilon}$}}

255: \newcommand{\gus}{\mbox{$\Gamma ^u \downarrow S_{\epsilon}$}}

256: \newcommand{\vt}{\mbox{$\overline \varphi _{\widetilde T}$}}

257: %\newcommand{\vvvv}{\mbox{$\varphi _{\widetilde T}$}}

258: \newcommand{\hh}{\mbox{${\bf H}^2$}}

259: \newcommand{\hhh}{\mbox{${\bf H}^3$}}

260: \newcommand{\lmS}{\mbox{$\Gamma ^s \downarrow S$}}

261: \newcommand{\lmU}{\mbox{$\Gamma ^u \downarrow S$}}

262: \newcommand{\lms}{\mbox{$\Gamma ^s \downarrow S_{\epsilon}$}}

263: \newcommand{\lmu}{\mbox{$\Gamma ^u \downarrow S_{\epsilon}$}}

264: \newcommand{\wall}{\mbox{$L \times \R$}}

265: \newcommand{\wa}{\mbox{$\widetilde {\alpha}$}}

266: \newcommand{\wb}{\mbox{$\widetilde {\beta}$}}

267: \newcommand{\wdd}{\mbox{$\widetilde {\delta}$}}

268: \newcommand{\lap}{\mbox{$\Gamma ^s \cup \Gamma ^u \cup Q$}}

269: \newcommand{\tz}{\mbox{$\widetilde T \times \{ 0 \}$}}

270: \newcommand{\es}{\mbox{${\cal E}^s_F$}}

271: \newcommand{\eu}{\mbox{${\cal E}^u_F$}}

272: \newcommand{\wt}{\mbox{$\widetilde T$}}

273: \newcommand{\wss}{\mbox{$\widetilde S$}}

274: \newcommand{\rs}{\mbox{$\widetilde \rho(\widetilde S)$}}

275: \newcommand{\trs}{\mbox{$\Theta(\widetilde \rho(\widetilde S))$}}

276: \newcommand{\wn}{\mbox{$\widetilde N$}}

277: \newcommand{\si}{\mbox{$S^2_{\infty}$}}

278: \newcommand{\su}{\mbox{$S^1_{\infty}$}}

279: \newcommand{\sss}{\mbox{$S^s_{\infty}$}}

280: \newcommand{\sus}{\mbox{$S^u_{\infty}$}}

281: \newcommand{\s}{\mbox{$S ^1$}}

282: \newcommand{\fol}{\mbox{$\cal F$}}

283: \newcommand{\gal}{\mbox{$\cal G$}}

284: \newcommand{\gp}{\mbox{$\cal G_+$}}

285: \newcommand{\gm}{\mbox{$\cal G_-$}}

286: \newcommand{\gn}{\mbox{$\widetilde {\cal G}$}}

287: \newcommand{\gmn}{\mbox{$\widetilde {\cal G}_-$}}

288: \newcommand{\gpn}{\mbox{$\widetilde {\cal G}_+$}}

289: \newcommand{\fn}{\mbox{$\widetilde {\cal F}$}}

290: \newcommand{\als}{\mbox{${\cal L} ^s$}}

291: \newcommand{\lss}{\mbox{${\cal L}^s_*$}}

292: \newcommand{\alu}{\mbox{${\cal L} ^u$}}

293: \newcommand{\lsd}{\mbox{${\cal L} ^s_d$}}

294: \newcommand{\lns}{\mbox{$\widetilde {\cal L} ^s$}}

295: \newcommand{\lnu}{\mbox{$\widetilde {\cal L} ^u$}}

296: \newcommand{\fs}{\mbox{${\cal F} ^s$}}

297: \newcommand{\fu}{\mbox{${\cal F} ^u$}}

298: \newcommand{\fss}{\mbox{${\cal F} ^{ss}$}}

299: \newcommand{\fuu}{\mbox{${\cal F} ^{uu}$}}

300: \newcommand{\fuo}{\mbox{${\cal F} ^{uu}_0$}}

301: \newcommand{\fnss}{\mbox{$\widetilde {\cal F} ^{ss}$}}

302: \newcommand{\fnuu}{\mbox{$\widetilde {\cal F} ^{uu}$}}

303: \newcommand{\fns}{\mbox{${\widetilde {\cal F}}^s $}}

304: \newcommand{\fnso}{\mbox{${\widetilde {\cal F}}^s_{\cal O} $}}

305: \newcommand{\fnuo}{\mbox{${\widetilde {\cal F}}^u_{\cal O} $}}

306: \newcommand{\fnsf}{\mbox{${\widetilde {\cal F}}^s_F $}}

307: \newcommand{\fnsg}{\mbox{${\widetilde {\cal F}}^s_G $}}

308: \newcommand{\fnuf}{\mbox{${\widetilde {\cal F}}^u_F $}}

309: \newcommand{\fnu}{\mbox{${\widetilde {\cal F}}^u $}}

310: \newcommand{\wv}{\mbox{${\overline {\varphi}}$}}

311: \newcommand{\oxi}{\mbox{${\overline {\xi}}$}}

312: \newcommand{\wvv}{\mbox{${\varphi}$}}

313: \newcommand{\fsS}{\mbox{${\cal F} ^s_S$}}

314: \newcommand{\fsG}{\mbox{${\cal F} ^s_G$}}

315: \newcommand{\fsE}{\mbox{${\cal F} ^s_E$}}

316: \newcommand{\fuG}{\mbox{${\cal F} ^u_G$}}

317: \newcommand{\fuS}{\mbox{${\cal F} ^u_S$}}

318: \newcommand{\fsT}{\mbox{${\cal F} ^s_T$}}

319: \newcommand{\fuT}{\mbox{${\cal F} ^u_T$}}

320: \newcommand{\fSS}{\mbox{$\widetilde{\cal F} ^s_{\cal S}$}}

321: \newcommand{\fUS}{\mbox{$\widetilde{\cal F} ^u_{\cal S}$}}

322: \newcommand{\fnuS}{\mbox{$\widetilde {{\cal F}}^u_S $}}

323: \newcommand{\fnsS}{\mbox{$\widetilde {{\cal F}}^s_S $}}

324: \newcommand{\ws}{\mbox{$\widetilde  W^s$}}

325: \newcommand{\wu}{\mbox{$\widetilde  W^u$}}

326: \newcommand{\hnt}{homotopically non trivial}

327: \newcommand{\up}{\mbox{${\cal U} _+$}}

328: \newcommand{\un}{\mbox{${\cal U} _-$}}

329: \newcommand{\Dp}{\mbox{$\Delta ^+$}}

330: \newcommand{\Dn}{\mbox{$\Delta ^-$}}

331: \newcommand{\fg}{\mbox{$\phi _g$}}

332: \newcommand{\h}{\mbox{$\bf H$}}

333: 

334:  

335: 

336: \def\@evenhead{\rm \leftmark \hfil \thepage}%        Left heading.

337: \def\chaptermark#1{\markboth {\sc {\ifnum \c@secnumdepth >\m@ne

338:       \@chapapp\ \thechapter. \ \fi #1}}{}}%

339: 

340: \catcode`\@=12

341: 

342: 

343: % Macros by Lee Mosher

344: %

345: %

346: %

347: % math symbols

348: %

349: \def\leftceiling{{\lceil}}

350: \def\rightceiling{{\rceil}}

351: \def\Reals{{\bf R}}

352: \def\reals{\Reals}

353: \def\Integers{{\bf Z}}

354: \def\integers{\Integers}

355: \def\rationals{{\bf Q}}

356: \def\Hyperbolic{{\bf H}}

357: \def\hyperbolic{\Hyperbolic}

358: \def\Euclidean{{\bf E}}

359: \def\euclidean{\Euclidean}

360: \def\Complex{{\bf C}}

361: \def\complex{\Complex}

362: \def\union{\cup}

363: \def\Union{\bigcup}

364: \def\intersect{\cap}

365: \def\intersection{\intersect}

366: \def\Intersection{\bigcap}

367: \def\inverse{{-1}}

368: \def\boundary{\partial}

369: \def\bd{\boundary}

370: \def\bdy{\boundary}

371: \def\suchthat{\bigm|}

372: \def\composed{\circ}

373: \def\comp{\composed}

374: \def\restrictedto{\bigm|}

375: \def\isomorphic{\approx}

376: \def\cross{\times}

377: \def\origin{{\bf O}}

378: \def\half{{1\over2}}

379: \def\inclusion{\hookrightarrow}

380: \def\infinity{\infty}

381: \def\concat{\ast}

382: \def\Sum{\sum}

383: \def\superset{\supset}

384: \def\from{\colon}

385: \def\homeo{\approx}

386: \def\disjunion{\coprod}

387: \def\tensor{\otimes}

388: 

389: %fix the next one

390: \def\directsum{{\bf +}}

391: 

392: %

393: %

394: % Math tokens of more than one letter

395: %

396: \def\func#1{\operatorname{#1}}

397: \def\Core{\func{Core}}

398: \def\Stable{\func{Stable}}

399: \def\Unstable{\func{Unstable}}

400: \def\cl{\func{cl}}

401: \def\closure{\cl}

402: \def\image{\func{image}}

403: \def\diam{\func{diam}}

404: \def\Isom{\func{Isom}}

405: \def\interior{\func{int}}

406: \def\Area{\func{Area}}

407: \def\area{\Area}

408: \def\Volume{\func{Vol}}

409: \def\volume{\Volume}

410: \def\Vol{\Volume}

411: \def\Length{\func{Length}}

412: \def\domain{\func{Domain}}

413: \def\range{\func{Range}}

414: \def\supp{\func{Supp}}

415: \def\source{\func{Srce}}

416: \def\target{\func{Targ}}

417: 

418: 

419: 

420: 

421: 

422: 

423: % Names with funny marks over the letters

424: \def\Mobius{M\"obius}

425: \def\Poincare{Poincar\'e}

426: \def\Teichmuller{Teichm\"uller}

427: 

428: 

429: 

430: 

431: %MosherMacros.sty}

432: 

433: \def\centeredepsfbox#1{\centerline{\epsfbox{#1}}}

434: 

435: \begin{document}

436: 

437: \title{Geometry of foliations and flows I: 

438: Almost transverse pseudo-Anosov flows and 

439: asymptotic behavior 

440: of foliations}

441: \author{S\'{e}rgio R. Fenley

442: \thanks{Reseach partially supported by NSF grants 

443: DMS-0296139 and DMS-0305313.}

444: \footnote{Mathematics Subject Classification.

445: Primary: 53C23, 57R30, 37D20; Secondary: 57M99, 53C12,

446: 32Q05, 57M50.}

447: }

448: \maketitle

449: 

450: \vskip .2in

451: 

452: {\small{

453: \centerline{\bf {Abstract}}

454: Let $\fol$ be a foliation in a 

455: closed  $3$-manifold with negatively curved fundamental group

456: and suppose that $\fol$ is almost transverse to a

457: quasigeodesic pseudo-Anosov flow.

458: We show that the leaves of the foliation

459: in the universal cover extend continuously to the sphere

460: at infinity, therefore the limit sets of the leaves are continuous images

461: of the circle.

462: One important corollary is that  if

463: $\fol$ is  a

464: Reebless, finite depth foliation in a hyperbolic manifold, then

465: it has the continuous extension property.

466: Such finite depth foliations 

467: exist whenever

468: the second Betti number is non zero.

469: The result also applies to other classes of foliations, including

470: a large class of foliations where all leaves

471: are dense and infinitely many examples with one sided

472: branching.

473: %The tools in the proof involve topological properties 

474: %of pseudo-Anosov flows and transverse/almost transverse

475: %foliations and also metric properties of flows.

476: One key tool is a detailed understanding

477: of asymptotic properties of almost pseudo-Anosov singular

478: $1$-dimensional foliations in the leaves of $\fol$ lifted

479: to the universal cover. 

480: }}

481: 

482: 

483: \section{Introduction}

484: 

485: A $2$-dimensional foliation in a $3$-manifold is called Reebless

486: if it does not have a Reeb component: a foliation of the solid

487: torus so that the boundary is a leaf and the interior is foliated

488: by plane leaves spiralling towards the boundary.

489: As such the boundary leaf does not inject in the fundamental

490: group level and is compressible. Novikov \cite{No} showed that 

491: Reebless foliations and the underlying manifolds have excellent

492: topological properties. This result was extended by Rosenberg

493: \cite{Ros}, Palmeira \cite{Pa}

494: and many others.

495: 

496: The goal of this article is to analyse geometric properties of

497: foliations. Let $\fol$ be a Reebless 

498: %transversely oriented

499: foliation in $M^3$ with negatively curved fundamental

500: group.

501: Reebless implies that $M$ is irreducible \cite{Ros}.

502: In this article we will not make use of Perelman's 

503: fantastic results \cite{Pe1,Pe2,Pe3}, which if confirmed

504: imply that the manifold is hyperbolic.

505: Reebless foliations exist for instance  whenever $M$ is irreducible,

506: orientable and

507: the second homology of $M$ is not finite \cite{Ga1,Ga3}.

508: They also exist in much more generality by work of Roberts 

509: \cite{Ro1,Ro2,Ro3}, Thurston \cite{Th5} and many others.

510: 

511: Let 

512: $M^3$ be closed, irreducible with negatively curved 

513: fundamental group.

514: The universal cover is canonically compactified with a sphere

515: at infinity (denoted by $\si$), with compactification

516: a closed ball \cite{Be-Me}. The covering transformations

517: act by homeomorphisms in the compactified space.

518: Let $\fn$ be the lifted foliation to the universal cover $\mi$.

519: The leaves of $\fn$ are topological planes \cite{No}

520: and they are properly embedded. 

521: Hence they only limit

522: in the sphere at infinity. For hyperbolic manifolds,

523: the relationship between objects

524: in hyperbolic $3$-space (isometric to $\mi$) and their

525: limit sets in the sphere at infinity is central to  

526: the theory of 

527: such manifolds \cite{Th1,Th2,Mar}. The same is true

528: if $\pi_1(M)$ is negatively curved \cite{Gr,Gh-Ha}.

529: There is

530: a metric in $M$ so that all leaves of $\fol$ are hyperbolic

531: (that is constant curvature $-1$) \cite{Ca} and so the universal

532: cover of each leaf of 

533: $\fol$ is isometric to the hyperbolic  plane ($\hh$).

534: The {\em continuous extension question} asks whether these

535: leaves extend continuously to the sphere at infinity,

536: that is: given the inclusion map from a leaf $F$ of $\fn$

537: to $\mi$ is there a continuous extension to a map

538: $F \cup \pin F$ to $\mi \cup \si$? 

539: Here $\pin F$ is the ideal boundary of $F$ which is homeomorphic

540: to a circle.

541: If this is true we say that $\fol$ has the 

542: {\em continuous extension property}.

543: In that case the 

544: restriction of the map to $\pin F$ expresses the limit

545: set of $F$ as the continuous image of a circle, showing

546: it is locally connected.

547: 

548: 

549: 

550: 

551: 

552: In this article we prove the continuous extension property

553: for a very large class of foliations.

554: A pseudo-Anosov flow is {\em almost transverse} to a foliation

555: if an appropriate blow up of the flow is transverse 

556: to the foliation.

557: A blow up is obtained by replacing a (possibly empty) collection

558: of singular orbits by a union of annuli.

559: Another property that is important is a metric property:

560: A flow is {\em quasigeodesic} if it is uniformly

561: efficient up to a bounded multiplicative

562: distortion in measuring distances in the universal cover.

563: This is extremely important for manifolds with

564: negatively curved fundamental group \cite{Th1,Th3,Gr}.

565: Our main result is the following:

566: 

567: 

568: \vskip .1in

569: \noindent

570: {\bf {Main theorem}} $-$ Let $\fol$ be a foliation in

571: $M^3$ closed, atoroidal.

572: Suppose that $\fol$ is almost

573: transverse to a quasigeodesic pseudo-Anosov flow $\Phi$,

574: which has some prong singularity (that is, not a topological

575: Anosov flow).

576: This implies that $M$ has

577: negatively curved

578: fundamental group. 

579: Then $\fol$ has the continuous extension property.

580: Therefore the limits sets of leaves of $\fn$ are locally 

581: connected.

582: The set of foliations almost transverse to a flow

583: is open in the Hirsch topology of foliations.

584: \vskip .1in

585: 

586: 

587: Notice that the hypothesis imply that $\fol$ is transversely

588: orientable.

589: Since $M$ has a singular pseudo-Anosov flow then

590: $M$ is irreducible and the stable/unstable foliations

591: of $\Phi$ split to genuine laminations in $M$.

592: A fundamental result of Gabai and Kazez \cite{Ga-Ka}

593: then implies that $M$ has negatively curved fundamental

594: group. For simplicity of statements we usually

595: use the group negative curvature hypothesis,

596: but in most places that could be substituted

597: by the atoroidal hypothesis.

598: 

599: %What is new in this result? The key hypothesis is

600: %that $\Phi$ is quasigeodesic. Here there are no restrictions

601: %on the metric behavior of the stable/unstable foliations

602: %The result also allows for almost transverse flows

603: %as opposed to transverse flows. This is necessary in

604: %important applications as the case of finite depth

605: %foliations shows, see below.

606: 

607: Notice that it is not necessary to assume 

608: that $\fol$ is

609: Reebless $-$ we prove that 

610: the condition of being almost transverse to a pseudo-Anosov

611: flow implies that $\fol$ is Reebless.

612: 

613: As a first consequence we prove the continuous

614: extension property for all Reebless finite depth foliations

615: in hyperbolic $3$-manifolds. 

616: Roughly a foliation is {\em finite depth} if all leaves

617: are proper and the leaves are perfectly fitted along

618: the cutting surfaces of a hierarchy of the manifold.

619: In particular there are

620: compact leaves.

621: These foliations exist whenever the second

622: homology is not finite.

623: 

624: \vskip .1in

625: \noindent

626: {\bf {Corollary A}} $-$ Let $\fol$ be a Reebless finite depth 

627: foliation in $M^3$ closed hyperbolic. Then $\fol$ has

628: the continuous extension property. In particular the

629: limit sets of the leaves are all locally connected.

630: \vskip .1in

631: 

632: 

633: 

634: Hence any orientable, hyperbolic $3$-manifold with non finite

635: second homology has such a foliation with the continuous

636: extension property. 

637: Conjecturally any closed, hyperbolic $3$-manifold

638: has a finite cover with positive first Betti number.

639: This would

640: imply there is always a foliation with the continuous extension

641: property in a finite cover.

642: The proof of corollary A is simple given previous results:

643: If necessary take a double cover and assume

644: that $\fol$ is transversely oriented. Then

645: Mosher and Gabai proved that such $\fol$ is almost transverse

646: to a pseudo-Anosov flow $\Phi$ \cite{Mo5}. 

647: We proved, jointly with Mosher that these

648: flows 

649: %almost transverse to finite depth

650: %foliations 

651: are quasigeodesic \cite{Fe-Mo}.

652: The main theorem then implies corollary A.

653: The result in \cite{Fe-Mo} is only for finite depth foliations: 

654: the proof depends heavily on the existence of a compact

655: leaf, $M$ being hyperbolic and the direct association with

656: a hierarchy.

657: %Here we assume that $M$ is hyperbolic as opposed to fundamental

658: %group being negatively curved.

659: %This is because, given that $\fol$ is

660: %finite depth, there is a compact leaf, which is an incompressible

661: %surface. The manifold is also irreducible \cite{Ros} and then

662: By Thurston's geometrization theorem 

663: \cite{Th1,Th2,Mor}

664: it suffices to assume that

665: $M$ is atoroidal.

666: 

667: 

668: 

669: \vskip .1in

670: \noindent

671: {\bf {Corollary B}} $-$ There are infinitely many 

672: foliations with all leaves dense which have the continuous

673: extension property. Many of these have one sided branching.

674: %These are not uniform foliations.

675: \vskip .1in

676: 

677: 

678: Foliations with all leaves dense can be obtained for example

679: starting with finite depth foliations and doing small perturbations 

680: $-$ keeping it still almost transverse to the same 

681: quasigeodesic pseudo-Anosov flow.

682: A construction is 

683: carefully explained by Gabai \cite{Ga3}, providing

684: infinitely many examples with dense leaves to which 

685: corollary B applies. 

686: The examples occur whenever the second Betti number

687: of $M$ is non zero.

688: In fact whenever a foliation $\fol$ satisfies the

689: hypothesis of the main theorem, then any $\fol'$ sufficiently

690: close to $\fol$ will also be transverse to the same

691: flow. By the main theorem again, 

692: $\fol'$  will have the continuous extension

693: property.

694: This perturbation feature of the main theorem is not

695: shared by any previous result

696: concerning the continuous extension property.

697: %, see

698: %the historical remarks section.

699: 

700: A foliation $\fol$ is ${\bf R}$-{\em covered} if 

701: the leaf space  of $\fn$ is homeomorphic to the

702: real numbers. Equivalently this leaf space is Hausdorff.

703: A foliation which is not $\rrrr$-covered has {\em branching},

704: that is there are non separated points in the leaf space.

705: This leaf space is oriented (being a simply connected, perhaps

706: non Hausdorff $1$-manifold) and there is a notion of branching

707: in the positive or negative directions. If it branches only

708: in one direction the foliation is said to have 

709: {\em one sided branching}.

710: Foliations with one

711: sided branching, where all leaves are dense and the

712: foliation is transverse to a suspension

713: pseudo-Anosov flow (which is quasigeodesic)  were constructed

714: by Meigniez \cite{Me}. This provides infinitely 

715: many examples with one sided 

716: branching to which corollary B applies.

717: 

718: 

719: A very important tool in the proof of the main theorem is

720: an analysis of the topological structure

721: of the pseudo-Anosov flow. 

722: Let $\Phi_1$ be the original pseudo-Anosov flow almost transverse

723: to $\fol$. To make the flow transverse to $\fol$ one

724: needs in general to blow up a collection of singular orbits

725: into a collection of flow saturated annuli so that

726: each boundary is a closed orbit of the new flow $\Phi$.

727: The blown up flow is called an {\em almost pseudo-Anosov flow}

728: (see section 3).

729: If $\wwp$ is the lifted flow to

730: the universal cover $\mi$ and $\oo$ is its orbit space, then

731: $\oo$ is homeomorphic to 

732: the plane $\rrrr^2$ \cite{Fe-Mo} $-$ this is true for

733: pseudo-Anosov and almost pseudo-Anosov flows. 

734: When one blows up some singular orbits into a collection of

735: joined annuli, the stable/unstable singular foliations also

736: blow up. The two new singular foliations $\Lambda^s, \Lambda^u$

737: are everywhere transverse

738: to each other

739: except at the singularities and the blown up annuli. The blown

740: up annuli are part of both singular foliations. 

741: Since

742: $\fol$ is transverse to the blown up foliations, then the

743: stable/unstable foliations $\Lambda^s, \Lambda^u$

744:  induce singular $1$-dimensional foliations

745: in the leaves of $\fol$ and hence also in the leaves

746: $\fn$. The behavior of this is

747: described in the following result, which is

748: of independent interest also: 

749: 

750: \vskip .1in

751: \noindent

752: {\bf {Theorem C}} $-$ Let $\fol$ be a foliation 

753: with hyperbolic leaves in $M^3$

754: closed. Let $\Phi_1$ be a pseudo-Anosov flow almost transverse

755: to $\fol$ and let $\Phi$ be a corresponding almost pseudo-Anosov flow

756: transverse to $\fol$. Let $\Lambda^s, \Lambda^u$ be

757: the stable/unstable $2$-dimensional

758: foliations of $\Phi$ and $\wls, \wlu$

759: the lifts to $\mi$. Given $F$ leaf of $\fn$,

760: let $\wlsf, \wluf$ be the induced singular 

761: $1$-dimensional foliations in $F$.

762: Then 

763: 

764: \begin{itemize}

765: 

766: \item

767: For every ray $l$ in a leaf of $\wlsf$ or $\wluf$,

768: then $l$ limits in a single point of $\pin F$. 

769: %The limit point usually depends on $l$.

770: 

771: \item

772: If 

773: the stable/unstable foliations $\wls, \wlu$ of $\Phi$ have 

774: Hausdorff leaf space, then the leaves of $\wlsf, \wluf$

775: are uniform quasigeodesics in $F$, the bound is independent

776: of the leaf. In general the leaves of $\wlsf, \wluf$ are not quasigeodesic.

777: 

778: \item

779: Any non Hausdorffness (of say $\wlsf$) 

780: is associated to a Reeb annulus

781: in a leaf of $\fol$ and when projected to $M$ it either projects

782: to or spirals

783: to a Reeb annulus. 

784: 

785: \item

786: The set of ideal points of

787: leaves of $\wlsf$ is dense in $\pin F$ and similarly for $\wluf$.

788: 

789: \item

790: If two rays of the same leaf of $\wlsf$ limit to the same ideal

791: point in $\pin F$ then this leaf of 

792: $\wlsf$ does not contain a singularity and the region in $F$

793: bounded by the leaf projects in $M$ to a set in a leaf

794: of $\fol$ which is either contained in or asymptotic to

795: a Reeb annulus.

796: \end{itemize}

797: 

798: \vskip .1in

799: This is a completely general result:

800: one does not need 

801: negatively curved fundamental group of $M$ or

802: any metric properties of $\Phi$.

803: Theorem C is a key tool used in

804: the proof of the main theorem.

805: 

806: 

807: 

808: The article is organized as follows:

809: In the next section we discuss previous results

810: on the continuous extension property and the strategy

811: of the proof of the main theorem and theorem C.

812: In section 3 we present basic definitions and

813: results concerning pseudo-Anosov flows and almost

814: pseudo-Anosov flows. In section 4 we analyse 

815: projections to the orbit space.

816: %set $\Theta(F)$ for leaves in the universal cover.

817: In sections 5 and 6 we analyse the singular foliations

818: $\wlsf, \wluf$ and asymptotic properties of their

819: leaves, proving theorem C.

820: In section 7 we prove the continuous extension property

821: $-$ the main theorem.

822: In the final section we comment on general relationships

823: between foliations and Kleinian groups.

824: 

825: In a subsequent article we analyse other important

826: consequences of quasigeodesic behavior for flows

827: and foliations \cite{Fe10}.

828: 

829: 

830: 

831: 

832: 

833: %Part of this article was done while the author was

834: %visiting Princeton University. The author thanks this 

835: %institution for its hospitality and in particular

836: %David Gabai.

837: 

838: 

839: \section{Historical remarks and strategy of proofs}

840: 

841: 

842: Here we review what is known about the continuous extension

843: property.

844: In a seminal work, Cannon and Thurston \cite{Ca-Th} proved

845: this property 

846: when $\fol$ is a fibration over

847: the circle. Previously Thurston showed that a fiberting 

848: manifold is hyperbolic when the monodromy of the fibration

849: is pseudo-Anosov \cite{Th1,Th3,Th4,Bl-Ca}. Since the fundamental group

850: of a leaf of $\fol$ is a 

851: normal subgroup of the fundamental group of $M$,

852: then every limit set of a leaf of $\fn$ is the whole

853: sphere. In this way Cannon and Thurston produced many examples of 

854: group invariant Peano curves.

855: 

856: Another extremely important class of foliations is the

857: following:

858: A foliation is {\em proper} if the  leaves never limit

859: on themselves $-$ this is in the foliation sense and

860: it means that a sufficiently small transversal to a given 

861: leaf only

862: meets the leaf in a single point.

863: In particular leaves are not dense.

864: In a proper foliation there are compact leaves which

865: are said to have {\em depth} $0$. The depth of a leaf is

866: inductively defined to be $i$ (for finite $i$) if $i-1$ is the maximum

867: of the depths of leaves in the (foliation) limit set of

868: the leaf. A foliation has {\em finite depth} if it is

869: proper and there is a finite upper bound to the depths

870: of all leaves.

871: 

872: Gabai proved that whenever a compact $3$-manifold $M$ is 

873: irreducible, orientable  and

874: the second homology group $H_2(M, \partial M, {\bf Z})$ is not 

875: finite,

876: then there is a Reebless finite depth, foliation associated

877: to each non torsion homology class \cite{Ga1,Ga3}. 

878: The foliation is directly associated to a hierarchy of the

879: manifold and as such is strongly connected with the topological

880: structure of the manifold.

881: These results had several fundamental consequences for the

882: topology of $3$-manifolds \cite{Ga1,Ga2,Ga3}.

883: If $M$ is hyperbolic (or atoroidal), then

884: one important question is whether these finite depth

885: foliations have the continuous extension property.

886: %Conjecturally every closed hyperbolic $3$-manifold

887: %has a finite cover with infinite homology, to which these

888: %constructions apply.

889: 

890: %We proved  that if $\fol$ has depth one and the monodromy

891: %is irreducible (similar to the fibration case and it uses

892: %end periodic homeomorphims of non compact surfaces), then

893: %$\fol$ has the continuous extension property \cite{Fe1}.

894: 

895: Subsequently 

896: Gabai and Mosher showed \cite{Mo5} that any Reebless finite

897: depth foliation in a closed, atoroidal $3$-manifold admits

898: a pseudo-Anosov flow $\Phi$ which is almost transverse

899: to it. Roughly a flow is {\em pseudo-Anosov} if it has transverse

900: hyperbolic dynamics $-$ even though it may have

901: finitely many singularities.

902: It has stable and unstable two dimensional foliations which in

903: general are singular.

904: The term {\em almost transverse} means that one may need to 

905: blow up one singular orbit (or more) into a finite

906: collection of joined annuli to make the flow transverse

907: to the foliation. 

908: See detailed definitions and comments in section 3.

909: Since $\fol$ has a compact leaf and $M$ is

910: atoroidal, then Thurston \cite{Th1,Th3} proved

911: that $M$ is in fact hyperbolic.

912: %See more about pseudo-Anosov flows transverse to

913: %foliations below.

914: 

915: We proved, jointly with Mosher,

916: that these pseudo-Anosov flows almost transverse to finite depth 

917: foliations

918: in hyperbolic $3$-manifolds are {\em quasigeodesic} \cite{Fe-Mo}.

919: This means that flow lines are uniformly efficient in

920: measuring distance in relative homotopy classes, or 

921: equivalently, uniformly efficient in measuring distance

922: in the universal cover.

923: This was first proved by Mosher \cite{Mo4} for a class

924: of flows transverse to some examples of depth one foliations

925: obtained by handle constructions.

926: Another concept is that of quasi-isometric behavior:

927: a foliation (perhaps singular) is {\em quasi-isometric} if

928: its leaves are uniformly efficient in measuring distance

929: in the universal cover. There are no non singular $2$ dimensional

930: quasi-isometric foliations in closed $3$-manifolds

931: with negatively curved fundamental group

932: \cite{Fe1}. 

933: As for {\underline {singular}} foliations the situation is

934: quite different and there are examples.

935: The stable/unstable singular foliations of the quasigeodesic

936: flows above may be quasi-isometric \cite{Fe6} and may

937: not \cite{Mo5,Fe6}. 

938: In general quasi-isometric behavior of $\Lambda^s$ (or $\Lambda^u$)

939: implies that $\Phi$ is quasigeodesic.

940: %It depends on whether the leaf spaces

941: %of the foliations (in the universal cover) are Hausdorff or

942: %not \cite{Fe4}.

943: 

944: 

945: If both the stable and unstable foliations are quasi-isometric

946: {\underline {and}}

947:  the flow is actually transverse (as opposed to being

948: almost transverse) to the finite depth foliation then

949: we proved in \cite{Fe6} that $\fol$ has the continuous extension

950: property.

951: We stress that this result only applies to finite depth foliations

952: $-$ the proof depends, amongst other things,  on induction in the depth.

953: % (see 

954: %theorem D below for a more general result).

955: To use this result we needed to check

956: the quasi-isometric behavior of $\Lambda^s, \Lambda^u$ and

957: transversality between $\fol$ and $\Phi$.

958: This was very tricky and we could only do that for some 

959: {\underline {depth one}} foliations.

960: More to the point, it is known that  these conditions do not

961: always hold for general finite depth foliations.

962: %As remarked before, in general

963: %both conditions may fail. 

964: Corollary A proves the continuous extension property

965: for all finite depth foliations: there are no restrictions

966: on the depth of the foliation, or about transversality

967: of the flow with the foliation or quasi-isometric behavior

968: of $\Lambda^s, \Lambda^u$.

969: 

970: 

971: \vskip .11in

972: The continuous extension property was also proved for

973: another class of foliations:

974: %We mention another class of foliations.

975: %A {\em slithering} of $M$ is a submersion map from the universal cover

976: %to the circle which is group equivariant. As such it

977: %defines a foliation in the manifold. This foliation $\fol$

978: %is $\rrrr$-covered, that is, the leaf space of $\fn$ is

979: %%homeomorphic to the real numbers. This is equivalent

980: %to having Hausdorff leaf space. 

981: %There are many examples

982: %of Anosov flows in hyperbolic $3$-manifolds, so that

983: %the stable and unstable foliations (non singular in this case)

984: %are $\rrrr$-covered \cite{Fe2}.

985: %Up to collapsing pockets of leaves 

986: %the slithering property is equivalent

987: A foliation $\fol$ is {\em uniform} if any two leaves

988: in the universal cover are a bounded distance apart,  the

989: bound depends on the individual leaves. 

990: Hence $\fol$ is $\rrrr$-covered.

991: Thurston \cite{Th5} proved

992: that there is a large class of

993: uniform foliations.

994: %$\rrrr$-covered Anosov foliations, branched covering

995: %constructions, etc.. 

996: %These foliations are ${\bf R}$-{\em covered}, that is,

997: %the leaf space  of $\fn$ is homeomorphic to the

998: %real numbers. Equivalently this leaf space is Hausdorff.

999: If in addition

1000: $\pi_1(M)$ is negatively

1001: curved, then 

1002: Thurston \cite{Th5} proved 

1003: there is a pseudo-Anosov flow

1004: transverse to $\fol$. From this  it is easy to 

1005: prove  that the flow has quasi-isometric stable/unstable

1006: foliations. In this case it also easily implies that the

1007: foliation $\fol$ has

1008: the continuous extension property.

1009: The arguments are a very clever generalization of the

1010: fibering situation.

1011: 

1012: %At this point it is important to stress 

1013: Notice that in all the previous

1014: results, there is a pseudo-Anosov flow $\Phi$

1015: {\underline {transverse}} to $\fol$ and so that the

1016: stable/unstable foliations of $\Phi$ are 

1017: quasi-isometric singular foliations.

1018: Both of these properties were crucial in all proofs.

1019: The main theorem implies all the previous results

1020: about the continuous extension property and

1021: it has the unique feature that it applies

1022: to an open set of foliations.

1023: 

1024: 

1025: \vskip .12in

1026: The main theorem 

1027: can potentially be widely applicable because of

1028: the abundance of pseudo-Anosov flows almost transverse

1029: to foliations: Thurston proved this for fibrations \cite{Th4}.

1030: It is also true for all

1031: $\rrrr$-covered foliations \cite{Fe7,Cal1}

1032: and Calegari proved it for all foliations with 

1033: one sided branching \cite{Cal2}, all minimal foliations \cite{Cal3}

1034: and many other foliations \cite{Cal3}.

1035: The missing ingredient is the quasigeodesic property of

1036: these pseudo-Anosov flows which is needed to apply

1037: the main theorem.

1038: In general the quasigeodesic property for a pseudo-Anosov

1039: flow (or an arbitrary flow) is very hard to obtain.

1040: This property is only known when the pseudo-Anosov

1041: flow is almost transverse to

1042: a foliation of one of the following types: 

1043: finite depth or uniform.

1044: One main goal in the study of a pseudo-Anosov flow $\Phi$ in

1045: $M$ with Gromov hyperbolic fundamental group is to decide

1046: whether $\Phi$ is quasigeodesic.

1047: There are many examples where it is not quasigeodesic \cite{Fe2}.

1048: 

1049: \vskip .1in

1050: Why almost transversality and not just transversality? 

1051: In many cases ($\rrrr$-covered, 

1052: one sided branching)  the pseudo-Anosov flow is actually

1053: transverse to $\fol$.

1054: But for finite depth foliations (which have

1055: two sided branching), there are many examples where

1056: it is impossible to make the pseudo-Anosov flow transverse

1057: to the foliation \cite{Mo5} and one can only get almost

1058: transversality. We will have more comments about this

1059: in section 3.

1060: Finite depth foliations are extremely important as they

1061: are strongly connected to the topology of the underlying

1062: manifold.

1063: Also, in some sense, foliations with two sided branching

1064: are probably more common than foliations which are

1065: either $\rrrr$-covered or with

1066: one sided branching. 

1067: 

1068: 

1069: 

1070: 

1071: We also remark that 

1072: in all previous results concerning the continuous extension

1073: property, theorem C was 

1074: a crucial property on which the whole analysis hinged.

1075: %For uniform foliations this is a very easy fact.

1076: %There is an 

1077: %analysis of $\wlsf, \wluf$ in \cite{Fe6} for 

1078: %finite depth foliations, with $\Lambda^s, \Lambda^u$ quasi-isometric

1079: %and the original flow $\Phi_1$ transverse to $\fol$.

1080: %Amongst other things the analysis is inductive on the depth

1081: %and so of course can only work for finite depth foliations.

1082: %In addition in that case one actually proves that

1083: %the leaves of $\wlsf, \wluf$ are uniform quasigeodesics and

1084: In the previous situations, the analysis of

1085: leaves of $\wlsf, \wluf$ was either trivial or 

1086: substantially easier.

1087: In particular in these situations the leaves of

1088: $\wlsf, \wluf$ were always uniform quasigeodesics, which

1089: simplified subsequent proofs considerably. Such is

1090: not the case in general.

1091: In particular if $\wls, \wlu$ do not have Hausdorff leaf

1092: space, then a priori the leaves of $\wlsf, \wluf$ do not have

1093: to be quasigeodesic.

1094: %this certainly is not true in many other situations.

1095: The proof of theorem $C$ works for any pseudo-Anosov

1096: flow almost transverse to a foliation with hyperbolic

1097: leaves.

1098: The proof uses the

1099: denseness of contracting directions for foliations

1100: as proved by Thurston \cite{Th6,Th7} when he introduced the

1101: universal circle for foliations $-$  even though we do not

1102: directly use the universal circle here.

1103: The basic idea is: if any ray 

1104: of (say) $\wlsf$ does not limit to a single

1105: point  in $\pin F$

1106: then it limits in a non trivial interval of $\pin F$.

1107: Zoom into this interval and analyse the situation 

1108: in the limit. This is actually the easiest statement to prove

1109: in theorem C.

1110: The facts about rays with same ideal point

1111: and non Hausdorffness are much trickier, but they will be essential

1112: in the analysis of the main theorem.

1113: The results of theorem C are also used in other contexts,

1114: for example to analyse rigidity of pseudo-Anosov flows

1115: almost transverse to a given foliation. This will be

1116: explored in a future article \cite{Fe10}.

1117: 

1118: 

1119: The proof of the main theorem has 2 parts: given a leaf $F$ of $\fn$,

1120: one first constructs an extension to the ideal boundary and

1121: then show it is continuous. To define the extension, one

1122: uses the foliations $\wlsf, \wluf$ as they hopefully

1123: define a basis neighborhood of an ideal point $p$ of $F$.

1124: The best situation is that the leaves

1125: of $\wls, \wlu$ which contain these leaves

1126: of $\wlsf, \wluf$ define basis neighborhoods of unique

1127: points in $\si$, hence defining the image of $p$ in $\si$.

1128: There are several difficulties here: first the leaves

1129: of $\wlsf, \wluf$ are not quasigeodesics in $F$, so much

1130: more care is needed. Another problem is that

1131: the foliations $\wls, \wls$ in general do not have

1132: Hausdorff leaf space. This keeps recurring

1133: throughout the proof. A further difficulty is that if

1134: intersections of leaves $L_i$ of say

1135: $\wls$ with a leaf $F$ of $\fn$ escape

1136: in $F$ as $i$ converges to infinity, it

1137: does not follow that the $L_i$ themselves escape compact

1138: sets in $\mi$.

1139: Consequently there

1140: are several cases to be analysed.

1141: 

1142: Another fact that is important for the analysis of the main

1143: theorem and theorem C is the following:

1144: Let $\Theta$ be the projection map from $\mi$ to $\oo$

1145: (projects flow lines to points).

1146: A leaf of $\fn$ 

1147: intersects an orbit of $\wwp$ at most once defining an injective

1148: projection of $F$ to $\Theta(F)$. The projection $\Theta(F)$ is equal

1149: to $\oo$ if and only if the foliation is $\rrrr$-covered.

1150: An important problem here is to determine the boundary 

1151: $\Theta(F)$ as a subset of $\oo$. This turns out to be

1152: a collection of subsets of stable/unstable leaves in $\oo$.

1153: This result is different than what happens

1154: for pseudo-Anosov flows transverse to foliations

1155: and its proof is much more delicate.

1156: This is analsyed in section 4.

1157: 

1158: 

1159: 

1160: \section{Preliminaries: Pseudo-Anosov flows and almost pseudo-Anosov flows}

1161: 

1162: 

1163: 

1164: %Pseudo-Anosov flows are a generalization of suspension flows of

1165: %pseudo-Anosov surface homeomorphisms. These flows behave much like

1166: %Anosov flows, but they may have finitely many singular orbits 

1167: %which are periodic and have 

1168: %a prescribed behavior. 

1169: 

1170: 

1171: Let $\Phi$ be a flow on a closed, oriented 3-manifold $M$. We say

1172: that $\Phi$ is a {\em pseudo-Anosov flow} if the following are

1173: satisfied:

1174: 

1175: %\begin{itemize}

1176: 

1177: %\item 

1178: - For each $x \in M$, the flow line $t \to \Phi(x,t)$ is $C^1$,

1179: it is not a single point,

1180: and the tangent vector bundle $D_t \Phi$ is $C^0$.

1181: 

1182: %\item 

1183: - There is a finite number of periodic orbits $\{ \gamma_i \}$,

1184: called {\em singular orbits}, such that the flow is 

1185: ``topologically" smooth off of the

1186: singular orbits (see below).

1187: 

1188: - The flowlines are tangent to two singular transverse

1189: foliations $\ls, \lu$ which have smooth leaves off of $\gamma_i$

1190: and intersect exactly in the flow lines of $\Phi$.

1191: These are like Anosov foliations off of the singular orbits.

1192: This is the topologically smooth behavior described above.

1193: A leaf containing a singularity is homeomorphic 

1194: to $P \times I/f$

1195: where $P$ is a $p$-prong in the plane and $f$ is a homeomorphism

1196: from $P \times \{ 1 \}$ to $P \times \{ 0 \}$.

1197: In a stable leaf, $f$ contracts towards towards

1198: the prongs and in an unstable leaf it expands away

1199: from the prongs.

1200: We restrict to $p$ at least $2$, that is, we do not allow

1201: $1$-prongs.

1202: 

1203: - In a stable leaf all orbits are forward asymptotic,

1204: in an unstable leaf all orbits are backwards asymptotic.

1205: 

1206: 

1207: Basic references for pseudo-Anosov flows are \cite{Mo3,Mo5}

1208: and for $3$-manifolds \cite{He}.

1209: 

1210: 

1211: \vskip .05in

1212: \noindent

1213: \underline {Notation/definition:} \ 

1214: The singular

1215: foliations lifted to $\mi$ are

1216: denoted by $\wls, \wlu$.

1217: If $x \in M$ let $W^s(x)$ denote the leaf of $\ls$ containing

1218: $x$.  Similarly one defines $W^u(x)$

1219: and in the

1220: universal cover $\ws(x), \wu(x)$.

1221: Similarly if $\alpha$ is an orbit of $\Phi$ define

1222: $W^s(\alpha)$, etc...

1223: Let also $\wwp$ be the lifted flow to $\mi$.

1224: 

1225: 

1226: 

1227: 

1228: \vskip .1in

1229: %Let $\Phi$ be a pseudo-Anosov flow in $M^3$ closed.

1230: We review the results about the topology of

1231: $\wls, \wlu$ that we will need.

1232: % which will be needed in the following

1233: %sections to prove the main theorem.

1234: We refer to \cite{Fe4,Fe6} for detailed definitions, explanations and 

1235: proofs.

1236: The orbit space of $\wwp$ in

1237: $\mi$ is homeomorphic to the plane $\rrrr^2$ \cite{Fe-Mo}

1238: and is denoted by $\oo \cong \mi/\wwp$. 

1239: Let $\Theta: \mi \rightarrow \oo \cong \rrrr^2$

1240: be the projection map. 

1241: %As the foliations $\fns, \fnu$

1242: %are invariant under $\wwp$, they 

1243: %induce singular, transverse $1$-dim foliations

1244: %$\fnso, \fnuo$ in $\oo$.

1245: %The singular points  of $\fnuo$ are the same as those

1246: %of $\fnso$.

1247: If $L$ is a 

1248: leaf of $\wls$ or $\wlu$,

1249: then $\Theta(L) \subset \oo$ is a tree which is either homeomorphic

1250: to $\rrrr$ if $L$ is regular,

1251: or is a union of $p$-rays all with the same starting point

1252: if $L$ has a singular $p$-prong orbit.

1253: %In particular every orbit in $L$ disconnects $L$.

1254: The foliations $\wls, \wlu$ induce $1$-dimensional foliations

1255: $\oos, \oou$ in $\oo$. Its leaves are $\Theta(L)$ as

1256: above.

1257: If $L$ is a leaf of $\wls$ or $\wlu$, then 

1258: a {\em sector} is a component of $\mi - L$.

1259: Similarly for $\oos, \oou$. 

1260: If $B$ is any subset of $\oo$, we denote by $B \times \rrrr$

1261: the set $\Theta^{-1}(B)$.

1262: The same notation $B \times \rrrr$ will be used for

1263: any subset $B$ of $\mi$: it will just be the union

1264: of all flow lines through points of $B$.

1265: 

1266: 

1267: \begin{define}

1268: Let $L$ be a leaf of $\wls$ or $\wlu$. A slice of $L$ is 

1269: $l \times \rrrr$ where $l$ is a properly embedded

1270: copy of the reals in $\Theta(L)$. For instance if $L$

1271: is regular then $L$ is its only slice. If a slice

1272: is the boundary of a sector of $L$ then it is called

1273: a line leaf of $L$.

1274: If $a$ is a ray in $\Theta(L)$ then $A = a \times \rrrr$

1275: is called a half leaf of $L$.

1276: %The closure is denoted by $\overline A = A \cup \gamma$ and 

1277: %its boundary is $\partial A = \gamma$.

1278: If $\zeta$ is an open segment in $\Theta(L)$ 

1279: it defines a {\em flow band} $L_1$ of $L$

1280: by $L_1 = \zeta \times \rrrr$.

1281: %\Theta^{-1}(\zeta)$.

1282: %Let $\overline L_1$ be the closure of $L_1$ in $\mi$.

1283: %If $\zeta$ is an open segment in $\Theta(L)$, then $\Theta^{-1}(\zeta)$

1284: %is called a segment flow band of $L$.

1285: Same notation for the foliations $\oos, \oou$ of $\oo$.

1286: \end{define}

1287: 

1288: 

1289: If $F \in \wls$ and $G \in \wlu$ 

1290: then $F$ and $G$ intersect in at most one

1291: orbit. 

1292: Also suppose that a leaf $F \in \wls$ intersects two leaves

1293: $G, H \in \wlu$ and so does $L \in \wls$.

1294: Then $F, L, G, H$ form a {\em rectangle} in $\mi$

1295: and there is  no singularity 

1296: in the interior of the rectangle \cite{Fe6}.

1297: There will be two generalizations of rectangles: 1) perfect fits $=$ rectangle

1298: with one corner removed 

1299: and 2) lozenges $=$ rectangle with two opposite corners removed.

1300: We will also denote by rectangles, perfect fits, lozenges

1301: and product regions the projection of these regions to

1302: $\oo \cong \rrrr^2$. 

1303: 

1304: 

1305: \begin{define}{(\cite{Fe2,Fe4,Fe6})}{}

1306: Perfect fits -

1307: Two leaves $F \in \wls$ and $G \in \wlu$, form

1308: a perfect fit if $F \cap G = \emptyset$ and there

1309: are half leaves $F_1$ of $F$ and $G_1$ of $G$ 

1310: and also flow bands $L_1 \subset L \in \wls$ and

1311: $H_1 \subset H \in \wlu$,

1312: so that 

1313: %$F_0$ is regular on the side containing $L$,

1314: %$G_0$ is regular on the side containing $H$ and:

1315: %

1316: %$$ \overline L_1 \cap \overline G_1 = \partial L_1 \cap \partial G_1,

1317: %\ \ \overline L_1 \cap \overline H_1 = \partial L_1 \cap \partial H_1,

1318: %\ \ \overline H_1 \cap \overline F_1 = 

1319: %\partial H_1 \cap \partial F_1,$$ 

1320: the set 

1321: 

1322: $$\overline F_1 \cup \overline H_1 \cup 

1323: \overline L_1 \cup \overline G_1$$

1324: 

1325: \noindent

1326: separates $M$ and forms an a rectangle $R$ with a corner removed:

1327: The joint structure of $\wls, \wlu$ in $R$ is that of

1328: a rectangle with a corner orbit removed. The removed corner

1329: corresponds to the perfect of $F$ and $G$ which 

1330: do not intersect.

1331: %`$${\rm with} \ \ \  \overline L_1 \cap \overline G_1 \ \not = \ \emptyset, 

1332: %\ \ \overline L_1 \cap \overline H_1  \ \not = \ \emptyset \ \ {\rm and}

1333: %\ \ \overline H_1 \cap \overline F_1 \ \not = \ \emptyset.$$

1334: %\noindent

1335: %Furthermore

1336: %$$\forall \ S \in \wlu, \ \ \ \   

1337: %S \cap L_1  \not = \emptyset \ \Rightarrow

1338: %S \cap F_1  \not = \emptyset \ \ \ (1) $$

1339: %$${\rm and} \ \ \forall \ E \in \wls, \ \ \ \ E \cap H_1  

1340: %\not = \emptyset \ \Rightarrow

1341: %E \cap G_1  \not = \emptyset \ \ \ (2).$$

1342: \end{define}

1343: 

1344: 

1345: 

1346: 

1347: We refer to fig. \ref{loz}, a for perfect fits.

1348: %Implications $(1), (2)$ force

1349: %equivalences (that is 

1350: %$S \cap L_1  \not = \emptyset \ \Leftrightarrow

1351: %S \cap F_1  \not = \emptyset$ and the same for (2)).

1352: %The set 

1353: %$\overline F_1 \cup \overline H_1 \cup 

1354: %\overline L_1 \cup \overline G_1$

1355: %separates $\mi$.

1356: %Let $A$ be the complementary region which does

1357: %not contain $F - F_1$ in its closure.

1358: %An important fact is that there are 

1359: %singularities of $\wwp$

1360: %in $A$.

1361: %Perfect fits produce ``ideal" rectangles, in the sense that

1362: %even though $F$ and $G$ do not intersect, there is

1363: %a product structure (of $\wls$ and $\wlu$) in the 

1364: %interior of $A$.

1365: There is a product structure in the interior of $R$: there are

1366: two stable boundary sides and two unstable one. An unstable

1367: leaf intersects one stable boundary side (not in the corner) if

1368: and only if it intersects the other stable boundary side

1369: (not in the corner).

1370: We also say that the leaves $F, G$ are {\em asymptotic}.

1371: 

1372: 

1373: %\begin{define}{\cite{Fe1,Fe4}}{}

1374: %Given $p \in \mi$ (or $p \in \oo$), and a half leaf $H$ 

1375: %of $\wu(p)$ defined by $\wwp_{\rrrr}(p)$, let 

1376: %

1377: %$$\uo(H) \ = \ \{ F \in \hs  \ | \ F \cap 

1378: %H \not = \emptyset \} \ \subset \ \hs.$$

1379: %

1380: %\noindent

1381: %Notice that $\ws(p) \not \in \uo(H)$.

1382: %Let  also

1383: %

1384: %$${\cal L}^u(H)  \ = 

1385: %\ \bigcup  \ \{ \ p \ \in \mi \ \ 

1386: %| \ \ p \ \in \ F \ \in \ \uo(H) \ \} \ \ \subset \ \ \mi.$$

1387: %

1388: %\noindent

1389: %Then ${\cal L}^u(H) \subset \mi$

1390: %and $\ws(p) \subset \partial {\cal L}^u(H)$.

1391: %Similarly define

1392: %$\so(L),

1393: %{\cal L}^s(L)$ for a stable half leaf $L$.

1394: %\end{define}

1395: 

1396: 

1397: \begin{figure}

1398: \centeredepsfbox{gm1.eps}

1399: \caption{a. Perfect fits in $\mi$,

1400: b. A lozenge, c. A chain of lozenges.}

1401: \label{loz}

1402: \end{figure}

1403: 

1404: 

1405: 

1406: 

1407: 

1408: \begin{define}{(\cite{Fe2,Fe4,Fe6})}{}

1409: Lozenges - A lozenge is a region of $\mi$ whose closure

1410: is homeomorphic to a rectangle with two corners removed.

1411: More specifically two points $p, q$ form the corners

1412: of a lozenge if there are half leaves $A, B$ of

1413: $\ws(p), \wu(p)$ defined by $p$

1414: and  $C, D$ half leaves of $\ws(q), \wu(q)$ so

1415: that $A$ and $D$ form a perfect fit and so do

1416: $B$ and $C$. The region bounded by the lozenge

1417: is $R$ and it does not have any singularities.

1418: %Let $p, q \in \mi$ and half leaves

1419: %$L_p, H_p$ of $\ws(p), \wu(p)$ defined by $\wwp_{\rrrr}(p)$,

1420: %half leaves $L_q, H_q$ of $\ws(q), \wu(q)$

1421: %defined by $\wwp_{\rrrr}(q)$

1422: %%so that:

1423: %

1424: %

1425: %$${\cal L}^u(L_p) \ \cap \ 

1426: %{\cal L}^s(H_q) \ \  = \ \ 

1427: %{\cal L}^u(L_q) \ \cap \ 

1428: %{\cal L}^s(H_p) \ \ \subset \ \mi$$

1429: %

1430: %

1431: %\noindent

1432: %Then this intersection

1433: %is called a lozenge ${\cal A}$

1434: %in $\mi$, see fig. \ref{loz}, b.

1435: %The corners of the lozenge are $\wwp_{\rrrr}(p)$ and

1436: %$\wwp_{\rrrr}(q)$ and ${\cal A}$ 

1437: %is a subset of $\mi$.

1438: %The sides of ${\cal A}$ are

1439: %$L_p, H_p, L_q, H_q$.

1440: The sides are not contained in the lozenge,

1441: but are in the boundary of the lozenge.

1442: See fig. \ref{loz}, b.

1443: \end{define}

1444: 

1445: 

1446: %Sometimes we also refer to $p$ and $q$ as corners of the lozenge.

1447: 

1448: There are no singularities in the lozenges,

1449: which implies that

1450: $R$ is an open region in $\mi$.

1451: There may be singular orbits

1452: on the sides of the lozenge and the corner orbits.

1453: %also may be singular.

1454: %The definition of a lozenge implies that

1455: %$L_p, H_q$ form a perfect fit and so do 

1456: %$L_q, H_p$.

1457: %This is an equivalent way to define a lozenge

1458: %with corners $\wwp_{\rrrr}(p), \wwp_{\rrrr}(q)$.

1459: 

1460: %Intuitively a lozenge is a rectangle with two 

1461: %opposite corners turned into ideal points:

1462: %each removed corner corresponds to a pair of leaves

1463: %(the sides of the lozenge) forming a perfect fit.

1464: 

1465: Two lozenges are {\em adjacent} if they share a corner and

1466: there is a stable or unstable leaf

1467: intersecting both of them, see fig. \ref{loz}, c.

1468: Therefore they share a side.

1469: A {\em chain of lozenges} is a collection $\{ \cc _i \}, 

1470: i \in I$, where $I$ is an interval (finite or not) in ${\bf Z}$;

1471: so that if $i, i+1 \in I$, then 

1472: ${\cal C}_i$ and ${\cal C}_{i+1}$ share

1473: a corner, see fig. \ref{loz}, c.

1474: Consecutive lozenges may be adjacent or not.

1475: The chain is finite if $I$ is finite.

1476: 

1477: 

1478: \begin{define}{}{}

1479: Suppose $A$ is a flow band in a leaf of $\wls$.

1480: Suppose that for each orbit $\gamma$ of $\wwp$ in $A$ there is a

1481: half leaf $B_{\gamma}$ of $\wu(\gamma)$ defined by $\gamma$ so that: 

1482: for any two orbits $\gamma, \beta$ in $A$ then

1483: a stable leaf intersects $B_{\beta}$ if and only if 

1484: it intersects $B_{\gamma}$.

1485: %for any 

1486: %\zeta \subset E \in \wls$ is a (possibly infinite) 

1487: %strong stable segment 

1488: %so that for each $p \in \zeta$ there is a half leaf $H_p$

1489: %of $\wu(p)$ defined by $\wwp_{\rrrr}(p)$ so that

1490: %

1491: %$$ \forall \ p, q \in \zeta, \ \ \uo(H_p) = \uo(H_q).

1492: %\ \ \  In \ that \ case \ let \ {\cal P} = \bigcup _{p \in \zeta} 

1493: %H_p.$$

1494: %

1495: %\noindent

1496: %Then ${\cal P} \subset \mi$ is called an unstable product

1497: %region with base segment $\zeta$. 

1498: %The base segment

1499: %is not uniquely determined by ${\cal P}$.

1500: %Similarly define negative unstable product regions 

1501: This defines a stable product region which is the union

1502: of the $B_{\gamma}$.

1503: Similarly define unstable product regions.

1504: \label{defsta}

1505: \end{define}

1506: 

1507: 

1508: %The main property of product regions is the following:

1509: %for any $F \in \wls$, $G \in \wlu$ so that 

1510: %$(i) \ F \cap A  \

1511: %\not = \ \emptyset \ \ {\rm  and} \ \ 

1512:  %(ii) \ G \cap A \ \not = \ \emptyset,

1513: %\ \ \ {\rm then} \ \ 

1514: %F \cap G \ \not = \ \emptyset$.

1515: There are no singular orbits of 

1516: $\wwp$ in $A$.

1517: 

1518: %One good way to visualize these objects in 

1519: %$\oo$ is  as follows.

1520: %Consider proper embeddings $\xi:U \rightarrow \oo$ of sets $U \subset 

1521: %\rrrr^2$ into $\oo$, sending the horizontal and vertical

1522: %foliations induced in $U$ to the stable and unstable

1523: %foliations in $\xi(U) \subset \oo$. Then

1524: %a proper embedding is associated to a 

1525: %rectangle $\xi(U)$ if $U = [0,1] \times [0,1]$.

1526: %A proper embedding is associated to a perfect fit

1527: %if $U$ is a rectangle without a

1528: %corner, that is, $U = [0,1] \times [0,1] - \{ (1,1) \}$. 

1529: %A lozenge is associated to the image

1530: %of a rectangle without two opposite

1531: %corners $U = [0,1] \times [0,1] - \{ (1,1), (0,0) \}$

1532: %(the lozenge is the interior of $\xi(U)$).

1533: %A stable product region is associated to the image 

1534: %of $U = [a,b] \times [0,\infty)$ (or $U = \rrrr \times [0,\infty)$

1535: %when the base segment is infinite)

1536: %and similarly for an unstable product region.

1537: %The important fact here is that there are no

1538: %singular orbits in any of these regions.

1539: %

1540: 

1541: We abuse convention and call 

1542: a leaf $L$ of $\wls$ or $\wlu$ is called {\em periodic}

1543: if there is a non trivial covering translation

1544: $g$ of $\mi$ with $g(L) = L$. This is equivalent

1545: to $\pi(L)$ containing a periodic orbit of $\Phi$.

1546: In the same way an orbit 

1547: $\gamma$ of $\wwp$

1548: is {\em periodic} if $\pi(\gamma)$ is a periodic orbit

1549: of $\Phi$.

1550: 

1551: 

1552: We say that two orbits $\gamma, \alpha$ of $\wwp$ 

1553: (or the leaves $\ws(\gamma), \ws(\alpha)$)

1554: are connected by a 

1555: chain of lozenges $\{ {\cal C}_i \}, 1 \leq i \leq n$,

1556: if $\gamma$ is a corner of ${\cal C}_1$ and $\alpha$ 

1557: is a corner of ${\cal C}_n$.

1558: 

1559: If ${\cal C}$ is a lozenge with corners $\beta, \gamma$ and

1560: $g$ is a non trivial covering translation 

1561: leaving $\beta, \gamma$ invariant (and so also the lozenge),

1562: then $\pi(\beta), \pi(\gamma)$ are closed orbits

1563: of $\wwp$ which are freely homotopic to the inverse of each

1564: other.

1565: 

1566: \begin{theorem}{\cite{Fe4,Fe6}}{}

1567: Let $\Phi$ be a pseudo-Anosov flow in $M^3$ closed and let 

1568: $F_0 \not = F_1 \in \wls$.

1569: Suppose that there is a non trivial covering translation $g$

1570: with $g(F_i) = F_i, i = 0,1$.

1571: Let $\alpha_i, i = 0,1$ be the periodic orbits of $\wwp$

1572: in $F_i$ so that $g(\alpha_i) = \alpha_i$.

1573: Then $\alpha_0$ and $\alpha_1$ are connected

1574: by a finite chain of lozenges 

1575: $\{ {\cal C}_i \}, 1 \leq i \leq n$ and $g$

1576: leaves invariant each lozenge 

1577: ${\cal C}_i$ as well as their corners.

1578: \label{chain}

1579: \end{theorem}

1580: 

1581: A chain from $\alpha_0$ to $\alpha_1$ is called

1582: {\em minimal} if all lozenges in the chain are distinct.

1583: Exactly as proved in \cite{Fe4} for Anosov flows,

1584: it follows that there

1585: is a unique minimal chain from $\alpha_0$ to $\alpha_1$

1586: and also all other chains have to contain all the lozenges

1587: in the minimal chain.

1588: 

1589: 

1590: 

1591: 

1592: 

1593: The main result concerning non Hausdorff behavior in the leaf spaces

1594: of $\wls, \wlu$ is the following:

1595: 

1596: \begin{theorem}{\cite{Fe4,Fe6}}{}

1597: Let $\Phi$ be a pseudo-Anosov flow in $M^3$. 

1598: Suppose that $F \not = L$

1599: are not separated in the leaf space of $\wls$.

1600: Then the following facts happen:

1601: 

1602: \begin{itemize}

1603: 

1604: \item

1605: $F$ is periodic and so is $L$.

1606: 

1607: \item

1608: Let $F_0, L_0$ be the line leaves of $F, L$ which

1609: are not separated from each other.

1610: Let $V_0$ be the sector of $F$ bounded by

1611: $F_0$ and containing $L$.

1612: Let $\alpha$ be the periodic orbit in $F_0$ and

1613: $H_0$ be the component of $(\wu(\alpha) - \alpha)$ 

1614: contained in $V_0$.

1615: Let $g$ be a non trivial covering translation

1616: with $g(F_0) = F_0$, $g(H_0) = H_0$ and $g$ leaves

1617: invariant the components of $(F_0 - \alpha)$.

1618: Then $g(L_0) = L_0$.

1619: 

1620: \item

1621: Let $\beta$ be the periodic orbit in $L_0$. As $g(\beta) = \beta$,

1622: $g(\alpha) = \alpha$, then  $\pi(\alpha), \pi(\beta)$ are 

1623: closed orbits of $\Phi$  which are  freely homotopic in $M$.

1624: By theorem \ref{chain} \   $F_0$ and $L_0$ are connected by

1625: a finite chain of lozenges 

1626: $\{ A_i \}, 1 \leq i \leq n$.

1627: %all contained in ${\cal L}^u(H_0)$k

1628: Consecutive lozenges are adjacent.

1629: They all intersect a common stable leaf $C$.

1630: There is an even number of lozenges 

1631: in the chain, see

1632: fig. \ref{pict}.

1633: 

1634: \item

1635: In addition 

1636: let ${\cal B}_{F,L}$ be the set of leaves non separated

1637: from $F$ and $L$.

1638: Put an order in ${\cal B}_{F,L}$ as follows:

1639: Put an orientation in the set of orbits of $C$ contained

1640: in the union of the lozenges and their sides.

1641: If $R_1, R_2 \in {\cal B}_{F,L}$ let $\alpha_1, \alpha_2$

1642: be the respective periodic orbits in $R_1, R_2$. Then

1643: $\wu(\alpha_i) \cap C \not = \emptyset$ and let 

1644: $a_i = \wu(\alpha_i) \cap C$.

1645: We define $R_1 < R_2$ in ${\cal B}_{F,L}$ 

1646: if $a_1$ precedes

1647: $a_2$ in the orientation of the set of orbits of 

1648: $C$.

1649: Then ${\cal B}_{F,L}$

1650: is either order isomorphic to $\{ 1, ..., n \}$ for some

1651: $n \in {\bf N}$; or ${\cal B}_{F,L}$ is order

1652: isomorphic to the integers ${\bf Z}$.

1653: 

1654: \item

1655: Also if there are $Z, S \in \wls$ so that

1656: ${\cal B}_{Z, S}$ is infinite, then there is 

1657: an incompressible torus in $M$ transverse to 

1658: $\Phi$. In particular $M$ cannot be atoroidal.

1659: Also if there are $F, L$ as above, then there are

1660: closed orbits $\alpha, \beta$ of $\Phi$ which

1661: are freely homotopic to the inverse of each other.

1662: 

1663: \item

1664: Up to covering translations,

1665: there are only finitely many non Hausdorff

1666: points in the leaf space of $\wls$.

1667: \end{itemize}

1668: \label{theb}

1669: \end{theorem}

1670: 

1671: 

1672: %Consider the chain ${\cal A}_i, 1 \leq i \leq n$ of

1673: %lozenges connecting $F_0$ and $L_0$ produced

1674: %by theorem \ref{theb}.

1675: %Let $\alpha, \beta$ be the corner orbits of ${\cal A}_1$.

1676: %Then $g({\cal A}_1) = {\cal A}_1$,

1677: %$g(\alpha) = \alpha$, $g(\beta) = \beta$.

1678: %Let $\alpha_1, \beta_1$ be the quotients of

1679: %$\alpha, \beta$ by $g$. Then $\alpha_1, \beta_1$

1680: %are periodic orbits of $\Phi$ - which may not

1681: %be indivisible.

1682: %Still their invariance by $g$ implies that $\alpha_1, \beta_1$.

1683: %are freely homotopic.

1684: %In fact suppose that $g$ acts as a contraction

1685: %on the set of orbits of $\ws(\alpha)$.

1686: %Then analysing the action of $g$ on the lozenge ${\cal A}_1$,

1687: %we see that $g$ acts as an expansion in the set

1688: %of orbits of $\ws(\beta)$. This implies that 

1689: %$\alpha_1$ is freely homotopic to the inverse of

1690: %$\beta_1$. This is fundamental for us:

1691: %

1692: %\vskip .03in

1693: %\noindent

1694: %{\underline {Conclusion}}:

1695: %If $\wls$ is 

1696: %not Hausdorff then there are 2 closed orbits $\gamma, \zeta$

1697: %of $\Phi$ which are freely homotopic to the inverse

1698: %of each other, $\gamma \cong \zeta^{-1}$.

1699: %\vskip .03in

1700: 

1701: Notice that ${\cal B}_{F,L}$ is a discrete set in this order.

1702: For detailed explanations and proofs, see

1703: \cite{Fe4,Fe6}. 

1704: 

1705: \begin{figure}

1706: \centeredepsfbox{gm2.eps}

1707: \caption{

1708: The correct picture between non separated

1709: leaves of $\wls$.}

1710: \label{pict}

1711: \end{figure}

1712: 

1713: 

1714: 

1715: \begin{theorem}{(\cite{Fe6})}{}

1716: Let $\Phi$ be a pseudo-Anosov flow. If there is

1717: a stable or unstable product region, then $\Phi$ is 

1718: topologically conjugate to a suspension Anosov flow.

1719: In particular $\Phi$ is non singular.

1720: \label{prod}

1721: \end{theorem}

1722: 

1723: 

1724: %Finally notice that product regions are very rare:

1725: %

1726: %\begin{theorem}{\cite{Fe4}}{}

1727: %Let $\Phi$ be a pseudo-Anosov flow in $M^3$.

1728: %If there is a product region

1729: %in $\mi$, then $\Phi$ is topologically conjugate to

1730: %a suspension Anosov flow.

1731: %\label{produ}

1732: %\end{theorem}

1733: 

1734: 

1735: \begin{proposition}{}{}

1736: Let $\varphi$ be a (topological) Anosov flow so that every leaf of

1737: its stable foliation $\wls$ intersects every leaf of its stable foliations

1738: $\wlu$. Then $\varphi$ is topologically conjugate to a suspension 

1739: Anosov flow. In particular $M$ fibers over the circle with

1740: fiber a torus and Anosov monodromy.

1741: \label{susp}

1742: \end{proposition}

1743: 

1744: \begin{proof}{}

1745: This result is proved by Barbot \cite{Ba1} when $\varphi$

1746: is a {\underline {smooth}} Anosov flow. That means it is $C^1$ and

1747: it has also {\underline {strong}} stable/unstable foliations and

1748: contraction on the level of tangent vectors along the flow.

1749: Here we only have the weak foliations and orbits being asymptotic

1750: in their leaves. 

1751: %Still all the work has been done and we just

1752: %use the pieces. 

1753: With proper understanding all the steps carry through to the

1754: general situation.

1755: 

1756: Lift to a finite cover where $\Lambda^s, \Lambda^u$ are transversely orientable.

1757: A cross section in the cover projects to a cross section in

1758: the manifold (after cut and paste following Fried \cite{Fr}) and

1759: so we can prove the result in the cover.

1760: 

1761: First, the flow $\varphi$

1762: is expanding: there is $\epsilon >  0$ so that no distinct 

1763: orbits are always less than $\epsilon$ away from each other. Inaba

1764: and Matsumoto

1765: then proved that this flow is a topological pseudo-Anosov flow \cite{In-Ma}.

1766: %He generalizes the ideas of Paternain ??? for homeomorphisms

1767: %of surfaces. 

1768: The main thing is the existence of a Markov partition

1769: for the flow. This implies that if $F$ is a leaf of $\wls$ which

1770: is left invariant by $g$, then there is a closed orbit of $\varphi$

1771: in $\pi(F)$ and all orbits are asymptotic to this closed orbit.

1772: Similarly for $\wlu$.

1773: 

1774: What this means is the following: consider the action of $\pi_1(M)$ 

1775: in the leaf space of $\wls$ which is the reals.

1776: Hence we have a group action in $\rrrr$. Let $g$ in $\pi_1(M)$

1777: which fixes a point. There is $L$ in $\wls$ with $g(L) = L$.

1778: So there is orbit $\gamma$ of $\widetilde \varphi$ with $g(\gamma) = \gamma$.

1779: Let $U$ be the unstable leaf of $\widetilde \varphi$ with $\gamma$ contained

1780: in $U$. Then $g(U) = U$. If $g$ is associated to the positive direction

1781: of $\gamma$ then $g$ acts as a contraction in the set of orbits of $U$

1782: with $\gamma$ as the only fixed point. Since every leaf

1783: of $\wlu$ intersects every leaf of $\wls$ then the set of orbits

1784: in $U$ is equivalent to the set of leaves of $\wls$.

1785: This implies the important fact:

1786: 

1787: \vskip .1in

1788: \noindent

1789: {\underline {Conclusion}} - If $g$ is in $\pi_1(M)$ has a fixed point

1790: in the leaf space of $\wls$ then it is of hyperbolic type and

1791: has a single fixed point.

1792: \vskip .1in

1793: 

1794: Using this topological characterization Barbot \cite{Ba1} showed that

1795: $G = \pi_1(M)$ is metabelian, in fact he showed that the commutator

1796: subgroup $[G,G]$ is abelian. In particular $\pi_1(M)$ is solvable.

1797: This used only an action by homeomorphisms in $\rrrr$ satisfying the

1798: conclusion above.

1799: Barbot \cite{Ba1} also proved 

1800: that the leaves of $\Lambda^s, \Lambda^u$ are

1801: dense in $M$.

1802: 

1803: Plante \cite{Pl1}, showed that if $\fol$ a minimal foliation

1804: in $\pi_1(M)$ solvable then $\fol$ is transversely affine:

1805: there is a collection of charts $f_i:U_i \rightarrow \rrrr ^2 \times \rrrr$,

1806: so that the transition functions are affine in the second coordinate.

1807: Using this Plante \cite{Pl1, Pl2} constructs a homomorphism

1808: 

1809: $$C: \ \pi_1(M) \ \ \rightarrow \ \ \rrrr$$

1810: 

1811: \noindent

1812: which measures the logarithm of how much distortion there is

1813: along an element of $\pi_1(M)$. This is a cohomology class.

1814: Every closed orbit $\gamma$ of $\varphi$ has a transversal fence which

1815: is expanding - this implies that $C(\gamma)$ is positive.

1816: Plante then refers to a criterion of Fried \cite{Fr} to conclude that

1817: $\varphi$ has a cross section and therefore it is easily

1818: seen that $\varphi$ is topologically conjugate

1819: to a suspension Anosov flow. 

1820: This finishes the proof of the proposition.

1821: \end{proof}

1822: 

1823: 

1824: 

1825: 

1826: \vskip .1in

1827: \noindent

1828: {\bf {\underline {Blown up orbits and almost pseudo-Anosov flows}}}

1829: \vskip .05in

1830: 

1831: We now describe almost pseudo-Anosov flows.

1832: The description is taken from \cite{Mo5}.

1833: First we describe the blow up of a singular orbit.

1834: Let $\gamma$ be a singular orbit of a pseudo-Anosov flow

1835: $\varphi$ and let $C$ be a small disk transverse to $\varphi$

1836: through a point $p$ of $\gamma$. The set  \ $\ws(\gamma)

1837: \cup \wu(\gamma)$ \ intersects $C$ in a tree $T$, which

1838: is made of a single common vertex and $2n$ prongs,

1839: if the stable/unstable leaves $\ws(\gamma), \wu(\gamma)$

1840: have $n$ prongs.

1841: The stable and unstable prongs alternate in $C$.

1842: Orient the prongs so that stable prongs point

1843: towards $p$ and unstable prongs point away from

1844: $p$. Hence $T$ is an oriented tree.

1845: Up to topological conjugacy we may assume in 

1846: polar coordinates centered at $p$ that 

1847: $T \ = \ \{ (r,\theta) \ | \ \theta = k\pi/n,

1848: k = 0,1, ..., 2n-1 \}$.

1849: 

1850: Choose a Poincar\'{e} section for $\gamma$, that is,

1851: a smaller disk $C'$ so that every point of $C'$ flows

1852: forward to a point in $C$. This  defines a continuous 

1853: map $f:C' \rightarrow C$ which fixes $p$ and is a

1854: homeomorphism into its image.

1855: This $f$ sends the corresponding pieces of

1856: $T$ into $T$,

1857: contracting stable direction and expanding unstable

1858: directions.

1859: There is a rotation or reflection $R$ in $C$ so

1860: that $R$ prescribes where the sectors of $C$ defined

1861: by $T$  go to under $f$.

1862: For instance $R$ may be a rotation by $2\pi/n$.

1863: 

1864: We first blow up $f$.

1865: Let $D$ be a small subdisk of $C'$ containing

1866: $p$ in its interior.

1867: Let $T^*$ be an oriented tree in $C$ which agrees

1868: with $T$ outside $D$ and which is invariant

1869: under $R$ as above and such that each vertex

1870: $v$ of $T^*$ is ``pseudo-hyperbolic", meaning

1871: as you go around the edges incident to $v$,

1872: the orientations of the edges of $T^*$

1873: alternate pointing toward and away from $v$.

1874: Each vertex of $T^*$ must have $2i$ edges for

1875: some $i \geq 2$.

1876: The point here is that $T^*$ is created from

1877: $T$ by replacing $p$ with a finite subtree.

1878: There are finitely many ways to choose

1879: $T^*$, up to compactly supported isotopy.

1880: The new edges of $T^*$ created

1881: by the blow up are called the finite edges of $T^*$.

1882: 

1883: 

1884: Given this data there is a $C^0$ perturbation 

1885: $f^*$ of $f$, and a continuous map $h: C \rightarrow C$

1886: so that 

1887: 

1888: \begin{itemize}

1889: 

1890: \item

1891: $f^*$ leaves $T^*$ invariant.

1892: 

1893: \item

1894: For each edge $E$ of $T^*$, the first return map of 

1895: $f^*$ acts as a translation on

1896: $int(E)$ moving points in the direction of the

1897: orientation on $E$.

1898: 

1899: \item 

1900: $h$ collapses the finite edges of $T^*$  to the

1901: point $p$ and $h$ is otherwise $1$ to $1$.

1902: 

1903: \item

1904: $h$ is a semiconjugacy from $f^*$ to $f$,

1905: that is, $f \circ h = h \circ f^*$.

1906: 

1907: \item

1908: $h$ is close to the identity map in the sup norm

1909: and $h$ equals the identity on 

1910: $C - D$.

1911: 

1912: \end{itemize}

1913: 

1914: 

1915: \noindent

1916: We say that  $f^*$ is obtained from

1917: $f$ by 

1918: {\em dynamically blowing up}

1919: the pseudo-hyperbolic fixed point $p$.

1920: Each choice of $T^*$ determines a unique

1921: $f^*$ up to conjugacy by isotopy.

1922: There are therefore finitely many ways to dynamically

1923: blow up a pseudo-hyperbolic fixed point,

1924: up to conjugation by isotopy.

1925: 

1926: Now define a dynamic blow up of $\gamma$ by altering

1927: the flow $\varphi$ near $\gamma$ as follows. Let $D'$ so that

1928: $f(D') \subset D$.

1929: Alter $\varphi$ so that the first return map

1930: $f: D' \rightarrow D$ is replaced

1931: by $f^*: D' \rightarrow D$.

1932: This has the effect of alterating the generating vector

1933: field of $\varphi$ inside the ``mapping torus"

1934: $T_g = \{ \varphi(x,s) \ | \ x \in D', 0 \leq s \leq t(x) \}$

1935: and leaving the vector field unaltered outside of

1936: $T_g$.

1937: Let $\varphi^*$ be the new flow.

1938: 

1939: 

1940: The orbit $\gamma$ gets blown up into a collection

1941: of flow invariant annuli. In each annulus $A$, the boundary

1942: components are closed orbits of $\varphi^*$

1943: which are isotopic to $\gamma$

1944: as oriented orbits.

1945: In the interior of $A$ orbits move from one boundary to the other,

1946: as given by the orientation of the corresponding

1947: edge of $T^*$.

1948: There is a blow down map $\xi:M \rightarrow M$ which is

1949: homotopic to the identity and isotopic to the

1950: identity in the complement of the annuli.

1951: It sends the collection of annuli into $\gamma$.

1952: The map $\xi$ sends orbits of $\varphi^*$ to orbits

1953: of $\varphi$ and preserves orientation.

1954: 

1955: 

1956: 

1957: 

1958: 

1959: 

1960: 

1961: 

1962: 

1963: \begin{define}{}{}

1964: Let $\varphi$ be a pseudo-Anosov flow in $M^3$ closed.

1965: Then $\varphi^*$ is an almost pseudo-Anosov flow associated to

1966: $\varphi$ if $\varphi^*$ is obtained from $\varphi$ by dynamically blowing

1967: some singular orbits of $\varphi$.

1968: \end{define}

1969: 

1970: 

1971: %into a collection

1972: %of flow annuli.

1973: %Specifically if $\gamma$ is such a singular orbit of $\Phi_1$,

1974: %then it blows up into a connected collection of annuli

1975: %$\{ A_i, 1 \leq i \leq n \}$, each of which is flow invariant.

1976: %The collection is embedded and the annuli have disjoint

1977: %interiors.

1978: %In each annulus the boundary components are closed orbits

1979: %of $\Phi$ isotopic to $\gamma$ as oriented orbits.

1980: %In the interior of each annulus all orbits are forward asymptotic to one

1981: %boundary component and backwards asymptotic to the other

1982: %one.

1983: %There is a blow down map $\xi: M \rightarrow M$, homotopic

1984: %to the identity and isotopic to the identity in the complement

1985: %of the $A_i$ and sending each connected collection

1986: %of $A_i$ into a periodic orbit of $\Phi_1$. The map

1987: %$\xi$ sends orbits of $\Phi$ to orbits of $\Phi_1$

1988: %preserving orientation.

1989: %\end{define}

1990: 

1991: The reason for considering almost pseudo-Anosov flows is

1992: as follows. All of the constructions of pseudo-Anosov flows

1993: transverse to foliations are in fact constructions of

1994: a pair of laminations $-$ stable and unstable $-$ which

1995: are transverse to each other and to the foliation

1996: \cite{Th4,Mo5,Fe7,Cal1,Cal2}.

1997: The intersection of the laminations is oriented producing

1998: a flow in this intersection.

1999: One then collapses the complementary regions to the

2000: laminations to produce transverse singular foliations and 

2001: a pseudo-Anosov flow.

2002: The transversality problem occurs in this last step, the

2003: blow down of complementary regions. In certain situations,

2004: for example for finite depth foliations, one cannot guarantee

2005: transversality of the flow and foliation after the blow down. 

2006: See extended explanations by Mosher in \cite{Mo5}.

2007: 

2008: The necessity of almost transversality as opposed to transversality

2009: was first discovered by Mosher in \cite{Mo1} for positioning

2010: surfaces with respect to certain ${\bf Z}$ invariant

2011: flows (in non compact manifolds).

2012: This was further explored in \cite{Mo2} where it was analysed

2013: when it is necessary to 

2014: blow up a pseudo-Anosov flow before

2015: it became transverse to a given surface.

2016: 

2017: 

2018: %We briefly explain

2019: %a possible problem. Mosher's construction \cite{Mo5} of flows (almost)

2020: %transverse to foliations is done inductively on the depth

2021: %of the leaves (starting with the top depth leaves), associated

2022: %to a sutured manifold hierarchy and the ensuing

2023: %foliations construction of Gabai. At each step there is

2024: %a foliation which is partially tangent/transverse to

2025: %the boundary and also two laminations (stable/unstable)

2026: %which are transverse to each other and to the foliation.

2027: %There is a flow in the intersection of the laminations

2028: %and a flow direction in ``periodic" leaves, since

2029: %all orbits in say a stable leaf are forward asymptotic.

2030: %The next step topologically involves glueing

2031: %two subsurfaces in the boundary in the construction

2032: %of the foliation and laminations/flow.

2033: %

2034: %One of the problems that can easily happen is the following.

2035: %Suppose the glueing is done along surface $S$ and after the

2036: %glueing there are closed orbits $\alpha, \beta$ of the flow,

2037: %which are oriented isotopic to the same simple closed

2038: %curve of $S$ and are in opposite sides of $S$, see

2039: %fig. \ref{prob}.

2040: %

2041: %

2042: \begin{figure}

2043: \centeredepsfbox{gm3.eps}

2044: \caption{

2045: Obstruction to transversality.

2046: Here $\alpha, \beta$ are isotopic closed orbits.

2047: If the flow is to be pseudo-Anosov, then one has to

2048: collapse $\alpha, \beta$ together. But then the collapsed

2049: flow will not be transverse to the foliation.}

2050: \label{prob}

2051: \end{figure}

2052: 

2053: 

2054: %In the resulting pseudo-Anosov flow, $\alpha, \beta$ will be

2055: %(oriented) freely homotopic to each other. By theorem 

2056: %\ref{chain} when lifted to

2057: %$\mi$ they are connected

2058: %by a finite chain of lozenges. This forces the existence

2059: %of another closed orbit $\alpha_1$ which is freely 

2060: %homotopic to the inverse of $\alpha$ (in the opposite corner

2061: %of a lozenge in $\mi$).

2062: %The problem is there is no guarantee such an orbit $\alpha_1$

2063: %will be produced in the inductive process. In order to

2064: %fix that, then in the collapsing step Mosher 

2065: %collapses $\alpha$ and $\beta$ into a single orbit.

2066: %This allows for the collapsed flow to be pseudo-Anosov.

2067: %Unfortunately the transversality is lost locally near this

2068: %region of $S$. There may be more collapsing forced by

2069: %the inductive process. 

2070: %In order to recover the transversality, in this particular

2071: %case one blows up the collapsed orbit into an embedded

2072: %annulus, with boundaries $\alpha, \beta$ and puts 

2073: %a flow going from one orbit to the other, crossing

2074: %$S$ in the correct direction. Since other collapsings

2075: %may be forced we may have a collection of annuli which

2076: %are joined together and collapse to a single periodic orbit.

2077: %

2078: %

2079: 

2080: \vskip .1in

2081: Given a dynamic blow up $\varphi^*$ of $\varphi$ with corresponding

2082: blow down

2083: map $\xi$, then $\xi^{-1}$ of $\Lambda^s, \Lambda^u$ are

2084: singular foliations, which now have some flow invariant

2085: annuli.

2086: These are the annuli that come from blowing up some 

2087: singular orbits $-$ they are called the

2088: {\em blown up annuli}.

2089: We still denote by $\ls, \lu$ the blown up stable/unstable foliations

2090: of an almost pseudo-Anosov flows. They are transverse

2091: to each other except at the blown up annuli. 

2092: The same notation is used for $\wls, \wlu$, etc..

2093: 

2094: Because $\varphi^*$ is transverse to the foliation $\fol$,

2095: then $\wls, \wlu$ are transverse to $\fn$ and induce

2096: singular $1$-dimensional foliations $\wlsf, \wluf$

2097: in any leaf $F$ of $\fn$.

2098: 

2099: The objects perfect fits, lozenges, product regions, etc.. all

2100: make sense in the setting of almost pseudo-Anosov flows: they

2101: are just the blow ups of the same objects for the corresponding

2102: pseudo-Anosov flows. Since the interior of these objects does

2103: not have singularities, the blow up operation does

2104: not affect these interiors. 

2105: There may be singular orbits in the boundary which get

2106: blown into a collection of annuli. All the results in

2107: this section still hold for almost pseudo-Anosov flows,

2108: with the blow up operation.

2109: For example if $F, L$ in $\wls$ are not separated from

2110: each other, then they are connected by an even number

2111: of lozenges all intersecting a common stable leaf.

2112: Since parts of the boundary of these may have been

2113: blown into annuli, there is not a product structure

2114: in the closure of the union of the lozenges, but there

2115: is still a product structure of $\wls, \wlu$ in the interior.

2116: 

2117: 

2118: 

2119: 

2120: 

2121: 

2122: 

2123: \section{Projections of leaves of $\fn$ to the orbit space}

2124: 

2125: \noindent

2126: {\bf {Notation}} $-$ In some of the proofs and arguments that

2127: follow we will be working with an almost pseudo-Anosov flow

2128: transverse to a foliation $\fol$. 

2129: In those cases, for notational simplicity 

2130: we denote the almost pseudo-Anosov flow by

2131: $\Phi$ and a corresponding blow down pseudo-Anosov

2132: flow by $\Phi_1$. 

2133: Hence $\Phi$ is some blow up of $\Phi_1$.

2134: This is different than the

2135: notation $\varphi, \varphi^*$ from the last section.

2136: 

2137: \vskip .1in

2138: Let then $\Phi$ be an almost pseudo-Anosov flow which is transverse

2139: to a foliation $\fol$. 

2140: This implies that $\fol$ is Reebless $-$ we provide

2141: a proof of this at the end of this section.

2142: An orbit of $\wwp$ intersects a leaf of $\fn$ at most once $-$

2143: because the leaves of $\fn$ are properly embedded and 

2144: $\wwp$ is transverse to $\fn$. 

2145: Hence the projection $\Theta: F \rightarrow \Theta(F)$ is injective.

2146: We want to determine

2147: the set of orbits a leaf of $\fn$ intersects $-$ 

2148: in particular we want to determine the boundary $\partial \Theta(F)$.

2149: As it turns out, $\partial \Theta(F)$ is composed of a disjoint

2150: union of slice leaves in $\oos, \oou$.

2151: 

2152: 

2153: We assume throughout this section that $\Phi$ is {\em blow down

2154: minimal} with respect to being transverse to $\fol$. This means

2155: that no blow down of some flow annuli of $\Phi$ produces a flow

2156: transverse to $\fol$.

2157: 

2158: Since $\Phi$ is transverse to $\fol$, there is $\epsilon > 0$ so that

2159: if a leaf $F$ of $\fn$ intersects an orbit of $\wwp$ at $p$ then 

2160: it intersects every orbit of $\wwp$ which passes $\epsilon$ near

2161: $p$ and the intersection is also very near $p$.

2162: To understand $\partial \Theta(F)$ one main ingredient is that 

2163: when considering pseudo-Anosov flows, then

2164: flow lines in the same stable leaf are forward asymptotic. So

2165: if $F$ intersects a given orbit in a very future time then it

2166: also intersects a lot of other orbits in the same stable

2167: since in future time

2168: they converge. In the limit this produces a stable boundary leaf

2169: of $\Theta(F)$. The blow up operation disturbs this: it is not 

2170: true that orbits in the same stable leaf of an almost pseudo-Anosov

2171: flow are forward asymptotic: when they pass arbitrarily near a blow

2172: up annulus the orbits are distorted and their distance

2173: can increase  enormously. This is the key difficulty in this section.

2174: Hence we first analyse the blow up operation more

2175: carefully.

2176: 

2177: \vskip .1in

2178: \noindent

2179: {\underline {Notation}} $-$ Given $\Phi$ an almost pseudo-Anosov flow,

2180: let 

2181: $\Phi_1$ be a corresponding pseudo-Anosov flow associated to $\Phi$. 

2182: The term $\ws(x)$ will denote the stable leaf of $\wwp$ or 

2183: $\wwp_1$, where the context will make clear which one it is.

2184: \vskip .1in

2185: 

2186: 

2187: Recall that $\pi: \mi \rightarrow M$ denotes the universal covering map.

2188: 

2189: 

2190: We will start with $\Phi_1$ and understand the blow up procedure.

2191: The blown up annuli come from singular orbits.

2192: The {\em lift annuli} are the lifts of blown up annuli

2193: to $\mi$. Their projections to $\oo$ are called {\em

2194: blown segments}.

2195: If $L$ is a blown up leaf of $\wls$ or $\wlu$ the components

2196: of $L$ minus the lift annuli are called the {\em prongs}.

2197: A {\em quarter} associated to an orbit $\gamma$ of $\wwp_1$

2198: is the closure of a connected component of $\mi - (\wu(\gamma)

2199: \cup \ws(\gamma))$. Its boundary is a union of $\gamma$ and

2200: half leaves in the stable and unstable leaves of $\gamma$.

2201: We will be interested in a neighborhood $V$ of $\gamma$ in

2202: this quarter which projects to $M$ near the closed orbit

2203: $\pi(\gamma)$. 

2204: We will understand the blow up in the projection of a

2205: quarter. Glueing up different quarter gives the overall picture

2206: of the blow up operation.

2207: In the projected quarter $\pi(V)$ in $M$ there

2208: is a cross section to the flow $\Phi_1$. The orbits across

2209: the cross section are determined by which stable and unstable

2210: leaf they are in. The return map on the stable direction

2211: is a contraction and an expansion in the unstable direction.

2212: Any contraction is topologically conjugate to say

2213: $x \rightarrow x/2$ and an expansion is conjugate to 

2214: $x \rightarrow 2x$. Hence the local return map is 

2215: topologically conjugate to 

2216: 

2217: 

2218: 

2219: $$

2220: \left[

2221: \begin{array}{rr}

2222:  1/2 & 0 \\ 0 & 2 \\ 

2223: \end{array}

2224: \right]$$

2225: 

2226: 

2227: \noindent

2228: a linear map. The whole discussion here is one of topological

2229: conjugacy. The flow is conjugate to 

2230: $(x,y,0) \ \rightarrow \ (2^{-t}x, 2^t y, t)$.

2231: %$$V_t \ \ = \ \ 

2232: %\left[

2233: %\begin{array}{rrr}

2234: % x \\ y \\ 0

2235: %\end{array}

2236: %\right]

2237: %\ \ \ {\rm goes \ to} \ \ \ 

2238: %\left[ \begin{array}{rrr}

2239: %2^{-t}x \\ 2^t y \\ t \\

2240: %\end{array}

2241: %\right]$$

2242: %

2243: %\noindent

2244: Think of the blow up annulus as the set of unit tangent vectors

2245: to $\gamma$

2246: associated to the quarter region. The flow in the

2247: annulus is given by the action of $D V_t$ on the tangent vectors.

2248: It has 2 closed orbits (the boundary ones corresponding to the stable

2249: and unstable leaves). The other orbits are asymptotic to the

2250: stable closed orbit in negative time and to the unstable

2251: closed orbit in positive time. This makes it into a continuous flow

2252: in this blown up part. 

2253: %See detailed explanation in \cite{Fr} or \cite{Ha-Th} (Fried, Handel-Thurston).

2254: For future reference recall this fact that in a blow up annulus the

2255: boundary components are orbits of the flow and in the interior

2256: the flow lines go from one boundary to the other without a Reeb annulus

2257: picture (there is a cross section to the flow in the annulus).

2258: Do this for each quarter region that is

2259: blown up. One can then glue up the 2 sides of the appropriate annuli

2260: because they are all of the same topological picture (using the

2261: standard model above). This describes the blown

2262: up operation in a quarter. There is clearly a blow down map which sends orbits

2263: of the blown up flow $\Phi$

2264: to orbits of $\Phi_1$ and collapses connected unions

2265: of annuli into a single $p$-prong singular orbit.

2266: 

2267: We quantify these: let $\epsilon$ very small so that any

2268: two orbits of $\Phi_1$ which are always less than $\epsilon$

2269: apart in forward time, then they are in the same stable leaf.

2270: Let ${\cal Z}$ the union of the singular orbits of 

2271: $\Phi_1$ which are blown up.

2272: Let $\epsilon' << \epsilon$ and let

2273: $U$ be the $\epsilon'$ tubular neighborhood of ${\cal Z}$.

2274: Let $U'$ (resp. $U$) be the $\epsilon'/2$ 

2275: (resp. $\epsilon'$)

2276: tubular neighborhood of ${\cal Z}$.

2277: Choose the blow up map to be the identity in the complement

2278: of $U'$, that is the blown up annuli are also contained in $U'$.

2279: The blow down map is then 

2280: an isometry of the Riemannian metric outside

2281: $U'$. Choose the blow down to move points very little in $U'$.

2282: Isotope $\fol$ so that it is transverse to the flow $\Phi$.

2283: %We choose a very small neighborhood $U$ of the union of the blown up

2284: %annuli and a Riemannian metric in $M$ so that the blown down operation 

2285: %is an isometry outside of this union. Choose the blow down to be

2286: %the identity outside this neighborhood $U$.

2287: We are now ready to analyse $\partial \Theta(F)$.

2288: 

2289: 

2290: \begin{proposition}{}{}

2291: Let $F$ in $\fn$. Then $\Theta(F)$ is an open subset of $\oo$.

2292: Any boundary component of $\Theta(F)$ is a slice of a leaf of 

2293: $\oos$ or $\oou$.

2294: If it is a slice of $\oos$, then as $\Theta(F)$ approaches $l$,

2295: the corresponding points of $F$ escape in the positive direction.

2296: Similarly for unstable boundary slices.

2297: \label{bounda}

2298: \end{proposition}

2299: 

2300: \begin{proof}{}

2301: First notice that since $F$ is transverse to $\wwp$ then $\Theta(F)$ is

2302: an open set. Hence $\partial \Theta(F)$ is disjoint from $\Theta(F)$.

2303: The important thing is to notice that the metric is the same outside

2304: the small neighborhood $U'$ of the blown up annuli. If two 

2305: points are in the same stable leaf, then their orbits under

2306: the blow down flow $\Phi_1$ are asymptotic in forward time.

2307: The same is true

2308: for $\Phi$, for big enough time if the point is outside $U$.

2309: This is because the points of the

2310: corresponding orbits of $\Phi_1$ will

2311: be {\underline {both}} outside $U'$ $-$ this is the reason for

2312: the construction of two neighborhoods $U', U$.

2313: The following setup will be used in {\underline {all}} cases.

2314: 

2315: \vskip .1in

2316: \noindent

2317: {\underline {Setup}} $-$

2318: Let $v$ in $\partial \Theta(F)$ and $v_i$ in $\Theta(F)$ with

2319: $v_i$ converging to $v$.

2320: Let $p_i$ in $F$ with $\Theta(p_i) = v_i$ and let $w$ in $\mi$

2321: with $\Theta(w) = v$. Let $D$ be any small disk in $\mi$

2322: transverse to $\wwp$ with $w$ in the interior of $D$. For $i$

2323: big enough $v_i$ is in $\Theta(D)$ so there are 

2324: $t_i$ real numbers with $p_i = \wwp_{t_i}(w_i)$

2325: and $w_i$ are in $D$. As

2326: $v$ is not in $\Theta(F)$, then $|t_i|$ grows without bound. Without loss

2327: of generality assume up to subsequence that $t_i \rightarrow +\infty$.

2328: Same proof if $t_i \rightarrow -\infty$.

2329: We will prove that there is a slice leaf $L$ of $\ws(w)$ so that 

2330: $\Theta(L) \subset 

2331: \partial \Theta(F)$ and $F$ goes up as it ``approaches" $L$.

2332: The stable/unstable leaves here are those of the almost pseudo-Anosov flow

2333: and they may have blown up annuli.

2334: 

2335: One transversality fact used here is the following:

2336: for each $\epsilon$ sufficiently small, there is $\epsilon' > 0$

2337: so that if $x, y$ are $\epsilon$ close in $M$ then

2338: the $\fol$ leaf through $x$ intersects the $\Phi$-orbit through

2339: $y$ in a point $\epsilon'$ close to $x$. 

2340: The $\epsilon'$ goes to $0$ as $\epsilon$ does. 

2341: 

2342: 

2343: 

2344: \vskip .1in

2345: \noindent

2346: {\underline {Case 1}} $-$ Suppose that $w$ is not in a blown up leaf.

2347: 

2348: First we show that we can assume no $w_i$ is in 

2349: $\ws(w)$. Otherwise up to subsequence assume all $w_i$ are in

2350: $\ws(w)$. The orbits through $w_i$ and $w$ start out very close

2351: and aside from the time they stay in $\pi^{-1}(U)$ they are always very close.

2352: %Since $F$ intersects $\wwr(w_i)$ in 

2353: %$\wwp_{t_i}(w_i)$ then $t_i \rightarrow \infty$. 

2354: Let $B$ be the component

2355: of the intersection 

2356: of $F$ with the flow band from $

2357: \wwr(w_i)$ to $\wwr(w)$ in the stable leaf $\ws(w)$,

2358: which contains $p_i$.

2359: Then $B$ does not intersect $\wwr(w)$ so it has

2360: to either escape up or down.

2361: If it escapes down it will have to intersect a small segment

2362: from $w_i$ to $w$ and hence so does $F$. For $i$ big enough

2363: $w_i$ is arbitrarily near $w$, so

2364: transversality of $\fol$ and $\Phi$ then

2365: implies that $F$ will intersect $\wwr(w)$ near $w$, contradiction

2366: see fig. \ref{neck}, a.

2367: 

2368: We now consider the case that $B$ escapes up. If the forward orbit

2369: through $w$ is not always in $\pi^{-1}(U)$ then at those times outside

2370: of $\pi^{-1}(U)$ it will be arbitrarily

2371: close to $\wwr(w_i)$ and transversality

2372: implies again that $F$ intersects $\wwr(w)$. If the forward orbit

2373: of $w$ always stays in $\pi^{-1}(U)$ the same happens after the blow down

2374: so the blow down orbit is in the stable leaf of the singular

2375: orbit which is being blown up. 

2376: This does not happen in case 1.

2377: %In this particular situation

2378: %the orbit through $w_i$ will also stay forever in $U$ once it

2379: %enters it and will be asymptotic to $\wwr(w)$. Transversality

2380: %rules this out. 

2381: 

2382: 

2383: We can now assume that all $v_i$ are in a sector of $\oos(v)$

2384: with $l$ the boundary of this sector and 

2385: $L = l \times \rrrr$,

2386: the line leaf of

2387: $\ws(w)$ which is the boundary of this sector.

2388: 

2389: Let now $q$ in $l$. We will show that $q$ is in $\partial \Theta(F)$

2390: so $l \subset \partial \Theta(F)$. There is a segment $[q,v]$ contained

2391: in $l$. 

2392: %This is disjoint from the projection of any blown up annuli.

2393: Choose $x$ in $L$ with $\Theta(x) = q$. Let $\alpha$ be a segment

2394: in $\ws(w)$ transverse to the flow lines and going from $x$ to $w$. 

2395: %Since

2396: %$\ws(w)$ has no blown up annulus we may choose a neighborhood

2397: %$N$ of $\alpha$ which is disjoint from the lift annuli.

2398: %This is because the blown annuli in $M$ form a compact set.

2399: Notice that $x, w$ are in $\mi$ and $q, v$ are in the orbit

2400: space $\oo$.

2401: Let $x_i$ converging to $x$ and

2402: $x_i$ in $\ws(w_i)$. We can do that since all $w_i$

2403: are in the same sector of $\ws(w)$.

2404:  Choose segments $\alpha_i$ from $x_i$ to $w_i$ in 

2405: $\ws(w_i)$ and transverse to the flowlines of $\wwp$ in $\ws(w_i)$.

2406: %For $i$ big enough $\alpha_i$ is contained in $N$ and

2407: %so $\alpha_i$ does not intersect lift annuli. This implies

2408: %that $\wwr(\alpha_i)$ is disjoint from lift annuli.

2409: 

2410: \vskip .1in

2411: \noindent

2412: {\underline {Claim}} $-$ For every orbit $\gamma$ 

2413: of $\wwp$ intersecting $\alpha_i$

2414: in $y$ then $\gamma$ intersects $F$ in $\wwp_s(y)$ where

2415: $s$ converges to $\infty$ as $i \rightarrow \infty$.

2416: 

2417: %The orbits through $y$ and $w_i$ will get arbitrarily close in forward time

2418: %whenever they are outside $\pi^{-1}(U)$. 

2419: %Because the $v_i$ are in a sector

2420: %of $\Theta^s(v)$ they are not in $\ws(v)$. 

2421: %Therefore the point $y$ is not in $\ws(w)$ and so its future orbit

2422: %will not stay in $U$ forever. 

2423: %For $i$ big enough $w_i$ is not in $\pi^{-1}(U)$???

2424: %FOr $i$ big enough the segments $\alpha_i$ are

2425: %disjoint from lift annuli. Otherwise $\alpha$ would

2426: %intersect a lift annulus because             

2427: %the blown annuli in $M$ form a compact set. Hence the

2428: %flow band $\wwr(\alpha_i)$ does not intersect lift annuli.

2429: 

2430: Suppose there is $a_0 > 0$ so that for some $i_0$ then

2431: 

2432: $$\wwp_{[a_0,t_i]}(w_i) \ \subset \ \pi^{-1}(U)

2433: \ \ \ {\rm for \ all} \ \ i \geq i_0$$

2434: 

2435: \noindent

2436: Then $\wwp_{[a_0,\infty)}(w)$ is contained in the closure

2437: of $\pi^{-1}(U)$.

2438: As seen before this implies that $w$ is in a blown up

2439: stable leaf, which is not the hypothesis of case 1.

2440: Therefore up to subsequence, 

2441: there are arbitrary

2442: big times $s_i$ between $0$ and $t_i$ so that $\wwp_{s_i}(w_i)$ is

2443: not in $\pi^{-1}(U)$. 

2444: Hence $\wwr(x_i)$ is very close to $\wwp_{s_i}(w_i)$

2445: and since $F$ cannot escape up or down then $F$ intersects 

2446: $\wwr(x_i)$. Hence the segment

2447: $[\Theta(x_i), v_i]$ of $\oos(v)$

2448: is contained in $\Theta(F)$ and

2449: so $[x,v]$ is contained in the closure of $\Theta(F)$. Also

2450: the time $s$ so that $\wwp_s(y)$ hits $F$ goes to $\infty$, hence

2451: $[x,v]$ cannot intersect $\Theta(F)$ $-$ else there would

2452: be bounded times where it intersects $F$, by transversality.

2453: We conclude that 

2454: $[x,v] \subset \partial \Theta(F)$, hence $l \subset \partial \Theta(F)$

2455: as desired.

2456: If there is a sequence $z_i$ in $F$

2457: escaping down with $\Theta(z_i)$ converging to 

2458: a point in $l$, then by connectedness

2459: there is

2460: one intersecting a compact middle region $-$ this would force

2461: an intersection of $F$ with $l \times \rrrr$ which is impossible.

2462: 

2463: This finishes the proof of case 1.

2464: In this case we proved there is a {\underline {line leaf}} $l$

2465: of $\Theta(L)$ with $l \subset \partial \Theta(F)$ and $F$ escapes

2466: up

2467: as $\Theta(F)$ approaches $l$.

2468: 

2469: 

2470: 

2471: \begin{figure}

2472: \centeredepsfbox{gm4.eps}

2473: \caption{

2474: a. A strangling neck is being forced,

2475: b. A slice in a leaf of $\oos$ or $\oou$.

2476: $x_i = \Theta(z_i)$.}

2477: \label{neck}

2478: \end{figure}

2479: 

2480: 

2481: 

2482: \vskip .1in

2483: \noindent

2484: {\underline {Case 2}} $-$ $w$ is in a blown up leaf, but

2485: $F$ does not intersect a lift annulus in $\ws(w)$.

2486: 

2487: %Here we split into several situation which together give the whole picture.

2488: %Some of the previous analysis carries over to this case as well.

2489: %Suppose first that $p \times \rrrr$ is not in  

2490: %a lift annulus. 

2491: %The analysis in case 1 can be used here.

2492: %We

2493: %use the same notation as in case 1, with $v_i, w_i, p_i = 

2494: %\wwp_{t_i}(w_i), t_i \rightarrow \infty$. 

2495: 

2496: Refer to the setup above.

2497: As before we first show we can assume $w_i$ are not in $\ws(w)$.

2498: Otherwise, up to subsequence assume all $w_i$ are in

2499: $\ws(w)$.

2500: Since $F$ does not intersect lift annuli in $\ws(w)$, then

2501: $w_i$ are all in prongs of $\ws(w)$. Up to subsequence

2502: we can assume they

2503: are all 

2504: in the same prong $C$ of $\ws(w)$ which has boundary an orbit

2505: $\gamma$ of $\wwp$.

2506: It follows that $w$ is in $\gamma$.

2507: All the orbits in $C$ are forward asymptotic to $\gamma$,

2508: even in the blown up situation.

2509: The strangling necks analysis of case 1 shows that $F$

2510: will be forced to intersect $\wwr(w)$.

2511: This cannot occur.

2512: 

2513: 

2514: Hence assume all $v_i$ are in a sector of $\oos(v)$ bounded

2515: by a line leaf $l$. Let $L$ be $l \times \rrrr$.

2516: %The half leaf $L$ does not contain any lift annulus.

2517: %If $w_i$ are in $\ws(p)$ then since $t_i \rightarrow \infty$,

2518: %then $F$ intersects $\wwr(p)$, contradiction. So assume $v_i$

2519: %are in a sector of $\ws(p)$. 

2520: Let $q$ be a point in $l$ and choose $x, \alpha,

2521: x_i$ and $\alpha_i$ as in the proof of case 1.

2522: Choose a small disc $D$ which is transverse to $\wwp$ and

2523: has $\alpha$ in its interior. For $i$ big enough  then

2524: $D$ intersects lift annuli  only in $\ws(x)$.

2525: This is because the union of the blown annuli forms a compact set

2526: in $M$, so  either $\alpha$ intersects a lift annulus,

2527: in which case there is no other lift annulus nearby

2528: or $D$ is entirely disjoint from lift annuli.

2529: %If the forward orbits of $w_i$ are always in $\pi^{-1}(U)$ then

2530: %it is in a prong of a blown up stable leaf and the proof

2531: %of case 1 applies otherwise 

2532: From now on the arguments of case 1 apply perfectly.

2533: This shows that 

2534: $\Theta(L)$ is contained in $\overline {\Theta(F)}$,

2535: it is disjoint from $\Theta(F)$ and so it is $\partial \Theta(F)$

2536: and $F$ escapes up as it approaches $L$.

2537: This finishes the proof of case 2.

2538: 

2539: %This gives a partial picture and not a slice leaf in

2540: %the boundary $\Theta(F)$. Now we need to understand what happens

2541: %when $F$ intersects lift annuli.

2542: 

2543: 

2544: 

2545: Now we need to understand what happens when $F$ intersects

2546: a lift annulus in general. We separate that in a special case.

2547: We need the following facts before addressing this case.

2548: A lift annulus $W$ through $b$ is contained in $\ws(b)$ and

2549: $\wu(b)$ so there is not stable/unstable flow directions

2550: in $W$. However there are still such directions

2551: in $\partial W$, because one attracts nearby orbits

2552: of $\wwp$ in $W$ and the other one repels nearby orbits

2553: in $W$. In

2554: this generalized sense the first one is stable and the

2555: second one is unstable. In this sense

2556: if $a$ is in an endpoint of a blown segment,

2557: then all local components of $\oos(a) - a, \oou(a) - a$ near $a$

2558: are either generalized stable or unstable. With

2559: this understanding there is an even number of such

2560: components and they alternate between generalized stable and unstable.

2561: Some local components of $\oos(a) - a$ are also local

2562: components of $\oou(a) -a$ if they are blown segments.

2563: One key thing to remember is that generalized 

2564: stable and unstable alternate.

2565: 

2566: \vskip .1in

2567: \noindent

2568: {\underline {Case 3}} $-$

2569: Suppose that $F$ intersects some lift annulus $A$ contained

2570: in $\ws(w)$.

2571: 

2572: Then 

2573: $F$ does not intersect both boundary orbits of $A$.

2574: Otherwise collapse $\pi(A)$ to a single orbit, still

2575: keeping $\Phi$ transverse to $\fol$. 

2576: This contradicts that $\Phi$ is blow down minimal with

2577: respect to $\fol$.

2578: Hence either $F \cap A$ is contained in

2579: the interior of $A$

2580: or it intersects only one boundary leaf.

2581: 

2582: 

2583: Assume without loss of generality that $F$ escapes 

2584: {\underline {up}} in one

2585: direction.

2586: % $-$ which then has to be getting closer to

2587: %he stable direction.

2588: This defines an orbit $\gamma$ of $\wwp$ with

2589: $a = \Theta(\gamma)$ in $\partial \Theta(F)$.

2590: The orbit $\gamma$ has to be in the boundary of the lift

2591: annulus $A$. This is because an interior orbit is asymptotic

2592: to both boundary orbits and hence would intersect $F$.

2593: We now look at the picture in $\oo$. Consider the stable leaf

2594: $\oos(a)$. Notice that $\Theta(F)$ intersects $\Theta(A)$. 

2595: %In 

2596: %$\oos(a)$ there

2597: %are two adjacent stable sides to $\Theta(A)$. 

2598: From the point

2599: of view of $\gamma$, orbits in $A$ move away from $\gamma$ in

2600: future time, that is $A$ is an unstable direction

2601: from $\gamma$. This means that $\Theta(A)$ is

2602: generalized unstable as seen from $a$.

2603: It follows that 

2604: there are two generalized stable sides of $\oos(a)$

2605: one on each side of $\Theta(A)$

2606: which are the closest to $\Theta(A)$.

2607: Choose one side, start at $a$ and follow along $\oos(a)$ if needed

2608: through blown segments and eventually into a prong in 

2609: $\oos(a)$ so as to

2610: produce a piece of a line leaf of $\oos(a)$ in that direction.

2611: This path is regular on the side associated

2612: to $\Theta(A)$ and defines a half leaf

2613: $l_1$ of $\oos(a)$. Similarly define

2614: $l_2$ in the other direction, see fig. \ref{neck}, b.

2615: Let $l$ be the union of $l_1$ and $l_2$. Then $l$ is a slice leaf

2616: of $\oos(a)$ but is not a line leaf since $\Theta(A)$ is in $\oos(a)$

2617: and is not in $l$.

2618: %Notice that $l$ is not a line leaf $-$ in order for that to

2619: %happen $l_1$ would have to be followed by $\Theta(A)$ for instance

2620: %and not by $l_2$.

2621: 

2622: 

2623: \vskip .1in

2624: \noindent

2625: {\underline {Claim}} $-$ $l$ is contained in $\partial \Theta(F)$

2626: and $F$ escapes positively as $\Theta(F)$ approaches $l$.

2627: 

2628: Let $b$ in $l_1$ with $b$ not in blown segment, that is,

2629: $b$ in a prong.

2630: Choose  $b_i$ in $\oou(b)$, with 

2631: $b_i \rightarrow b$ and in that component of $\oo - l$.

2632: Let $D$ be an embedded disc in $\mi$ which is transverse to $\wwp$ and

2633: projects to $\oo$ to a neighborhood of the arc $\xi$ in $l_1$ 

2634: from $a$ to $b$.

2635: Let $y_i$ in $D$ with 

2636: $\Theta(y_i) = b_i$, $y_i \rightarrow y$ with $\Theta(y) = b$.

2637: Assume that $y$ is not in $\pi^{-1}(U)$.

2638: Choose $b$ so that it is not in the unstable leaf of one singular orbit,

2639: hence $\wu(y)$ does not contain lift annuli.

2640: In addition choose $y_i$ so that $\ws(y_i)$ does not contain lift annuli either.

2641: 

2642: Choose points $u_j$ in $F \cap A$ so that 

2643: $\Theta(u_j) = a_j$ converges to $a$.

2644: For each $j$ the set $\Theta(F)$ contains a small neighborhood

2645: $V_j$ of $\Theta(u_j)$ with $V_j$ converging to $a$ when

2646: $j$ converges to infinity. 

2647: %Notice that  

2648: %$F$ intersects $A$ transversely. 

2649: %From $u_j$ slide along $F$ until one gets to the complement of $\pi^{-1}(U)$.

2650: %This defines an arc $\beta_j$ in $F$. 

2651: Notice that $a$ is not in $V_j$ as $a$ is not in $\Theta(F)$.

2652: The leaves $\oos(b_i)$ are getting closer 

2653: and closer to $l_1$ and $\Theta(A)$.

2654: For $j$ fixed there is $i$ big enough so that

2655: $\oos(b_i)$ intersects $V_j$. Let

2656: 

2657: $$z_i \ \in \ F \cap \ws(y_i) \ \ {\rm with} \ \ 

2658: \Theta(z_i) \in V_j$$

2659: 

2660: \noindent

2661: here $i$ depends on $j$.

2662: %leaves $\ws(y_i)$ are getting closer and closer to 

2663: %the blown up strip $A$.

2664: %Then for $i, j$ big enough the arc

2665: %$\beta_j$

2666: %will

2667: %intersect $\ws(y_i)$ producing $z_i$ in $F$ with 

2668: %$z_i$ in $\ws(y_i)$.  

2669: Let $z_i = \wwp_{t_i}(r_i)$ with $r_i$ in $D$. 

2670: %Then $t_i$ converges to 

2671: %infinity.

2672: By choosing $j$ and $i$ converging

2673: to infinity we get that 

2674: %If we fix $j$ big enough and allow $i$ to grow then we 

2675: %can assure that 

2676: $\Theta(z_i)$ converges to $a$ and we can ensure that the

2677: arc of  $D \cap \ws(y_i)$  between $r_i$ and $y_i$ is converging

2678: to the arc $\eta$ of $\ws(\gamma) \cap D$ with $\Theta(\eta) = \xi$.

2679: We can also choose $V_j$ small enough so that $t_i$ converges to

2680: $+\infty$.

2681: %in $l_1$.

2682: %Now from $y_i$ go to the point

2683: %$w_i$ in $\wwr(z_i)$ in this segment. 

2684: 

2685: The orbits $\wwr(y_i), \wwr(r_i)$ are very close 

2686: in the forward direction as long

2687: as they are outside $\pi^{-1}(U)$. Since $\ws(y_i)$ does not

2688: contain lift annuli then for times $s$

2689: converging to infinity $\wwp_s(y_i)$ is not in $\pi^{-1}(U)$.

2690: %Also $\wwp_s(y_i)$ is not in $\pi^{-1}(U)$ for $s$ small

2691: %as $y$ is not in this set.

2692: Consider the flow band $C$ in $\ws(y_i)$

2693: between $\wwr(r_i)$ and $\wwr(y_i)$. 

2694: The leaf $F$ intersects $\wwr(r_i)$ in $\wwp_{t_i}(r_i)$ with 

2695: $t_i$ converging to infinity. 

2696: Then an analysis exactly as in case 1 considering strangling necks

2697: and the arcs $B$ in that proof, shows that $F \cap \ws(y_i)$ 

2698: cannot escape up 

2699: before intersecting $\wwr(y_i)$. 

2700: 

2701: Suppose that $F$ escapes down before intersecting $\wwr(y_i)$.

2702: We show that this is impossible.

2703: Since $F \cap \ws(y_i)$ has points $z_i$ in the forward

2704: direction from $D$ and points in the backwards direction

2705: from $D$ it follows that

2706: $F \cap \ws(y_i)$ must intersect  $D$ in at least a

2707: point $q_i$. Up to subsequence we may assume that $q_i$

2708: converges to $q$ in $\ws(y)$.

2709: This will be an iterative process. 

2710: Let $u_1 = q$.

2711: It is crucial

2712: to notice that in the flow band of $\ws(y)$ between $\wwr(y)$ and

2713: $\gamma$ the flow lines tend to go closer to $\gamma$,

2714: that is, either they project to closed orbits freely

2715: homotopic to $\pi(\gamma)$ or they are asymptotic

2716: to one of these orbits moving closer to $\gamma$.

2717: We now consider the component

2718: of $F \cap \ws(y)$ containing $u_1$ and follow

2719: it towards $\gamma$. 

2720: This component does not intersect $\gamma$ and by

2721: the above it can only escape down in $\ws(y)$.

2722: As it escapes down it produces points $c_i$ in $\ws(r_i)$ 

2723: and as before produces points $c'_i$ in $D$, which

2724: up to subsequence converge to $c$ in $D \cap F$.

2725: By construction $c$ is not $u_1$ and its

2726: orbit is closer to $\gamma$.

2727: Let $u_2 = c$. 

2728: %Notice that $u_2$ is different than $u_1$

2729: %by construction.

2730: We can iterate this process. 

2731: Notice the $u_i$ cannot accumulate in $D$, or else

2732: all the corresponding points of $F$ are in a compact

2733: set of $\mi$.

2734: On the other hand the process does not terminate.

2735: This produces a contradiction.

2736: 

2737: 

2738: 

2739: The contradiction shows that in fact  

2740: the arc

2741: $\Theta(C)$ is in $\Theta(F)$ which implies that 

2742: $\xi = \Theta(\eta)$ is

2743: contained in $\overline {\Theta(F)}$. As the time to hit $F$

2744: from $D$ grows with $i$, this shows that $\Theta(F)$ does

2745: not intersect $\xi$ and hence $\xi$ is contained in $\partial \Theta(F)$.

2746: As $b$ is arbitrary this shows that $l \subset \partial \Theta(F)$ and

2747: $F$ escapes up as $\Theta(F)$ approaches $l$.

2748: This finishes the analysis of case 3.

2749: 

2750: 

2751: 

2752: %\vskip .1in

2753: %For $i$ big enough the region in $\ws(y_i)$ between $\wwr(y_i)$ and

2754: %$\wwr(z_i)$ does not intersect a blown up annulus, because

2755: %in $M$ the collection of blown up annuli is compact and they

2756: %are flow invariant.

2757: %Therefore $\ws(y_i)$ is not always in $\pi^{-1}(U)$ and whenever it

2758: %is not, it will be very close to $\wwr(z_i)$. Now an analysis as in

2759: %case 1, cannot escape up or down applies and it follows

2760: %that the segment in $\oos(x_i)$ between $x_i$ and a point

2761: %very close to $v$ is in $\Theta(F)$ with times very positive.

2762: %These orbits are

2763: %asymptotic in future time. As $z_i$ is outside of

2764: %$U$, then so is the corresponding point of $\wwr(y_i)$ and

2765: %so the entire segment from $z_i$ to $x_i$ is in $\Theta(F)$. As

2766: %$x_i$ converges to $x$ and $z_i$ converges to $v$, it follows

2767: %that $l_1 \subset \overline \Theta(F)$. Since the times

2768: %go to infinity then $l_1 \subset \partial \Theta(F)$. 

2769: %Similarly for $l_2$. So $l = l_1 \cup l_2$ is contained

2770: %in $\partial \Theta(F)$. 

2771: 

2772: 

2773: 

2774: 

2775: %\vskip .1in

2776: %We continue the analysis of case 2.1.

2777: %The segment $\Theta(A)$ in $\oo$ has boundary two points

2778: %$v, c$. We now look at the behavior of $\Theta(F)$ beyond $c$.

2779: %If $F$ does not intersect $c \times \rrrr$ we produce

2780: %more boundary $\partial \Theta(F)$, in this case $F$

2781: %going down as it approaches

2782: %$c \times \rrrr$. Let $b$ in $c \times \rrrr$.

2783: %If $F$ does intersect $\wwr(b)$ then it can intersect

2784: %other lift annuli $B_1,...B_n$ with a boundary component

2785: %in $\wwr(b)$. They cannot intersect the other boundary

2786: %component of $B_i$ producing stable or unstable boundary

2787: %as in the beginning of case 3.1.

2788: 

2789: %This analysis shows also that the point $p$ in $\partial \Theta(F)$

2790: %cannot be an orbit in the interior of a lift annulus.

2791: %This is because of the transversality of $\fn$ and the lift

2792: %annuli, so if $p$ corresponds to an interior orbit

2793: %then $F$ in fact intersects the interior of the lift annulus.

2794: %As seen in the beginning analysis of case 2.1 then $\Theta(F)$

2795: %can only have boundary points in the boundary points of the

2796: %projection of the lift annulus.

2797: 

2798: 

2799: %If on the other hand the $w_i$ escape down, then

2800: %$l = l_1 \cup l_2$ is also a slice in the boundary and $l$ escapes

2801: %down. Now if $F$ intersects an annulus, then in at least one

2802: %direction 

2803: %one obtains a boundary component of $\Theta(F)$ as we desired.

2804: 

2805: %If we go in the other direction with respect to blown 

2806: %$F$ can intersect at most

2807: %one more blown annulus, forming a chain of 2. Whenever

2808: %it intersects another blown annulus we produce more 

2809: %$\partial \Theta(F)$ either up or down.

2810: 

2811: %GIven the analysis above if $F$ intersects a blown up 

2812: %annulus with $v$ in it, then $v$ is in the boundary

2813: %of the annulus, there is a slice $l$ in $\oos(v)$

2814: %with $l \subset \partial \Theta(F)$ and $F$ goes

2815: %up as it approaches $l$.

2816: 

2817: 

2818: 

2819: \vskip .1in

2820: \noindent

2821: {\underline {Case 4}} $-$ $w$ is in a blown up stable leaf and

2822: $F$ intersects some lift annulus $A$ in $\ws(w)$.

2823: 

2824: The difference from case 3 is that in case 3 we obtained a slice

2825: boundary $l$ of $\Theta(F)$ $-$ but in our situation 

2826: we do not yet know

2827: if it contains $\Theta(w)$ and  whether it is a stable or

2828: unstable. Here we prove it is a stable slice and it contains

2829: $\Theta(w)$.

2830: 

2831: Recall the setup: $v = \Theta(w)$ is in $\partial \Theta(F)$

2832: and there are $v_i$ in $\Theta(F)$ with $v_i$ converging  to 

2833: $v$ and with $p_i$ in $(v_i \times \rrrr) \cap F$.

2834: Also $p_i = \wwp_{t_i}(w_i)$ with $w_i$ converging to $w$

2835: in $\mi$ and $t_i$ converging to infinity.

2836: %In case 3 we assume that $F$ intersects lift annulus $A$ in

2837: %$\ws(w)$.

2838: Let $\xi$ be the blown segment $\Theta(A)$.

2839: 

2840: The analysis of case 3 shows that $\Theta(F)$ contains 

2841: the interior of $\Theta(A)$.

2842: Suppose first that 

2843: $v$ is in $\xi$. Then $v$ is in the boundary of 

2844: $\xi$ and by case 3 again $F$ escapes up

2845: or down when $\Theta(F)$ approaches a slice

2846: which contains $v$.

2847: If it escapes up, then the slice is a stable slice

2848: and we obtain the desired result in this case.

2849: We now show that $F$ does not escape down. 

2850: Let $l$ be the unstable slice in $\partial \Theta(F)$

2851: associated to this. Then $l$ cuts in half a small

2852: disk neighborhood of $v$ in $\oo$.

2853: The set $\Theta(F)$ intersects only one component

2854: of the complement, the one which intersects $\xi$.

2855: As $F$ escapes down when $\Theta(F)$ approaches $l$, then

2856: for all points in $\Theta(F)$ near enough $v$ the corresponding

2857: point in $F$ is flow backwards from $D$.

2858: This contradicts the

2859: fact that $t_i$ is converging to infinity.

2860: Therefore $F$ cannot escape down as it approaches $l$.

2861: 

2862: We can now assume that $v$ is not in $\xi$.

2863: By changing $\xi$ if necessary assume that $\xi$ is the

2864: blown segment in $\oos(v)$  intersected by $\Theta(F)$ which is

2865: closest to $v$.

2866: Let $z$ be the endpoint of $\xi$ separating the rest of $\xi$

2867: from $v$ in $\oos(v)$. 

2868: 

2869: 

2870: We first show that $z$ is not in $\Theta(F)$. Suppose

2871: that is not the case and let $b$ the intersection point

2872: of 

2873: $z \times \rrrr$ and $F$.

2874: Since $\xi$ is the last blown segment of

2875: $\oos(v)$ between $\xi$ and $v$ intersected by

2876: $\Theta(F)$ and $\Theta(F)$ contains an open neighborhood

2877: of $z$, it follows that $v$ is in a prong $B$ of

2878: $\oos(v)$ with endpoint $z$. 

2879: Let $\tau$ be the component of $F \cap \ws(b)$ containing

2880: $b$. Since $F$ does not intersect

2881: $v \times \rrrr$ then it escapes. As the region between

2882: $\wwr(b)$  and $z \times \rrrr$ (should it be

2883: $v \times \rrrr$ instead of $z \times \rrrr$? previous

2884: report???)

2885: is a prong,

2886: then $F$ cannot escape up. As seen in the arguments

2887: for case 3, $F$ cannot escape down either.

2888: This shows that $z$ cannot be in $\Theta(F)$.

2889: 

2890: %Let $\beta$ be

2891: %the segment of $\oos(v)$ from $z$ to $v$.

2892: %Let $D$ be a disk in $\mi$ transverse to $\wwp$

2893: %so that $\Theta(D)$ is a small neighborhood of $\beta$.

2894: %Since $\beta$ is in the closure of a prong of $\oos(v)$ then 

2895: %$\oos(v_i)$ limits on $\beta$ as $i$ converges

2896: %to infinity $-$ whether $v_i$ is in $\oos(v)$ or not.

2897: %So for big enough $i$ there are 

2898: %

2899: %$$r_i \ \in \ D \cap \ws(w_i), \ \ \ \Theta(r_i) \rightarrow z

2900: %\ \ {\rm as} \ \ i \rightarrow \infty$$

2901: %

2902: %\noindent

2903: %Let $\alpha_i$ be the segments in $\oos(v_i)$ between 

2904: %$v_i$ and $\Theta(r_i)$. As seen in case 1 we can choose

2905: %$r_i$ so that for $i$ big enough $\alpha_i$ does not

2906: %intersect blown segments. Then the forward part of

2907: %the orbits $\wwr(r_i)$ and $\wwr(w_i)$ are very close

2908: %while outside $\pi^{-1}(U)$. As seen in 

2909: %the proof of case 1 there are $s_i$ converging

2910: %to infinity with $\wwp_{s_i}(r_i)$ in $F$. But this

2911: %contradicts the existence of bounded $a_i$ with

2912: %$\wwp_{a_i}(r_i)$ in $F$ near $b$. 

2913: 

2914: It follows that $F$ escapes either up or down as $\Theta(F)$

2915: approaches $z$. 

2916: Suppose first that it escapes up.

2917: Then we are in the situation

2918: of case 3 and we produce a stable slice $l$ in

2919: $\partial \Theta(F)$ with $F$ going up as $\Theta(F)$ approaches

2920: $l$. 

2921: If $v$ is not in $l$ then $l$ separates $v$ from

2922: $\Theta(F)$. This contradicts $v_i$ in $\Theta(F)$

2923: with $v_i$ converging to $v$. Hence $v$ is in $l$

2924: with $F$ escaping up as $\Theta(F)$ approaches $l$.

2925: This is exactly what we want finishing the analysis

2926: in this case.

2927: 

2928: The last situation is $F$ escaping down in $A$ as $\Theta(F)$

2929: approaches $z$. By case 3 there is a slice leaf $l$ in

2930: $\oou(z)$ with $l$ contained in $\partial \Theta(F)$ and

2931: $F$ escaping down as $\Theta(F)$ approaches $l$.

2932: We want to show that this case cannot happen.

2933: Notice that the blown segments 

2934: of $\oos(z)$ are exactly the

2935: same as the blown segments of $\oou(z)$. The sets $\oos(z), \oou(z)$

2936: differ exactly in the prongs and as they go around the collection

2937: of blown segments. 

2938: The collection of all prongs in $\oos(z), \oou(z)$

2939: also alternates between stable and unstable as it goes around

2940: the union of the blown segments.

2941: 

2942: Suppose first that $v$ is in $l$.

2943: This contradicts $F$ escaping down and $t_i \rightarrow \infty$.

2944: Finally suppose that $v$ is not in $l$.

2945: We claim that in this case $l$ separates $v$ from $\Theta(F)$.

2946: Let $\alpha$ be the path in $\oos(v)$ from $z$ to $v$.

2947: If $\alpha$ only intersects $l$ in $z$, then the separation

2948: property follows because

2949: $l_1$ and $l_2$ contain the local components

2950: of $\oos(z) \cup \oou(z) - z$ which are closest to $\Theta(A)$.

2951: %see fig. \ref{bar}, a. 

2952: This was part of the construction

2953: of $l$ in case 3.

2954: Here the $\xi$ is generalized stable at $z$

2955: and $l_1, l_2$ are generalized unstable at $p$. The path from $z$ to 

2956: $v$ in $\oos(v)$ cannot start in $\xi$ or $l_1$ or $l_2$, hence

2957: $l$ separates $\Theta(F)$ from $v$.

2958: 

2959: %\blankfig{bar}{1.1}{a. The case $\alpha \cap l$ is only $z$,

2960: %b. The case when $\alpha \cap l$ is a nontrivial subsegment.}

2961: 

2962: If on the other hand $\alpha \cap l = \delta$ is not

2963: a single point, then it is a union of blown segments.

2964: Let $u$ be the other endpoint of $\delta$. By regularity

2965: of $l_1$ and $l_2$ on the $\Theta(F)$ side it follows that each

2966: blown up segment in $\delta$ has flow direction away from

2967: $z$. Hence $\delta$ is generalized stable at $u$.

2968: Therefore the closest component of

2969: $\oos(u) \cup \oou(u) - u$ on the

2970: $\Theta(F)$ side is generalized unstable and that is contained

2971: in $l$.

2972: %, see fig. \ref{bar}, b.

2973: In this case it also follows that $l$ separates $v$ from

2974: $\Theta(F)$. As seen before this is a contradiction.

2975: 

2976: This finishes the proof of proposition \ref{bounda}

2977: \end{proof}

2978: 

2979: %Suppose first there is no blown segment in $\oos(v)$ between

2980: %$\xi$ and $v$.

2981: %Then if $v$ is in a blown up segment if follows that

2982: %$v$ is in $l$ and $F$ ha

2983: %Then if $v$ is in a blown up segment if follows that

2984: %$v$ is in $l$. But notice that $F$ escapes down as $\Theta(F)$

2985: %approaches $l$ and so $\Theta(F)$ does not intersect any other

2986: %%component of $\oo - l$.

2987: %Since $F$ escapes down and $t_i$ converge to infinity this

2988: %is a contradiction.

2989: %The other option is that $v$ is in a prong $B$ of $\oos(v)$

2990: %with endpoint $z$, see fig. \ref{bar}, a.

2991: %Locally near $\xi \cup B$ there is a sector $W$ of $\oo$ that $\Theta(F)$

2992: %intersects. There is $B'$ which is either a prong of $\oou(v)$

2993: %or a blown segment of $\oou(v)$ which cuts the sector, separating

2994: %$\xi$ from $v$, see fig. \ref{bar}, a.

2995: %Then $\xi \cup B'$ is contained in $l$ and hence in $\partial \Theta(F)$

2996: %and this shows that $v$ cannot be in the closure

2997: %of $\Theta(F)$. This is a contradiction.

2998: %

2999: %

3000: %

3001: %The other option is that there is a blown segment

3002: %$\xi_1$ in $\oos(v)$ separating $\xi$ from $v$, see 

3003: %fig. \ref{bar}, b. Let $W$ be the corresponding sector

3004: %as the situation above. Notice there can be no unstable

3005: %prong of $\oou(v)$ entering $w$ or the separation argument

3006: %above applies. Then as seen in case 3.1 it follows that

3007: %$\xi_1$ is also contained in $l$, that is $\xi_1$ is

3008: %in $\partial \Theta(F)$ and $F$ escapes down.

3009: %Let $z_1$ be the other endpoint of $\xi_1$. 

3010: %Then we can restart the analysis with $z_1$ instead

3011: %of $z$. Induction shows this cannot happen.

3012: %

3013: %This finishes the proof of proposition????

3014: %

3015: 

3016: 

3017: 

3018: 

3019: 

3020: 

3021: 

3022: %

3023: %

3024:  %with $p$ in $\partial \Theta(F)$

3025: %and $F$ not intersecting any blown up annulus in $\ws(v)$.

3026: %Assume now that $F$ does not intersect a blown up

3027: %annulus with $v$ in it. Then locally $F$ is in a sector

3028: %of $\ws(p)$. If $v$ is in the interior of a blown up

3029: %annulus with direction $a_2 \rightarrow a_1$. We follow

3030: %backwards from $a_2$ along the blown up annuli $A_1, A_2$

3031: %until one gets to a half leaf $L_1$. Here $F$ does not

3032: %intersect $A_1$. If it intersects $A_2$ then

3033: %it does not intersect the contracting end so it

3034: %escapes down. By the analysis done previously (Case 2.1)

3035: %$F$ escapes down along a slice leaf of $\oou(a_2)$.

3036: %This contradicts $F$ limiting up to $v$.

3037: %In this case we have a continuous set of leaves, not

3038: %just a sequence of leaves.

3039: %

3040: %Finally we check whether $F$ intersects $L_1$. 

3041: %Since it does not intersect the boundary it escapes and

3042: %it needs to escape down.

3043: %As in the proof of case 2.1 it produces a line leaf in $\oos(p)$

3044: %towards which $F$ escapes down, contradiction.

3045: %Either separates the $v_i$ or escapes up.

3046: %

3047: %We obtain that $F$ does not intersect that half leaf on

3048: %that side and neither in the other side as well.

3049: %Suppose the unstable leaf coming into $a_1$ is a blown

3050: %up annulus and $F$ intersects it. Then it goes up

3051: %as it nears $a_1$ and the slice leaf $l_1 \cup l_2$ is in

3052: %the boundary of $F$ as required. Otherwise now 

3053: %$\oos(p) \cap F = \emptyset$ and since there is no

3054: %blown up annulus separating any 2 points in this

3055: %line leaf of $\ws(p)$ we obtain that this line leaf

3056: %of $\oos(p)$ is contained in $\partial \Theta(F)$

3057: %and $F$ is going up as it approaches this line leaf.

3058: %

3059: %If now there are $p_i$ in $F$ with $\Theta(p_i) \rightarrow v$

3060: %and $p_i = \wwp_{t_i}(w_i)$ with $w_i \rightarrow w$ in 

3061: %$\mi$ and $t_i \rightarrow -\infty$, then the whole

3062: %analysis produces an unstable slice in $\oos(v)$ 

3063: %which is entirely in the boundary of $\Theta(F)$. These

3064: %two possibilities are incompatible. Either the slice

3065: %in $\oou(F)$ cuts the slice of $\oos(F)$ into 2 pieces

3066: %in different components of $\mi - R$ impossible since they

3067: %have to limit in 2 sides. The other option is that both

3068: %of them are on the same side. But then the stable slice cuts

3069: %$F$ away from $\partial \Theta(F)$ as seen in the unstable

3070: %slice. Hence we can only have $t_i \rightarrow \infty$.

3071: %Also $F$ really escapes up as it approaches this slice.

3072: %

3073: %This finishes the proof of proposition boundary of $\Theta(F)$.

3074: %%\end{proof}

3075: 

3076: 

3077: 

3078: 

3079: 

3080: %\section{Projections}

3081: %\label{proj}

3082: %

3083: %Given $F$ a leaf of $\gn$ we want to determine the set of orbits of 

3084: %$wwp$ it intersects of $\oo(F)$.

3085: %

3086: %\begin{lemma}{}{}

3087: %Given $F$ in $\gn$, it intersects an orbit of $\wwp$ at most once.

3088: %This means that $\Theta: R \rightarrow \oo$ is injective

3089: %for any $F$ in $\gn$.

3090: %\end{lemma}

3091: %

3092: %\begin{proof}{}

3093: %Otherwise one can perturb a segment of $\wwp$ jointly with a segment in

3094: %$F$ to produce a closed transversal to $\gn$.

3095: %\end{proof}

3096: %

3097: %\begin{proposition}{}{}

3098: %Let $F$ in $\gn$ and $a$ a point in the boundary of $\Theta(F)$

3099: %(that is $\partial \Theta(F)$. Then either there is a subleaf $l$ of

3100: %$\wu(a)$ or $\ws(a)$ so that $\Theta(l)$ is contained in

3101: %$\partial \Theta(F)$ and $l$ does not intersect $F$.

3102: %\end{proposition}

3103: %

3104: %\begin{proof}{}

3105: %Let $a_i$ be points in $\oo$, with $a_i$ in $\Theta(F)$ and $a_i$ converges

3106: %to $a$.  Consider a small disk $D$ in $\mi$ transverse to $\wwp$

3107: %with $a$ in $D$. Let $b_i$ in $D$ with $\theta(b_i) = a_i$. 

3108: %Since $a_i$ are in $\Theta(F)$, there are points $p_i$ in $F$ with

3109: %$p_i$ in $\wwp_{t_i} (b_i)$. 

3110: %Up to a subsequence suppose that $t_i$ converges to $+\infty$.

3111: %

3112: %\vskip .1in

3113: %\noindent

3114: %{\underline {Case 1}} - Suppose that $v$ is not in the lift of

3115: %a blown up annulus.

3116: %

3117: %Hence $\ws(v), wu(v)$ neatly divides the neighborhood of 

3118: %$v$ in sectors. Up to subsequences assume all $v_i$ are

3119: %in the closure of the same sector.  If $v_i$ is in $\ws(v)$, then

3120: %since $t_i \rightarrow \infty$ it follows that $v_i$ is very 

3121: %close to $\wwp_{\rrrr}(v)$ so as $F$

3122: %is transverse to $\wwp$ then $F$ intersects $\wwp_{\rrrr}(v)$

3123: %contradiction. 

3124: %

3125: %If now $r$ is in $l$ and

3126: %$\wu(r) \cap ws(l)$ intersects for $i$ big enough in a point the orbit

3127: %of $b_i$. For $i$ big enough???

3128: %$\Phi_{t_i}(b_i)$ is very close to $wwp_{t_i}(v_i)$ so

3129: %$\wwp_{t_i}(b_i)$ is in $F$. Therefore $r$ is in the closure

3130: %$\Theta(F)$ in $\oo$. 

3131: %

3132: %If on the other hand $r$ is in $\Theta(F)$ there is a small disk $D$ in

3133: %$F$ transverse to $\wwp$ so that $r$ is in $D$. Then for $i$ big enough

3134: %$b_i$ is in $D$ contradiction to the previous lemma.

3135: %

3136: %\vskip .1in

3137: %\noindent

3138: %{\underline {Case 2}} - 

3139: %The point $r$ is in a blown up annulus

3140: %and $r$ is an interior point.

3141: %

3142: %Notice holonomy is as picture in figure 1.

3143: %Notice in the interior of the picture is a product of the

3144: %leaf shoots throuh, hence this case cannot happen.

3145: %

3146: %\vskip 1.5in

3147: %\centerline{figure 1 will go here}

3148: %

3149: %\end{proof}

3150: 

3151: This has an important consequence that will be used extensively

3152: in this article.

3153: 

3154: \begin{proposition}{}{}

3155: Let $F$ in $\fn$ and $L$ in $\wls$ or $\wlu$.  Then the intersection

3156: $F \cap L$ is connected.

3157: \label{conn}

3158: \end{proposition}

3159: 

3160: \begin{proof}{}

3161: By transversality of $\fol$ and

3162: $\Phi$, the intersection $C = \Theta(F) \cap \Theta(L)$

3163: is open in $\Theta(L)$.

3164: Suppose there are 2 disjoint components $A, B$ of $C$.

3165: Then there is $v$ in $\partial A$ with $v$ separating

3166: $A$ from $B$. There are $v_i$ in $A$ with $v_i$ converging

3167: to $v$. By the previous proposition $F$ escapes up or down

3168: in $A \times \rrrr$ as $\Theta(F)$ approaches $v$.

3169: Assume wlog that $F$ escapes up. Then there is a slice

3170: leaf $l$ of $\oos(v)$ with $l \subset \partial \Theta(F)$

3171: and $F$ escapes up as $\Theta(F)$ approaches $l$.

3172: Since $l$ and $\Theta(F)$ are disjoint then $B$ is

3173: disjoint from $l$. In addition $v$ separates

3174: $B$ from $A$ in $\Theta(L)$. It follows from the construction

3175: of the slice $l$ as being the closest to $A$, that

3176: $l$ separates $A$ from $B$. Hence $\Theta(F)$ cannot

3177: intersect $B$, contrary to assumption.

3178: This finishes the proof.

3179: \end{proof}

3180: 

3181: 

3182: 

3183: 

3184: As promised, we now prove that $\fol$ being almost 

3185: transverse to a pseudo-Anosov flow implies that $\fol$

3186: is Reebless.

3187: 

3188: \begin{proposition}{}{}

3189: Let $\fol$ be a foliation almost transverse to a pseudo-Anosov

3190: flow  $\Phi_1$ and transverse to a corresponding almost pseudo-Anosov

3191: flow $\Phi$. Then $\fol$ is Reebless.

3192: \end{proposition}

3193: 

3194: \begin{proof}{}

3195: Suppose that $\fol$ is not Reebless and consider a Reeb component

3196: which is a solid torus $V$ bounded by a torus $T$.

3197: Assume that the flow $\Phi$ is incoming along $T$.

3198: 

3199: Recall that  there are some singular orbits of $\Phi_1$

3200: which blow up into a collection of flow annuli of

3201: $\Phi$. Suppose that $V$ intersects one of these annuli

3202: $A$. Then since $\Phi$ is incoming along $T$, the torus

3203: $T$ cannot intersect the closed orbits in $\partial A$.

3204: Hence it intersects the interior of $A$, say in

3205: a point $p$ and the forward orbit of $p$ will

3206: limit in a closed orbit which is contained in

3207: the interior of $V$.

3208: 

3209: If on the other hand $V$ does not intersect these blown

3210: annuli then the blow down operation does not affect

3211: the flow in $V$. That means we can assume that $\Phi_1$ is

3212: equal to $\Phi$ in $V$. Since orbits of $\Phi_1$

3213: are trapped inside $V$

3214: once they enter $V$, the shadow lemma for pseudo-Anosov

3215: flows \cite{Han,Man,Mo1}, shows that 

3216: there is also a periodic orbit of $\Phi_1$ (and hence

3217: also of $\Phi$) in $V - T$.

3218: Notice that the shadow lemma is for pseudo-Anosov flows

3219: and not for almost pseudo-Anosov flows and that is why

3220: we split the analysis into 2 cases.

3221: 

3222: In any case there is a closed orbit $\gamma$ of $\Phi$ contained

3223: in the interior of $V$.

3224: Consider the generalized stable/unstable local

3225: leaves at $\gamma$. Since $\Phi$ is incoming along $T$,

3226: the generalized unstable leaves have to be contained

3227: in $V$. We eventually obtain that a whole half leaf

3228: of $W^u(\gamma)$ is contained in $V$. 

3229: A lift $\widetilde V$ to $\mi$ is homeomorphic to $D^2 \times

3230: \rrrr$, because closed orbits of $\Phi$ are not null homotopic.

3231: The procedure above produces a half leaf of $\wu(\widetilde \gamma)$

3232: contained in $\widetilde V$. This contradicts the

3233: fact that $\wu(\widetilde \gamma)$ is properly 

3234: embedded \cite{Ga-Oe}.

3235: This shows that $\fol$ is Reebless.

3236: \end{proof}

3237: 

3238: 

3239: 

3240: 

3241: \section{Asymptotic properties in leaves of the foliation}

3242: \label{asym}

3243: 

3244: Let $\Phi$ be an almost pseudo-Anosov flow transverse

3245: to a foliation $\fol$ with hyperbolic leaves. Let

3246: $\Lambda^s, \Lambda^u$ be the singular foliations

3247: of $\Phi$. Given leaf $F$ of $\fn$ let $\wlsf, \wluf$

3248: be the induced one dimensional singular foliations in $F$.

3249: In this section we study asymptotic properties of 

3250: rays in $\wlsf$.

3251: First we mention a result of Thurston \cite{Th5,Th7} concerning

3252: contracting directions, which for convenience we

3253: state for 3-manifolds:

3254: 

3255: \begin{theorem}{}{(Thurston)} 

3256: Let $\fol$ be a codimension one foliation with hyperbolic

3257: leaves in $M^3$ closed. Then for every $x$ in any leaf

3258: $F$ of $\fn$ 

3259: and every $\epsilon > 0$ there is a dense set of

3260: geodesic rays 

3261: of $F$ starting at $x$ such that:

3262: for any such ray $r$

3263: there is a transversal $\beta$ to $\fn$ starting at $x$ so

3264: that any leaf $L$ intersecting $\beta$ and any

3265: $y$ in $r$, then the distance between $y$ and $L$ is

3266: less than $\epsilon$.

3267: If there is not a holonomy invariant transverse measure

3268: whose support contains $\pi(F)$ then one can show

3269: that the directions are actually contracting, that is:

3270: if $y$ escapes in $r$ then the distance between $y$ and

3271: $L$ converges to $0$.

3272: Finally if $\pi(F)$ is not closed one can choose the 

3273: $\beta$ above to have $x$ in the interior.

3274: \end{theorem}

3275: 

3276: There is a carefully written published version of this

3277: result in \cite{Ca-Du}.

3278: The directions above where distance to nearby $L$ goes to

3279: $0$ are called contracting directions. The other ones

3280: where distance is bounded by $\epsilon$

3281: are called $\epsilon$ non expanding directions.

3282: 

3283: The goal of this section is to show that given a leaf

3284: $L$ of $\fn$ and a ray $l$ of ${\widetilde \Lambda}^s_L$, 

3285: then $l$ converges to a single point in $\pin L$.

3286: We first analyse 

3287: the non $\rrrr$-covered case.

3288: %if $l$ does not 

3289: %converge to a single point in $\pin L$, this forces

3290: %$\fol$ to be $\rrrr$-covered. 

3291: The proof is very involved and is done by way 

3292: of contradiction.

3293: Later we deal

3294: with the $\rrrr$-covered situation.

3295: This result is a natural extension of a result by

3296: Levitt  \cite{Le} who proved that if

3297: $\gal$ is a foliation with prong singularities

3298: in a closed hyperbolic surface, then in the universal

3299: cover, an arbitrary ray converges to a single point in

3300: the circle at infinity.

3301: The situation for non compact leaves of foliations

3302: is much more delicate.

3303: 

3304: 

3305: 

3306: 

3307: 

3308: \begin{proposition}{}{}

3309: Let $\Phi$ be an almost pseudo-Anosov flow transverse to

3310: a foliation $\fol$ in $M^3$ closed and $\fol$ with hyperbolic

3311: leaves. 

3312: %Let $\varphi$ be the blow up flow transverse to

3313: %$\la$. 

3314: Suppose that $\fol$ is not $\rrrr$-covered.

3315: Given a leaf $L$ of $\fn$ and an arbitrary ray $l$ 

3316: in a leaf of ${\widetilde \Lambda}^s_L$ or

3317: ${\widetilde \Lambda}^u_L$ 

3318: then $l$ limits to a single

3319: point in $\pin L$. The limit depends on the ray $l$.

3320: \label{as1}

3321: \end{proposition}

3322: 

3323: \begin{proof}{}

3324: We do the proof for 

3325: ${\widetilde \Lambda}^s_L$.

3326: Let $\epsilon$ positive so that if $p$ in $\mi$ is

3327: less than $\epsilon$ from a leaf $F$ of $\fn$, then

3328: the flow line through $p$ intersects $F$ less than

3329: $2 \epsilon$ away from $p$.

3330: Let $l$ be a ray in ${\widetilde \Lambda}^s_L$.

3331: Because $\fol$ and $\Phi$ are transverse,

3332: $L$ is properly embedded in $\mi$ and leaves of $\wls$ are

3333: properly embedded, it follows that $l$ is a properly embedded

3334: ray in $L$. Therefore it can only limit in 

3335: $\pin L$. 

3336: 

3337: Suppose by way of contradiction that $l$ 

3338: limits on 2 distinct points $a_0, b_0$ in $\pin L$.

3339: Fix $p$ a basepoint in $L$. 

3340: Since

3341: $l$ limits in $a_0, b_0$, there are compact arcs $l_i$ of

3342: $l$ with endpoints which converge to $a_0, b_0$ respectively

3343: in

3344: $L \cup \pin L$ and so that the distance from $l_i$ to $p$ in

3345: $L$ converges to infinity. Also we can assume that the

3346: $l_i$ converges to a segment $v$ in $\pin L$, where $v$ connects

3347: $a_0, b_0$. This is in the Hausdorff topology of closed sets

3348: in $L \cup \pin L$,  which is a closed disk.

3349: 

3350: The key idea is to bring this situation to a compact part of $\mi$.

3351: Choose a sequence $p_i$ at bounded distance

3352: from points in $l_{k_i}$ so that that $p_i$ converges

3353: to a point $a$ in the interior of $v$. 

3354: The bound depends on the sequence.

3355: Up to subquence

3356: assume that there are convering translations $g_i$ in 

3357: $\pi_1(M)$ and a point $p_0$ in $\mi$ 

3358: so that $g_i(p_i)$ converges to 

3359: $p_0$ in $\mi$.

3360: 

3361: We claim that the set of possible limits $p_0$ 

3362: obtained as above projects

3363: to a sublamination of $\fol$.

3364: Clearly if $g_i(p_i)$ converges to $p_0$ and $q$ is in the same 

3365: leaf $L_0$ of $\fn$ as $p_0$, then the distance 

3366: from $p_0$ to $q$ is finite 

3367: and there are $q_i$ in

3368: $L$ with $d_L(q_i, p_i)$ bounded and

3369: $g_i(q_i)$ converging to $q$. Also $q_i$ converges to $a$ in $\pin L$.

3370: In addition if a sequence of such limits $c_j$ converges

3371: to $c_0$ then a diagonal process shows that $c_0$ is also

3372: obtained as a single limit.

3373: This proves the claim.

3374: Choose a minimal sublamination $\ccl$.

3375: 

3376: 

3377: 

3378: A leaf $F$ of $\fn$ is isometric to the hyperbolic plane. A 

3379: {\em wedge} $W$ in $F$ with corner $b$ and ideal set 

3380: an interval $B \subset \pin F$

3381: is the union of the rays in $F$ from $b$ with ideal point in $B$.

3382: The angle of the wedge is the angle that the boundary rays of

3383: $W$ make at $b$.

3384: For any such sequence $p_i$ as above, then

3385: the visual angle at $p_i$ subintended by the arc $v$ in $\pin L$ grows

3386: to $2 \pi$. Therefore the angle of wedge with corner

3387: $p_i$ and ideal set 

3388: $\pin L - v$ converges to $0$.

3389: This is called the {\em bad wedge}.

3390: 

3391: 

3392: 

3393: Assume up to subsequence that $g_i(p_i)$ is converging

3394: to $p_0$ in a leaf $L_0$ of $\fn$ and that the 

3395: directions of the

3396: bad wedges with corners $g_i(p_i)$ in $g_i(L)$ are converging 

3397: to the direction $r_0$  of $L_0$. Let $c$ be the ideal

3398: point of $r_0$ in $\pin L_0$.

3399: 

3400: Suppose first that $\pi(L_0)$ is not compact $-$ we shall

3401: see briefly that this is in fact always the case.

3402: Thurston's theorem shows that the set of two sided contracting

3403: directions (or $\epsilon$ non

3404: expanding directions) in $L_0$ is dense

3405: in $\pin L_0$. 

3406: We will use these to transport a lot of the structure

3407: of ${\widetilde \Lambda}^s_{L_0}$ to nearby leaves.

3408: %Choose $F_1, F_2$ on both sides of

3409: %$L_0$ so that there is a big disk $D$ in

3410: %$L_0$ which is less than $\epsilon$ from both

3411: %$F_0$ and $F_1$.

3412: %Very near $r_0$.

3413: Choose $s_0, s_1$ to be rays in $L_0$ defining

3414: contracting directions

3415: (or $\epsilon$ non expanding directions)

3416: very near $r_0$ so that together they form a small wedge

3417: $W$ in $L_0$ with corner $p_0$.

3418: There is an interval of leaves near $L_0$ so that

3419: any such leaf $V$ is less than $\epsilon$ away from 

3420: $s_0, s_1$. Then a flow line  of $\wwp$ through any point

3421: in $s_0$ or $s_1$ intersects $V$ less than $2 \epsilon$

3422: away. So $s_0$ flows to a curve in $V$, where we can

3423: assume it has geodesic curvature very close to $0$,

3424: if $\epsilon$ is sufficiently small. It is therefore 

3425: a quasigeodesic with a well defined ideal point.

3426: The same happens for $s_1$ and the flow images $u_0, u_1$

3427: of $s_0, s_1$ in $V$ define a generalized wedge $W'$

3428: in $V$.

3429: %For $i$ big enough $g_i(p_i)$ is very close to $p_0$ and

3430: %$g_i(L)$ is in this interval so $s_0, s_1$ flow into

3431: %$g_i(L)$ with images 

3432: The ideal points $e_0, e_1$ of $u_0, u_1$  are

3433: close and bound an interval $I$ which is 

3434: almost all of $\pin g_i(L)$.

3435: 

3436: 

3437: By construction $g_i(l)$ is a

3438: ray which limits in an interval of $\pin g_i(L)$ which

3439: contains $I$ in its interior if

3440: $i$ is big enough. 

3441: There are then  subarcs $\tau_j$ of $g_i(l)$ with endpoints

3442: $a_j, b_j$ in $u_0, u_1$ respectively so that $a_j$ converges

3443: to $e_0$ and $b_j$ converges to $e_1$ and $\tau_j$ converges

3444: to $I$, see fig. \ref{inde}.

3445: Here $i$ is fixed and $j$ varies. 

3446: Since $a_j, b_j$ are in $u_0, u_1$ then they flow 

3447: (by $\wwp$) to points

3448: in $L_0$. The images in $L_0$ are in the same leaf of  $\wls$.

3449: By proposition \ref{conn} these images 

3450: are in the same leaf of

3451: $\widetilde \Lambda^s_{L_0}$.

3452: Hence the whole segment $\tau_j$   flows into

3453: $L_0$.  

3454: 

3455: 

3456: \begin{figure}

3457: \centeredepsfbox{gm5.eps}

3458: \caption{

3459: Transporting the

3460: structure between leaves $g_i(L)$ and $L_0$.}

3461: \label{inde}

3462: \end{figure}

3463: 

3464: 

3465: 

3466: The point $p_0$ flows into $p'$ in $g_i(L)$ under the flow.

3467: The arc $\tau_j$ together with 

3468: subarcs or $u_0, u_1$ from $a_j, b_j$ to $p'$ 

3469: bound a disc $D_j$ in $g_i(L)$. The arguments above show

3470: that the boundary of $D_j$ flows into $L_0$ producing

3471: a curve in $L_0$ bounding 

3472: a disc $B_j$. The segments of $\wwp$ connecting points

3473: in $\partial D_j$ to points in

3474: $\partial B_j$ produce an annulus $C_j$. Then $D_j \cup C_j

3475: \cup B_j$ is an embedded sphere in $\mi$ and hence

3476: bounds an embedded ball. 

3477: Since orbits of $\wwp$ are properly embedded in $\mi$,

3478: it follows that all orbits of $\wwp$ intersecting

3479: $D_j$ will also intersect $B_j$. Hence there is product

3480: flow in this ball. Since this is true for all $j$ then

3481: the union of the $D_j$ flows into $L_0$.

3482: The union of the $D_j$ is the closure of $g_i(L) - W'$.

3483: The image is contained in the closure of $L_0 - W$

3484: in $L_0$ $-$ call the closure $J$. 

3485: 

3486: We claim

3487: that the image is in fact $J$.

3488: All the $\tau_j$ are in the same leaf of $\wls$ and hence

3489: all their flow images  in $L_0$ also are. 

3490: Since rays of $\widetilde \Lambda^s_{L_0}$ are

3491: properly embedded in $L_0$ then 

3492: when $j$ converges to infinity the images of $\tau_j$

3493: in $L_0$ escape

3494: compact sets. This shows the claim.

3495: Therefore the flow produces a homeomorphism between

3496: the closure of $L_0 - W$ and the closure of 

3497: $g_i(L) - W'$. Clearly the same is true for any leaf

3498: in the interval

3499: associated to the contracting (non expanding) directions

3500: $s_0, s_1$. 

3501: In particular we have the following conclusions:

3502: 

3503: \vskip .1in

3504: \noindent

3505: {\bf {Conclusion}} $-$ In any limit leaf $L_0$ with 

3506: a limit direction $r_0$ of bad wedges the following

3507: happens: Let $c$ be the ideal point of $r_0$ and

3508: $A$ a closed interval of $\pin L_0 - \{ c \}$.

3509: Then there is a leaf $l$ of $\widetilde {\Lambda}^s_{L_0}$

3510: with compact subsegments $l_i$ so that the endpoints

3511: of $l_i$ converge to the endpoints $a, b$ of

3512: $A$ and $l_i$ converges to $A$.

3513: In particular $l_i$ escapes compact sets.

3514: There are also subsegments $v_i$ with both endpoints

3515: converging to $a$ and so that $v_i$ converges to 

3516: sets in $\pin L_0$ which contain $A$.

3517: Finally for sufficiently near leaves 

3518: there is a wedge in $L_0$ which forms

3519: a product flow region with these nearby leaves.

3520: \vskip .1in

3521: 

3522: To get the second assertion above just follow $l$ beyond the 

3523: endpoint of $l_i$ near $b$ until it returns near $a$

3524: again.

3525: As a preliminary step to obtain proposition \ref{as1} we prove

3526: the following:

3527: 

3528: 

3529: 

3530: 

3531: 

3532: 

3533: 

3534: \begin{lemma}{}{}

3535: %Either $\fol$ is $\rrrr$-covered or

3536: For any limit $g_i(p_i)$ converging to $p_0$,

3537: the distinguished direction of the bad wedge associated

3538: to $g_i(p_i)$ converges to a single direction at $p_0$.

3539: In the second case

3540: this direction varies continuously with the leaves in $

3541: \widetilde {\ccl}$.

3542: \label{single}

3543: \end{lemma}

3544: 

3545: 

3546: \begin{proof}{}

3547: Suppose there are subsequences $q_i, p_i$ converging to 

3548: points in (interior) $v$ with $g_i(p_i), h_i(q_i) \rightarrow p_0

3549: \in L_0 \in \fn$,

3550: but the directions of the wedges converge to $r_0, r_1$

3551: distinct geodesic rays in $L_0$.

3552: We will first show that there is an interval of leaves

3553: of $\fn$  so that the flow $\wwp$ is a product flow 

3554: in this region.

3555: 

3556: Using the limit direction $r_0$ we produce a wedge

3557: $W$ in $L_0$ so that the closure of $L_0 - W$

3558: is part of a product flow region with nearby leaves

3559: of $\fn$.

3560: Using the other limit direction $r_1$

3561: we produce a flow product region associated to

3562: another wedge region $W_*$ disjoint from $W - p_0$.

3563: Together they produce a global product structure

3564: of the flow in a neighborhood of $L_0$.

3565: 

3566: 

3567: 

3568: 

3569: 

3570: 

3571: 

3572: 

3573: %there are contracting directions (or non expanding

3574: %directions $s_0, s_1$ so that for all sufficiently near to them

3575: %$F$ is $\epsilon$ near $L$ and so flow lines intersecting

3576: %$L_0$ in $s_0, s_1$ will intersect $F_1$ as well. Furthermore since

3577: %$g_i(l_k)$ (here with $k >> i$) is an arc very close to $\pin L_0 - 

3578: %\overline{s_0 s_1}$ then this also goes through from $L_0$ to $L_1$.

3579: 

3580: %The set of points obtained as above forms a sublamination of $\Gamma$.

3581: %Clearly id $g_i(p_i)$ converge to $p_0$ and $q$ is in the same 

3582: %leaf $L_0$ as $p$, then the distance is finite and there are $q_i$ in

3583: %$L$ with $d_L(q_i, p_i)$ bounded and

3584: %$g_i(q_i)$ converging to $q$. Also $q_i$ converges to $v$ in $L$.

3585: %So the set of limits is a sublamination. We extract a minimal sublamination

3586: %$\lal$ out of it.

3587: 

3588: %The region bounded by $g_i(l_k)$ and compact arcs of $s_1, s_0$

3589: %from $p_0$ to $s_1 \cap g_i(l_k)$, and $s_0 \cap g_i(l_k)$ has boundary

3590: %which goes through from $L_0$ to $L_1$. In $g_i(L)$ it also bounds

3591: %a disk $D$. Then $D_0 \cup W \cup D_1$ is a sphere which 

3592: %bounds a ball $B$ in $\mi$. Since orbits of $\widetilde \varphi$ are

3593: %properly embedded, it 

3594: %follows that every orbit intersecting $D_0$ must intersect

3595: %$D_1$ and vice versa. 

3596: 

3597: %The only thing missing from this 

3598: %picture is the small wedge defined by $s_0$ and

3599: %$s_1$. But if we have another sequence converging to another direction

3600: %it will eat up the wedge. 

3601: 

3602: %Notice this depends only on the 

3603: %picture in $L_0$ and not in the limiting leaves.

3604: %%

3605: %\vskip .1in

3606: %\noindent

3607: %{\bf {Conclusion}} - 

3608: 

3609: This shows that there is a neighborhood $N$ of $L_0$ in the

3610: leaf space of $\fn$ so that

3611: the flow is a product flow in $N$.

3612: In particular there is no non Hausdorffness of $\fn$ in

3613: this neighborhood.

3614: This is a very strong property as we shall see below. It 

3615: implies a global product structure of the flow.

3616: 

3617: Notice that the structure of $\widetilde \Lambda^s_{g_i(L)}$ in

3618: $g_i(L) - W'$ flows over to $L_0$. In particular there are 

3619: many rays of $\widetilde \Lambda^s_{L_0}$ which do not

3620: have a single limit in $\pin L_0$. This implies that

3621: $\pi(L_0)$ is not compact. This is because Levitt \cite{Le}

3622: proved that given any singular foliation with prong singularities

3623: in a closed hyperbolic surface $R$, then 

3624: the rays of the lift to $\widetilde R$ all have unique

3625: limit points in the ideal boundary.

3626: This shows that the minimal lamination $\ccl$ is not a compact leaf

3627: and hence it has no compact leaves.

3628: 

3629: 

3630: \vskip .1in

3631: Consider the neighborhood $N$ as above.

3632: Consider the translates $g(N)$ where $g$ runs through all elements

3633: of the fundamental group. 

3634: Let $P$ be the component of the union containing $N$. 

3635: It is easy to see that the

3636: set $P$ is precisely invariant: if $g$ is in $\pi_1(M)$

3637: and $g(P)$ intersects $P$ then $g(P)$ is equal to $P$.

3638: In addition $\fol$ restricted to $P$ has leaf space

3639: homeomorphic to $\rrrr$ because of the product flow

3640: property.

3641: We are assuming that $N$ is open.

3642: 

3643: Suppose first that $P$ is not all  of $\mi$, hence

3644: $\partial P$ is a non empty collection of leaves of $\fn$.

3645: Let $C$ be the projection of $P$ to $M$.

3646: Then $C$ is open, saturated by leaves of $\fol$.

3647: Notice that $g(P)$ does not intersect $\partial P$ for

3648: any $g$ in $\pi_1(M)$ for otherwise $g(P)$ intersects

3649: $P$ and so $g(P) = P$. It follows that $\pi(\partial P)$

3650: is disjoint from $C$ hence $C$ is a proper open, foliated

3651: subset of $M$.

3652: 

3653: Dippolito \cite{Di} developed a theory of such open, saturated

3654: subsets. Let $\overline C$ be the metric completion of $C$.

3655: There is an induced foliation in $\overline C$, which

3656: we will also denote by $\fol$. Then

3657: 

3658: $$\overline C \ \ = \ \ V \ \cup \ \bigcup_1^n V_i$$

3659: 

3660: \noindent

3661: where $V$ is compact and may be all of $\overline C$.

3662: Each nonempty $V_i$ is an $I$-bundle over a non compact

3663: surface with boundary, so that $\fol$ is a foliation

3664: transverse to the $I$-fibers. Each component of

3665: the intersection $\partial V_i \cap V$ is an annulus

3666: (or Moebius band) with induced foliation transverse

3667: to the $I$ fibers.

3668: In our situation with $\Phi$ transverse to the flow,

3669: if $V$ is not $\overline C$, we can choose $V$

3670: big enough so that the flow is transverse to $\fol$ 

3671: in each $V_i$ and induces an $I$-fibration there.

3672: 

3673: 

3674: Let $\widetilde R$ be a component of $\partial P$ and

3675: $R$ the projection to $M$, so $R$ is component

3676: of $\partial C$.

3677: %Suppose first that $R$ is closed. We show this

3678: %is impossible, basically using holonomy.

3679: Parametrize the leaves of $\fn$ in $P$ as

3680: $F_t, 0 < t < 1$ with $t$ increasing with

3681: flow direction. A leaf in the boundary of $P$ which

3682: is the limit of leaves in $P$ which are limiting from

3683: the positive side above has to be the

3684: limit of $F_t$ as $t$ goes to $0$: 

3685: Suppose that $S$ is in the boundary of $P$ and there

3686: are $x_i$ in $F_{t_i}$ with $t_i$ converging

3687: to $t_0 > 0$ 

3688: and $x_i$ converging to $x$ in $S$.

3689: Then $S$ and $F_{t_0}$ are not separated from

3690: each other.

3691: For $i$ big enough the flow line

3692: through $x_i$ will intersect $S$ 

3693: and therefore this flow line will not intersect

3694: $F_{t_0}$. This contradicts the fact

3695: that  $F_{t_o}$ and $F_{t_i}$ 

3696: have a flow product structure.

3697: 

3698: Suppose then that $\widetilde R$ is a limit of $F_t$

3699: where $t$ converges to $0$.

3700: Suppose first that $R$ is compact. 

3701: %Let $x$ be

3702: %a point in $\widetilde R$ and $\gamma$ be the

3703: %flow line of $\wwp$ through $p$. Then $\gamma$ intersects

3704: %$F_t$ for $t$ close to $0$ hence intersects all $F_t$.

3705: Suppose there are $t_i$ converging to $0$ so that

3706: $F_{t_i}$ are in $\widetilde {\ccl}$. Then since

3707: $\ccl$ is a closed subset of $M$ it follows that

3708: $\widetilde R$ is in $\widetilde {\ccl}$ and so

3709: $R$ is in $\ccl$. But $R$ is closed, contradicting

3710: the fact that $\ccl$ has no closed leaves.

3711: There is then $a > 0$ which is the smallest $a$ 

3712: so that $F_a$ is in $\widetilde {\ccl}$ $-$ notice

3713: that $\widetilde {\ccl}$ has leaves in $P$.

3714: For any $g$ in $\pi_1(R)$ then $g(N) \cap N$ is

3715: not empty hence $g(N) = N$. It follows that

3716: $g(F_a) = F_b$ for

3717: some $b$. If $b$ is not $a$ then by taking

3718: $g^{-1}$ if necessary we may assume that $b < a$.

3719: But as $F_b$ is in $\widetilde {\ccl}$, this contradicts

3720: the definition of $a$. Hence $g(F_a) = F_a$ for

3721: any $g$ in $\pi_1(R)$. 

3722: This implies that $\pi(F_a)$ is a closed surface,

3723: again contradiction.

3724: 

3725: We conclude that $R$ is not compact, hence it eventually

3726: enters some $V_i$ (the point here is that $V$ is

3727: not $\overline C$). 

3728: The flow restricted to any component of $\partial V_i

3729: \cap \overline C$ goes from one component to the

3730: other in the annulus. This implies that all $\pi(F_t)$ 

3731: intersect this annulus. 

3732: There is then a leaf $B$ of $\ccl$ which enters

3733: $V_i$. Going deeper and deeper in this non compact $I$-bundle

3734: will produce a limit point which is not in $C$.

3735: This shows the very important fact that $\ccl$ is

3736: not contained in $C$ and therefore

3737: 

3738: $${\cal E} \ = \ \ccl  \cap   (M - C) \ \not  =  \ \emptyset$$

3739: 

3740: 

3741: \noindent

3742: In addition ${\cal E}$ is not equal to $\ccl$ 

3743: since 

3744: $\ccl$ has leaves in $C$

3745: and

3746: $(M - C)$ is closed.

3747: Hence ${\cal E}$ is a non trivial, proper sublamination

3748: of  $\ccl$. This contradicts the fact that $\ccl$ is

3749: a minimal lamination.

3750: 

3751: This shows that the assumption $P \not = \mi$ is

3752: impossible. Hence $P = \mi$, which implies the

3753: flow $\wwp$ produces a global product picture

3754: of $\fn$ and in particular $\fol$ is $\rrrr$-covered,

3755: contrary to assumption.

3756: 

3757: This shows 

3758: %that if $\fol$ is not $\rrrr$-covered, then

3759: the limits of the bad wedges are unique directions

3760: in the limit leaves. It also shows that they vary

3761: continuously from leaf to leaf, for otherwise one

3762: obtains bad wedges in very near leaves which have

3763: definitely separated directions. The same proof above

3764: then applies.

3765: This finishes the proof of lemma \ref{single}.

3766: \end{proof}

3767: 

3768: 

3769: %If $g_1(N)$ intersects $g_2(N)$, then

3770: %their union is a region where there is a product flow. Obviously the

3771: %foliation restricted to these unions has Hausdorff leaf space, homeomorphic

3772: %to an interval in the reals.

3773: %Notice that the structure in any of these product regions, has the same structure

3774: %as the original interval.

3775: %This means that the two ideal points are distinguished and in a product 

3776: %region this carries through. 

3777: %Let $P$ be a maximal product region. Consider the top boundary.

3778: %If the top has more than one leaf in the limit, let $F_1, F_2$ be 

3779: %leaves in the limit. Take a lot of contracting directions in $F_i$

3780: %evenly spaced in $F_i$. The theory of transverse directions

3781: %shows that there is a unique direction associated to $b_1$ in 

3782: %$F_1$ and similarly for $b_2$, see figure ????

3783: %

3784: %\vskip 1.5in

3785: %\centerline{Figure 3, the non Hausdorff situation}

3786: %

3787: %If they are distinct then the previous argument shows that all of 

3788: %$F_1$ flows into $Z$. If they are the same, we get that

3789: %more and more of $F_1$ flows in to $Z$ as $Z$ gets closer

3790: %and closer to $F_1$. Since the $N$ is a product region, then

3791: %all of $F_1$ flows into $Z$ for all $Z$ and vice versa. 

3792: %

3793: %This shows that there can be no other leaf of $\lal$ non

3794: %separated from $F_1$ and the flow is a product in

3795: %$N \cup F_1$. 

3796: %

3797: %	If the distinguished points in $F_1$ are distinct, we can

3798: %iterate this process. There is a small neighborhood of $F_1$ in

3799: %the leaf space where we can extend the product picture with 2 distinct ideal

3800: %points. 

3801: %

3802: %\vskip .2in

3803: %MINIMALITY - aside from previous page?????

3804: %

3805: %We take $\lal^*$ to be a minimal sublamination.

3806: %

3807: %\vskip .1in

3808: %\noindent

3809: %{\underline {Claim}} - $\lal^*$ is not a closed leaf.

3810: %

3811: %If $\lal^*$ is a closed surface $R$, then $\wls_R$ is a singular

3812: %foliation in a closed surface. Levitt \cite{Le} has studied

3813: %this in detail and showed that every leaf converges to

3814: %a single point in $\pin R$. This cannot occur in our situation.

3815: %\vskip .1in

3816: %

3817: %

3818: %Going in the positive direction, this process may either take up

3819: %all the leaves in the positive direction. This proves that $\fol$

3820: %above $L_0$ is $\rrrr$-covered. This implies that $\fol$ in

3821: %an $\rrrr$-covered foliation and finishes the proof in this

3822: %case.

3823: %Otherwise there is a maximal $N_1$ with boundaries $F_0, F_1$

3824: %where the ideal points are identified.

3825: %The interval defined by $F_0, F_1$ and $N_1$ is precisely invariant:

3826: %either $g(I) = I$ or $g(I) \cap I = \emptyset$. For otherwise there is

3827: %$g$ in $\pi_1(M)$ with $g(F_0)$ in $N_1$ 	or $g(F_1)$ in $N_1$.

3828: %

3829: %%	Either the top leaf has a unique bad point or there are at	least 2. 

3830: %%If there are at least 2, we can continue the process, otherwise

3831: %it is precisely invariant. Then $N_1$ projects to 

3832: %an open foliated subset $O$ of $M$ with 

3833: %$\pi(F_0), \pi(F_1)$ its boundaries (may be the same). We can eliminate

3834: %$O \cap \lal*$ from $\lal*$ to obtain another smaller sublamination 

3835: %- but this contradicts minimality of $\lal*$.

3836: %So this option cannot occur.

3837: %

3838: %The leaves are in this $I$-bundle and since the $I$-bundle

3839: %is not compact the leaves limit somewhere in the manifold

3840: %(not themselves). This implies that $\lal* - O$ is non empty.

3841: %Then $\lal* - O \cap \lal*$ is a nontrivial sublamination.

3842: %This is a contradiction.

3843: %

3844: %This finishes the proof of the lemma.

3845: %%\end{proof}

3846: 

3847: \vskip .1in

3848: \noindent

3849: {\underline {Continuation of the proof of proposition \ref{as1}}}

3850: 

3851: %Regardless of whether $\fol$ is $\rrrr$-covered or not

3852: %we get the sequences of bad wedges converging to directions

3853: %in the limit leaves. If $F$ is such a leaf with

3854: %limit direction associated to ideal point $a$ then for

3855: %any closed interval $I$ in $\pin F - \{ a \}$ there

3856: %is a ray $l$ of $\wlsf$ with subarcs $l_i$ which limit

3857: %exactly on $I$.

3858: 

3859: %We continue the proof of the proposition, assuming that $\fol$

3860: %is not $\rrrr$-covered. 

3861: By the previous lemma

3862:  we know that limit directions of

3863: bad wedges are unique and they vary continuously in leaves

3864: of $\widetilde {\ccl}$.

3865: These unique directions are distinguished in their

3866: respective leaves.

3867: 

3868: We first show that any complementary region of $\ccl$

3869: (if any)

3870: is an $I$-bundle with a product flow.

3871: %and use the previous proof to obtain a contradiction.

3872: %Let $A$ be a complementary region of $\lal*$ and let $B$

3873: %be a boundary leaf of $A$.

3874: 

3875: 

3876: Lift to a double cover if necessary to assume that $M$

3877: is orientable. Assume this is the original foliation

3878: $\fol$, flow $\Phi$, etc..

3879: Let $Z$ be a leaf of $\widetilde {\cal L}$. Since $Z$

3880: has a distinguished ideal point, then the fundamental

3881: group of $\pi(Z)$ can be at most ${\bf Z}$. Since there

3882: is a transverse flow and $M$ is orientable this implies

3883: that $\pi(Z)$ is either a plane or an annulus.

3884: 

3885: Let $U$ be a complementary region of $\ccl$ with

3886: boundary leaves $R_1, R_2, R_3$, etc..

3887: As explained before the completion of $U$ has

3888: a compact thick part and the non compact 

3889: arms which are in thin, $I$-bundle

3890: regions. 

3891: Suppose first that $R_1$ is a plane.

3892: There is a big disk $D$ so that $R_1 - D$ is contained

3893: in the thin arms and flows across $U$ to another boundary

3894: components of $U$. By connectedness it flows into a single

3895: boundary component $R_2$ of $U$. Then $\partial D$ flows

3896: into a curve $\gamma$ in $R_2$ which is null homotopic

3897: in $M$. 

3898: The flow segments in $M$ produce an annulus

3899: $C$ in the completion of $U$. Since $\fol$ is Reebless

3900: then $\gamma$ bounds a disk $D'$ in $R_2$ and so $R_2$

3901: is a plane. The union $D \cup C \cup D'$ is an

3902: embedded sphere in $M$ which bounds a ball $B$.

3903: %As explained before 

3904: Since orbits of $\wwp$ are properly embedded in $\mi$,

3905: it follows that 

3906: the flow has to a product flow

3907: in $B$ as well. 

3908: This shows that flow is a product

3909: in the completion of $U$.

3910: 

3911: Suppose now that each $R_i$ is an annulus. Let

3912: $F$ be a lift of $R_1$ to $\mi$ with $F$ in the boundary 

3913: of a component $\widetilde U$ of $\pi^{-1}(U)$.

3914: In $R_1$ there are two disjoint open annuli $A_1, A_2$ contained

3915: in the thin arms so that $B = R_1 - (A_1 \cup A_2)$

3916: is a closed annulus in the core.

3917: Then $A_1, A_2$ flow into two annuli leaves

3918: $R_2, R_3$ in the boundary of $U$.

3919: Lifting to $F = \widetilde R_1$ we see leaves

3920: of $\wlsf$ limiting in an interval of $\pin F$ with very

3921: small complement (near the distinguished ideal

3922: point of $F$). This implies they will have

3923: points in the lifts $\widetilde A_1, \widetilde A_2$

3924: of $A_1, A_2$ to $F$.

3925: This shows that $\widetilde A_1, \widetilde A_2$

3926: are in the same leaf of $\fn$.

3927: This implies that $R_2 = R_3$. 

3928: In the same way a half of the infinite strip $\widetilde B$

3929: flows into $\widetilde R_2$. Since $B$ is compact,

3930: then all of $B$ flows into $R_2$.

3931: %The flow segments from

3932: %of $\partial A_1, \partial A_3$ to $R_2$ produce

3933: %two annuli $C_1, C_2$ in the completion of $U$.

3934: %The boundaries of $C_1, C_2$ in $R_2$ are freely

3935: %homotopic since $R_2$ is an annulus, hence they

3936: %jointly bound an annulus $C_3$. Then $B \cup C_1 

3937: %\cup C_2 \cup C_3$ is an embedded $2$-torus in

3938: %the closure of $U$. As $M$ is atoroidal

3939: %then this torus either bounds a solid torus

3940: %or has to be contained in a $3$-ball. The second

3941: %option is disallowed since the annulus $B$ is

3942: %$\pi_1$-injective in $M$.

3943: %It follows that the $2$-torus bounds

3944: %a solid torus $V$,

3945: %which must be contained

3946: %in $U$. 

3947: This implies that the region $U$ is an $I$-bundle.

3948: It is also easy to show that the flow is a product

3949: in this $I$-bundle.

3950: 

3951: This implies that we can collapse this complementary region

3952: along flow lines to completely eliminate it.

3953: This is because even in the universal cover we are 

3954: eliminating product regions of the flow and the

3955: asymptotic behavior is still preserved in the remaining

3956: regions.

3957: This can be done to all complementary regions and

3958: therefore we can assume there are no complementary

3959: regions, that is $\ccl = \fol$ or that $\fol$ is minimal.

3960: 

3961: 

3962: %Suppose now that $\fol$ is not $\rrrr$-covered.

3963: Let $F_1, F_2$ be leaves of $\fn$ which are not

3964: separated from each other. Consider leaves $F$ of 

3965: $\fn$ which are very close to points in both $F_1$

3966: and $F_2$. As stated in the conclusion in the beginning

3967: of the proof of this theorem,

3968: there is a wedge of $F$ which flows into $F_1$

3969: and similarly for $F_2$. Hence there are half planes

3970: $E_1, E_2$ of $F$ which flow into $F_1, F_2$.

3971: As $F_1, F_2$ are not separated this implies that

3972: $E_1, E_2$ are disjoint. Fix a point $w$ in $F$ and

3973: a big enough radius $r$ so that the disk $D$ of radius

3974: $r$ around $w$ intersects both $E_1, E_2$.

3975: Again as seen in the conclusion above

3976: there is an arc $l$ in a leaf of $\wlsf$

3977: so that both endpoints of $l$ are outside $D$ and in $E_1$

3978: and so that $l$ is entirely outside $D$ and as seen

3979: from $p$ the visual measure of $l$ is almost $2 \pi$.

3980: This implies that $l$ intersects $E_2$. Since the

3981: endpoints of $l$ are in $E_1$, which flows to $F_1$,

3982: then proposition \ref{conn} implies that the whole arc $l$ flows

3983: into $F_1$. The points of $l$ in $E_2$ will also flow

3984: to $F_2$. This is a contradiction.

3985: 

3986: This contradiction 

3987: %shows that $\fol$ has to be $\rrrr$-covered

3988: finishes the proof of proposition \ref{as1}

3989: \end{proof}

3990: 

3991: Next we analyse the $\rrrr$-covered situation which has interest

3992: on its own:

3993: 

3994: \begin{theorem}{}{}

3995: Let $\fol$ be an $\rrrr$-covered foliation and $\Phi$ be a pseudo-Anosov flow

3996: almost transverse to $\fol$. Then $\Phi$ is actually transverse to $\fol$.

3997: In addition for any leaf $F$ of $\fn$ and for any ray $l$ in $\wlsf$

3998: it converges to a unique ideal point in $\pin F$.

3999: The limit usually depends on $l$.

4000: \label{as2}

4001: \end{theorem}

4002: 

4003: \begin{proof}{}

4004: If $\Phi$ is not transverse to $\fol$, let 

4005: $\Phi^*$ be an almost 

4006: pseudo-Anosov flow which is transverse to $\fol$ and

4007: is a blow up of $\Phi$. 

4008: Notice this is not the same notation as in proposition \ref{as1}

4009: $-$ here we prove $\Phi$ is equal to $\Phi^*$.

4010: There is flow annulus $A$ of $\Phi^*$

4011: with closed orbits $\gamma_1, \gamma_2$ in

4012: the boundary, so that $A$ blows down to a single orbit of $\Phi$.

4013: 

4014: The foliation induced by $\fol$ in $A$ 

4015: has leaves which spiral to at least one

4016: boundary component $-$ which they do not intersect.

4017: Lifting this picture to the universal cover one

4018: obtains an orbit of $\wwp^*$ which does not intersect

4019: every leaf of $\fn$. This means that the flow $\wwp^*$

4020: is not {\em regulating} for $\fn$ \

4021: \cite{Th6, Th7}. 

4022: We also say that $\Phi^*$ does not regulate $\fol$.

4023: In \cite{Fe9} we analysed a similar situation and proved 

4024: the following: if $\Upsilon$ is a pseudo-Anosov flow transverse

4025: to an $\rrrr$-covered foliation and $\Upsilon$ is not regulating,

4026: then $\Upsilon$ is an $\rrrr$-covered

4027: Anosov flow.

4028: The same arguments work with an almost pseudo-Anosov flow

4029: transverse to an $\rrrr$-covered foliation.

4030: This shows that $\Phi^*$ is an $\rrrr$-covered Anosov flow

4031: and has no (topological) singularities.

4032: In particular $\Phi^*$ is equal to $\Phi$, that is

4033: the original flow is already transverse to $\fol$.

4034: This proves the first assertion of the theorem.

4035: 

4036: %The same article shows that if $\Phi$ is not regulating, then the intersection

4037: %of $\fs$ and $\fol$ is always a quasigeodesic (verify this part????).

4038: 

4039: %From now on, we assume that $\Phi$ is 

4040: %regulating for $\fol$. 

4041: Assume by way of contradiction that there

4042: is $L'$ in $\widetilde {\Lambda}^s$ and $l$ in 

4043: $\widetilde {\Lambda}^s_{L'}$ which 

4044: does not converge to a single point in $\pin L'$.

4045: As in the proof of theorem \ref{as1} we construct a minimal

4046: sublamination $\ccl$ of $\fol$ such that:

4047: for every $L$ in $\widetilde {\ccl}$ there is

4048: an ideal point $u$ in $\pin L$ so that for every closed

4049: segment $J$ in $\pin L - \{ u \}$ there is a ray

4050: $l$ of $\widetilde {\Lambda}^s_L$ which has subsegments

4051: limiting to $J$..

4052: As shown in the proof of theorem \ref{as1}, 

4053: $\ccl$ cannot be

4054: a compact leaf.

4055: 

4056: %In fact if $\fol$ has a compact leaf then every minimal set is a compact

4057: %leaf: use Brittenham's result \cite{Br1}

4058: %to obtain the result on a manifold with boundary.

4059: 

4060: %So we can assume that 

4061: %Hence $\ccl$ does not have compact leaves.

4062: %Under these conditions we proved in \cite{Fe4} that

4063: %there is a unique minimal set of $\fol$, which must

4064: %be $\ccl$.

4065: %Also the complementary regions of $\ccl$ are $I$-bundles

4066: %where the flow is a product.

4067: %Hence we can collapse all complementary regions of $\ccl$

4068: %to obtain a new foliation which is minimal, but still transverse to the

4069: %flow. The asymptotic properties of are the same, so we assume from

4070: %now on that this is the original foliation, that is that $\fol$ is minimal.

4071: 

4072: %Let then $l$ in $\widetilde g$ and $l$ ray of $\wls_l$ with

4073: %$l$ limiting. As before produce $p_i$ in $L$ and $g_i(p_i)$

4074: %converging to $p_0$ and $L*$ and every leaf?????

4075: 

4076: %\vskip .1in

4077: %\noindent {\underline {Claim}} - If $u \in \pin F$ is an ideal point

4078: %of shrinking behavior then the same occurs in every other leaf.

4079: %

4080: %For any $u_1, u_2$ sufficiently near $u$ there is a single small

4081: %intervalin $\pin F$ connecting $u_1, u_2$. If $S$ is any other

4082: %leaf there is a homeomorphism $\pin F$ to $\pin S$.

4083: %For any $S'$ between $F$ and $S$, there is a contracting

4084: %direction in the middle of $S$. Now

4085: %cover $I$ by these intervals. Only finitely many to go from $\pin F$ to

4086: %$\pin S$.  So the limiting behavior comes through the intersection of these.

4087: %Hence through $u_1, u_2$. Since $u_1, u_2$ are arbitrary, the whole picture goes

4088: %through.

4089: %

4090: %\vskip .1in

4091: %\noindent {\underline {Case 1}} - There are $a, b$ distinct in $\uu$

4092: %which are ideal points of shrinking neighborhoods.

4093: %

4094: %

4095: %

4096: %

4097: %

4098: %Then there is a covering translation $g$ leaving invariant a leaf $F \in \wl$

4099: %with ideal points which are not of $a, b$. 

4100: %Suppose $b$ is not ideal point. COnsider the sequence of points

4101: %associated to $b$ in $L$, where we assume that they

4102: %are not separated by the ideal points of $g$.

4103: %Choose one subarc intersection with axis which is

4104: %very far away both endpoints and does not intersect

4105: %$\cal A$ in the interior and the distance between intersections

4106: %is less than the translation length of $\cal A$. 

4107: %Let $e_1, e_2$ be the endpoints of this arc.

4108: %Then there is an integer $n$ with $g^n(a')$ in the open

4109: %arc $(e_1, e_2)$. But then $g^n(e_2)$ is on the other side

4110: %and $g^n(a)$ intersects $a$ transversely which

4111: %is a contradiction, see fig 4?????

4112: %\vskip 1.5in

4113: %\centerline{figure 4, intersections being forced}

4114: 

4115: 

4116: Suppose first that every leaf of $\fol$ is a plane.

4117: Then Rosenberg \cite{Ros} proved that $M$ is the $3$-dimensional

4118: torus $T^3$. 

4119: This manifold is a Seifert fibered space. In this case

4120: Brittenham \cite{Br1} proved that an essential lamination

4121: is isotopic to one which is either vertical (a union

4122: of Seifert fibers) or horizontal (transverse to the 

4123: fibers). So after isotopy assume $\ccl$ has one

4124: of these types. If $\ccl$ has a vertical leaf $B$, then

4125: geometrically it is a product of the reals with the circle.

4126: Hence it is an Euclidean  leaf and in the universal

4127: cover it has polynomial growth of area.

4128: If $\ccl$ has

4129: a horizontal leaf $B$, then

4130: because the fibration is a product,

4131: there is a projection to a $T^2$ fiber, which distorts

4132: distances by a bounded amount. Again the same growth properties

4133: hold.  But the leaves of $\fol$ are hyperbolic, which is 

4134: a contradiction. We conclude that $M$ cannot be $T^3$.

4135: 

4136: 

4137: 

4138: 

4139: 

4140: %\vskip .1in

4141: %noindent

4142: %\underline {Case 2}} - In this case $a$ in $\uu$ is uniquely determined

4143: %nd it is invariant under the group action.

4144: 

4145: 

4146: 

4147: 

4148: %\blankfig{cros}{1.1}{Forcing the crossing of leaves.}

4149: 

4150: Let then $F$ in $\widetilde {\ccl}$ with $\pi(F)$ not simply 

4151: connected. Let $g$ in $\pi_1(M)$ non trivial with $g(F) = F$

4152: and $\xi$ be the axis of $g$ in $F$. At least one ideal point

4153: of $\xi$, call it $u$, is not the direction of a fixed 

4154: limit of bad wedges. Then as explained before there is a 

4155: ray $l$ of $\wlsf$ and segments $l_i$ 

4156: of $l$, bounded by $a_i, b_i$ both points

4157: in $\xi$, so that $l_i$ escapes compact sets

4158: and converges to a non trivial segment in $\pin F$.

4159: We may assume that $l_i \cap \xi = \{ a_i, b_i \}$ and also

4160: that all $l_i$ are in the same side of $\xi$.

4161: Let $e_0$ be the translation length of $g$ in $F$.

4162: 

4163: If the distance from $a_i$ to $b_i$ along $\xi$ is bigger than $e_0$ then

4164: this produces a contradiction as follows:

4165: There is an integer $n$ so that $g^n(a_i)$ is  in the 

4166: open segment $(a_i,b_i)$ of $\xi$ and

4167: and $g^n(b_i)$ is outside of the closed segment $[a_i,b_i]$.

4168: Since the arc $l_i$ only intersects $\xi$ in $a_i, b_i$, then 

4169: $l_i$, together with $[a_i,b_i]$

4170: bounds a closed disk in $F$ and $g^n(a_i)$ is in $(a_i,b_i)$.

4171: But $g^n(b_i)$ is outside and $g^n(l_i)$ is also on this side

4172: of $\xi$, so this produces

4173: a transverse self intersection of $\wlsf$.

4174: If $g^n(l_i)$ is contained in the leaf $v$ which contains $l_i$,

4175: then $g^n(v) = v$ and this produces infinitely many

4176: singularities in $v$, which is impossible. Hence $g^n(l_i)$ is

4177: not in $v$ and the transverse intersection is impossible.

4178: The same arguments deal with the case that $l_i$ intersects

4179: $\xi$ in other points besides $a_i, b_i$.

4180: 

4181: %First i%f $l_i$ intersects $a$ in the

4182: %other end we can map in the middle and again get a

4183: %self intersection.

4184: 

4185: We conclude that the distance in $\xi$ from $a_i$ to $b_i$ is bounded.

4186: %But $d_{l_i}(b_i,c_i)$ is not. 

4187: Up to subsequence we may assume there are integers $n_i$

4188: so that 

4189: $g^{n_i}(a_i)$ converges to $a_0$ and $g^{n_i}(b_i)$ converges to $b_0$,

4190: both limits in $\xi$ of course.

4191: Since the lengths of $g^{n_i}(l_i)$ are converging to infinity,

4192: it follows that $a_0, b_0$ are not in the same leaf of $\wlsf$.

4193: By proposition \ref{conn} it follows that $a_0, b_0$ are not

4194: in the same leaf of $\wls$. But for each $i$, the pair

4195: of points $g^{n_i}(a_i), g^{n_i}(b_i)$ is in the same leaf

4196: of $\wls$. This implies that the leaf space of $\wls$ is

4197: not Hausdorff.

4198: 

4199: First of all this implies that $\Phi$ is regulating for $\fol$,

4200: for otherwise the aforementioned result from \cite{Fe9}

4201: shows that $\Phi$ is an $\rrrr$-covered Anosov flow $-$ in

4202: particular $\wls$ has Hausdorff leaf space.

4203: Also by theorem \ref{theb} the fact that $\wls$ has non Hausdorff

4204: leaf space implies that there are closed orbits $\alpha, \beta$

4205: of $\Phi$ so that $\alpha$ is freely homotopic to the inverse

4206: of $\beta$.

4207: Let $h$ be a covering translation associated to $\alpha$ and

4208: $\widetilde \alpha$, $\widetilde \beta$ lifts of $\alpha, \beta$

4209: to $\mi$ which are left invariant by $h$. Without loss of generality

4210: assume that $h$ acts in $\widetilde \alpha$ sending points forwards.

4211: As $\alpha \cong \beta^{-1}$ this implies that $h$ acts

4212: on $\widetilde \beta$ taking points backwards. But since

4213: both of them intersects all leaves of $\fn$ (by the regulating

4214: property) then

4215: as seen from $\widetilde \alpha$ the translation $h$ acts

4216: increasingly in the leaf space of $\fn$, with opposite

4217: behavior when considering $\widetilde \beta$.

4218: This is a contradiction, which shows that this cannot happen.

4219: This finishes the proof of theorem \ref{as2}.

4220: \end{proof}

4221: 

4222: 

4223: %Because the lengths of the $l_i$ between $b_i$ and $c_i$ 

4224: %grow without bound, then $b_0, c_0$ are not in the same

4225: %leaf. This implies that the leaf space of

4226: %$\wl$ is not Hausdorff. This implies that there are closed

4227: %orbits $\alpha, \beta$ of $\Phi$ in $M$ which are freely

4228: %homotopic to the inverse of each other. But since

4229: %$\Phi$ is regulating, this is impossible. This finishes

4230: %the proof of the theorem.

4231: 

4232: Proposition \ref{as1} and theorem \ref{as2} imply the following:

4233: 

4234: \begin{corollary}{}{}

4235: Let $\Phi$ be an almost pseudo-Anosov flow transverse to

4236: a foliation $\fol$ with hyperbolic leaves in $M^3$ closed.

4237: For any leaf $L$ of $\fn$ and any ray $l$ of $\wlsf$

4238: or $\wluf$, then $l$ converges to a single point in $\pin L$.

4239: \end{corollary}

4240: 

4241: 

4242: 

4243: \vskip .1in

4244: \noindent

4245: {\bf {Remark}} - Group invariance and compactness of $M$ are both

4246: essential here.  For example start with a nicely behaved singular foliation

4247: of $\hh$, so that all rays converge. It could be a foliation by geodesics

4248: or for instance the lift of the stable singular foliation associated

4249: to a suspension. 

4250: Fix a base point $p$. Now rotate the leaves at a distance $d$ of $p$ by

4251: an angle $d$. In this situation all rays limit in all points of $\pin L$,

4252: in fact they spiral indefinitely into it.

4253: Another operation is to fix a ray through $p$ and then distort the rest more

4254: and more one way and the other way. Here we have the leaves getting

4255: closer and closer to segments in $\pin F$ which 

4256: are complementary to the ideal point associated to the

4257: ray.

4258: 

4259: 

4260: 

4261: 

4262: \section{Properties of leaves of $\wlsf, \wluf$ and their ideal points}

4263: 

4264: In this section $\Phi$ is an almost pseudo-Anosov flow

4265: transverse to a foliation $\fol$. 

4266: As in the previous section

4267: there is no restriction on $M$ here.

4268: In the previous section we proved that

4269: for any ray $r$ of a leaf of $\wlsf$ or $\wluf$, then

4270: it has a unique ideal point in $\pin F$. The notation for this ideal

4271: point will be $r_{\infty}$.

4272: We now analyse further properties of leaves of $\wlsf$ and their

4273: ideal points.

4274: Analogous results hold for $\wluf$.

4275: 

4276: First we want to show that if $E$ is a fixed

4277: leaf of $\wls$ (or $\wlu$)

4278: then the ideal points in $\pin F$ of rays of $E \cap F$

4279: vary continuously with $F$.

4280: In order to do that we first

4281: put a topology on the union of ideal boundaries of an interval

4282: of leaves. Let $p$ in $F$ leaf of $\fn$ and $\tau$ a transversal

4283: to $\fn$ with $p$ in the interior.

4284: For any $L$ in $\fn$ intersecting $\tau$, the ideal

4285: boundary is in 1-1 correspondence with the unit

4286: tangent bundle to $L$ at $\tau \cap L$:

4287: ideal points correspond to rays in $L$ starting at $L \cap \tau$.

4288: This is a homeomorphism.

4289: This puts a topology in 

4290: 

4291: $$ {\cal A}  \ \ = \ \ 

4292: \cup \ \{ \pin L \ \ | \ \ L \cap \tau \not = \emptyset \}$$

4293: 

4294: \noindent

4295: making it into an annulus homeomorphic to 

4296: $\cup  \ \{ T^1_q \fn, q \in \tau \}$ as a subspace

4297: of the unit tangent bundle of $M$.

4298: This topology in ${\cal A} $ is independent of the choice of transversal

4299: $\tau$. The following definition/result is proved

4300: in \cite{Fe7} or \cite{Cal1}.

4301: 

4302: \begin{define}{(markers)}{}

4303: Given a foliation $\fol$ by hyperbolic leaves

4304: of $M^3$ closed, then  there is $\epsilon > 0$ so

4305: that:

4306: Let $v$ be a geodesic ray in a leaf $F$ so that 

4307: it is associated to a contracting (or $\epsilon$ non

4308: expanding) direction of $F$.

4309: For any leaf $L$ sufficiently near $F$, then

4310: all the points of $v$ flow into $L$ and define

4311: a curve denoted by $v_L$. 

4312: Then $v_L$ has a unique ideal point denoted by $a_L$. The

4313: union $m$ of the $a_L$ is called a marker and is

4314: a subset of ${\cal A}  = \cup \ \{ \pin L \}$. 

4315: Then $m$ is an embedded curve in ${\cal A} $ in the topology

4316: defined above.

4317: \end{define}

4318: 

4319: 

4320: In addition the markers are dense in ${\cal A} $ in the following sense:

4321: Let $z$ in $\pin F$ and $a_i, b_i$ in $\pin F$ which are 

4322: in markers associated to contracting (non expanding) directions

4323: on a fixed side of $F$. Suppose that the sequence of open 

4324: intervals

4325: $(a_i, b_i)$ in $\pin F$ contains $z$ and converges

4326: to $z$ as $i$ converges to infinity.

4327: Let $\alpha_i, \beta_i$ be the markers in that

4328: side of $\pin F$ containing $a_i, b_i$ respectively.

4329: Let $L_i$ in $\fn$ be a sequence of leaves converging

4330: to $F$ and on that side of $F$ so that $\pin L_i$ intersects

4331: both $\alpha_i$ and $\beta_i$.

4332: In 

4333: the annulus ${\cal A} $ of circles at infinity, consider the

4334: rectangle $R_i$ bounded by $(a_i, b_i)$ in $\pin F$,

4335: the parts of $\alpha_i, \beta_i$ between $\pin F$ and

4336: $\pin L_i$ and the small segment in $\pin L_i$ bounded

4337: by \ $\pin L_i \cap \alpha_i$ \ and \ $\pin L_i \cap \beta_i$.

4338: Then the sets $R_i$ converge to $z$  as $i$ converges

4339: to infinity. This is proved in \cite{Fe7}.

4340: 

4341: 

4342: From now on the $\epsilon$ is chosen small enough

4343: to also satisfy

4344: the conclusions of the definition above and also that

4345: any set in $\mi$ of diameter less than $10 \epsilon$

4346: is in a product box of $\fn$ and $\wwp$.

4347: Given a curve $\zeta$ in a leaf $F$ with starting

4348: point $p$ and limiting on a unique point $q$ in $\pin F$,

4349: let $\zeta^*$ denote the geodesic ray of $F$ with

4350: same starting and ideal points.

4351: 

4352: 

4353: 

4354: 

4355: 

4356: 

4357: \begin{lemma}{}{}

4358: Let $E$ be a leaf of $\wls$ and $p$ the starting point of the ray $r$ of 

4359: $E \cap F$.

4360: Assume that $r$ does not have any singularity.

4361: For any $L$ near $F$, then $E \cap L$ has a ray

4362: $r_L$ which is near $r$. The ideal points

4363: of $r_L$ in $\pin L$ vary continuously with $L$

4364: in the topology of ${\cal A}$ defined above.

4365: \label{cont}

4366: \end{lemma}

4367: 

4368: \begin{proof}{}

4369: We do the proof for say the positive side of $F$.

4370: We consider $r$ without singularity or else we would

4371: have to check the 2 exterior rays in $\wlsf$ emanating from

4372: $p$. We can always get a subray of $r$ which has

4373: no singularities.

4374: 

4375: Let $u = r_{\infty}$.

4376:  Choose 

4377: contracting (or $\epsilon$ non expanding) directions  

4378: in both sides of $u$, with ideal points very close to $u$.

4379: Let them be defined by geodesic rays $r_0, r_1$ starting at $p$.

4380: There is $\tau$ a small flow segment starting at $p$ and

4381:  in that side of $F$  so that for any $L$ intersecting

4382: $\tau$, then $L$ is asymptotic to $F$ along the

4383: $r_0, r_1$ rays, or at least always $\leq \epsilon$ from $F$.

4384: Hence $r_0, r_1$ flow along $\wwp$ to $L$. Let 

4385: $s_0, s_1$ be the flow images in $L$. 

4386: The $\epsilon$ is also chosen small enough so that 

4387: $s_0, s_1$ have geodesic curvature very small

4388: (this $\epsilon$ depends only on $M$ and $\fol$).

4389: In particular the curves $s_0, s_1$ 

4390: are a small bounded distance (depending only on $\epsilon$) 

4391: from

4392: the corresponding geodesic arcs $s^*_0, s^*_1$.

4393: Let the 

4394: ideal points of $s_0, s_1$  in $\pin L$

4395: be denoted by  $v_0, v_1$

4396: and let $J_L$ be the small closed interval

4397: in $\pin L$ bounded by $v_0, v_1$.

4398: Then $v_0, v_1$ 

4399: are in the markers

4400: associated to $r_0, r_1$ respectively and so they vary continuously

4401: with $L$.

4402: 

4403: 

4404: \begin{figure}

4405: \centeredepsfbox{gm6.eps}

4406: \caption{

4407: Leaf in wedge defined by markers.}

4408: \label{bet}

4409: \end{figure}

4410: 

4411: 

4412: 

4413: Consider $\xi = E \cap L$ and the rays $l$ of

4414: $\xi$ starting at $\tau \cap L$ and

4415: containing some points which flow back to points

4416: in $r$. It may be that $\xi$ has singularities $-$

4417: even if $r$ does not $-$ but there are only

4418: finitely many such rays.

4419: We want to prove that the ideal point of any such 

4420: is in $J_L$.

4421: As the rectangles $R_i$ defined above converge to 

4422: $u$ in 

4423: ${\cal A}$ this will prove the continuity property of the lemma.

4424: 

4425: 

4426: Choose $d > 0$ so that outside of a disk $D$

4427: of radius $d$ in $F$, then

4428: $r$ is in the small wedge $W$ of $F$ defined by $r_0, r_1$,

4429: see fig. \ref{bet}.

4430: Choose $\tau$ small enough so that if $L$ intersects $\tau$, then

4431: the entire disk $D$ is

4432: $\epsilon$ near $L$.

4433: Let $V$ be the closure in $F$ of $W - D$.

4434: The boundary $\partial V$ consists of subrays 

4435: of $r_0, r_1$ and an arc in $\partial D$. Therefore

4436: all points in $\partial V$ are less than $\epsilon$

4437: from $L$ and flow to $L$ under $\wwp$ with image

4438: a curve $\gamma$. This curve contains subrays of 

4439: $s_0, s_1$ and it is properly embedded in $L$.

4440: Points of $F$ near $\partial V$ also flow to $L$ so there

4441: is a unique component $U$ of $L - \gamma$ which has

4442: some points flowing back to points in $V$.

4443: We want to show that the ray $l$ is eventually

4444: contained in $U$.

4445: 

4446: Let $r_{init}$ be the subarc of $r$ between $p$ and

4447: the last point $c_0$ of $r$ in $D$.

4448: As $p$ and $c_0$ flow into $L$, then proposition \ref{conn}

4449: shows that the entire arc $r_{init}$ flows into $L$

4450: and let $\delta$ be its image in $L$.

4451: As $r$ is singularity free, then so is $\delta$ and

4452: hence $\delta$ is contained in any ray $l$ of $E \cap L$

4453: in that direction.

4454: After $c_0$ the curve $r$ enters $V$ and so $l$ must

4455: enter $U$ after $\delta$.

4456: If after that the ray $l$ exits $U$ then it must

4457: cross $\partial U = \gamma$ in some point, call it $c_1$.

4458: But $c_1$ flows back to $F$ and one can apply

4459: proposition \ref{conn} again in the backwards direction

4460: to show that $c_1$ has to flow to a point in $r$.

4461: This contradicts the choice of $c_0$.

4462: 

4463: This shows that $l$ is eventually entirely contained

4464: in $U$ and therefore $l_{\infty}$ is a point

4465: in $J_L$. This shows the continuity property 

4466: as desired and finishes the proof of the lemma.

4467: \end{proof}

4468: 

4469: Now we have a property which will be crucial to a lot of our analysis.

4470: 

4471: \begin{proposition}{}{}

4472: Suppose that $\fol$ is not topologically conjugate to the

4473: stable foliation of a suspension Anosov flow.

4474: Then the set of ideal points of rays of $\wlsf$ is dense in

4475: $\pin F$.

4476: \end{proposition}

4477: 

4478: \begin{proof}{}

4479: Suppose that there is $F$ in $\fn$ so that the set of

4480: ideal points in $\wlsf$ is not dense in $\pin F$.

4481: Let $J$ be an open interval in $\pin F$ free of such

4482: ideal points. Choose $p_i$ in $F$, $p_i$ converging

4483: to a point in $J$. 

4484: The visual angle of $J$ as seen from $p_i$ converges

4485: to $2 \pi$, so the complementary wedge $W_i$

4486: with corner  $p_i$ has angle which converges to

4487: zero. Up to subsequence assume that 

4488: $g_i(p_i)$ converges to $p_0$ in a leaf $L$ of $\fn$

4489: and the small wedges $g_i(W_i)$ converge to a geodesic

4490: ray $s$ in $L$ with ideal point $z$.

4491: 

4492: \vskip .15in

4493: \noindent

4494: {\underline {Claim }} $-$ In $L$ all the rays of $\wlsl$ converge

4495: to $z$.

4496: 

4497: %This is very similar to previous arguments. 

4498: Suppose there is

4499: $x$ different from $z$ which is an ideal point of a ray

4500: $r$ in $\wlsl$. Then $r$ is contained in $\ws(c_0)$ for

4501: some $c_0$ in $\mi$  and

4502: for $g_i(F)$ sufficiently near $L$ then $\ws(c_0)$ intersects

4503: $g_i(F)$. Any 

4504: ray of $\ws(c_0) \cap g_i(F)$ which is near $r$ will have ideal point

4505: near $x$ in the topology of corresponding annulus ${\cal A}$

4506: of ideal circles near $\pin L$. This is a consequence of the previous

4507: lemma.

4508: But $g_i(W_i)$ converges to $r$ in this topology of ${\cal A}$,

4509: so the sets $g_i(\pin F - J)$ converge to $z$ in ${\cal A}$.

4510: There are no ideal points of leaves of 

4511: $\widetilde {\Lambda}^s_{g_i(F)}$ in $g_i(J)$.

4512: This contradicts the fact that the ideal points above are

4513: very near $x$ and proves the claim.

4514: \vskip .1in

4515: 

4516: The proof of the proposition

4517: is similar to that of theorem \ref{as1}.

4518: As in that theorem consider the set of possible limits

4519: $g_i(p_i)$ as above. This projects to a lamination in $M$

4520: and let ${\cal L}$ be a minimal sublamination.

4521: The claim shows that each leaf of $\wcl$ has a distinguished

4522: ideal point towards which all rays of $\wlsl$ converge.

4523: The arguments in the claim also prove that if $\tau$ is

4524: a transversal to $\fn$, then the ideal points of

4525: leaves of $\wcl$ intersecting $\tau$ vary continuously

4526: in the corresponding ideal annulus. Because of the

4527: distinguished ideal point property, then each leaf of

4528: ${\cal L}$ has fundamental group at most ${\bf Z}$.

4529: If needed lift to a double cover so that all leaves

4530: of $\fol$ are orientable. Hence a leaf of ${\cal L}$ is

4531: either a plane or an annulus.

4532: 

4533: 

4534: Consider a complementary component $U$ of ${\cal L}$

4535: and a boundary leaf $A$ of $U$. If $A$ is a plane then

4536: as in the proof of theorem \ref{as1},

4537: the region $U$ is an $I$-bundle over $A$ and the flow

4538: $\Phi$ is a product in $U$. This region can be

4539: collapsed away.

4540: 

4541: Suppose now that $A$ is an annulus.

4542: Assume that flow lines through $A$ flow into $U$.

4543: Again we want to show that $U$ is a product region.

4544: As in the proof of theorem \ref{as1} let $A_1, A_2$ be

4545: two noncompact, disjoint annuli in $A$ with $A - (A_1 \cup A_2)$

4546: a compact annulus and $A_1, A_2$ contained in the 

4547: thin, $I$-bundle region. Then $A_1, A_2$ flow entirely

4548: into leaves $B$ and $C$ in $\partial U$. 

4549: %If $B = C$ then

4550: %as in theorem \ref{as1}, the region $U$ is a product with

4551: %product flow.

4552: Suppose first that $B, C$ are different. Lift to the

4553: universal cover to produce lifts $\widetilde U, \widetilde

4554: A, \widetilde A_1, \widetilde A_2, \widetilde B, \widetilde C$.

4555: Then $\widetilde A_1, \widetilde A_2$ are disjoint half

4556: planes of $\widetilde A$ which flow positively respectively

4557: into $\widetilde B$ and $\widetilde C$. 

4558: Let $g$ be the generator of the isotropy group of $\widetilde A$,

4559: which has fixed points in $z, x$ where $z$ is the

4560: distinguished ideal point in $\widetilde A$.

4561: The argument will show there is a leaf in

4562: $\widetilde {\Lambda}^s_{\widetilde A}$ which also has

4563: ideal point in $x$, contradiction.

4564: 

4565: From a point in 

4566: $\widetilde A_1$ draw a geodesic segment of 

4567: $\widetilde A$ to a point

4568: in $\widetilde A_2$. Let $p$ be the first point

4569: of this segment which does not flow positively into 

4570: $\widetilde B$.

4571: Then $\Theta(p)$ is in the boundary of $\Theta(\widetilde B)$.

4572: Also  points in the segment near $p$ flow to $\widetilde B$

4573: in positive time, hence there is a slice leaf $l$ of

4574: $\oos(\Theta(p))$  which is in the boundary of $\Theta(\widetilde B)$.

4575: Notice that every point in $l$ is a limit of points in $\Theta(B)$ on

4576: that side.

4577: The set $(l \times \rrrr)$ intersects $\widetilde A$ in at least

4578: $p$: if

4579: $l$ is contained in $\Theta(\widetilde A)$ then it 

4580: generates a properly embedded copy of the reals

4581: in a leaf $s$ of $\widetilde {\Lambda}^s_{\widetilde A}$

4582: otherwise the part that is contained  in $\Theta(\widetilde A)$

4583: also does. 

4584: Every point of $s$ is a limit of points  that flow

4585: %of $\widetilde A_1$ on

4586: %that side of $s$, so that these near points flow 

4587: positively into $\widetilde B$. Therefore no point in $s$

4588: can flow positively in $\widetilde C$ or else we would have

4589: points flowing both in $\widetilde B$ and $\widetilde C$.

4590: 

4591: 

4592: This shows that the leaf $s$ of $\widetilde {\Lambda}^s_{\widetilde A}$

4593: is a bounded distance from the axis $r$ of $g$.

4594: Iterate $s$ by powers of $g$ acting with $z$ as an expanding

4595: fixed point.

4596: The iterates $g^n(s)$ with $n > 0$ are all distinct. 

4597: Either they are all nested or they are disjoint.

4598: If they are not nested since they all have to be in a bounded

4599: distance neighborhood of the axis of $g$ and have both

4600: endpoints in $z$, then eventually they will have two 

4601: points which are far along the leaf, but close in $\widetilde A$.

4602: By Euler characteristic reasons, this would force

4603: a center or one prong singularity, which is impossible.

4604: Hence they are nested, increasing and 

4605: they limit to a leaf of $\widetilde {\Lambda}^s_{\widetilde A}$

4606: which has ideal limit points in $z$ and $x$. This is a contradiction.

4607: This shows that $B = C$.

4608: In fact the same arguments show that all of the points

4609: in $A$ flow into $B$, since that happens for the complement

4610: of a compact annulus in $A$ and then the arguments above 

4611: apply here.

4612: Hence $U$ is a product region.

4613: Therefore

4614: we can collapse $\fol$

4615: to a minimal foliation. 

4616: 

4617: %There is a subannulus which is the core. The rest is in the interstitial

4618: %bundle. An annulus flows to a leaf $A$. From $B$ we consider all the

4619: %flow lines to $A$. It contains a half plane and is disjoint from the

4620: %other half plane.

4621: %If the other side also maps to $A$ then if everything maps to

4622: %$A$ we are done. Else it does not map to $A$. The boundary of the set

4623: %where it does not map to $A$. If a leaf of $\wls_B$ does not

4624: %map to $A$ entirely it has to stop somewhere. It 

4625: %cannot stop in $A$ going down, since $B$ is below it.

4626: %Then $B$ would escape. So it escapes going up,

4627: %see fig. \ref{esc}.

4628: %

4629: %\blankfig{esc}{1.}{Escaping of nearby leaves}

4630: %

4631: %Since stable this only happens if the $B$ intersects the orbit

4632: %$C$ but not $A$. Then $A$ escapes up and there is a stable leaf

4633: %in $\partial \Theta(A)$. This intersects $B$ $-$ so it is

4634: %entirely contained in $B$ not going to $A$ $-$ every point

4635: %is in $\partial \Theta(A)$ so it is in the middle strip following

4636: %the geodesic. So we produce a leaf $l$ of $\wls_B$ entirely

4637: %in  a bounded neighborhood of a geodesic. Iterate by $g$ and

4638: %take a limit $-$ they have to be nested, otherwise violate

4639: %transversality $-$ do not connect the same leaf. So the

4640: %boundary of these is a  leaf with ideal points $p$ and $q$.

4641: %But since nested this clearly cannot happen.

4642: %Also because outer leaf is contained in the region, we conclude

4643: %that all complementary regions are products. 

4644: 

4645: 

4646: 

4647: 

4648: \vskip .1in

4649: As in theorem \ref{as1}

4650: we can then show that $\fol$ is $\rrrr$-covered. Suppose this

4651: is not the case and let $F_1, F_2$ be non separated leaves.

4652: Let $L_i$ in $\fn$ leaves converging to both $F_1, F_2$.

4653: Let $u_1, u_2$ be the distinguished ideal points

4654: in $\pin F_1, \ \pin F_2$  respectively.

4655: Let $a_1, b_1$ be points in $\pin F_1$ very near 

4656: $u_1$ and on opposite sides of $u_1$ and which are

4657: in markers associated to contracting or $\epsilon$

4658: non expanding directions in $F_1$ associated to the $L_i$

4659: side. 

4660: Let $r_1$ be the geodesic in $F_1$ with ideal

4661: points $a_1, b_1$.

4662: Similarly for $F_2$ producing $a_2, b_2, r_2$.

4663: For $i$ big enough $L_i$ is at most $\epsilon$ far

4664: from all points in $r_1, r_2$. Therefore $r_1$ flows (by $\wwp$)

4665: into a curve $s_1$ in $L_i$ and $r_2$ flows into

4666: $s_2$.  This implies that $s_1, s_2$ are disjoint

4667: in $L_i$.

4668: Also $s_1$ has ideal points $a'_1, b'_1$

4669: which are in markers containing $a_1, b_1$ respectively

4670: (this is using a transversal to $\fn$ through a point

4671: in $F_1$). Similarly $s_2$ has ideal points $a'_2, b'_2$

4672: in markers containing $a_2, b_2$

4673: (using transversal to $\fn$ through a point in $F_2$).

4674: As $s_1, s_2$ are disjoint then $a'_1, b'_1$ do not

4675: link $a'_2, b'_2$ in $\pin L_i$, see fig. \ref{link}.

4676: 

4677: 

4678: 

4679: 

4680: \begin{figure}

4681: \centeredepsfbox{gm7.eps}

4682: \caption{

4683: Pushing ideal points near.}

4684: \label{link}

4685: \end{figure}

4686: 

4687: 

4688: 

4689: The ideal point $a'_1$ cannot be in a marker to $\pin F_1$

4690: and to $\pin F_2$ at the same time since they are non

4691: separated leaves. Hence the points $a'_1, b'_1,

4692: a'_2, b'_2$ are all distinct.

4693: Let $J_1$ be the interval of $\pin L_i$ bounded by

4694: $a'_1, b'_1$ and not containing the other points

4695: and similarly define $J_2$.

4696: For simplicity we are ommitting the dependence of $J_1, J_2$ 

4697: on $L_i$ (or on $i$).

4698: Now consider $E$ a leaf of $\wls$ intersecting $F_1$.

4699: Then $E \cap F_1$ has a ray with ideal point $u_1$,

4700: which is in the interval $(a_1,b_1)$ of $\pin F_1$.

4701: The proof of lemma \ref{cont} shows that if

4702: $L_i$ is close enough to $F_1$ then the ideal

4703: points of the corresponding rays of $(E \cap L_i)$ 

4704: have to be in $J_1$. In the same way using $F_2$ one

4705: shows that the distinguished ideal point has to be

4706: in $J_2$. Since $J_1, J_2$ are disjoint, this is

4707: a contradiction. This shows that $\fol$ is

4708: $\rrrr$-covered.

4709: 

4710: 

4711: 

4712: 

4713: 

4714: 

4715: Since $\fol$ is $\rrrr$-covered then theorem \ref{as2} implies

4716: that $\Phi$ can be chosen to be a pseudo-Anosov flow.

4717: %In addition it is regulating, for otherwise the intersection

4718: %of flow lines and leaves are geodesics??????.

4719: 

4720: Also as $\fol$ is $\rrrr$-covered we can choose a transversal

4721: $\tau$ intersecting all the leaves of $\fn$. This shows that

4722: the union of all the circles at infinity has a natural topology

4723: making it into a cylinder ${\cal A}$. This situation of

4724: $\rrrr$-covered foliations is carefully analysed in \cite{Fe7}.

4725: The fundamental group of $M$ acts in ${\cal A}$ by

4726: homeomorphisms.

4727: The union of the distinguished

4728: ideal points of leaves of the distinct leaves of $\fn$

4729: is a continuous curve $\zeta$ in ${\cal A}$ which

4730: is group invariant.

4731: 

4732: Suppose first that $\fol$ admits a holonomy invariant

4733: transverse measure. Since $\fol$ is minimal then

4734: the transverse measure has full support.

4735: Under these conditions Imanishi \cite{Im} proved

4736: that $M$ fibers over the circle

4737: with fiber a closed surface.

4738: In addition $\fol$ is approximated arbitrarily near

4739: by a a fibration.

4740: The pseudo-Anosov flow is also transverse to these

4741: nearby fibrations and so the same situation occurs

4742: for the fibrations: there is a global invariant

4743: curve in the cylinder at infinity. Since now

4744: there are compact leaves, this is impossible.

4745: 

4746: We conclude that there is no holonomy invariant

4747: transverse measure. Therefore Thurston's theorem

4748: shows the existence of contracting directions

4749: and not just $\epsilon$ non expanding directions.

4750: So the markers are associated to contracting directions.

4751: If $\zeta$ intersects a marker $m$, that corresponds

4752: to a direction in a leaf of $\fn$ which is contracting.

4753: Under the flow $\wwp$ this

4754: gets reflected in the contracted leaves nearby, that

4755: is the marker is contained in $\zeta$.

4756: Since $\fol$ is minimal and $\zeta$ is $\pi_1(M)$ invariant,

4757: this shows that the entire curve $\zeta$ is

4758: a marker associated to  contracting directions.

4759: The results from \cite{Fe7} apply here, in particular

4760: lemma 3.17  through proposition 3.21 of \cite{Fe7}:

4761: they show that no other direction in $\fn$ (outside

4762: of $\zeta$) is a contracting direction.

4763: %This implies that no other direction in leaves

4764: %of $\fn$ can be a contracting direction.

4765: By Thurston's theorem again, there would be a

4766: holonomy invariant transverse measure, contradiction.

4767: 

4768: Therefore $\zeta$ has no contracting directions.

4769: The same analysis of \cite{Fe7} now shows that 

4770: %This shows that $\zeta$ cannot cross from one side

4771: %of the marker to the other side, that is, no transversal

4772: %intersection with markers. 

4773: %This curve $\zeta$ cannot cross any marker why???

4774: for any leaf $F$ in $\fn$

4775: and every direction other than the distinguished direction,

4776: then it is a contracting direction.

4777: In fact it is a contracting direction with any other leaf

4778: of the foliation.

4779: 

4780: 

4781: 

4782: This is a very interesting situation. Let $a_F$ be the

4783: distinguished ideal point of $F$ leaf of $\fn$.

4784: Consider a one dimensional 

4785: foliation in $\mi$ whose leaves are geodesics in leaves $F$ of $\fn$ 

4786: and which have one ideal point $a_F$.  Let $\widetilde \xi$ be the

4787: flow which is unit speed tangent to this foliation and moves

4788: towards the ideal point $a_F$. 

4789: 

4790: This is a flow in $\mi$. Clearly in each leaf of $\fn$, it is a

4791: smooth flow. If $q_i$ in $L_i$ of $\fn$

4792: converge to $q$ in $L$, then the

4793: geodesics of $L_i$ with ideal point $a_{L_i}$ converge to the 

4794: geodesic through $q$ in $L$ 

4795: with ideal point $a_L$. This is 

4796: because the ideal points $a_F$ vary continuously with $F$

4797: and $q_i$ converges to $q$ $-$ this is the local

4798: trivialization of the union of the circles at infinity

4799: using the tangent bundles to a transversal.

4800: Hence $\widetilde \xi$ varies continuously.

4801: % and clearly

4802: %is group equivariant (this is where the hypothesis is used).

4803: 

4804: Since $\zeta$ is group invariant, this induces a flow 

4805: in $M$, which is tangent

4806: to the foliation $\fol$. Clearly it is smooth along the leaves of

4807: $\fol$ and usually just continuous in the transverse direction.

4808: 

4809: This flow is a topological Anosov flow: the stable

4810: foliation is just the original foliation $\fol$. 

4811: The unstable foliation: Let $p$ in leaf $L$ of $\fn$, let

4812: $\gamma$ be the flow line of $\widetilde \xi$ through $p$.

4813: Then $\gamma$

4814: has positive ideal point  $a_L$ and negative ideal point

4815: $v$.  As explained above $v$ is in a marker $m$ which is associated

4816: to a contracting direction and so that $m$

4817: intersects

4818: all ideal circles. For each $F$ in $\fn$, let $m_F$ be

4819: the intersection of $m$ and $\pin F$. Let $\gamma_F$ be

4820: the geodesic in $F$ with ideal points $a_F$ and $m_F$.

4821: Let $E_p$ be the union of these $\gamma_F$. Then all

4822: orbits of $\widetilde \xi$ in $E_p$ are backwards

4823: asymptotic by construction. By construction the $E_p$ are

4824: either disjoint or equal as $p$ varies in $\mi$ and they

4825: form a group invariant foliation in $\mi$. This is the

4826: unstable foliation.

4827: Hence $\xi$ is a topologically Anosov flow. 

4828: Notice that in the universal cover 

4829: every stable leaf intersects every unstable leaf

4830: and vice versa.

4831: 

4832: By proposition \ref{susp} it follows that $\xi$ is topologically

4833: conjugate to a suspension Anosov flow.

4834: The foliation $\fol$ is then topologically conjugate

4835: to the stable foliation of this flow.

4836: This finishes the proof of this proposition.

4837: \end{proof}

4838: 

4839: %As seen in the previous section this implies that the flow

4840: %is topologically conjugate to a suspension Anosov flow.

4841: %Now suppose there is a unique idealpoint. Then this induces an invariant

4842: %vertical line. This implies that $\Gamma$ is conjugate to the stable

4843: %foliation of a suspension Anosov flow $-$ but we will not need that.

4844: %Now consider a covering translation $g$ with an invariant leaf

4845: %$F$ and axis $a$ and translation length $r$. As seen before there

4846: %are arcs $l_i$ with endpoints in $a$ and limiting more and more on

4847: %big segments in $\pin F$. We assume the intersection of $l_i$ and

4848: %$a$ is just two points $b_i, c_i$ in the boundary of $l_i$. 

4849: %\vskip .1in

4850: %

4851: %Let $\gamma$ be a closed orbit of $\Phi$ with a lift

4852: %$\widetilde \gamma$ to $\mi$ and $g(\widetilde \gamma)

4853: %= \widetilde \gamma$ generator of the isotropy group of $\widetilde

4854: %\gamma$. As explained in the proof of proposition ??? $\fol$

4855: %is topologically conjugate to the stable foliation of

4856: %a suspension Anosov flow. Since $g$ acts freely in

4857: %%the leaf space of $\wl$, then it is associated to

4858: %an element of $\pi_1(T)$, where $T$ is a fiber for the

4859: %fibration. Therefore acts by an isometry of the hyperbolic

4860: %structure in $M$. But the return map is a contraction, at least

4861: %along $\ws(\widetilde \gamma) \cap F$. So we obtain

4862: %a contradiction. This finishes the proof of the result.

4863: 

4864: 

4865: \vskip .1in

4866: \noindent

4867: {\bf {Remark}} $-$ The hypothesis is necessary. Suppose that

4868: $\fol$ is the stable foliation of a suspension Anosov flow,

4869: $\xi$ so that it is transversely orientable. Perturb the flow

4870: slightly so that flow lines are still tangent to the original

4871: {\underline {unstable}} foliation of  $\xi$. The new flow, 

4872: call it

4873: $\Phi$ is transverse to $\fol$, it 

4874: has the same unstable foliation as $\xi$ but different

4875: stable foliation. 

4876: The flow $\Phi$ is not regulating for $\fol$. The intersections

4877: of leaves of $\wls$ with leaves $F$ of $\fn$ are all horocycles

4878: with the same ideal point which is the positive ideal point

4879: of flow lines in $F$. So the ideal points of rays of leaves of

4880: $\wlsf$ are not dense in $\pin F$. Notice  these leaves

4881: are not quasigeodesics in $F$ either.

4882: This example is studied in detail in section 7 of \cite{Fe9}.

4883: \vskip .1in

4884: 

4885: Now we want to study metric properties of  slices of 

4886: leaves of $\wlsf$.

4887: The best metric property such leaves could have is that they

4888: are {\em quasigeodesic}: this means that length along the

4889: curve is at most a bounded multiplicative distortion

4890: of length in the leaf $F$ of $\fn$ \cite{Th1,Gr,Gh-Ha,CDP}. 

4891: If the bound is $k$ then we say the curve is a $k$-quasigeodesic.

4892: Since $F$ is hyperbolic this would imply that such leaves

4893: (the non singular ones) are a bounded distance from

4894: true geodesics. 

4895: Very unfortunate for us, this is not true in general. 

4896: But there are still some good properties. 

4897: %We start

4898: %with the case that $\wls$ has Hausdorff leaf space,

4899: %in which case the quasigeodesics property holds.

4900: 

4901: Let $\hhs$ be the leaf space of $\wls$ and $\hhu$ be

4902: the leaf space of $\wlu$.

4903: Clearly since $\hhs$ may be non Hausdorff, it could

4904: be that some $\wlsf$ does not have Hausdorff leaf space. 

4905: This easily would imply that the slices of $\wlsf$ are

4906: not uniformly quasigeodesic \cite{Fe1}.

4907: This in fact occurs, see Mosher \cite{Mo1,Mo5}. Still it could be

4908: that given a ray in $\wlsf$, it is a quasigeodesic $-$

4909: with the quasigeodesic constant depending on the particular

4910: ray. We are not able to prove this and we cannot conjecture

4911: what happens in generality. But we are able to

4912: prove a weaker property, which will be enough for our purposes.

4913: If $r$ is a ray in a leaf of $\wlsf$, recall that 

4914: $r^*$ is the unique

4915: geodesic ray in $F$ with same starting point as $r$ and same

4916: ideal point. We would like to prove that $r, r^*$ are a bounded

4917: distance apart, but we do not know if that is true. But 

4918: we can prove the following important property:

4919: 

4920: \begin{lemma}{}{}

4921: There is $\delta_0 > 0$ so that for any $F$ in $\fn$ and any ray $r$ in

4922: a leaf of $\wlsf$, then given any segment of length $\delta_0$ in

4923: $r^*$, there is a point in this segment which is less than

4924: $\delta_0$ from $r$ in $F$. That implies that $r^*$ is in the neighborhood

4925: of radius $2 \delta_0$ of $r$ in $F$.

4926: \label{boun}

4927: \end{lemma}

4928: 

4929: \begin{proof}{}

4930: This means that 

4931: %$r$ is roughly near $r^*$, that is 

4932: $r^* \subset N_{2 \delta_0}(r)$ in $F$. We do not know if the converse

4933: holds. Suppose the lemma is not true. Then there are $F_i$ leaves

4934: of $\fn$, $r_i$ rays of $\wlsfi$

4935: and $p_i$ in $r^*_i$ so that $B_{2i}(p_i)$ (in $F_i$) does not

4936: intersect $r_i$. There is one 

4937:  side of $r^*_i$ in $F_i$ so that $r_i$ goes

4938: around that side, see fig. \ref{aroun}, a.

4939: Let $q_i$ inside a half

4940: disk of $B_{2i}(p_i)$ 

4941: with $B_i(q_i)$ tangent to $r^*_i$ and

4942: $\partial B_{2i}(p_i)$, see fig. \ref{aroun}, a.

4943: 

4944: 

4945: \begin{figure}

4946: \centeredepsfbox{gm8.eps}

4947: \caption{

4948: a. Limits of points, b. Going around disks

4949: in $F_i$,

4950: c The picture in $L$.}

4951: \label{aroun}

4952: \end{figure}

4953: 

4954: 

4955: 

4956: As usual up to subsequence there are $g_i$ in $\pi_1(M)$ with

4957: $g_i(q_i)$ converging to $q_0$ in $L$ leaf of $\fn$

4958: and so that the geodesic

4959: segments $\zeta_i$ from $g_i(q_i)$ to $g_i(p_i)$ in $F_i$

4960: converge to a geodesic

4961: ray $s$ in $L$. Choose two markers with 

4962: points $u_0, u_1$ in $\pin L$ very close to $s_{\infty}$

4963: and on opposite sides of it.

4964: The markers are associated to the side of $L$

4965: where the $g_i(F_i)$ are limiting to.

4966: Let $s_0, s_1$ be the geodesic rays of $L$ starting at $q_0$ and

4967: with ideal points $u_0, u_1$.

4968: For $i$ big enough $g_i(F_i)$ is $\epsilon$ close to both

4969: $s_0$ and $s_1$ and so these two rays flow (under $\wwp$)

4970: to curves $s'_0, s'_1$ in $g_i(F_i)$. 

4971: The ideal points $u'_0, u'_1$ of $s'_0, s'_1$ are in the markers

4972: above.

4973: 

4974: For $i$ big enough the ray $g_i(r_i)$ has a subray which goes 

4975: around $g_i(B_i(q_i))$ in $g_i(F_i)$ and has ideal

4976: point in the small segment of $\pin g_i(F_i)$ defined

4977: by $u'_0, u'_1$, see fig. \ref{aroun}, b.

4978: Since $s'_0, s'_1$ flows back to $L$ this figure flows

4979: back to $L$ producing a ray $l_i$ of $\wlsl$

4980: which goes around a big disk in $L$ centered at $q_0$

4981: and has ideal point in the small segment bounded by

4982: $u_0, u_1$, see fig. \ref{aroun}, c.

4983: As $i$ goes to infinity, these $l_i$ escape

4984: to infinity in $L$ because bigger and bigger disks

4985: in $g_i(F_i)$ flow to $L$. 

4986: This implies that there is no ideal point of a ray

4987: of $\wlsl$ outside the small segment of $\pin L$

4988: bounded by $u_0, u_1$. This contradicts

4989: the previous proposition that such ideal points are

4990: dense in $\pin L$.

4991: 

4992: This finishes the proof of the lemma.

4993: \end{proof}

4994: 

4995: \begin{lemma}{}{}

4996: The limit points 

4997: of rays of $\wlsf$ 

4998: vary continuously in $\pin F$

4999: except for the

5000: non Hausdorffness in the leaf space of $\wlsf$.

5001: \label{encon}

5002: \end{lemma}

5003: 

5004: \begin{proof}{}

5005: Suppose that $p_i$ converges to $p$ in $F$, with respective rays

5006: $r_i$ converging to the ray $r$ of $\wlsf$. 

5007: Let $l$ be the leaf of $\wlsf$ through $p$.

5008: Up to subsequence

5009: assume the $r_i$ are all in the same sector of $l$ defined by $p$

5010: and that they form a nested sequence of rays.

5011: Then the ideal ponts 

5012: $(r_i)_{\infty}$ form a monotone sequence

5013: in $\pin F$. Perhaps some

5014: ideal points are the same.

5015: If $(r_i)_{\infty}$ does not converge to $r_{\infty}$ there

5016: is an interval $v$ in $\pin F$, between the limit and $r_{\infty}$.

5017: Since the ideal points are dense in $\pin F$, there is $w$ leaf

5018: of $\wlsf$ with $w_{\infty}$ in $v$. Therefore there is $l'$ not

5019: separated from $l$ with $r_i$ converging to $l'$ as well.

5020: In this fashion we can go from $l$ to $l'$. 

5021: This shows that if there is no leaf of $\wlsf$ non separated from

5022: $l$ in that side and in the direction the rays $r_i$ go, then

5023: the limit points vary continuously.

5024: 

5025: We analyse a bit further the non Hausdorffness.

5026: In the setup above there are subrays of $r_i$ with points

5027: converging to a point in $l'$ and we can restart the

5028: analysis with $l'$ instead of $l$.

5029: If there are finitely many leaves non separated from $l$ and

5030: $l'$ we can assume that $l, l'$ are consecutive. Then

5031: they have subrays which share an ideal point. 

5032: If $m$ is the last leaf non separated from $l, l'$ 

5033: in the direction the rays $r_i$ go to, then

5034: there is a ray $\zeta$ of $m$ so that there are subrays

5035: of $r_i$ with points converging to a point in $\zeta$

5036: and $(r_i)_{\infty}$ converges to $\zeta_{\infty}$.

5037: If there are infinitely many such leaves non separated

5038: from $l$, then we can order them as $\{ l_j \}, j \in {\bf N}$ 

5039: all in the

5040: direction the rays $r_i$ go to. The ideal points of

5041: \ $l_j$ \ form a monotone sequence in $\pin F$ which converge

5042: to a point $u$ in $\pin F$. The arguments above show

5043: that $(r_i)_{\infty}$ converges to $u$.

5044: \end{proof}

5045: 

5046: Our next goal is to analyse the 

5047: non Hausdorffness in the leaf space of $\wlsf$.

5048: We also want to understand when can the ideal points of two

5049: different rays of 

5050: the same leaf of $\wlsf$ be the same.

5051: A {\em Reeb annulus} is an annulus $A$ with a foliation so that

5052: the boundary components are leaves and every leaf in the

5053: interior is a topological line which spirals towards 

5054: the two boundary components in the same direction.

5055: In the universal cover the lifted foliation does

5056: not have Hausdorff leaf space.

5057: The lifted foliation to the universal cover is 

5058: called a {\em Reeb band}.

5059: 

5060: \begin{define}{(spike region)}{}

5061: A stable spike region in a leaf $F$ of $\fn$

5062: is a closed $\wlsf$ saturated set ${\cal E}$

5063: satisfying:

5064: 

5065: 

5066: \begin{itemize}

5067: 

5068: \item

5069: There are

5070: finitely many  boundary leaves of ${\cal E}$

5071: which are line leaves of $\wlsf$.

5072: The ideal points of consecutive rays in the boundary of ${\cal E}$

5073: are the same, otherwise they are distinct (like an ideal

5074: polygon). 

5075: 

5076: 

5077: \item

5078: The region ${\cal E}$ is a bounded distance

5079: from the ideal polygon with these vertices. The bound

5080: is not universal in $\fn$. 

5081: 

5082: \item

5083: There is an ideal

5084: point  $z$ of  ${\cal E}$

5085: so that every leaf in the interior of ${\cal E}$ has both ideal

5086: points equal to $z$. In addition the leaves in the

5087: interior are nested.

5088: The finitely many leaves in the boundary

5089: are all non separated from each other and they are limits

5090: of the interior leaves.

5091: 

5092: \item

5093: There is no singularity of $\wlsf$ in the interior of

5094: ${\cal E}$. If there is a singularity of $\wlsf$ 

5095: in a boundary leaf $\tau$ of ${\cal E}$ then 

5096: the interior of ${\cal E}$ is contained in the sector

5097: defined by the line leaf $\tau$.

5098: \end{itemize}

5099: 

5100: \noindent

5101: Similarly define an unstable spike region. A spike region is either

5102: a stable or unstable one.

5103: \end{define}

5104: 

5105: 

5106: \begin{proposition}{}{}

5107: Let $E$ be a leaf in $\fn$ and $\upsilon$ a slice of a leaf

5108: $\upsilon_0$ of $\wlse$.

5109: Suppose that both ideal points of $\upsilon$ are the same.

5110: Then $\upsilon$ is contained in 

5111: the interior of a stable spike region $B$ of $E$.

5112: %, so that

5113: %$B$ has 2 distinct ideal points in $\pin E$ and $B$ is a bounded

5114: %distance from the geodesic connecting these ideal

5115: %points. In addition the foliation in $B$ is 

5116: %a Reeb band.

5117: In addition either $B$ projects to a Reeb annulus in a leaf

5118: of $\fol$ or for any two consecutive rays in $\partial B$,

5119: the region between them projects to a set asymptotic to

5120: a Reeb annulus in a leaf of $\fol$.

5121: Similarly for $\wluf$.

5122: \label{band}

5123: \end{proposition}

5124: 

5125: \begin{proof}{}

5126: We do the proof for $\wlsf$.

5127: Let $\upsilon$ be a slice as above with ideal point $x$ in 

5128: $\pin E$. 

5129: Let $C$ be the region bounded by $\upsilon$ in $E$ which

5130: only limits in $x$.

5131: First we assume 

5132: assume that $\upsilon$ is a line leaf of some leaf of

5133: $\wlsf$.

5134: %$\upsilon_0$,

5135: %since any prong of $\upsilon_0$ which enters $C$ will

5136: %have ideal point $x$.

5137: We will show that 

5138: the region $C$ as it approaches $x$, projects to a set

5139: in $M$ which limits to a Reeb annulus in  a leaf of $\fol$.

5140: The process will be done in a series of steps.

5141: The proof of this proposition is very long with

5142: several intermediate results and lemmas.

5143: 

5144: Choose $z_0$ in $\upsilon$ and let $e_1, e_2$ be the rays

5145: of $\upsilon$ defined by $z_0$.

5146: Let $\zeta^*$ be the geodesic ray of $E$ starting at

5147: $z_0$ and with ideal point $x$. Then $\zeta^*$ is 

5148: contained in the $2 \delta_0$ neighborhood of $e_1$ or $e_2$,

5149: where $\delta_0$ is the constant of lemma

5150: \ref{boun}. It follows that 

5151: we can choose $p_i, q_i$ in $e_1, e_2$ respectively 

5152: with $p_i, q_i$ converging to $x$ in $E \cup \pin E$

5153: and also $d_E(p_i, q_i) < 4 \delta_0$. 

5154: Let $e^i_1$ be the subray of $e_1$ starting at $p_i$

5155: and $e^i_2$ the subray of $e_2$ starting at $q_i$.

5156: Up to subsequence  there are $p_0, q_0$ in $\mi$ and

5157: are $g_i$ in $\pi_1(M)$ with $g_i(p_i)$, $g_i(q_i)$

5158: converging to $p_0$, $q_0$ respectively. 

5159: The distance condition implies 

5160: $p_0, q_0$  are in the same leaf of $\fn$, let $F$ be this leaf.

5161: Then $g_i(E)$ converges to $F$

5162: and perhaps other leaves as well.

5163: 

5164: For $i$ big enough

5165: the flowlines of $\wwp$ through $g_i(p_i), g_i(q_i)$

5166: go through to $u_i$ and $v_i$ in $F$. Also $u_i \rightarrow p_0,

5167: v_i \rightarrow q_0$.

5168: If the leaf of $\wlsf$ through $p_0$ contains $q_0$ then 

5169: for $i$ big enough the arcs 

5170: in leaves of $\wlsf$ from $u_i$ to $v_i$

5171: will have bounded length and

5172: bounded diameter. The same will happen for the arcs of

5173: of $g_i(\upsilon)$ between

5174: $g_i(p_i)$ and $g_i(q_i)$, contradiction.

5175: Hence $p_0, q_0$ are not in the same leaf of $\wlsf$. Let

5176: $l$ be the leaf of $\wlsf$ through $p_0$ and $r$ be the one

5177: through $q_0$. Let $L, R$ be leaves of $\wls$ containing $l$ and

5178: $r$ respectively.

5179: Since the intersection of

5180: a leaf of $\wls$ with $F$ is connected, then $L$ and $R$ are

5181: distinct and also are not separated from each other

5182: in the leaf space of $\wls$.

5183: 

5184: 

5185: 

5186: 

5187: 

5188: %Fix $i$ big enough so that $g_i(p_i)$ is very close

5189: %to $p_0$ and $g_i(q_i)$ very close to $q_0$.

5190: %Let $A$ be the leaf of $\wls$ containing $g_i(p_i), g_i(q_i)$.

5191: %The index $j$ will denote an arbitrary integer much bigger

5192: %than $i$ so that $g_j(p_j)$ is much closer to $p_0$ and similarly

5193: %for $g_j(q_j)$.

5194: %Choose $\gamma$ a segment connecting $p_0$ in $L$ to $q_0$ 

5195: %in $R$ and intersecting $L$, $R$ only in these points and

5196: %also transverse to $\fn$ near $p_0, q_0$. Since $L, R$

5197: %are non separated we can choose $\gamma$ to be transverse

5198: %to $\fn$ except at one tangency.

5199: %The point $g_j(p_j)$ is in $g_j \circ g^{-1}_i(A)$.

5200: %For simplicity let $h = g_j \circ g^{-1}_i$

5201: %(which really depends on $j$).

5202: %We want to analyse $h^{-1}(L), h^{-1}(R)$ which are non

5203: %separated from each other, see fig. \ref{conv}.

5204: %Our goal is to show that $h(L) = L$ and $h(R) = R$.

5205: %

5206: %\blankfig{conv}{1.1}{Non separated leaves and rigidity.}

5207: %

5208: %

5209: %Now fix $j$.

5210: %We can choose $\gamma$ so that $g_j(p_j) = h(g_i(p_j))$ and

5211: %$g_j(q_j) = h(g_i(q_j))$ are in $\gamma$.

5212: %They are closer to $p_0, q_0$ than $g_i(p_i), g_i(q_i)$ are.

5213: %Figure \ref{conv} is a representation of this situation

5214: %in the universal cover $\mi$.

5215: %In the leaf $A$, the orbits intersecting $\gamma$, that is,

5216: %those containing $g_i(p_i)$ and $g_i(q_i)$ separate

5217: %the orbits containing $g_i(p_j), g_i(q_j)$ $-$ reflecting

5218: %what happens in $\alpha$.

5219: %

5220: %\vskip .1in

5221: %\noindent

5222: %{\underline {Claim 1}} $-$ The leaves $h^{-1}(L), h^{-1}(R)$

5223: %are non separated from $L$ and $R$.

5224: %

5225: %First of all we show that $R$ does not intersect $h^{-1}(\gamma)$

5226: %except perhaps at $h^{-1}(R)$ and $h^{-1}(L)$.

5227: %Suppose not. Then since $h(A)$ is closer to $R$ then

5228: %$A$ is, then $h^{-1}(R)$ intersects $h^{-1}(\gamma)$

5229: %either between $h^{-1}(R \cap \gamma)$ and 

5230: %$g_i(q_j)$

5231: %or between $h^{-1}(L \cap \gamma)$ and 

5232: %$g_i(p_j)$.

5233: %If $R$ intersects $h^{-1}(\gamma)$ between $h^{-1}(L \cap \gamma)$

5234: %and $g_i(p_j)$, then because of the ordering of orbits

5235: %in $A$ and in $R$ it follows that $R$ has to

5236: %intersect $\gamma$ between $g_i(p_i)$ and $L \cap \gamma$.

5237: %But then $R, L$ intersect a common transversal, contradiction

5238: %to these leaves being non separated in the leaf space.

5239: %If $R$ intersects $h^{-1}(\gamma)$ between $h^{-1}(R \cap \gamma)$

5240: %and $g_i(q_j)$ then since $h^{-1}(L)$ is non separated

5241: %from $h^{-1}(R)$, it follows that $R$ separates 

5242: %$h^{-1}(L)$ from $A$. This reduces to the first

5243: %situation.

5244: %

5245: %Let $\xi_1$ be the ideal point of the ray of $l$ which is

5246: %the limit of the rays associated to $e_1$ and $\xi_2$

5247: %the ideal point of $r$ the one corresponding to $e_2$.

5248: 

5249: For simplicity assume that the leaves of $\wls$ through

5250: $u_i$ form a nested collection with  $i$.

5251: 

5252: \vskip .1in

5253: The first goal is to show that we can choose $l, r$ line leaves

5254: of $\wlsf$ as above so that they also share an ideal

5255: point.

5256: Let $\beta_i$ be a ray in the leaf of $\wlsf$ through

5257: $u_i$ starting at $u_i$ and containing points in the 

5258: flowlines which go to the ray

5259: $g_i(e^i_1)$. Similarly let $\gamma_i$ be a subray

5260: in the same leaf starting at $v_i$ and associated with 

5261: the ray $g_i(e^i_2)$.

5262: Let ${\cal C}_1$ \ (resp. ${\cal C}_2$) be the collection of line leaves of 

5263: $\wlsf$ that $\beta_i$ \ (resp. $\gamma_i$) converges to, including the

5264: ray of $l$ \ (resp. $r$).

5265: Let ${\cal C}$ be the collection of all line leaves of $\wlsf$ which

5266: are non separated from $l, r$. 

5267: Then ${\cal C}$ contains ${\cal C}_1$ and ${\cal C}_2$. 

5268: For any element $\tau$ in ${\cal C}$, let

5269: $B(\tau)$ be the leaf of $\wls$ containing it.

5270:  All of the $B(\tau)$ are not separated

5271: from each other, and they are in the set of leaves ${\cal B}$ of $\wls$

5272: non separated from both $L, R$.

5273: By theorem \ref{theb}, the set ${\cal B}$ has a linear order,

5274: making it order isomorphic to either ${\bf Z}$ or a finite

5275: set. This induces an order in ${\cal C}$ where we can choose

5276: this so that an arbitrary element of ${\cal C}_1$ is

5277: bigger than any element in ${\cal C}_2$.

5278: 

5279: 

5280: \begin{figure}

5281: \centeredepsfbox{gm9.eps}

5282: \caption{

5283: a. Non Hausdorffness in the limit,

5284: b. Showing $\xi_1 = \xi_2$.}

5285: \label{spre}

5286: \end{figure}

5287: 

5288: 

5289: 

5290: If there are finitely many elements in ${\cal C}_1$ let 

5291: $l'$ be the last one and let $\xi_1$ be the ideal point of

5292: the ray of $l'$ corresponding to the direction of the

5293: rays $\beta_i$.

5294: Otherwise the ideal points of the leaves

5295: in ${\cal C}_1$ form a weakly monotone sequence in $\pin F$

5296: and let $\xi_1$ be the limit of this sequence.

5297: Similarly define $\xi_2$ associated to $r$, see fig. \ref{spre},

5298: a.

5299: 

5300: Fix a basepoint $x_0$ in $F$.

5301: The first thing to prove is the following:

5302: 

5303: \begin{lemma}{}{}

5304: $\xi_1 = \xi_2$.

5305: \end{lemma}

5306: 

5307: \begin{proof}{}

5308: Suppose by way of contradiction that this is not true.

5309: Choose 2 markers very near $\xi_1$ bounding

5310: an interval $J_1$ in $\pin F$ with $\xi_1$ in the

5311: interior and similarly choose markers near $\xi_2$ and

5312: interval $J_2$ so that $J_1, J_2$ are disjoint.

5313: Let $W_1$ be the wedge of $F$ centered at the point

5314: $x_0$ with

5315: ideal set $J_1$ and $W_2$ the wedge of $F$ centered

5316: also at $x_0$ with ideal set $J_2$. For i big enough

5317: both boundaries of $W_1$ and $W_2$ flow into

5318: $g_i(E)$. 

5319: 

5320: Suppose first that there is a last leaf $l'$ in ${\cal C}_1$.

5321: Then $l'$ has a ray which is eventually contained in

5322: a strictly smaller wedge $W'_1$, since 

5323: the ideal point of $l'$ is  $\xi_1$.

5324: %in a strictly smaller subwedge $W'_1$ of $W_1$ $-$ since its

5325: %ideal point is $\xi_1$.

5326: Now choose a big disk $D$ of $F$ centered in $x_0$ .

5327: Let $N_1$ be the closure of $W_1 - D$.

5328: Choose $D$ big

5329: enough so that $l'$ enters $N_1$ through $\partial D$

5330: and is then entirely in $W'_1$. 

5331: For $i$ big enough $\beta_i$ will be close to $l'$ for 

5332: a long distance. By lemma \ref{encon} the ideal 

5333: points of $\beta_i$ converge to $\xi_1$ as $i$ converges

5334: to infinity, since $l'$ is the last leaf non separated

5335: from $l$ in that side. The ideal point is in the limit

5336: set of the subwedge $W'_1$. If the rays $\beta_i$ keep

5337: exiting $W_1$ then since they are trapped by $l'$ and

5338: $\beta_{i_0}$ (for some $i_0$), it follows that  the

5339: sequence \ $\beta_i$ \ 

5340: %they will have to intersect a compact

5341: %part of $\partial W_1$. Then the sequence $\{ \beta_i \}$ 

5342: has additional limits besides the leaves in ${\cal C}_1$,

5343: contradiction.

5344: Therefore for big enough $i$, the $\beta_i$ enters $N_1$ through

5345: $\partial D$ and stays in $N_1$ from then on.

5346: 

5347: We want to get the same result when ${\cal C}_1$ is infinite.

5348: In that case let \ $\{ \nu_j, \ j \in {\bf N} \}$ 

5349: \ be the leaves in

5350: ${\cal C}_1$ ordered with same ordering as in ${\cal C}_1$

5351: and $\nu_1 = l$.

5352: %$ in the same leaf as $l$.

5353: Since the leaves $v_i$ are non separated from each other

5354: then they cannot accumulate anywhere in $F$ as $i \rightarrow \infty$

5355: and the leaves $\nu_j$ escape compact sets as $j$ grows.

5356: The ideal points of $\nu_j$

5357: are also converging to $\xi_1$. By 

5358: density of ideal points of $\wlsf$ in $\pin F$ the

5359: leaves 

5360: $\nu_j$

5361: cannot be getting closer to non trivial intervals in $\pin F$.

5362: This implies that there is $j_0$ so that for 

5363: 

5364: $$j \geq j_0, \ \ 

5365: \nu_j \ \ {\rm  is \ very \ close  \ to } \ \ 

5366: \xi_1 \ \ {\rm in} \   F \cup \pin F$$

5367: 

5368: \noindent

5369: and so contained in $W_1$.

5370: Now an argument entirely similar as in the case ${\cal C}_1$

5371: finite implies that for $i$ big enough then $\beta_i$ has

5372: subrays entirely contained in $N_1$.

5373: The same holds for $\gamma_i$ producing subrays entirely

5374: contained in the corresponding set $N_2$ $-$ the disk 

5375: $D$ may need to be bigger to satisfy all these conditions.

5376: 

5377: There is $a_1 > 0$ and 

5378: $i_0$ so that for $i \geq  i_0$ then except for the initial

5379: segment of length $a_1$ then  $\beta_i$ is entirely

5380: contained in $N_1$ and similarly for $\gamma_i$ and $N_2$.

5381: Choose $k_0$ big enough so that $D$ is $\epsilon$ close

5382: to $g_k(E)$ for any $k \geq k_0$.

5383: Then $D$ flows in $g_k(E)$ under $\wwp$

5384: and so do $\partial W_1, \partial W_2$.

5385: For $i$ bigger than both $i_0, k_0$

5386: the ray $\beta_i$ flows into the ray $g_i(e^i_1)$

5387: (notice these do not

5388: have singularities). 

5389: The ray $g^i(e^i_1)$ has to be in the generalized wedge

5390: which is bounded by the image of $\partial W_1$ in $g_i(E)$.

5391: Similarly for $\gamma_i$. 

5392: This argument is done in lemma \ref{cont}.

5393: These two generalized wedges

5394: have disjoint ideal sets in $\pin g_i(E)$. Therefore

5395: $g_i(e^i_1)$ and $g_i(e^i_2)$ do not have the same

5396: ideal points. This is a contradiction

5397: because $e_1, e_2$ have the same ideal point in $\pin E$. 

5398: 

5399: This proves that $\xi_1 = \xi_2$.

5400: \end{proof}

5401: 

5402: 

5403: \noindent

5404: {\underline {Continuation of the proof of proposition \ref{band}}}

5405: 

5406: 

5407: The first part of the proof was this: in $E$ zoom in

5408: towards an ideal point $x$ of  

5409: $\pin E$ and use covering translations $g_i$ of $\mi$

5410: to map back these points near a point in $\mi$ which

5411: is in a leaf $F$.

5412: We will redo this process starting with $F$.

5413: By taking translates of $F$ we will limit to a leaf $F^*$.

5414: The difference is that now we have leaves

5415: of $\wlsf$ which are non separated from each other.

5416: These non separated leaves are much better suited

5417: to perturbation arguments as seen below.

5418: 

5419: The lemma shows that $\xi_1 = \xi_2$ and this

5420: implies that the ideal points

5421: of $\beta_i, \gamma_i$ are all the same and equal to $\xi_1$.

5422: %Next we want to show that under these conditions then

5423: %${\cal C}_1$ has to be finite and similarly for ${\cal C}_2$.

5424: Let $\xi = \xi_1$.

5425: The $\beta_i, \gamma_i$ are rays in leaves of

5426: $\wlsf$ and contained in $F$.

5427: Let $\mu$ be the geodesic ray in $F$ starting at $x_0$ 

5428: (the basepoint in $F$) with

5429: ideal point $\xi$.

5430: Since $(\beta_i)_{\infty} = (\gamma_i)_{\infty} = \xi$,

5431: then lemma \ref{boun} implies that for $z$ in

5432: $\mu$ far enough from $p_0$, we can choose

5433: a point in $\beta_i$ which is less than 

5434: $2\delta_0$ away from $z$ in $F$.

5435: Call this point $b_i(z)$. Similarly define $c_i(z)$ in $\gamma_i$.

5436: This is for any $i$ in ${\bf N}$. 

5437: 

5438: For each $z$ we may take a subsequence

5439: of the $b_i(z)$ which converges in $F$ and 

5440: the limit is denoted by $b(z)$. Similarly define $c(z)$.

5441: By definition of ${\cal C}_1$ the point $b(z)$ has

5442: to be in one of the leaves of ${\cal C}_1$ and

5443: similarly for $c(z)$.

5444: The $b(z), c(z)$ are not uniquely defined and most likely

5445: do not vary continuously with $z$.

5446: 

5447: 

5448: \begin{lemma}{}{}

5449: There is at least one element  $\zeta$ of ${\cal C}_1$ which

5450: has ideal point $\xi$.

5451: Similarly for ${\cal C}_2$.

5452: \end{lemma}

5453: 

5454: \begin{proof}{}

5455: If there are finitely elements in ${\cal C}_1$ then the

5456: last one satisfies this property.

5457: Suppose then there are infinitely many elements

5458: in ${\cal C}_1$.

5459: As $z$ varies in $\mu$, then so does $b(z)$. If 

5460: there are $z$ escaping in $\mu$ so that $b(z)$ is

5461: in the same element $\zeta$ of ${\cal C}_1$ then

5462: $\zeta$ has an appropriate ray with ideal

5463: point $\xi$. In this case we are done.

5464: 

5465: Otherwise we can find $z_k$ in $\mu$ converging to $\xi$

5466: so that $b(z_k)$ are in leaves $\nu_{m(k)}$ of ${\cal C}_1$

5467: which are all distinct. We can choose $z_k$ so that

5468: the $m(k)$ increases with $k$.

5469: In the same way we have $c(z_k)$ in 

5470: distinct elements of ${\cal C}_2$.

5471: Let 

5472: 

5473: $$B_k \ = \ \ws(b(z_k)), \ \ C_k \ = \ \ws(c(z_k)),

5474: \ \ \ {\rm both \ in} \  \ {\cal B}$$

5475: 

5476: \noindent

5477: Recall that ${\cal B}$ is the set of leaves of $\wls$

5478: non separated from both $L, R$.

5479: As the length from $b(z_k)$ to $c(z_k)$ in $F$

5480: is bounded by $4\delta_0$, then up to subsequence assume

5481: $\pi(b(z_k)), \pi(c(z_k))$ converge in $M$.

5482: For $n, k$  big enough there is $h_{nk}$ covering

5483: translation of $\mi$ so that $h_{nk}(b(z_n))$ is very close

5484: to $b(z_k)$ and $h_{nk}(c(z_n))$ is very close to 

5485: $c(z_k)$. Suppose $n >> k$, let $h = h_{nk}$ for simplicity.

5486: Then $B_k$ has a point $b(z_k)$  very close to 

5487: $h(b(z_n)) \in  h(B_n)$ and similarly $c(z_k)$ in 

5488: $C_k$ very close to $h(c(z_n)) \in h(C_n)$.

5489: But $B_k$ is non separated from $C_k$ and

5490: similarly for $h(B_n), h(C_n)$, so the only

5491: way this can happen is that 

5492: 

5493: $$h(B_n) \ = \ B_k, \ \ \ \ h(C_n) \ = \ C_k$$

5494: 

5495: \noindent

5496: This implies that $h$ sends the set of leaves non

5497: separated from $B_i, C_i$ to itself, that is $h$ acts

5498: on the set ${\cal C}$ and therefore acts on ${\cal B}$ as

5499: well. 

5500: Notice that $B_k < B_n$ in the order of ${\cal B}$ 

5501: because $n > k$ 

5502: and $C_k \geq C_n$ (the $C_k$ could be all the same, but

5503: if they are not then they decrease in the order).

5504: Since $h(B_n) = B_k$ then

5505: $h$ acts as a decreasing translation in the ordered

5506: set ${\cal B}$.

5507: But since $h(C_n) = C_k$ then $h$ acts as a non

5508: decreasing translation. 

5509: These two facts are incompatible.

5510: 

5511: This implies that we have to have at least one element

5512: in ${\cal C}_1$ with ideal point $\xi$.

5513: The same happens for ${\cal C}_2$.

5514: This finishes the proof of the lemma.

5515: \end{proof}

5516: 

5517: %\noindent

5518: %{\bf {Remark}} $-$ 

5519: %As will be seen from future arguments, there is only

5520: %one element in ${\cal C}_1$ and similarly for 

5521: %${\cal C}_2$. These two line leaves have

5522: %an ideal point $\xi$.

5523: %\vskip .1in

5524: 

5525: 

5526: Since the sequence $\{ \beta_i \}$ 

5527: also converges to the leaf $\zeta$ of $\wlsf$ we can rename

5528: the objects and assume that $l = \zeta$ and $p_0$ is

5529: a point in $l$. This can be accomplished by

5530: choosing different points $p_i$ in the

5531: ray $e_1$. Similarly do the same thing in the

5532: other direction. We state this conclusion:

5533: 

5534: \vskip .1in

5535: \noindent

5536: {\bf {Conclusion}} $-$  There are $p_i, q_i$ in $e_1, e_2$

5537: respectively, escaping these rays, so that 

5538: $d_E(p_i, q_i) < 4 \delta_0$

5539: and there are covering translations $g_i$ so that:

5540: $g_i(p_i)$ converges to $p_0$, 

5541: \ $g_i(q_i)$ converges to $q_0$,

5542: both in $F$ and in rays $l, r$ of $\wlsf$.

5543: Also $l, r$ converge to the same ideal point

5544: $\xi$ in $\pin F$ and $l, r$ are not separated

5545: in the leaf space of $\wlsf$.

5546: \vskip .1in

5547: 

5548: %The nearby leaves $l_i$ converge to $l$.

5549: %If their limit points do not converge to $l$ then there is

5550: %a leaf $l'$ non separated from $l$. We change this accordingly,

5551: %assuming that $l$ is the last leaf non separated from 

5552: %and in that side. Then the ideal points

5553: %of $l_i$ are eventually in $\overline W_1$ and the

5554: %piece of $W_1$ enters through $\partial D$ and are entirely

5555: %contained in $W_1$. This means that same happens for nearby

5556: %leaves. Hence in the near leaf to $F$,

5557: %$\ws(q_i)$ also has a part very near $l_i \cap D$

5558: %and then it enters $W_1$. Since it can only exit $W_1$ through

5559: %$\partial W_1$ and both sides carry over to $F$, then

5560: %this cannot happen. So the ideal point of $l_i$ in

5561: %$g_i(A)$ is in the associated set to $W_1$

5562: %and similarly for $r_i$. Hence they are different,

5563: %contradiction.

5564: %This proves the claim.

5565: %

5566: %Since $\xi_1 = \xi_2$ and $l$ is non separated from $e$,

5567: %we obtain that $l_i$ is a collection of leaves

5568: %in $F$ all with ideal points $\xi$ and limiting to $l, r$

5569: %with ideal points $\xi$.

5570: 

5571: 

5572: 

5573: 

5574: We will continue this perturbation approach. 

5575: We want to show that the region in $F$ ``between"

5576: $l$ and $r$  projects to a Reeb annulus of $\fol$ in $M$.

5577: Let then $z_i$ in $l$ converging to $\xi$ and $w_i$ in

5578: $r$ converging to $\xi$, so that $d_F(z_i,w_i)$ is always

5579: less than $4 \delta_0$. Up to subsequence assume there

5580: are $h_i$ covering translations

5581: with 

5582: 

5583: $$h_i(z_i) \rightarrow z_0, \ \ \ h_i(w_i) \rightarrow w_0$$

5584: 

5585: \noindent

5586: %Therefore $\ws(z_0), \ws(w_0)$ are not separated from

5587: %each other. 

5588: Notice that $h_i(L), h_i(R)$ are 

5589: non separated from each other and $h_i(L) \rightarrow \ws(z_0), 

5590: \ h_i(R) \rightarrow \ws(w_0)$.

5591: The argument in the previous lemma then implies

5592: that $h_i(L) = h_j(L), \ h_i(R) = h_j(R)$ for all $i, j$

5593: at least equal to some $i_0$.

5594: Discard the first $i_0$ terms

5595: and postcompose $h_i$ with $(h_{i_0})^{-1}$ 

5596: (that is $(h_{i_0}^{-1} \circ h_i)$)

5597: and assume that this is the original sequence $h_i$.

5598: This implies that

5599: %the limits and

5600: %they have to be very near these, it follows that 

5601: %$h_i(L) = \ws(z_0)$ for $i$ big enough. So pull

5602: %back by $h_i$ (here $i$ is fixed) to suppose 

5603: $h_i(L) = L,

5604: h_i(R) = R$ for all $i$.

5605: So the $h_i$ are all in the intersection of

5606: the isotropy groups of $L$ and $R$. This group is

5607: generated by a covering translation

5608: $h$.

5609: Therefore there are $n_i$ with $h_i = h^{n_i}$. 

5610: Since 

5611: $h_i(z_i) \rightarrow z_0$ and the $\{ z_i, \ | \ i 

5612: \in {\bf N} \}$ do not

5613: accumulate in $\mi$ then 

5614: $|n_i| \rightarrow \infty$.

5615: In addition since $L, R$ are not separated from each other,

5616: then $h$ preserves each individual line leaf, slice and

5617: possible lift annulus of $L$.

5618: 

5619: Let $F^*$ be the leaf of $\fn$ containing $z_0, w_0$. Then

5620: $h_i(F)$ converges to $F^*$.

5621: Since $z_i$ is in $L$ and $h_i(L) = L$ then $h_i(z_i)$ is

5622: in $L$ and so $z_0$ is in $L$. It follows that $L$ intersects

5623: $F^*$.

5624: 

5625: Up to subsequence and perhaps taking the inverse of $h$,

5626: assume that $n_i$ converges to $+\infty$.

5627: If $h(F) = F$, then since $h(L) = L$ this produces

5628: a closed leaf of $\Lambda^s \cap \fol$

5629: in $\pi(F)$. Similarly 

5630: $h(R \cap L) = R \cap L$ so produces another closed leaf in

5631: $\pi(F)$ and together bound an annulus in $\pi(F)$ with a sequence of

5632: leaves of $\Lambda^s \cap \pi(F)$

5633: converging to the boundary leaves.

5634: By Euler characteristic reasons, there can be no singularities

5635: inside the annulus, so we conclude that the annulus in $\pi(F)$

5636: has a Reeb foliation.

5637: 

5638: Now assume that $h(F)$ is not equal to $F$.

5639: Let $\hp$ be the leaf space of $\fn$. This is a one dimensional

5640: manifold, which is simply connected, but usually not

5641: Hausdorff \cite{Ba2}. The element $h$ acts on $\hp$. An analysis of

5642: group actions on simply connected non Hausdorff spaces

5643: was iniatilly done by Barbot in

5644: \cite{Ba2} and subsequently in \cite{Ro-St,Fe8}.

5645: One possibility

5646: is that $h$ acts freely in $\hp$. Then $h$ has an axis $\tau$

5647: in $\hp$ which is invariant under $h$.  In general this

5648: axis is not properly embedded, see \cite{Fe8}.

5649: Since all the $h^{n_i}(F)$ intersect a common transversal,

5650: then the analysis in \cite{Ba2} shows

5651: that $F$ has to be in the axis of $h$ and 

5652: $h^n(F)$ converges to a collection of non separated

5653: leaves. In this case we get that $F^*$ and $h(F^*)$

5654: are non separated from each other.

5655: 

5656: 

5657: The other situation is that $h$ has fixed points in $\hp$.

5658: In general the set of fixed points of $\hp$ is not a closed

5659: set, but the set of points $z$ in $\hp$ so that $z$ and

5660: $h(z)$ are not separated in $\hp$ is a closed subset

5661: $Z$ of $\hp$ \cite{Ba2,Ro-St}.

5662: None of the images of $F$ under $h$ can be in $Z$,

5663: so $F$ is in a component of $\hp - Z$.

5664: Then $h$ permutes these components. In addition

5665: $h$ preserves an orientation in $\hp$ $-$ since

5666: $\fol$ is transversely orientable.

5667: Since

5668: $h^{n_i}(F)$ all intersect a common transversal then

5669: they have all to be in the same component $U$ of $\hp - Z$.

5670: Let $i_0$ be the smallest positive integer

5671: so that $h^{i_0}(U) = U$.

5672: It follows that all $n_i$ are multiples of $i_0$.

5673: Since $h^{n_i}(F)$ converges to $F^*$ then

5674: the leaf $F^*$ is in the boundary of the component

5675: $U$ and $h^{i_0}(F^*) = F^*$. 

5676: %Therefore the same

5677: %argument as above can be applied to show that 

5678: %there is a Reeb annulus in $\pi(F^*)$.

5679: 

5680: The only remaining case to be analysed

5681: is that $h$ acts freely and $h^n(F)$ converges

5682: to $F^*$ with $h(F^*)$ non separated from $F^*$.

5683: In this particular case we prove this is not

5684: possible, that is:

5685: 

5686: %If $h(F)$ is not equal $F$, then  as $h^{n_i}(z_i) \rightarrow z_0$,

5687: %it follows that

5688: %$h^{n_i}(F)$ converges to $F^*$ which contains $z_0$ and 

5689: %also $w_0$.

5690: %Up to subsequence assume that

5691: %$n_i = a \ (mod \ n_1)$ for all $i \geq 2$.

5692: %Let $S = h^{n_2}(F)$ and $f = h^{n_1}$. Then

5693: %

5694: %$$h^{n_i}(F) \ = \ h^{n_i - n_2} h^{n_2}(F) \ = \ h^{m_i n_1}(S)

5695: %\ = \ f^{m_i}(S)$$

5696: %

5697: %\noindent

5698: %with $m_i$ converging to $\infty$.

5699: %%For simplicity assume that $F = S$ and consider $f^{m_i}(F)$.

5700: %Then $S$ is in a local axis of $f = h^{n_1}$

5701: %and $f^n(S)$ converges to $F^*$ when $n$ converges to infinity.

5702: %Local axis means that $f(S)$ separates $S$ from $f^2(S)$.

5703: %Since $S$ is a translate of $F$, the structure of $\wlsf$

5704: %gets translated into $S$.

5705: 

5706: \vskip .1in

5707: \noindent

5708: {\bf {Claim}} $-$ $h(F^*) = F^*$.

5709: 

5710: Suppose this is not true. 

5711: The leaves $h(F^*), F^*$ are 

5712: not separated in $\hp$. 

5713: %the leaf space of $\fn$, see

5714: %fig. \ref{jump}, a.

5715: This implies that $\Theta(F^*)$ and $\Theta(h(F^*))$ are

5716: disjoint subsets of $\oo$, see fig. \ref{jump}.

5717: Therefore there are boundary leaves separating them.

5718: But $L$ intersects both $F^*$ and $h(F^*)$

5719: as $L$ intersects $F^*$ and is invariant under $h$.

5720: Therefore both 

5721: $\Theta(F^*)$ and $\Theta(h(F^*))$ intersect

5722: the same stable leaf $\Theta(L)$.

5723: 

5724: 

5725: Suppose that there is a stable boundary component of

5726: $\Theta(F^*)$ separating it from $\Theta(h(F^*))$.

5727: Then it has to be  a slice of $\Theta(L)$ as this

5728: set intersects both of them. It would not be a line

5729: leaf of $\Theta(L)$.

5730: But as remarked before, $h$ leaves invariant all the

5731: slices, line leaves and lift annuli of $L$ and

5732: this contradicts $\Theta(h(F^*))$

5733: % = \Theta(h^{n_1}(F^*))$ 

5734: being disjoint from $\Theta(F^*)$.

5735: This implies there is

5736: an {\underline {unstable}} boundary component

5737: of $\Theta(F^*)$ separating it from $\Theta(h(F^*))$,

5738: see fig. \ref{jump}.

5739: 

5740: 

5741: 

5742: \begin{figure}

5743: \centeredepsfbox{gm10.eps}

5744: \caption{

5745: Contradiction in the orbit space $\oo$.}

5746: \label{jump}

5747: \end{figure}

5748: 

5749: 

5750: 

5751: 

5752: %In the same way $L$ intersects $f^2(F^*)$ as well and so on.

5753: %Notice that $F^*$ does not intersect any periodic orbit in

5754: %$L$. 

5755: %This is because such orbits are invariant under $f$ and then

5756: %$F^*, f(F^*)$ would intersect the same orbit of

5757: %$\wwp$ and be separated.

5758: %%In addition Also $f$ leaves the sectors associated to the 

5759: %%periodic orbit invariant. This implies that line leaves

5760: %are $f$ invariant ???? 

5761: %In addition $F^*$ cannot intersect a lift

5762: %annulus $A$ in $L$: if it intersects $A$ then it intersects

5763: %all orbits of $\wwp$ in the interior of $A$.

5764: %But these sets are invariant by $f$

5765: %and $\Theta(F^*)$, $\Theta(f(F^*))$ are disjoint, contradiction.

5766: %

5767: %Now $L$ intersects $\Theta(F^*)$, $\Theta(f(F^*))$,

5768: %$\Theta(f^2(F^*))$, etc.. Similarly $\Theta(R)$

5769: %also intersects these same leaves. 

5770: 

5771: In the same way $\Theta(R)$ intersects both $\Theta(F^*)$ and

5772: $\Theta(h(F^*))$.

5773: Let $L_i = \ws(u_i)$. Recall from the

5774: beginning of the proof of  proposition \ref{band}

5775: that $u_i, v_i$ are  points in $F$ with $u_i$ converging

5776: to $p_0$ in $L$ and $v_i$ converging to $q_0$ in $R$.

5777: Then 

5778: $\Theta(L_i)$ converges to 

5779: $\Theta(L) \cup \Theta(R)$ (maybe other leaves as well).

5780: So $\Theta(L_i)$ intersects $\Theta(F^*)$ and $\Theta(h(F^*))$ 

5781: for $i$ big enough.

5782: The intersection of $\Theta(L_i)$ with 

5783: at least one of $\Theta(F^*)$ or $\Theta(h(F^*))$ 

5784: cannot be connected,

5785: see fig. \ref{jump}.

5786: This contradicts  propostion \ref{conn}.

5787: This contradiction implies that

5788: $h(F^*) = F^*$ and proves the claim.

5789: 

5790: \vskip .1in

5791: %This now has other consequences:

5792: So far we have  proved the following:

5793: in any case there is $i_0$ a positive integer

5794: so that if $f = h^{i_0}$ then

5795: $f(F^*) = F^*$. As $f(L) = L$ then

5796: $f(F^* \cap L) = F^* \cap L$

5797: and similarly $f(F^* \cap R) = F^* \cap R$.

5798: This produces an annulus $B$ in 

5799: $\pi(F^*)$ with a Reeb foliation.

5800: The region of $F^*$ bounded by $F^* \cap R$ and

5801: $F^* \cap L$ bounds a band $B$

5802: which is a bounded distance from a geodesic in

5803: $F^*$ and projects to a Reeb annulus in a leaf of $\fol$.

5804: 

5805: 

5806: 

5807: 

5808: But to prove proposition \ref{band},

5809: we really want these facts for $F$ and not just $F^*$.

5810: That is, we want a region in $\pi(E)$ which spirals

5811: towards a Reeb annulus.

5812: This turns out to be true: $\pi(E)$ has points converging

5813: to $\pi(F)$ and $\pi(F)$ has points converging to a Reeb

5814: annulus in $\pi(F^*)$. Since the annulus is compact,

5815: it turns out the second step is unnecessary.

5816: This depends on an analysis of holonomy of the foliation

5817: $\fol$ near the annulus in $\pi(F^*)$ as explained below.

5818: 

5819: 

5820: \vskip .1in

5821: \noindent {\bf {Claim}} $-$ The point

5822: $\pi(p_0)$ of $\pi(F)$ is in the boundary of

5823: a Reeb annulus of $\fol$ contained in $\pi(F)$.

5824: This implies that $F = F^*$.

5825: 

5826: %In fact $\pi(F)$ has

5827: %an annulus with a Reeb foliation.

5828: %

5829: %In $\pi(F)$ we consider the direction $\delta$

5830: %of collapsing of ideal

5831: %points. A closed curve which is oriented. Consider

5832: %its holonomy of $\fol$ in the direction

5833: %of $\delta$.

5834: 

5835: The point $z_0$ is in $F^* \cap L$. Then $\pi(z_0)$ is

5836: in $\pi(F^* \cap L) = \alpha$ which is a closed curve

5837: since $h^{i_0}$ leaves invariant both $F^*$ and $L$

5838: and their intersection is connected.

5839: Previous arguments in the proof imply that for

5840: $i$ big enough $h_i(z_i)$ is in the same local

5841: sheet of $\wls$ as $z_0$.

5842: Hence the points $\pi(z_i)$ are in $W^s(\pi(z_0)) = \pi(L)$

5843: and converge to $\pi(z_0)$. This shows that $\pi(F \cap L)$

5844: is asymptotic to $\alpha$ in the direction corresponding

5845: to the projection of the direction of escaping $z_i$ in

5846: the ray of $F \cap L$.

5847: Namely $\alpha$ has contracting holonomy (of $\fol$) in

5848: the side the $\pi(z_i)$ are converging to and eventually

5849: $\pi(z_i)$ is in the domain of contraction of $\alpha$.

5850: 

5851: This means that the direction of $F$ associated to the ideal

5852: point $\xi$ is a contracting direction towards

5853: $F^*$. The rays in the leaves

5854: $F^* \cap L, \ F^* \cap R$ in $F^*$ are

5855: a bounded distance from a geodesic ray in $F^*$ with

5856: same ideal point. The contraction above implies

5857: that the corresponding rays $F \cap L, \ F \cap R$ of $F$ 

5858: are also a bounded distance from a ray in $F$ with ideal

5859: point $\xi$.

5860: 

5861: 

5862: Now recall the points $p_i$ in $E$. We have $g_i(p_i)$

5863: very close to $p_0$ in the leaf $l$ of $\wlsf$.

5864: Also $\pi(l)$ is eventually in a region contracting

5865: towards a Reeb annulus of $\fol$.

5866: Hence if $i$ is big enough the $g_i(p_i)$ will also

5867: be in this region. The leaf of $\Lambda^s \cap \fol$

5868: through $\pi(p_i)$ will

5869: be contracted towards the Reeb annulus in that direction.

5870: This implies that the limit of the $\pi(p_i)$ is

5871: already in a Reeb annulus, consequently the limit

5872: of the $g_i(p_i)$ is

5873: already in a Reeb band. 

5874: %

5875: %

5876: %

5877: %\vskip .1in

5878: %%\noindent

5879: %%{\underline {Case 1}} $-$

5880: %$\delta$ has contracting holonomy.

5881: %

5882: %When we are moving in $\pi(F^*)$ in the direction

5883: %of $p$ we get arbitrarily close to $\pi(F^*)$ in that

5884: %direction. It will eventually get into the

5885: %domain of contraction. As the $\pi(A)$ is limiting

5886: %on $\pi(F)$ once it gets enough to $\pi(p_0)$, it

5887: %will also follow $\pi(F)$ until it hits this domain

5888: %of contraction. From then on it will spiral towards $\pi(F^*)$.

5889: %Hence $\pi(F^*) = \pi(F)$ and it already has

5890: %the Reeb annulus. We conclude that $F^* = F$ in this

5891: %case.

5892: %

5893: %\vskip .1in

5894: %From now on assume that $F^* \not = F$.

5895: %

5896: %\vskip .1in

5897: %\noindent

5898: %{\underline {Case 2}} $-$ There are $y_i \rightarrow 0$ with

5899: %$v(y_i) = y_i$. 

5900: %

5901: %Since $\pi(F)$ gets close enough to $\delta$ in this direction

5902: %it is either $y_i$ or between $y_i$ and $y_{i+1}$ for some

5903: %$i$. In either case moving in this direction

5904: %will make it limit either on $z(y_i)$ or $z(y_{i+1})$ and

5905: %not limit on $\pi(F^*)$. This case is impossible.

5906: %

5907: %\vskip .1in

5908: %\noindent

5909: %{\underline {Case 3}} $-$ $\delta$ has expanding holonomy.

5910: %

5911: %So $\delta^{-1}$ has contracting holonomy. Then $\pi(F)$

5912: %limits to $\pi(F^*)$ in the opposite direction. Since $f(L) = L$

5913: %and we are moving in the direction away from $\pi(F^*)$ we cannot have

5914: %$\pi(F)$ limiting on $\pi(F^*)$. We conclude that this case cannot

5915: %happen either.

5916: %

5917: %The final conclusion is that cases 2) and 3) do not occur, so

5918: %case 1) has to occur, which in fact implies that $F = F^*$.

5919: %

5920: %

5921: %Since the strip in $\pi(F)$ limits in $B$ and the strip

5922: %in $\pi(A)$ limits in $\pi(F)$ 

5923: 

5924: It now follows that 

5925: $\pi(F) = \pi(F^*)$.

5926: That means that the second perturbation procedure (from points

5927: in $F$ to points in $F^*$) in fact does not

5928: produce any new leaf. 

5929: This implies that up to covering translations then

5930: the leaf $E$ is asymptotic to $F$ in the direction of

5931: the ideal point $x$ in $\pin E$.

5932: Let $V$ be the region of $E$ bounded by $\upsilon$ with

5933: ideal point $x$. Then outside of a compact part

5934: it projects very near a Reeb annulus in $\pi(F)$ and so 

5935: this tail of $V$

5936: has no singularity of the foliation $\wlse$.

5937: By Euler characteristic reasons it follows that

5938: the interior of 

5939: $V$ has no singularities in the compact part of $V$ also.

5940: In fact the arguments show that the tail of $V$ flows

5941: into the {\underline {interior}} of a Reeb band 

5942: in a nearby leaf $U$ of $\fn$.

5943: Then leaves of $\wlse$ near $\upsilon$ on the outside of

5944: $V$ will also flow to interior of the Reeb band  in $U$.

5945: Therefore there are appropriate 2 rays on the outside 

5946: so that they will be in the same leaf of $\widetilde \Lambda^s_U$

5947: and hence in the same leaf of $\wls$.

5948: It follows that in $E$ the leaf $\upsilon$ is also approximated

5949: on the outside by a leaf which has a line leaf with 

5950: both ideal points the same. This implies that $\upsilon$

5951: has no singularities.

5952: 

5953: 

5954: 

5955: 

5956: 

5957: So far we proved the following:

5958: 

5959: 

5960: \vskip .1in

5961: \noindent

5962: {\bf {Conclusion}} $-$ Let $\upsilon$ be a slice of 

5963: $\wlse$ with 

5964: two rays converging to the same ideal point $x$ of

5965: $\pin E$  and $V$ is the region of $E$ bounded by $\upsilon$.

5966: Then $\upsilon$ has no singularity of $\wlse$ and neither does $V$.

5967: Also $\pi(V)$ is either contained in or asymptotic

5968: to a Reeb annulus in a leaf of $\fol$ and so $E$

5969: is asymptotic to a Reeb band in a leaf $F$ in

5970: the direction $x$.

5971: \vskip .1in

5972: 

5973: \begin{figure}

5974: \centeredepsfbox{gm11.eps}

5975: \caption{

5976: a. $l_i$ converging to non separated

5977: leaves $e_L, e_Z, e_Y, e_R$ of $\wlse$, b. Nested families

5978: and identifications of ideal points.}

5979: \label{nonsep}

5980: \end{figure}

5981: 

5982: 

5983: \noindent

5984: {\underline {Continuation of the proof of proposition \ref{band}}}.

5985: 

5986: What we want to prove is that in $E$ itself the region

5987: $V$ is contained in the interior of a spike region.

5988: Notice it is not true in general that $\pi(V)$ is

5989: contained in a Reeb annulus, only that it is

5990: asymptotic to a Reeb annulus. For instance start

5991: with a leaf of $\fol$ having a Reeb annulus and blow that

5992: into an I-bundle. Then produce holonomy associated

5993: to the core of the Reeb annulus. Then one produces

5994: Reeb bands asymptotic to but not contained in Reeb annuli.

5995: 

5996: Since $V$ is asymptotic to the Reeb band in $F$, it turns

5997: out that (after rearranging by covering translations) that

5998: $E$ intersects both $L$ and $R$ leaves of $\wls$.

5999: Their intersection produces two leaves $e_L, e_R$ of 

6000: $\wlse$ which are not separated from each other and which

6001: have the same ideal point $x$. There are then leaves

6002: $l_i$ of $\wlse$ all with ideal point $x$ and which 

6003: converge to $e_L \cup e_R$. This follows from

6004: the fact that in $F$ the same is true and $E$ is

6005: asymptotic to $F$ in that direction, plus the connectivity

6006: of the intersection of $E$ with leaves of $\wls$.

6007: 

6008: Now the sequence $l_i$ can converge to other

6009: leaves as well, all of which will be non separated

6010: from $e_L, e_R$. 

6011: The set of limits is an ordered set and the any other 

6012: leaf is 

6013: between $e_L$ and $e_R$.

6014: By theorem \ref{theb}

6015: there are only finitely many of them.

6016: We refer to

6017: fig. \ref{nonsep}, a, where for simplicity we consider there

6018: are 4 leaves in the limit: $e_L, e_Y, e_Z, e_R$ contained

6019: in leaves $L, Z, Y, R$ of $\wls$. These leaves of $\wls$

6020: are non separated from each other and form an ordered set.

6021: Let $\xi$ be the region of

6022: $E$ which is the  union of the region bounded by all the $l_i$

6023: plus the boundary leaves, which are non separated from $e_L, e_R$.

6024: Clearly every leaf in the interior has ideal point $x$ and

6025: has no singularity. We want to show that $\xi$ is

6026: a spike region.

6027: 

6028: 

6029: 

6030: 

6031: 

6032: Any two consecutive leaves 

6033: of $\partial \xi$ in this ordering will have rays

6034: with same ideal point and leaves $l_i$ converging to them.

6035: This situation is important on its own and is analysed in

6036: the following proposition:

6037: 

6038: 

6039: \begin{proposition}{}{}

6040: Suppose $v_1, v_2$ are non separated leaves in $\wlsg$ for some $G$ 

6041: leaf of $\fn$.

6042: Suppose there are no leaves non separated from $v_1, v_2$ in

6043: between them. Then

6044: the corresponding rays 

6045: of $v_1, v_2$ 

6046: have the same ideal point in $\pin G$. In addition

6047: they are a bounded distance from a geodesic ray of $G$ with same ideal point.

6048: In $M$ this region either projects to or is asymptotic to a Reeb

6049: annulus.

6050: \label{nh}

6051: \end{proposition}

6052: 

6053: \begin{proof}{}

6054: We do the essentially the

6055: same proof as in the case of leaves of $\wlsg$

6056: with same ideal points, except that we

6057: go in the direction of the non Hausdorfness. 

6058: Because there are no non separated leaves in between

6059: $v_1, v_2$, then the corresponding rays have the same ideal

6060: point.

6061: Choose $w_i, y_i$ in these rays of $v_1, v_2$ and escaping

6062: towards the ideal point and so that $d_G(w_i,y_i)$ is

6063: less than $4 \delta_0$.

6064: We do the limit analysis 

6065: using $f_i(w_i), f_i(y_i)$ converging in $\mi$. Because 

6066: $v_1, v_2$ are non separated it follows that $f_i(w_i), f_j(w_j)$

6067: are in the same stable leaf (of $\wls$) for $i, j$ big enough.

6068: Hence we can readjust so that they are all in the same

6069: stable leaf and similarly for $f_i(y_i)$. The same arguments

6070: as before show that that region of $G$ between $v_1, v_2$ projects

6071: in $M$ to set in a leaf of $\fol$

6072: which is either contained in or asymptotic to

6073: a Reeb annulus. The results follow.

6074: In general nothing can be said about the other direction

6075: in the leaves

6076: $v_1, v_2$: in particular it does not follow at all 

6077: that the other rays of $v_1, v_2$ have to have

6078: the same ideal point.

6079: \end{proof}

6080: 

6081: 

6082: 

6083: 

6084: %That is $\pi(F)$ has a Reeb annulus which the strip

6085: %in $\pi(E)$ limits to.

6086: %This implies that $h(F) = F$ 

6087: %(because $h^{n_1}(F) = F$ and the argument above with

6088: %$\Theta(L) \cap \Theta(F^*)$ not empty).

6089: %So now $g_i(A)$ limits to $F$. The limiting is obtained in

6090: %the converging direction. Since $h(F) = F$ then $F$ does

6091: %not intersect

6092: %the annuli or periodic orbits.

6093: %But

6094: %

6095: %$$\ws(\widetilde \alpha) \cap F \not = \emptyset,

6096: %\ws(\widetilde \alpha') \cap F \not = \emptyset,

6097: %\wu(\widetilde \beta) \cap F \not = \emptyset$$

6098: %

6099: %\noindent

6100: %

6101: %We analyse the situation in the Reeb annulus $A$ in $\pi(F)$ which

6102: %is associated to the limiting behavior of $\pi(V)$ ($V$ is

6103: %the region in $E$ bounded by a leaf $\upsilon$ with same

6104: %ideal points in $\pin E$).

6105: %The holonomy of $\Lambda^s$ in the positive flow 

6106: %direction (the flow $\Phi$)

6107: %is expanding $-$ since transverse to $\Lambda^s$ we

6108: %have unstable leaves. One boundary component of $A$

6109: %is in $\pi(F \cap L)$, hence is isotopic to a closed

6110: %orbit of $\Phi$ in $\pi(L)$.

6111: %This closed curve inherits an orientation from the positive

6112: %direction of that orbit. This is called the {\em flow direction}.

6113: %Similarly for the other component of $\partial A$ which 

6114: %is in $\pi(F \cap R)$.

6115: %Because of the above fact, the spiralling of the interior

6116: %leaves of the Reeb annulus $A$ limits to the boundary

6117: %leaves in the direction opposite to the flow direction,

6118: %see fig. \ref{direc} a.

6119: %Keeping track of directions of spiraling and holonomy

6120: %contraction will be crucial for all the analysis here.

6121: %

6122: %

6123: %\blankfig{direc}{1.1}{a. Direction of spiraling in the

6124: %leaves of the Reeb annlus $A$ in $\pi(F)$ is opposite

6125: %to the flow direction and also is the same

6126: %as the direction of contracting holonomy transverse

6127: %to $\fol$, b. The direction of holonomy in $\pi(G)$ is

6128: %the same as the flow direction and opposite

6129: %to the spiraling in the Reeb annulus in $\pi(G)$, c. The direction

6130: %of holonomy contraction in $\pi(H)$ is opposite

6131: %to the flow direction and the same as the direction

6132: %of spiraling in the Reeb annulus in $\pi(H)$.}

6133: %

6134: %

6135: %On the other hand the region in $\pi(E)$ associated to $\pi(V)$

6136: %limits to the Reeb annulus $A$ in $\pi(F)$. This limits in

6137: %the direction associated to the identified ideal points

6138: %of the leaf $\pi(v)$ $-$ this direction also corresponds

6139: %to the opposite of the flow direction, see fig. \ref{direc}, a.

6140: %Hence $\fol$ has contracting holonomy along $A$ in the side

6141: %$\pi(V)$ is limiting on and in the direction opposite to

6142: %the flow direction.

6143: %

6144: %

6145: %We will now continue this analysis along the leaves of $\wlse$

6146: %non separated from $e_L$ and in the direction from $e_L$ towards

6147: %$e_R$ passing through $e_Z, e_Y, etc..$.

6148: %Recall that $h$ is the generator of the isotropy group

6149: %of $L$ and $R$.

6150: %What we just did was to analyse the situation in $e_L$ as

6151: %one converges to the ideal point $x^*$. 

6152: %Next look at the other ray of $e_L$ converging to the ideal

6153: %point $x_1$ in $\pin E$.

6154: %This ray of $e_L$ is non separated from a ray of $e_Z$ 

6155: %which also has ideal point $x_1$ and both are a bounded

6156: %distance from a geodesic ray in $E$. By proposition \ref{nh},

6157: %the region between these rays projects in $M$ to a

6158: %set spiraling to a Reeb annulus $B$ in a leaf of

6159: %$\fol$. Lifting to $\mi$ this produces a leaf $G$

6160: %towards which $E$ is asymptotic to in that direction

6161: %and so that $B$ is contained in $\pi(G)$.

6162: %The proof of proposition \ref{nh} also shows that

6163: %the ray of $\pi(e_L)$ in question spirals towards

6164: %a closed curve $\sigma_1$ of $\pi(L)$ . 

6165: %Also $\sigma_0$ is isotopic to $\sigma_1$ in $\pi(L)$,

6166: %which then implies that $G$ is invariant by $h$ as well.

6167: %

6168: %Now the spiraling of this ray of $\pi(e_L)$ towards

6169: %$\sigma_1$ is in the direction opposite of the

6170: %spiraling of (the other ray of) $\pi(e_L)$ towards

6171: %$\sigma_0$ in $\pi(F)$. This is because 

6172: %$\sigma_0, \sigma_1$ are both closed curves

6173: %in the stable leaf 

6174: %

6175: %The spiraling direction in $\pi(F)$ lifts to the 

6176: %direction contrary to the flow direction, see fig. \ref{direc}, c.

6177: %The annulus in $\ws(\alpha)$ from $\pi(F)$ to $\alpha$

6178: %intersect $\pi(F)$ and all orbits flow asymptotically

6179: %to $\alpha$. Therefore if there is no other closed leaf

6180: %of $\fol$ in this annulus the leaves spiral towards $\pi(F)$

6181: %in the positive flow direction, opposite to the spiralling

6182: %direction, contradiction.

6183: %

6184: %

6185: %

6186: %

6187: %This means that in $A$ we are getting closer to $F$ as one goes

6188: %in the opposite direction of the spiralling. This implies that

6189: %the picture in $A$ is as desired in the proposition.

6190: %If there is another compact leaf then $\pi(A)$

6191: %has this region between two flow homeomorphic annuli,

6192: %see fig. \ref{direcs}, a.

6193: %Therefore it also has the same picture as in $F$.

6194: %If $\pi(A)$ is on the negative side of $F$,

6195: %then it intersects the $\wu(\widetilde \beta)$.

6196: %Notice $\pi(\beta)$ is freely homotopic to

6197: %the inverse of $\pi(\alpha)$. Hence now the 

6198: %spiralling direction of $\pi(F)$ agrees with the flow direction

6199: %of $\beta$.

6200: %Now $\pi(A)$ spirals towards $\pi(F)$ in the opposite direction

6201: %of the spiralling direction. This produces the same picture as above.

6202: %

6203: %\blankfig{direcs}{1.1}{a. Spiralling when directions go, b. The other case.}

6204: %

6205: %

6206: %In either case we produce the same picture as desired. This finishes

6207: %the proof of the proposition.

6208: %\end{proof}

6209: %

6210: %Just to clarify what we achieved: if $A$ has 2 rays with identified ideal points,

6211: %then it is non singular and they are part of a non Hausdorff strip which 

6212: %is a bounded distance from a geodesic. In $M$ either $\pi(C)$

6213: %is a Reeb annulus, or spirals towards a Reeb annulus in both directions.

6214: %This is the only thing that can happen when ideal points agree.

6215: %

6216: %What about non Hausdorff behavior? A lot can be said about that

6217: %too $-$ one has the following result:

6218: %

6219: 

6220: Given this last proposition then for any two consecutive

6221: rays in $\partial {\cal E}$ it follows that they are a bounded

6222: distance from a geodesic ray in $E$. 

6223: All that is needed to show that ${\cal E}$ is

6224: a spike region is to prove that the ideal

6225: points of the rays in the boundary are distinct except

6226: for consecutive rays.

6227: 

6228: Suppose there are other identifications of ideal points of

6229: leaves in the boundary of ${\cal E}$. Then there is at least one

6230: line leaf $\tau$ in the boundary of ${\cal E}$ so that

6231: $\tau$ has identified ideal points.

6232: Our analysis so far shows that $\tau$ is in the interior

6233: of another region similar to the one constructed above

6234: so that all leaves have just one common ideal point.

6235: Since the $l_i$ limit on $\tau$, then the ideal

6236: point of $\tau$ has to be $x$. In addition the leaves in

6237: this new region have to be nested. But if the $l_i$ together

6238: with $\tau$ are a nested family of leaves of $\wlsf$, 

6239: then the $\tau$ is outside the $l_i$ hence the region

6240: in $E$ bounded by $\tau$ enclosed the whole region

6241: ${\cal E}$, see fig. \ref{nonsep}, b.

6242: There is at least one other leaf $\tau'$ in

6243: $\partial {\cal E}$. The same arguments we used for $\tau$

6244: can be applied to $\tau'$. But it is impossible that

6245: the $l_i$ are also nested with the $\tau'$, see fig. 

6246: \ref{nonsep}, b.

6247: 

6248: 

6249: This shows that the ideal points of ${\cal E}$ are

6250: distinct except as mandated by consecutive rays. In addition

6251: any line leaf in the boundary of ${\cal E}$ has distinct

6252: ideal points and rays which are a bounded distance from

6253: geodesic rays. It follows that the whole leaf is

6254: a bounded distance from a geodesic in $E$.

6255: This shows that 

6256: ${\cal E}$ is a spike region.

6257: This finishes the proof of proposition \ref{band}.

6258: \end{proof}

6259: 

6260: Finally in the case $\wls$ has Hausdorff leaf space

6261: one can say much, much more about metric properties

6262: of leaves of $\wlsf$:

6263: 

6264: \begin{proposition}{}{}

6265: Suppose that $\Phi$ is an almost pseudo-Anosov flow transverse

6266: to a foliation $\fol$ with hyperbolic leaves.

6267: Suppose that $\wls$ has Hausdorff leaf space.

6268: %and that $\fol$ is not

6269: %topologically conjugate to the stable foliation of a suspension

6270: %Anosov flow . 

6271: Then there is $k_0 > 0$

6272: so that for any $F$ leaf of $\wls$, then the slice leaves 

6273: of $\wlsf$ are uniform $k_0$-quasigeodesics.

6274: \end{proposition}

6275: 

6276: \begin{proof}{}

6277: If there is a leaf $F$ of $\fn$ and a slice leaf of $\wlsf$

6278: with only one ideal point, then the proof of proposition

6279: \ref{band} shows that there are leaves of $\wls$ non

6280: separated from each other. This is impossible.

6281: 

6282: Suppose now that for any integer $i$, there are 

6283: $x_i$ in $\mi$, $x_i$ in leaves $F_i$ of $\fn$ with

6284: $x_i$ in line leaves $l_i$ of $\widetilde {\Lambda}^s_{F_i}$

6285: with distance

6286: from $x_i$ to $l^*_i$ in $F_i$ going to infinity.

6287: Here $l^*_i$ is the geodesic in $F_i$ with same ideal

6288: points as $l_i$.

6289: Up to covering translations assume $x_i$ converges

6290: to $x$.

6291: Also assume all $x_i$ are in the same sector 

6292: of $\wls$ defined by $x$. Since $l_i$

6293: converges to $l$, the arguments

6294: in lemmas \ref{encon} and \ref{cont} would show that

6295: the ideal points of $l$ are the same. 

6296: This was just disproved above.

6297: 

6298: Given that, the line leaves are within some global distance

6299: $a_0$ 

6300: of the respective

6301: geodesics in their leaves.

6302: It is well known that these facts imply that

6303: the slice leaves of $\wlsf$ are uniform quasigeodesics.

6304: For a proof of this well known fact see 

6305: for example \cite{Fe-Mo}.

6306: \end{proof}

6307: 

6308: 

6309: \section{Continuous extension of leaves}

6310: 

6311: The purpose of this section is to prove the main theorem: \

6312: the continuous extension property for leaves of foliations

6313: which are almost transverse to quasigeodesic singular pseudo-Anosov

6314: flows in atoroidal $3$-manifolds. As seen

6315: before this implies that $M$ has negatively curved

6316: fundamental group.

6317: 

6318: 

6319: Suppose first that $\Phi$ is an almost pseudo-Anosov flow

6320: which is transverse to a foliation $\fol$ 

6321: with hyperbolic leaves

6322: in a

6323: general closed $3$-manifold $M$.

6324: Given a leaf $F$ of $\fn$

6325: we introduce geodesic ``laminations" in $F$  

6326: coming from $\wlsf, \wluf$.

6327: We only work with the stable foliation, similar results hold for the

6328: unstable foliation.

6329: Assume that a leaf $l$ of $\wlsf$ is not singular. If both ideal

6330: points are the same let $l^*$ be empty. Otherwise let $l^*$ be the geodesic

6331: with same ideal points as $l$. If $l$ is singular, then no line leaves

6332: of $l$ 

6333: have the same ideal point by proposition \ref{band}. For each line

6334: leaf $e$ 

6335: of $l$

6336: let $e^*$ be the corresponding geodesic and $l^*$ their

6337: union. Let now $\tau^s_F$ be the union of these geodesics

6338: of $F$.

6339: Leaves of $\wlsf$ do not have transverse intersections and

6340: therefore the same happens for leaves of $\tau^s_F$.

6341: 

6342: Suppose that $\wlsf$ has non separated leaves 

6343: $l, v$ which are not in the boundary of a spike region.

6344: Then there are $l_i$ converging to $l \cup v$ (and maybe

6345: other leaves as well),

6346: but $l^*_i$ 

6347: does not converge to $l^*$ or $v^*$. Notice none of the limit leaves

6348: can have identified ideal points, because then they

6349: would be in the interior of a spike region (proposition 

6350: \ref{band}) and have a neighborhood which is product

6351: foliated.

6352: Let $\otsf$ be the closure

6353: of $\tau^s_F$. Then $\otsf$ is a geodesic lamination in $F$.

6354: Similarly define $\tau^u_F$, $\overline \tau^u_F$.

6355: In a complementary region $U$ 

6356: of $\otsf$ associated to non Hausdorffness, there

6357: is one boundary component which is added (a leaf of $\otsf - \tau^s_F$)

6358: and which is the limit of the $l^*_i$ as above.

6359: All of the other boundary

6360: leaves of the region

6361: are associated to the non separated leaves

6362: of $\wlsf$ and are in $\tau^s_F$.

6363: 

6364: 

6365: \begin{lemma}{}{}

6366: The new leaves in $\otsf$ (that is those

6367: in $\otsf$ - $\tau^s_F$) come from non Hausdorff

6368: pairs $(l,v)$ 

6369: of $\wlsf$

6370: as in the description above.

6371: \end{lemma}

6372: 

6373: \begin{proof}{}

6374: Let $e_i$ in $\tau^s_F$ converging to $e$ not in $\tau^s_F$.

6375: Then choose $l_i$ line leaves in $\wlsf$ with

6376: $e_i = l^*_i$. Given $u$ a point in $e$, there is

6377: $u_i$ in $l^*_i$ very close to $u$. Then there are

6378: $p_i$ in $l_i$ which are $2 \delta_0$ close to $u_i$. 

6379: Up to subsequence

6380: assume that $p_i$ converges to $p_0$ and

6381: let $l$ be the line leaf of $\wlsf$ that the sequence

6382: $l_i$ converges to. 

6383: Since the $l^*_i$ converges to $e$ which is not in

6384: $\otsf$ and $l^*$ is in $\otsf$, it follows that 

6385: $l^*_i$ does not converge to $l^*$.

6386: By lemma  \ref{encon} this is associated to 

6387: a non Hausdorff situation:

6388: $l_i$ converging to $l$ and other leaves as well and $l^*$ is

6389: the added leaf associated to this non Hausdorfness. This finishes

6390: the proof of the lemma.

6391: \end{proof}

6392: 

6393: \begin{lemma}{}

6394: The complementary regions of $\otsf$ are 

6395: ideal polygons associated to singular leaves and non

6396: Hausdorff behavior of $\wls$.

6397: If $M$ is atoroidal 

6398: then these regions

6399: are finite sided ideal polygons.

6400: \end{lemma}

6401: 

6402: \begin{proof}{}

6403: Let $x$ be in a complementary region $U$ of $\otsf$.

6404: Let $e$ be a leaf in the boundary $\partial U$.

6405: Let $I$ be the interval of $\pin F - \partial e$ containing

6406: other ideal points of $U$.

6407: Suppose first that $e$ is an actual leaf of $\tau^s_F$,

6408: which comes from a line leaf $l$ of $\wlsf$.

6409: It may be that $l$ is contained in a singular leaf $z$ of

6410: $\wlsf$ which

6411: is singular on the $x$ side.  This means that $z$ has ideal

6412: points in $I$.

6413: In that case $x$ is in the complementary region obtained

6414: by splitting $z$. This region must be $U$.

6415: Otherwise $l$ is not singular on the side containing

6416: $x$ and we may assume there

6417: are $l_i$ leaves of $\wlsf$

6418: with ideal points in the closure of $I$ in $\pin F$,

6419: with $l_i$ converging to $l$. If

6420: the ideal points of $l_i$ converge to that of $l$ then eventually

6421: $l^*_i$ separates $x$ from $e$ and $x$ is

6422: not in the complementary region $U$ $-$ impossible. 

6423: Hence the ideal points of $l_i$ do not

6424: converge to $\partial e$ and there is non Hausdorfness and

6425: a complementary region in that side of $l$. Then $x$ needs to be

6426: in this complementary region (which is $U$) and $e$ is a boundary

6427: leaf of $U$ which comes from a line leaf of $\tau^s_F$. 

6428: 

6429: Suppose now that $e$ is an added leaf.

6430: There are $l_i$ 

6431: leaves of $\wlsf$ with $e_i = l^*_i$ converging to $e$ on the side

6432: opposite to $x$, otherwise $x$ is not in $U$.

6433: Then $l_i$ converges to more than one leaf of $\wlsf$ 

6434: producing non Hausdorff behavior

6435: and a complementary region with $e$ in its boundary.

6436: The $x$ is in the region

6437: associated to this non Hausdorff behavior, so

6438: the complementary region must be $U$. 

6439: 

6440: If there is a complementary region of $\otsf$ with infinitely

6441: many sides then it is associated to non Hausdorff behavior

6442: and so there are leaves $l_i$ of $\wlsf$ converging to infinitely

6443: many distinct leaves of $\wlsf$.

6444: Then there is $L$ leaf of $\wls$ which is non separated from

6445: infinitely many other leaves. Theorem \ref{theb} implies that

6446: there is a ${\bf Z} \oplus {\bf Z}$ subgroup

6447: of $\pi_1(M)$, contradiction.

6448: % and hence $\pi_1(M)$ is not negatively 

6449: %curved, contradiction. 

6450: This finishes the proof.

6451: %%converging to $l$ on the other side. This produces

6452: %a complementary region and $x$ needs to be in these.

6453: \end{proof}

6454: 

6455: We now turn to the continuous extension property.

6456: %A preliminary analysis of the continuous extension property

6457: %was done in \cite{Fe6}

6458: %in the case that $\Lambda^s, \Lambda^u$ where quasi-isometric

6459: %singular foliations, $\fol$ is a finite depth foliation,

6460: %and $\Phi$ is a pseudo-Anosov {\underline {transverse}} to $\fol$. 

6461: %Under these conditions it was shown in \cite{Fe6}

6462: %that leaves of $\wlsf, \wluf$ are 

6463: %uniform quasigeodesics in their respective leaves $F$.

6464: %Here we are analysing a much more general situation: 

6465: %in particular there are examples where

6466: %$\wls, \wlu$ have non Hausdorff leaf space \cite{Mo5,Fe6}

6467: %and so 

6468: %$\wlsf, \wluf$  can have non Hausdorff leaf space, immediately implying

6469: %that their leaves cannot be uniform quasigeodesics.

6470: %In addition the results here apply to general foliations, for instance

6471: %to foliations with dense leaves, foliations

6472: %with one sided branching, etc.. The main theorem is:

6473: 

6474: 

6475: \begin{theorem}{(Main theorem)}{}

6476: Let $\fol$ be a foliation in $M^3$ closed, atoroidal.

6477: %with

6478: %$\pi_1(M)$ negatively curved.

6479: Suppose 

6480: that $\fol$ is almost transverse to a 

6481: quasigeodesic, singular pseudo-Anosov flow $\Phi_1$

6482: and transverse to an associated almost pseudo-Anosov flow $\Phi$.

6483: Singular means $\Phi_1$ is not a topological Anosov flow.

6484: Then for any leaf $F$ of $\fn$, the inclusion map

6485: $\Psi: F \rightarrow \mi$ extends to a continuous map

6486: 

6487: $$\Psi: \ F \cup \pin F \ \rightarrow \ \mi \cup \si$$

6488: 

6489: \noindent 

6490: The map

6491: $\Psi$ restricted to $\pin F$, gives a continuous parametrization

6492: of the limit set of $F$, which is then locally connected.

6493: \label{exten}

6494: \end{theorem}

6495: 

6496: \begin{proof}{}

6497: The hypothesis imply that $\pi_1(M)$ is negatively curved.

6498: Difficulties in the proof  of this result are that

6499: $\wls, \wlu$ may have non Hausdorff leaf space \cite{Mo5,Fe6}

6500: and so 

6501: $\wlsf, \wluf$  can have non Hausdorff leaf space. This

6502: implies

6503: that the leaves of $\wlsf, \wluf$ 

6504: cannot be uniform quasigeodesics in $F$.

6505: In addition the leaves of $\wls, \wlu$ are not quasi-isometrically

6506: embedded in $\mi$.

6507: The proof is done in two steps: first we define an extension and then

6508: we show that it is continuous.

6509: 

6510: The proof will fundamentally use the fact that $\Phi_1$ is

6511: a quasigeodesic pseudo-Anosov flow.

6512: It was proved in \cite{Fe-Mo} that this implies

6513: that $\Phi$ is a quasigeodesic flow as well.

6514: From now on we use the stable/unstable foliations

6515: $\wls, \wlu$ of $\wwp$.

6516: First we need to review some facts about quasigeodesic 

6517: almost pseudo-Anosov flows. 

6518: If $\gamma$ is an orbit of $\wwp$ then it is

6519: a quasigeodesic and hence has unique distinct

6520: ideal points $\gamma_-$ and $\gamma_+$ 

6521: in $\si$ corresponding to

6522: the positive and negative flow directions \cite{Th1,Gr,Gh-Ha,CDP}.

6523: Hence given $x$ in $\mi$ define

6524: 

6525: $$\eta_+(x) \ = \ \gamma_+, \ \ \ \eta_-(x) = \ \gamma_-,

6526: \ \ \ \eta_+(x) \ \not = \ \eta_-(x),$$

6527: 

6528: 

6529: 

6530: \noindent

6531: where $\gamma$ is the $\wwp$ flowline through $x$.

6532: If $L$ is a leaf of $\wls$ or $\wlu$ and $a$ is a limit point

6533: of $L$ in $\si$, then there is an orbit $\gamma$ of $\wwp$ 

6534: contained in $L$ with either $\gamma_- = a$ or $\gamma_+ = a$,

6535: that is, any limit point of $L$ is a limit point of one

6536: of its flow lines \cite{Fe2,Fe6}.

6537: Also any such $L$ in $\wls$ is Gromov negatively curved

6538: %in the large 

6539: \cite{Gr,Gh-Ha,Fe2,Fe6} and

6540: has an intrinsic ideal boundary $\partial L$ 

6541: consisting of a single forward ideal point and distinct negative

6542: ideal points for each flow line

6543: \cite{Fe2,Fe6}. The set $L \cup \pin L$ is a natural

6544: compactification of $L$ in the Gromov sense.

6545: For instance if $L$ is a non singular leaf,

6546: then $L \cup \pin L$ is a closed disk. In this

6547: case the foliation by flow lines in $L$ is equivalent

6548: to the foliation in $\hh$ by geodesics sharing a fixed

6549: point in $\su$.

6550: 

6551: 

6552: A very important fact for us is that  the inclusion

6553: 

6554: $$\kappa: L \rightarrow \mi \ \ \ \ 

6555: {\rm extends \  to \ a \ continuous \

6556: map} \ \ \ \  \kappa: L \cup \partial L \rightarrow \mi \cup \si,$$

6557: 

6558: \noindent

6559: \cite{Fe2,Fe6}.

6560: This all follows from the fact that $\Phi$ is quasigeodesic.

6561: This works for any $L$ in $\wls$ or $\wlu$.

6562: If $L$ is in $\wls$

6563: there is a unique distinguished ideal point in $\si$ denoted

6564: by $L_+$

6565: which is the forward limit point in $\si$ of any flow line

6566: in $L \subset \mi$.

6567: If in addition $\Lambda^s$ is a quasi-isometric singular

6568: foliation, then the extension $\kappa$ is always a homeomorphism

6569: into its image, but this is not true if $\Lambda^s$ is

6570: not quasi-isometric.

6571: Similarly for $L$ in $\wlu$.

6572: 

6573: Throughout the proof we fix a unique identification

6574: of $\mi \cup \si$ with the closed unit ball in

6575: ${\bf R}^3$.

6576: The Euclidean metric in this ball induces the visual distance

6577: in $\mi \cup \si$.

6578: Then $diam(B)$ denotes the diameter in this distance for

6579: any subset $B$ of $\mi \cup \si$.

6580: A notation used throughout here is the following: 

6581: if $A$ is a subset of a leaf $F$ of $\fn$, then $\overline A$

6582: is its closure in $F \cup \pin F$. 

6583: %If $A$ is a subset

6584: %of $\mi$, then $\overline A$ is its closure in $\mi \cup \si$.

6585: %Whenever needed we will make explicit where the closure

6586: %is being taken.

6587: 

6588: 

6589: 

6590: We now produce an extension $\Psi: \pin F \rightarrow \si$.

6591: 

6592: \vskip .1in

6593: \noindent

6594: {\underline {Case 1}} $-$ Suppose that $v$ in $\pin F$

6595:  is not an ideal point

6596: of a ray in $\wlsf$ or in $\wluf$.

6597: 

6598: Since $\pi_1(M)$ is negatively curved, then

6599: complementary regions of $\otsf$ are finite sided ideal

6600: polygons.

6601: Hence there are $e_i$ in $\otsf$ 

6602: so that $\{ e_i \cup \partial e_i \}, \ i \in {\bf N}$ define

6603: a neighborhood basis of $v$ (in $F \cup \pin F$) 

6604: and $\{ e_i \}$ forms a nested

6605: sequence.

6606: Here $\partial e_i$ are the ideal points of $e_i$ in $\pin F$.

6607: We say that the $\{ e_i \}$ define a neighborhood basis at $v$.

6608: Assume that no two $e_i$ share an ideal point $-$

6609: possible because of hypothesis.

6610: If $e_i$ is in $\otsf - \tau^s_F$ then it is the limit of leaves

6611: in $\tau^s_F$ and by adjusting the sequence above we can

6612: assume that $e_i$ is always in $\tau^s_F$.

6613: Let $l_i$ in $\wlsf$ with $l^*_i = e_i$

6614: and $L_i$ leaves of $\wls$ with $l_i \subset L_i$.

6615: %By changing $l_i$ slightly we may assume that the

6616: %$L_i$ are not singular leaves and similarly for the $B_i$.

6617: 

6618: Similarly there are $c_i$ in $\otuf$ defining a neighborhood

6619: basis of $v$.

6620: Up to subsequence we may assume that $e_1, c_1, e_2, c_2$, etc..

6621: are nested and none of them have any common ideal points

6622: (in $F \cup \pin F$)

6623: and $c_i$ is in $\tau^u_F$.

6624: Let $b_i$ in $\wluf$

6625: with $b^*_i = c_i$ and $B_i$ leaves of $\wlu$ with $b_i \subset B_i$.

6626: %We may assume that $l_i$ are not singular. Also $b_i$ in $\wluf$

6627: %with $b^*_i = c_i$.

6628: %and $B_i$ in

6629: %$\wlu$ with $b_i \subset B_i$.

6630: 

6631: 

6632: 

6633: 

6634: 

6635: At this point we need the following result:

6636: 

6637: \begin{lemma}{}{}

6638: Let $L$ leaf of $\wls$, $B$ leaf of $\wlu$ and $F$ leaf of $\fn$

6639: so that $F$ intersects both $L$ and $B$: $l = L \cap F,

6640: \ b = F \cap B$. Suppose that $b$ and $l$ are disjoint

6641: in $F$. Then $L$ does not intersects $B$ in $\mi$.

6642: \end{lemma}

6643: 

6644: \begin{proof}{}

6645: Suppose not. Recall that $\Theta(L), \Theta(B)$ are

6646: finite pronged, non compact

6647: trees and they intersect in a compact subtree.

6648: The union is also a finite pronged tree.

6649: In addition $\Theta(L \cap B)$ is connected.

6650: The sets $\Theta(l), \Theta(b)$ are disjoint in this union.

6651: Let $x$ be a boundary point of $\Theta(l)$ which is

6652: either in $\Theta(L \cap B)$ or separates $\Theta(L \cap B)$

6653: from $\Theta(l)$ in this union, see fig. \ref{inter}, a.

6654: 

6655: 

6656: \begin{figure}

6657: \centeredepsfbox{gm12.eps}

6658: \caption{

6659: a. Obstruction to intersections of leaves,

6660: b. The case of $F$ escaping up.}

6661: \label{inter}

6662: \end{figure}

6663: 

6664: 

6665: 

6666: Let $\gamma = x \times \rrrr$, an orbit of $\wwp$.

6667: The first possibility is that $F$ escapes up

6668: as $\Theta(F)$ approaches $x$.

6669: Then $\gamma$ is a repelling orbit with respect to the

6670: $\Theta(l)$ side, see fig. \ref{inter}, b

6671: and $\gamma$ is in the boundary of a lift annulus

6672: $A$.

6673: This means that $\Theta(l)$ is a generalized unstable prong

6674: from the point of view of $x$.

6675: By proposition \ref{bounda} there is a stable slice

6676: $r$ of $\oos(x)$ with $r$ contained in $\partial \Theta(F)$

6677: and $F$ escapes up as $\Theta(F)$ approaches $r$, see

6678: fig. \ref{inter}, a.

6679:  The two sides of $r$ are the closest

6680: generalized prongs to $\Theta(l)$ on either side of $\Theta(l)$.

6681: This implies that $r$ separates $\Theta(b)$ from $\Theta(F)$

6682: see fig. \ref{inter}, a.

6683: Then $\Theta(b)$ cannot be contained in $\Theta(F)$, contradiction.

6684: 

6685: The second option is that $F$ escapes down as $\Theta(F)$ 

6686: approaches $x$ along $\Theta(l)$.

6687: Here there is a slice $r$ of $\oou(x)$ with $r$ contained

6688: in $\partial \Theta(F)$ and the closest to $\Theta(l)$ 

6689: on both sides of $\Theta(l)$. 

6690: Either $\Theta(b) \subset r$ or $r$ 

6691: separates $\Theta(b)$ from $\Theta(F)$. In any case

6692: $\Theta(b)$ does not intersect $\Theta(F)$, again

6693: a contradiction.

6694: This finishes the proof of the lemma.

6695: \end{proof}

6696: 

6697: 

6698: \vskip .1in

6699: \noindent

6700: {\underline {Claim}} $-$ Both $L_i$ and $B_i$ escape in $\mi$.

6701: 

6702: Notice $e_i \cap c_j = \emptyset$ for any $i, j$.

6703: If $l_i \cap b_j$ is non empty with $j > i$, then the nesting

6704: property above implies that $b_{i+1}, b_{i+2}, ..., b_j$

6705: all have to intersect. Since there is a global upper

6706: bound on the number of prongs of leaves of $\wls, \wlu$, this

6707: can happen for only finitely many times. Up to taking a 

6708: further subsequence we may assume that all the $l_i, b_j$

6709: are disjoint. 

6710: 

6711: 

6712: The lemma shows that 

6713: $L_i \cap B_j$ is empty for any $i, j$, and they

6714: form nested sequences of leaves in $\mi$.

6715: Suppose that the sequence $\{ L_i \}$ does not escape

6716: compact sets. Then there is $L$ in $\wls$ which is a limit

6717: of $L_i$ (and possibly other leaves as well).

6718: Let $\alpha$ be an orbit in $L$ which is not in

6719: a lift annulus. Then $\wu(\alpha)$ is transverse

6720: to $L$ in $\alpha$ and hence intersects $L_i$ for

6721: $i$ big enough. Since the $L_i, B_j$ are nested

6722: this would force $\wu(\alpha)$ to intersect

6723: $B_j$ for $j$ big enough, contradiction.

6724: It follows that both $L_i$ and $B_j$ escape compact sets

6725: as $i, j \rightarrow \infty$.

6726: 

6727: Let $r$ be a geodesic ray in $F$ with ideal point $v$.

6728: For each $i$, there is a subray of $r$ contained in

6729: the component of $F - l_i$ which is in a small neighborhood

6730: of $v$. Hence $\Psi(r)$ has a subray which is contained 

6731: in the corresponding component $V_i$ of $\mi - L_i$.

6732: These components $V_i$ form a nested sequence. The

6733: ray $\Psi(r)$ can only limit in the limit set of $V_i$.

6734: We need the following lemma which will be 

6735: a key tool 

6736: throughout the proof.

6737: 

6738: \begin{lemma}{(basic lemma)}{}

6739: Let $\{ Z_i \}$ be a sequence of leaves or line leaves

6740: or slices or any flow saturated sets in 

6741: leaves of either in $\wls$ or

6742: $\wlu$ (not all leaves $Z_i$

6743: need to be in the same singular foliation).

6744: If the sets $Z_i$ escape compact sets in $\mi$, then

6745: up to taking a subsequence $\overline Z_i$ converges 

6746: to a point in $\si$.

6747: \label{narrow}

6748: \end{lemma}

6749: 

6750: \begin{proof}{}

6751: Let $Y_i$ be the leaf of $\wls$ or $\wlu$ which contains $Z_i$.

6752: Up to subsequence assume $Y_i \in \wls$.

6753: The statement is equivalent to $diam(Z_i)$ converges to $0$.

6754: Otherwise up to subsequence we can assume $diam(Z_i) > a_0$

6755: for some $a_0$ and all $i$ and hence no subsequence

6756: can converge to a single point in $\si$.

6757: %Suppose then that $diam(Z_i) > a_0$ for all 

6758: %$i$ and $a_0 > 0$.

6759: %Up to subsequence assume $Z_i$ are (say) all stable and

6760: %that $(Z_i)_+$ converges to $p$ a point in $\si$.

6761: %For big enough $i$,

6762: Then there is $p_i$ in $Z_i$ with visual distance from

6763: $p_i$ to $(Y_i)_+$ is bigger than $a_0/2$.

6764: Notice that $(Y_i)_+$ is a point in $\overline Z_i$.

6765: Let $\gamma_i$ the orbit of $\wwp$ through $p_i$.

6766: If $(\gamma_i)_-$ is very close to $(\gamma_i)_+ = (Y_i)_+$

6767: then the geodesic with these ideal points has

6768: very small visual diameter. Since $\gamma_i$ is a global

6769: bounded distance from this geodesic \cite{Gr,Gh-Ha,CDP},

6770: the same

6771: is true for $\gamma_i$ contradiction to the choice of $p_i$.

6772: Hence the geodesic above intersects a fixed

6773: compact set in $\mi$ and so does $\gamma_i$. This

6774: contradicts the fact that $Z_i$ escape compact

6775: sets in $\mi$ and finishes the proof.

6776: \end{proof}

6777: 

6778: 

6779: 

6780: 

6781: %\vskip .1in

6782: %\noindent

6783: %{\underline {Claim}} $-$ 

6784: We claim that the limit sets of $V_i$ above shrink to a single

6785: point in $\si$.

6786: The limit sets form a weakly monotone decreasing sequence, because

6787: the $L_i$ are nested and so are the $V_i$.

6788: If the limit set does not have diameter going to zero,

6789: then there are points in the limit set of $L_i$ which

6790: are at least $ 2 \delta_1$ apart for some fixed

6791: $\delta_1 > 0$.

6792: By the previous lemma the $L_i$ cannot escape 

6793: compact sets in $\mi$, contradiction.

6794: %So one of them is at least $\delta_0$ from $L_+$.

6795: %There is a flow line $\gamma_i$ of $L$ with these ideal points

6796: %and since it is a uniform quasigeodesic, it is a bounded

6797: %distance from a geodesic of $\mi$ connecting these ideal points.

6798: %Then $\gamma_i$ has to intersect a fixed compact set in $\mi$.

6799: %This contradicts the fact that $L_i$ escapes compact sets

6800: %in $\mi$.

6801: Since the limit sets of $V_i$ shrinks to a point in $\si$, let

6802: $\Psi(v)$ be this point. Clearly $\Psi(r)$ limits

6803: to this point and so does $\Psi(r')$ for any other geodesic

6804: ray $r'$ in $F$ with ideal point $v$. 

6805: 

6806: 

6807: 

6808: \vskip .15in

6809: \noindent

6810: {\underline {Case 2}} $-$ Suppose that $v$ is an ideal point

6811: of a leaf of $\wlsf$ or $\wluf$.

6812: 

6813: Let $l$ be a ray in say $\wlsf$ which limits on $v$

6814: and $r$ a geodesic ray on $F$ with ideal point $v$.

6815: Then $l$ is contained in $L$ leaf of $\wls$. 

6816: Either $\Theta(l)$ escapes in $\Theta(L)$ or limits

6817: to a point $x$ in $\Theta(L)$.

6818: 

6819: Consider the first case. Then in the intrinsic geometry of

6820: $L$, the ray $l$ converges to the positive ideal point

6821: of $L$, hence in $\mi \cup \si$, the image $\Psi(l)$ converges

6822: to $L_+$.

6823: In the other option let $\beta = x \times \rrrr$,

6824: an orbit of $\wwp$.

6825: As $l$ escapes in $F$ then in $L$ it either escapes up or

6826: down. If it escapes down then it converges to the

6827: negative ideal point of $\beta$ in $L \cup \pin L$

6828: and hence $\Psi(l)$ converges to $\beta_-$. Otherwise

6829: $l$ escapes up in $L$ as $\Theta(l)$ approaches

6830: $x$. In this case $\beta$ is in the boundary of

6831: a lift annulus and $l$ converges to the positive ideal

6832: point in $L \cup \pin L$ and so $\Psi(l)$ converges to $L_+$ 

6833: again.

6834: 

6835: The remaining case is that $\Theta(l)$ escapes in $\Theta(L)$.

6836: Then as seen in $L \cup \pin L$ the ray $l$ converges

6837: to the positive ideal point $p$ of all flow lines in $L$.

6838: Hence $\Psi(l)$ converges to $\kappa(p) = L_+$.

6839: Let $\Psi(v)$ be the limit point in any case.

6840: Similarly if $l$ is a ray of $\wluf$.

6841: 

6842: 

6843: 

6844: Every point in $r$ it is $2 \delta_0$ close

6845: to a point in $l$ in $F$, hence the limit of $\Psi(r)$ in

6846: $\mi \cup \si$ is the same as that of

6847: $l$. If $l'$ is another ray 

6848: of $\wlsf$ or $\wluf$ converging to $v$,

6849: then it will have

6850: points boundedly close to $r$ which escape in 

6851: $l'$ and therefore $\Psi(l')$ 

6852: has the same ideal point

6853: in $\si$. 

6854: %Similarly if the point is an ideal

6855: %point of leaves of $\wluf$. 

6856: Therefore $\Psi(v)$ is well defined.

6857: 

6858: This finishes the construction of the extension

6859: of $\Psi$ to $\pin F$.

6860: 

6861: \vskip .1in

6862: \noindent

6863: {\underline {Proof of continuity of the extension}} $-$

6864: 

6865: \vskip .1in

6866: \noindent

6867: {\underline {Case 1}} $-$  $v$ is not an ideal point of 

6868: a ray in $\wlsf$ or $\wluf$.

6869: 

6870: Let $r$ be a geodesic ray in $F$ with ideal point $v$.

6871: Recall the extension construction. 

6872: There are $l_i$ in $\wlsf$ shrinking to $v$

6873: in $F \cup \pin F$ and similarly $b_i$ in $\wluf$, 

6874: assumed to be nested with the $l_i$.

6875: Let $\{ l^*_i \}$ define a neighborhood basis of $v$ in

6876: $F \cup \pin F$.

6877: Let $L_i$ in $\wls$ with 

6878: $l_i \subset L_i$, and $b_i \subset B_i \in \wlu$ 

6879: as in the construction

6880: case 1. Then as seen in the construction, the 

6881: $L_i, B_i$ escape in $\mi$.

6882: Let $U_i$ be the component of $F - l_i$ containing a subray of

6883: $r$ and $V_i$ the component of $\mi - L_i$ containing $U_i$.

6884: Notice that $\Psi(U_i) \subset V_i$.

6885: Let now $z$ in $\overline U_i$ with the closure taken in 

6886: $F \cup \pin F$ and $\overline V_i$ the closure of $V_i$

6887: in $\mi \cup \si$.

6888: Then $\overline U_i$ is a neighborhood of $v$ in $F \cup  \pin F$.

6889: If $z$ is in $\Psi(\overline U_i)$ then using either of the

6890: constructions in the extension part shows that $z$ is

6891: a limit of points in $\Psi(U_i) \subset V_i$.

6892: %If $z$ is in $V_i$ then its flow line has both ideal

6893: %points in $\overline V_i \cap \si$, which has diameter

6894: %decreasing to $0$ with $i$.

6895: %A geodesic of $\mi$ with these ideal points

6896: %is in a set of $\mi \cup \si$ of small visual diameter.

6897: %Then since the orbits of $\wwp$ are uniform quasigeodesics

6898: %they are a bounded distance from these geodesics. This shows

6899: As seen in the construction arguments

6900: the diameter of $\overline V_i$ in the visual distance

6901: is converging to $0$. Hence we obtain continuity of $\Psi$ at

6902: $v$.

6903: This finishes the proof in this case.

6904: 

6905: \vskip .1in

6906: \noindent

6907: {\underline {Case 2}} $-$ $v$ is an ideal point of 

6908: a ray of $\wlsf$ or $\wluf$.

6909: 

6910: This case is considerably more complicated, with several possibilities.

6911: 

6912: \vskip .1in

6913: \noindent

6914: {\underline {Case 2.1}} $-$

6915: $v$ is an  ideal point of $\wlsf$ but not of $\wluf$

6916: (or vice versa).

6917: 

6918: Suppose the first option occurs.

6919: There is $l$ ray in $\wlsf$ with ideal 

6920: point $v$. 

6921: We may assume that $l$ is not in a leaf of $\wlsf$ with

6922: same ideal points. Otherwise we can choose $l$ to be one

6923: of the boundary leaves of the corresponding spike region.

6924: Since $v$ is not an ideal point of $\wluf$,

6925: there are $g_i$ line leaves

6926: in $\wluf$ defining a basis neighborhood system at $v$.

6927: Let $g_i$ be contained in $G_i$ leaves of $\wlu$.

6928: Let $L$ in $\wls$ containing $l$.

6929: If $G_i$ escapes in $\mi$ as $i \rightarrow \infty$, then as

6930: seen in case 1, we are done.

6931: Let then $G_i$ converge to the finite set of leaves

6932: 

6933: $${\cal V} \ =  \ H_1 \cup H_2 .... \cup H_m \ \ \ {\rm leaves

6934: \ of } \ \ \wlu$$

6935: 

6936: \noindent

6937: We can assume that $G_i \cap l \not = \emptyset$ for all $i$,

6938: $G_i$ is non singular and the sequence $\{ G_i \}, i \in {\bf N}$

6939: is nested.

6940: 

6941: 

6942: \vskip .1in

6943: \noindent

6944: {\underline {Case 2.1.1}} $-$ Suppose that $L$ intersects

6945: ${\cal V}$, say $L \cap H_1 \not = \emptyset$.

6946: 

6947: Then $l$ escapes down as $\Theta(l)$ 

6948: approaches $\Theta(L \cap H_1)$.

6949: Otherwise $L \cap H_1$ is in the boundary of a lift annulus

6950: $A$ and $l$ has a subray contained in this lift annulus.

6951: But then $A$ is also contained in the unstable leaf $\wu(L \cap H_1)$

6952: and so $G_i$ cannot intersect $l$, contradiction.

6953: As $l$ escapes down in $L$, then

6954: the ideal point of $\Psi(l)$ is

6955: $(L \cap H_1)_-$ which is equal to $(H_1)_-$, the negative

6956: ideal point of $H_1$. 

6957: 

6958: Since the values of $\Psi(p)$ for $p$ in $\pin F$ are obtained

6959: as limits of values in $\Psi(F)$, then we only need to show that

6960: if $z_k$ is in $F$ and $z_k$ converges to $p$ as $k \rightarrow \infty$,

6961: then $\Psi(z_k)$ converges to $\Psi(p)$. Suppose this is not

6962: the case.

6963: 

6964: %\blankfig{2.1}{1.1}{a. The leaves $\{ g_t \}$ converge

6965: %to $v$ in $F \cup \pin F$, b. $L$ intersects $H_1$, which

6966: %is in the limit of the $G_t$.}

6967: 

6968: By taking a subray if necessary, we may assume that $l$ 

6969: does not intersect a lift annulus and hence it is transverse

6970: to the unstable foliation $\wluf$ in $F$.

6971: Parametrize the leaves of $\wluf$

6972: intersected by $l$ as $\{ g_t, \  t \in \rrrr_+ \}$,

6973: contained in $G_t \in \wlu$

6974: (by an abuse of notation think of the $G_i$ as

6975: a discrete subcollection of the $G_t, t \in \rrrr_+$). Let 

6976: 

6977: $${\cal U}  \ = \ \bigcup_{t>0} G_t$$

6978: 

6979: \noindent

6980: No $g_t$ (or leaf of $\wluf$) 

6981: has ideal point $v$ in $\pin F$.

6982: As $\{ g_i \}, i \in {\bf N}$ converges to $v$ in $F \cup \pin F$, then

6983: $g_t$ escapes compact sets in $F$ as

6984: $t \rightarrow \infty$ and the ideal points of $g_t$ converge

6985: to $v$ on either side of $v$. 

6986: %If ideal points do not

6987: %converge to $v$ then since ideal points of leaves of $\wluf$

6988: %are dense in $\pin F$, there will be leaf $g$ in the

6989: %limit of the $g_t$. 

6990: %%If there is a leaf $g$ in the limit of the $g_t$, then

6991: %Then since $\pi_1(M)$ is negatively

6992: %curved there can only be finitely many leaves in

6993: %the limit and consecutive leaves share an ideal

6994: %point, because of the denseness again. It would then

6995: %follow that some limit leaf has to have ideal point $v$,

6996: %contradiction.

6997: %It now follows that 

6998: Up to subsequence assume that all of the elements

6999: of the sequence $\{ z_k \}$ 

7000: are either entirely contained

7001: in ${\cal U}$ or disjoint from ${\cal U}$.

7002: 

7003: \vskip .1in

7004: \noindent

7005: {\underline {Situation 1}} $-$ Suppose that $z_k$ is not in ${\cal U}$ 

7006: for any $k$.

7007: 

7008: Since $z_k$ is very close to $v$ in the compactification

7009: $F \cup \pin F$

7010: and $g_t$ converges to $v$ in $F \cup \pin F$ when $t \rightarrow \infty$,

7011: then there are $t, s$

7012: with $z_k$ between $g_t$ and $g_s$ (in $F$). Notice $z_k$ is not in

7013: any of them. Now there is a unique time $t_k$ so that exactly at that

7014: time $\Psi(z_k)$ switches from being in one side of $G_t$  in $\mi$

7015: to the other

7016: (equivalently compare the $z$ and $g_t$ in $F$).

7017: In particular, either there is a line leaf $L_{t_k}$

7018: of $G_{t_k}$ which separates $\Psi(z_k)$ from all the other 

7019: $G_t$, see fig. \ref{sepa}, a, or there is a leaf $L_{t_k}$

7020: non separated from $G_{t_k}$ with $\Psi(z_k)$ either in $L_{t_k}$

7021: or $L_{t_k}$ separates $\Psi(z_k)$ from all $G_t$,

7022: see fig. \ref{sepa}, b. This can be seen

7023: in the leaf space of $\wlu$, which is a non Hausdorff tree 

7024: \cite{Fe8,Ga-Ka,Ro-St}.

7025: 

7026: 

7027: \begin{figure}

7028: \centeredepsfbox{gm13.eps}

7029: \caption{

7030: a. Line leaf separating points, b. Non separated

7031: leaf separating points.}

7032: \label{sepa}

7033: \end{figure}

7034: 

7035: 

7036: 

7037: \vskip .1in

7038: \noindent

7039: {\underline {Claim}} $-$ In the Gromov-Hausdorff topology of

7040: closed sets of 

7041: $\mi \cup \si$, the sets $\overline L_{t_k}$ 

7042: converge to $(H_1)_-$ as

7043: $k \rightarrow \infty$.

7044: 

7045: %\noindent

7046: %Proof of the claim $-$ 

7047: If $L_{t_k}$ is a line leaf of 

7048: $G_{t_k}$, then $(L_{t_k})_- = (G_{t_k})_-$. If

7049: $L_{t_k}$ is not separated from $G_{t_k}$ then also

7050: $(L_{t_k})_- = (G_{t_k})_-$. 

7051: This is because there are $E_i$ leaves of $\wlu$ with

7052: $E_i$ converging to $L_{t_k} \cup G_{t_k}$.

7053: So there are $x_i, \ y_i$ in $E_i$ with 

7054: $x_i \rightarrow x, \ y_i \rightarrow y$ and $x \in L_{t_k}, \

7055: y \in G_{t_k}$. Then 

7056: 

7057: 

7058: $$\eta_-(x_i) \ \rightarrow \eta_-(x) \ = \ \eta_-(L_{t_k}), \ \ \ 

7059: \eta_-(y_i) \ \rightarrow \eta_-(y) \ = \ \eta_-(G_{t_k})

7060: \ \ \ \ {\rm and} \ \ \eta_-(x_i) \ = \ \eta_-(y_i).$$

7061: 

7062: 

7063: \noindent

7064: The last equality occurs

7065:  because $x_i, y_i$ are in the same unstable

7066: leaf $E_i$.

7067: Therefore $(L_{t_k})_-$

7068: converges to $(H_1)_-$ when $k \rightarrow \infty$. 

7069: Suppose that $\overline L_{t_k}$ does not 

7070: converge to $(H_1)_-$ in $\mi \cup \si$.

7071: Since

7072: 

7073: $$(L_{t_k})_- \ \ {\rm converges \ to} \ \ (H_1)_-,$$

7074: 

7075: \noindent

7076: then lemma \ref{narrow} shows that $L_{t_k}$ does not

7077: escape compact sets in $\mi$. So it limits to some $u$ in

7078: $\mi$ and up to subsequence  we may assume there 

7079: $u_k$ in $L_{t_k}$ with $u_k$ converging

7080: to $u$.

7081: The first possibility is that the $L_{t_k}$ are subsets of 

7082: %Suppose there are $u_k$ points in

7083: %$L_{t_k}$ with the sequence 

7084: %$u_{t_k}$ not converging to $(H_1)_-$. Then up to subsequence

7085: %assume $u_k$ converges to $u$ a point in $\si$

7086: % distinct from $(H_1)_-$.

7087: %Hence $\wwp_{\rrrr}(v_k)$ does not escape $\hhh$ (because

7088: %of the quasigeodesic property). Hence up to another subsequence

7089: %we obtain

7090: %

7091: %$$\wwp_{\rrrr}(v_k) \ \ {\rm converges \ to} \ \ \wwp_{\rrrr}(v)$$

7092: %

7093: %\noindent

7094: %in the compact open topology of closed sets in $\hhh$.

7095: %In the first case $v_k$ are in 

7096: the leaves $G_{t_k}$. This implies that $\wwp_{\rrrr}(u)$

7097: is in the limit of the sequence of leaves $G_{t_k}$ (in $\mi$), so

7098: it is contained in ${\cal V}$.

7099: The second possibility is $L_{t_k}$ non separated from $G_{t_k}$ so

7100: $L_{t_k}$ is between $G_{t_{k-1}}$ and

7101: $G_{t_{k+1}}$ hence $u$ is again in the limit of the $G_t$ so

7102: $u$ is in ${\cal V}$.

7103: %the same property for $v$ holds.

7104: %holds. 

7105: The leaves $H_j$ in ${\cal V}$ are non singular in the

7106: side the $G_t$ are limiting on, 

7107: %But since

7108: %$v$ is in $H$ and it is non singular in that region, there

7109: so there is a neighborhood of $u$ on that side of $H_j$ which has

7110: no singularities hence the $u_k$ will be in ${\cal U}$ for

7111: $k$ big enough. This contradicts the hypothesis

7112: in this case.

7113: 

7114: This shows that 

7115: $\overline L_{t_k}$ converges to $(H_1)_-$ 

7116: in $\mi \cup \si$.

7117: Also $L_{t_k}$ 

7118: either contains $\Psi(z_k)$

7119: or separates it from a base point in $\mi$. It follows that

7120: $\Psi(z_k)$ converges to $(H_1)_-$, which is what we wanted to prove.

7121: This finishes the analysis in situation 1.

7122: 

7123: \vskip .1in

7124: \noindent

7125: {\underline {Situation 2}} $-$ For all $k$ assume that $\Psi(z_k)$ is

7126: in ${\cal U}$.

7127: 

7128: Let $t_k$ with $\Psi(z_k)$ in $G_{t_k}$,

7129: hence $z_k$ is in $G_{t_k} \cap F = g_{t_k}$. 

7130: Then $(\Psi(z_k))_- = (G_{t_k})_-$ converges to $(H_1)_-$.

7131: We want to show that $\Psi(z_k)$ converges to $(H_1)_-$.

7132: Otherwise there is $q$  in $\si$

7133: different from $(H_1)_-$ and a subsequence, still denoted by

7134: $\Psi(z_k)$ so that

7135: $\Psi(z_k)$ converges

7136: to $q$.

7137: As in the claim of situation 1 above, 

7138: $\wwp_{\rrrr}(z_k)$ does not escape compact sets

7139: in $\mi$ and there is $z$ in $\mi$, so that up to another

7140: subsequence, we may 

7141: assume that $\wwp_{\rrrr}(\Psi(z_k))$ converges to $\wwp_{\rrrr}(z)$.

7142: Since $\Psi(z_k)$ is in $G_{t_k}$ then $z$ is

7143: in ${\cal V}$, say $z$ is in $H_j$.

7144: Let $p = \Theta(z)$.

7145: At this point notice that $F$ does not intersect any 

7146: leaf $H_i$ in ${\cal V}$.

7147: If it did, say in $w$ then $F$ intersects the nearby leaves

7148: $G_t$ (for any $t$ big enough) near $w$. This would

7149: imply $F \cap G_t = g_t$ 

7150: does not escape compact sets in $F$,

7151: % hence $g_t$

7152: %does not converge to $p$, 

7153: contradiction.

7154: Therefore 

7155: %as $\Theta(\Psi(z_k))$ are in $\Theta(F)$ and

7156: %$\Theta(p)$ is not $\Theta(F)$

7157: % and in addition $\Theta(H_j)$ does

7158: %not intersect $\Theta(F)$ then: $\Theta(H_j)$ has points

7159: $p = \Theta(z)$ is in $\partial \Theta(F)$.

7160: Let $x_k$ in $g_{t_k} \cap l$.

7161: Then

7162: $\Theta(x_k)$ converges to a point in $\Theta(H_1 \cap L)$.

7163: There are segments $b_k$ in 

7164: $F \cap G_{t_k} = g_{t_k}$ from $x_k$

7165: to $z_k$. Then $\Theta(b_k)$ converges to a ray in

7166: $\Theta(H_1)$

7167: and a ray in $\Theta(H_j) \subset \oou(p)$ and possibly

7168: other unstable leaves. Then there is a ray in $\oou(p)$

7169: contained in $\partial \Theta(F)$. This implies that

7170: $F$ escapes down as $\Theta(F)$ approaches this ray of $\Theta(H_j)$.

7171: Hence $\Psi(z_k)$ is getting closer to $z_-$ which is $(H_j)_-$,

7172: which is also equal to $(H_1)_-$.

7173: This is what we

7174: wanted to prove anyway.

7175: 

7176: This finishes the proof of case 2.1.1,

7177: that is, when $L$ intersects ${\cal V}$.

7178: 

7179: \begin{lemma}{}{}

7180: Let $A$ in $\wlu$, $B$ in $\wls$ satisfying:

7181: there are

7182: $R_i$ leaves of $\wlu$ intersecting $B$ 

7183: with $R_i$ converging to $A$

7184: and $R_i \cap B$ escaping compact sets in $B$.

7185: Then $A_-$ is equal to $B_+$.

7186: \label{conperf}

7187: \end{lemma}

7188: 

7189: \begin{proof}{}

7190: Since $R_i$ converges to $A$ then $(R_i)_-$ converges

7191: to $A_-$.

7192: Also $R_i$ intersects $B$ so

7193: $(R_i)_- = (R_i \cap B)_-$. As $R_i \cap B$ escapes compact

7194: sets in $B$ then in the intrinsic geometry of $B$,

7195: the $R_i \cap B$ converges

7196: to the positive ideal point of $B$. This implies that

7197: $(R_i \cap B)_-$ converges to $B_+$. This implies the result.

7198: \end{proof}

7199: 

7200: 

7201: \noindent

7202: {\underline {Case 2.1.2}} $-$ $L$ does not intersect 

7203: ${\cal V}$.

7204: 

7205: Then $\Theta(l)$ escapes in $\Theta(L)$ and so $\Psi(l)$ converges to

7206: $L_+$. By the previous lemma,

7207: this is also equal to $(H_1)_-$. From this point on,

7208: the proof is the same

7209: as in case 2.1.1.

7210: This finishes the proof of case 2.1.

7211: 

7212: \vskip .2in

7213: \noindent

7214: {\underline {Case 2.2}} $-$ $v$ is an ideal point of both $\wlsf$ 

7215: and $\wluf$.

7216: 

7217: \vskip .1in

7218: \noindent

7219: {\underline {Case 2.2.1}} $-$ For any ray $l$ of 

7220: $\wlsf$ and $e$ of $\wluf$

7221: with $l_{\infty} = e_{\infty} = v$, then $l$ does not intersect

7222: $e$. 

7223: 

7224: 

7225: Let $l', e'$ be rays as above. 

7226: We may assume that $l', e'$ do not have any singularities.

7227: Parametrize the leaves of $\wlsf$

7228: intersecting $e'$ as $\{ l_t, t \geq 0 \}$ 

7229: where $l_t \cap e'$ converges to $v$

7230: in $F \cup \pin F$

7231: as $t$ converges to infinity.

7232: 

7233: %We will prove continuity of $\Psi$

7234: %at $v$ from one side of $e'$. The other side is

7235: %treated in the same fashion.

7236: Since $l'$ limits on $v$ and is disjoint from $e'$, then $l'$ 

7237: is on a side

7238: defined by $e'$. We will prove continuity of

7239: $\Psi$ at $v$ from the other side

7240: of $e'$. 

7241: The point $p_t = l_t \cap e'$ disconnects $l_t$. For simplicity

7242: we only consider those $l_t$ with $l_t \subset L_t \in \wls$ and

7243: $L_t$ non singular. Let $l^1_t$ be the component of $(l_t - p_t)$ in 

7244: the $e'$ side containing $l'$ 

7245: union with $p_t$. Let $l^2_t$ be the other component

7246: of $(l_t - p_t)$ union with $p_t$, see fig. \ref{onsid}, a.

7247: 

7248: The $l^1_t$ are rays (here we use $L_t$ non singular - but this

7249: is just a technicality) and $(l^1_t)_{\infty}$ are not equal $v$ by

7250: hypothesis. They cannot escape compact sets of $F$ since

7251: $l'$ with ideal point $v$ is on that side of $e'$.

7252: %Also they are bounded by $l'$. 

7253: Hence as $t$ converges

7254: to infinity $l^1_t$ converges to a leaf $l$ of $\wlsf$

7255: with a ray (also denoted by $l$) with ideal point $v$ and

7256: maybe some other leaves as well. The leaf

7257: $l$ either shares a subray with $l'$ or separates $l'$ from $e'$

7258: Let $e' \subset E$ leaf of $\wlu$ and $l \subset L$, leaf of $\wls$.

7259: 

7260: 

7261: \vskip .1in

7262: \noindent

7263: {\underline {Case 2.2.1.1}} $-$ $l^2_t$ escapes in $F$

7264: as $t \rightarrow \infty$.

7265: 

7266: Let $b_t$ be the ideal point of $l^2_t$. Then $b_t \not = v$.

7267: Let $L^2_t$ be the union of $\wwp_{\rrrr}(p_t)$ and the component

7268: of $L_t - \wwp_{\rrrr}(p_t)$ containing $l^2_t$. If $L^2_t$ escapes

7269: in $\mi$, then the arguments in case 1 show 

7270: continuity of 

7271: $\Psi$ at $v$ in

7272: the side of $e'$ not containing $l'$.

7273: 

7274: Now assume that $L^2_t$ converges to $R_1 \cup ... \cup R_m$

7275: leaves of $\wls$

7276: with union ${\cal R}$.

7277: Notice $F$ may intersect some of these leaves or not.

7278: If $\Theta(\Psi(p_t))$ does not escape in $\Theta(E')$,

7279: then one of the $R_i$, call it $R_1$, is a leaf

7280: intersecting $E'$.

7281: As seen in the arguments for case 2.1.1, $F$ escapes

7282: up in this direction so $\Psi(p_t)$ converges to

7283: $(R_1)_+$.

7284: If $\Theta(\Psi(p_t))$ escapes in $\Theta(E')$, then

7285: lemma \ref{conperf} shows that $\Psi(p_t)$ also

7286: converges to $(R_1)_+$.

7287: This is equal to $(R_j)_+$ for any $j$.

7288: 

7289: Suppose there are $t_k \rightarrow \infty$ and 

7290: $z_k$ in $l^2_{t_k}$ with $\Psi(z_k)$ not converging

7291: to $(R_1)_+$. 

7292: Here there is no need to assume that $L_{t_k}$ is non singular.

7293: Then there $q$ in $\si$, \ $q \not = (R_1)_+$

7294: and a subsequence still denoted by $\Psi(z_k)$,

7295: so that $\Psi(z_k)$ converges to $q$.

7296: As before there is $z$ in $\mi$ so that up to another 

7297: subsequence, still denoted by $\wwp_{\rrrr}(z_k)$,

7298: then $\wwp_{\rrrr}(z_k)$  converges

7299: to $\wwp_{\rrrr}(z)$ and hence $z$ is in ${\cal R}$, say in

7300: $R_i$.

7301: %As in case ???

7302: Then $\wwp_{\rrrr}(z_k)$ are near 

7303: $\wwp_{\rrrr}(z)$ and since a ray of $\Theta(R_i)$ is in

7304: $\partial \Theta(F)$, then this is stable boundary. 

7305: So $F$ escapes up as $\Theta(F)$ approaches $\Theta(z)$ and hence 

7306: $\Psi(z_k)$ converges to $(R_i)_+$. 

7307: This is equal to $(R_1)_+$.

7308: The arguments of Case 2.1.1, situation 1 then show

7309: continuity of $\Psi$ at $v$ on this side

7310: of $e'$. This finishes the analysis of case 2.2.1.1.

7311: 

7312: 

7313: 

7314: \begin{figure}

7315: \centeredepsfbox{gm14.eps}

7316: \caption{

7317: a. Convergence on one side,

7318: b. Case 2.2.1.2 - intersection of leaves.}

7319: \label{onsid}

7320: \end{figure}

7321: 

7322: 

7323: 

7324: 

7325: \vskip .1in

7326: \noindent

7327: {\underline {Case 2.2.1.2}} $-$ The \ $l^2_t$ \ limit to $r$ in $F$

7328: as $t \rightarrow \infty$.

7329: 

7330: Choose the leaf $r$ with a ray which has ideal point $v$.

7331: Then the leaves $r, l$ are not separated from each other in the leaf

7332: space of $\wlsf$. 

7333: %Proposition \ref{band} shows that the region bounded by these 

7334: Since $r, l$ have ideal point $v$ and there is no leaf

7335: of $\wlsf$ non separated from $r, l$ and between them,

7336: proposition \ref{nh} shows that the region bounded by these

7337: rays of $r, l$ with ideal point $v$ projects in $M$ to a set

7338: asymptotic to a Reeb annulus.

7339: It follows that in $F$ this region is a bounded 

7340: distance from

7341: a geodesic ray with ideal point $v$. 

7342: Now we restart

7343: the process with the ray $r$ of $\wlsf$ instead of $e'$ of $\wluf$.

7344: Let $\{ b_t, t \geq 0 \}$ be a parametrization of the 

7345: leaves of $\wluf$ 

7346: through the corresponding points $x_t$ of $r$.

7347: %see fig. \ref{nonsep1}, a.

7348: If the  components of $(b_t - x_t)$ on the side opposite

7349: of $e'$ escapes compact sets in $F$, then the analysis

7350: of case 2.2.1.1 shows continuity of $\Psi$ at $v$ in that

7351: side of $r$. Since $r$ and $e'$ are a bounded distance

7352: from each other in $F$, this shows continuity of $\Psi$ at

7353: $v$ on that side of $e'$.

7354: 

7355: Otherwise this process keeps being repeated.

7356: Let $A_0 = L$, $A_1$ be the leaf of $\wls$ containing $r$.

7357: If the process above does not stop, we keep producing $A_i$ in $\wls$,

7358: so that they all disjoint and $A_i$ is non separated from

7359: $A_{i+1}$. By theorem \ref{theb} 

7360: up to covering translations there are only finitely many

7361: leaves of $\wls$ which are not separated from some other leaf of

7362: $\wls$.

7363: There is then $m > n$ and $h$ covering translation with

7364: $h(A_n) = A_m$. Let $f$ be the generator of the joint

7365: stabilizer of $A_0, A_1$. 

7366: This is non trivial by theorem \ref{theb}.

7367: Then $f$ preserves all the prongs

7368: of $A_1$ and therefore leaves invariant all the $A_i$.

7369: Hence $h^{-1} f h(A_n) = A_n$ and so $h^{-1} f h = f^a$ for

7370: some integer $a$. This implies there is a ${\bf Z} \oplus {\bf Z}$

7371: in $\pi_1(M)$, see detailed arguments in \cite{Fe8}.

7372: This is a contradiction.

7373: 

7374: There is then a last leaf $l_y$ (of $\wlsf$ or $\wluf$)

7375: obtained from this process. The arguments of case 2.2.1.1 show

7376: continuity of $\Psi$ at $v$ on the other side of $l_y$.

7377: The region between $e'$ and $l_y$ is composed of a finite

7378: union of regions between non separated rays of $\wlsf$ or

7379: $\wluf$. They are all a bounded distance from a geodesic

7380: ray with ideal point $v$, so the whole region also

7381: satisfies this property.  It follows that this region

7382: can only limit in $\Psi(v)$ as well and this proves

7383: continuity of $\Psi$ at $v$ in that side of $e'$.

7384: 

7385: 

7386: An entirely similar analysis shows continuity of $\Psi$ at

7387: $v$ from the side of $l'$ not containing $e'$.

7388: 

7389: \vskip .1in

7390: 

7391: What remains to be analysed is the region of $F$ 

7392: {\underline {between}} the rays $l'$ and $e'$.

7393: Consider first the case that there is pair of non separated

7394: leaves in the chain from $l'$ to $e'$.

7395: Then as seen before the region between $l'$ 

7396: and  $e'$

7397: %there is non Hausdorfness involved, this region

7398: is a bounded distance (the bound is not uniform) from a geodesic

7399: ray with ideal point $v$. 

7400: This is not the case a priori if

7401: there is no non Hausdorfness involved. 

7402: In this case the region between $l'$ and $e'$ may not have

7403: bounded thickness in $F$ and hence it is unclear whether

7404: its image under $\Psi$ can only limit in $\Psi(v)$.

7405: We analyse this case now.

7406: 

7407: In this last case 

7408: parametrize the leaves of $\wluf$ intersecting the ray 

7409: $l$ of $\wlsf$ as \ $\{ e_t \ | \ t \geq 0 \}$.

7410: Since $l_t$ converges to $l$, then for big enough

7411: $t$, the leaves $l_t, e_t$ intersect $-$ let $u_t$ be

7412: their intersection point, see fig. \ref{onsid}, b.

7413: Now define $l^*_t$ to be

7414: the component of $l_t - u_t$  intersecting $e$ and $e^*_t$ 

7415: the component of

7416: $e_t - u_t$ intersecting $l$. 

7417: Since $e'$ is on that side of $l$, the $e_t$ cannot escape

7418: and converge to a leaf $e$ of $\wluf$ with an ideal point

7419: $v$. Let $e \subset E$  leaf of $\wlu$.

7420: 

7421: Recall that $L_t$ is the leaf of $\wls$ containing $l^*_t$

7422: and similarly let $E_t$ be the leaf of $\wlu$ containing $e_t$.

7423: Let $L^*_t$ be the component

7424: of $L_t - \wwp_{\rrrr}(u_t)$ containing $l^*_t$

7425: and similarly define $E^*_t$.

7426: In this remaining case the $l^*_t$ escape in $F$ and

7427: so do the $e^*_t$. Hence

7428: $\mu_t \ = \ l^*_t \cup \{ u_t \} \cup e^*_t$

7429: defines a shrinking neighborhood system of

7430: $v$ in $F \cup \pin F$. Consider the set

7431: 

7432: $$B_t \ = \ L^*_t \cup \wwr(u_t) \cup E^*_t$$

7433: 

7434: We want to show that $\overline B_t$ converges

7435: to $L_+$ in the topology of closed sets of $\mi \cup \si$.

7436: 

7437: First consider $L^*_t \cap E$ which intersects $F$

7438: in $(l^*_t \cap e)$. If $L^*_t \cap E$ does not

7439: escape compact sets in $E$ then it limits to an orbit

7440: $\gamma$ contained in a leaf $H$ of $\wls$.

7441: Then $L, H$ are not separated from each other.

7442: But for $t$ big enough then $E_t$ is near enough

7443: $E$ and will intersect $H$ as well. This contradicts

7444: $E_t \cap L$ is not empty and $L, H$ non separated.

7445: Hence $L^*_t \cap E$ escapes in $E$ and similarly

7446: $E^*_t \cap L$ escapes in $L$. 

7447: Hence $L, E$ form a perfect fit.

7448: This implies that $L_+ = E_-$.

7449: Also $\Psi(e)$ limits to $E_-$ and $\Psi(l)$ limits

7450: to $L_+ = E_-$.

7451: 

7452: The set $\overline L^*_t$ contains 

7453: $(L^*_t \cap E)_+$ and this converges to $E_-$

7454: when $t \rightarrow \infty$.

7455: This is because $(L^*_t \cap E)$ escapes in $E$.

7456: If $\overline L^*_t$ does

7457: not converge to $E_-$ in $\mi \cup \si$, then

7458: we find $t_k \rightarrow \infty$ and $x_k \in

7459: L^*_{t_k}$ with $x_k$ converging to $x$ not equal to $E_-$.

7460: Since $(x_k)_+ = (L_{t_k})_+$ converges to $E_-$,

7461: then up to subsequence assume

7462: $\wwr(x_k)$ converges to $\wwr(z)$ for some $z$ in $\mi$.

7463: Then $z$ is in a leaf $H$ of $\wls$ which is non

7464: separated from $L$.

7465: %, see fig. \ref{fini}.

7466: 

7467: %\blankfig{fini}{1.1}{Non separated leaves in $\mi$.}

7468: 

7469: %Since all the $x_k$ are in the component of $\mi - E_t$

7470: %containing $E$ then $H$ is separated from $L$ by $E$, see

7471: %fig. \ref{fini}.

7472: The leaf $H$ does not intersect $F$, because $l^*_t$

7473: escapes in $F$ by hypothesis in this final situation.

7474: It follows that $\Theta(H)$ has a ray contained

7475: in $\partial \Theta(F)$ and so this is stable boundary

7476: of $\Theta(F)$. Hence $F$ escapes up as $\Theta(F)$

7477: approaches $\Theta(H)$ and consequently 

7478: $\Psi(x_k)$ limits to $H_+ = L_+ - = E_-$

7479: $-$ which is what we wanted anyway.

7480: This shows that $\overline L^*_t$ converges

7481: to $E_-$ in $\mi \cup \si$.

7482: 

7483: Analysing the sets $E^*_t$ in the same manner we obtain

7484: that $\overline E^*_t$ converges to $L_+$ as $t \rightarrow \infty$

7485: as well.

7486: This implies that $\overline B_t$ converges to 

7487: $L_+ = \Psi(v)$. Since $B_t \cap F = \mu_t$ and the

7488: $\mu_t$ define a neighborhood basis of $v$ in $F \cup \pin F$,

7489: this shows continuity of $\Psi$ at $v$. 

7490: This finishes the proof of case 2.2.1.2 and hence of

7491: case 2.2.1.

7492: 

7493: 

7494: 

7495: 

7496: \vskip .2in

7497: \noindent

7498: {\underline {Case 2.2.2}} $-$ There are rays $l$ of $\wlsf$ and

7499: $e$ of $\wluf$ starting at $u_0$ and having the ideal point $v$.

7500: 

7501: We will first prove continuity on the side of $e$ not containing

7502: a subray of $l$. There will be an iteration of steps.

7503: Before we start the analysis we want to get rid of some

7504: problems as described now. Suppose that there are $\alpha_0, 

7505: \beta_0$ leaves of $\wlsf$ (or leaves of $\wluf$)

7506: which have non separated rays converging to $v$ in $\pin F$

7507: and on that side of $e$.

7508: Suppose there are infinitely many of these on that side of

7509: $e$. Let them be $\alpha_i, \beta_i$ and $G_i$ in

7510: $\wls$ containing $\alpha_i$. Each region $B$ between $\alpha_0$

7511: and any $\alpha_i$ is a bounded distance from a geodesic

7512: ray in $F$ with ideal point $v$. The image $\Psi(B)$ then

7513: can only limit in $\Psi(v)$.

7514: If the $G_i$ do not escape in $\mi$ then they converge

7515: to a leaf $G$ of $\wls$.

7516: Let $A$ be an unstable leaf intersecting $G$ tranversely.

7517: For $i$ big enough then $A$ intersects $G_i$ transversely,

7518: which is impossible, as it would intersect

7519: $\alpha_i$ and $\beta_i$ and these are not separated.

7520: Hence the 

7521: the $G_i$ escapes in $\mi$. Then  as seen in case 1,

7522: there is continuity of $\Psi$ at $v$ in that side of $\alpha_1$.

7523: 

7524: Another situation is when there are leaves $\alpha_i$ in

7525: that side of $e$ with two rays with ideal point $v$.

7526: Then they are in the interior of a spike region $B$ with

7527: one boundary $g$ with ideal point $v$. If there are infinitely

7528: many of these, where none of the $\alpha_i$ are

7529: nested with each other, then

7530:  let $G_i$ in $\wls$ containing $\alpha_i$.

7531: As in the previous paragraph,

7532: the $G_i$ have to escape in $\mi$ and we have continuity

7533: in that side of $\alpha_1$.

7534: 

7535: Therefore we can assume there are only finitely many

7536: occurrences of spike regions or non separated leaves

7537: with ideal point on this side of $e$. 

7538: If there is any of these let $e_0$ be the last ray

7539: in that side coming from such occurrences. 

7540: Otherwise let 

7541: $e_0$ be the ray given $e$ by the hypothesis in this case.

7542: For simplicity assume that $e_0$ is a ray in $\wluf$,

7543: the other case being similar. Let $e_0 \subset E_0 \in \wlu$.

7544: 

7545: 

7546: Parametrize the ray of $e_0$ as 

7547: \ $\{ p_t \ | \ t \geq 0 \}$ \

7548: with $p_t$ converging to $v$ as $t \rightarrow \infty$. Let $l_t$ be

7549: the leaf of $\wlsf$ through $p_t$ and $L_t$ in $\wls$

7550: with $l_t \subset L_t$. If $L_t$ escapes $\mi$ as

7551: $t \rightarrow \infty$ then as seen before we have continuity 

7552: of $\Psi$ at

7553: $v$ in that side of $e_0$.

7554: So now suppose that $L_t$ converges to $A_1 \cup .... A_m$,

7555: leaves of $\wls$. 

7556: %Then $F$ escapes up as $\Theta(F)$ approaches $\Theta(A_i)$

7557: %for any $i$.

7558: This case is considerably more involved, with several 

7559: possibilities.

7560: 

7561: \vskip .1in

7562: \noindent

7563: {\underline {Claim}} $-$ $\Psi(e_0)$ converges to $(A_i)_+$ 

7564: (notice the  $(A_i)_+, 1 \leq i \leq m$ are all equal).

7565: 

7566: If $E_0$ intersects some $A_i$, say $A_1$, then

7567: as seen in case 2.1.1,

7568: $F$ escapes positively along $\Psi(e_0)$ as $\Theta(F)$

7569: approaches $A_1$. This implies that

7570: $\Psi(e_0)$ converges to $(E_0 \cap A_1) _+ = (A_1)_+$.

7571: If $E_0$ does not intersect any 

7572: $A_i$ then $\Psi(e_0)$ converges 

7573: to $(E_0)_- = (A_1)_+$. This proves the claim.

7574: 

7575: Let $l^1_t$ be the component of $(l_t - p_t)$ in the side

7576: of $e_0$ we are considering.

7577: We are really interested

7578: in the behavior for $t \rightarrow \infty$, so we may

7579: assume $p_t$ is not singular and there is only one

7580: such component.

7581: 

7582: Suppose first that no $l^1_t$ has a ray with

7583: ideal point $v$ and that

7584: $l^1_t$ escapes in $F$ as $t \rightarrow \infty$.

7585: In this case it is easy to show continuity 

7586: of $\Psi$ at $v$ and in this side of $e_0$:

7587: Suppose there are $x_i$ in $l^1_{t_i}$ with $t_i \rightarrow \infty$

7588: and $\Psi(x_i) \not \rightarrow (A_i)_-$.

7589: Since $(x_i)_+$ converges to $(A_i)_+$ then up to subsequence

7590: assume that $(x_i)_- \rightarrow b \not = (A_i)_+$.

7591: Up to subsequence $\wwr(x_i) \rightarrow \wwr(x)$. Then $x$

7592: is in some $A_i$ say $x \in A_2$. But $F$ escapes positively

7593: as 

7594: $\Theta(F)$ approaches $\Theta(A_2)$,

7595: so $\Psi(x_i) \rightarrow (A_i)_+$,  as we wanted.

7596: Then as in case 2.1.1 this implies continuity.

7597: 

7598: 

7599: There are 2 other options: 1) There is no $t$ with $l^1_t$

7600: with an ideal point $v$ and $l^1_t$ does not escape in $F$; and

7601: 2) There is $t$ with $l^1_t$ having ideal point $v$.

7602: These two options interact and intercalate in appearance

7603: as explained below:

7604: 

7605: 

7606: \vskip .1in

7607: \noindent

7608: {\underline {Situation 1}} $-$  There is no $t$ with $l^1_t$

7609: with ideal point $v$ and $l^1_t$ does not escape in $F$.

7610: 

7611: %Otherwise the limit leaf would be in the interior

7612: %of a spike region and so would $l_t$ for $t$ big,

7613: %contradiction to identified ideal points.

7614: There could be several leaves of $\wlsf$ in the limit of $l^1_t$ 

7615: as $t \rightarrow \infty$ 

7616: but there is a single leaf, call it $g$ with ideal point $v$.

7617: If there is more than one such leaf with ideal point $v$, then

7618: there would have to be one with

7619: two rays with ideal point $v$. This leaf would be in a spike region

7620: and it is separated from any other leaf in $\wlsf$,

7621: contradiction.

7622: Let $g$ be contained in a leaf 

7623:  $G$ of $\wls$.

7624: %The arguments of case 2.2.1.2 show that if $B$ is the region in $F$ 

7625: %between $e_0$ and $g$,  then $\Psi(B)$ can 

7626: %only limit at $\Psi(v) = (A_i)_+$.

7627: 

7628: 

7629: \begin{figure}

7630: \centeredepsfbox{gm15.eps}

7631: \caption{

7632: Some limits in $F$, b. The picture in $\mi$.}

7633: \label{remo}

7634: \end{figure}

7635: 

7636: 

7637: Parametrize the ray $g$ as \ $\{ q_t \ | \ t \geq 0 \}$,

7638: with $q_t \rightarrow v$ as $t \rightarrow \infty$.

7639: Let $s_t$ be the unstable leaf of $\wluf$ through

7640: $q_t$. 

7641: Let $s^1_t$ be the component of $(s_t - q_t)$ on the

7642: side of $g$ opposite to $e_0$ and $s^2_t$ the other

7643: component. Then $s^2_t$ cannot have ideal point

7644: $v$: for $t$ big enough it intersects $l^1_t$,

7645: see fig. \ref{remo}, a.

7646: Then $s^2_t$ converges to $e_0$.

7647: By hypothesis there are no more occurrences of non separated

7648: leaves of $\wlsf$

7649: with ideal point $v$ on that side of $e_0$,

7650: which implies that $s^1_t$ 

7651: cannot limit to a leaf of $\wluf$ at $t \rightarrow \infty$

7652: (it would distinct but non separated from $e_0$).

7653: Hence the $s^1_t$ have to escape compact sets

7654: in $F$. 

7655: If $s^1_t$ does not have

7656: an ideal point at $v$ for any $t$, then 

7657: the previous analysis shows continuity of $\Psi$ at $v$ in

7658: that side of $g$.

7659: As in case 2.2.1.2

7660: if $B$ is the region between $g$ and $e_0$ then

7661: $\Psi(B)$ can only limit in $\Psi(v)$.

7662: 

7663: Hence assume 

7664: there is some $t_0$ so that $s^1_{t_0}$ has ideal point $v$,

7665: see fig. \ref{remo}, a.

7666: Then for $t$ bigger than $t_0$ all ideal points of $s^1_t$

7667: are $v$.

7668: Let $s^1_{t_0}$ be contained in a leaf $S$

7669: of $\wlu$ and $s_t$ contained in $S_t$ leaf of $\wlu$.

7670: Since

7671: 

7672: $$l^1_t \ \rightarrow \ g, \ \ \ s^2_t \ \rightarrow \ e_0

7673: \ \ \ \ {\rm when } \ \ t \rightarrow \infty,$$

7674: 

7675: 

7676: $${\rm then} \ \ L_t \ \rightarrow \ G, \ \ \ 

7677: S_t \ \rightarrow \ E_0, \ \ \ {\rm when} \ \ 

7678: t \rightarrow \infty.$$

7679: 

7680: \noindent

7681: It follows that $E_0, G$ form a perfect fit,

7682: see fig. \ref{remo}, b.

7683: Hence $(E_0)_- = G_+$.

7684: If $\Theta(s^1_{t_0})$ is a ray in $\Theta(S)$ then

7685: $\Psi(s^1_{t_0})$ converges to $S_-$.

7686: But $\Psi(s^1_{t_0})$ also converges to 

7687: 

7688: $$\Psi(v)  \ = \ (E_0)_- \ = \ G_+ \ = \ (G \cap S)_+.$$

7689: 

7690: \noindent

7691: Let $\gamma_0 = G \cap S$, an orbit of $\wwp$ in $G$.

7692: The above equations imply that 

7693: 

7694: $$(\gamma_0)_+ \ = \ (G \cap S)_+  \ =  \

7695: \Psi(v) \ = \ S_- \ = \ 

7696: (\gamma_0)_-,$$

7697: 

7698: 

7699: \noindent

7700: which is a contradiction.

7701: Hence $\Theta(s^1_t)$ is not a ray and has an endpoint

7702: $x_1$ in $\Theta(S)$. Let $\gamma_1 = x_1 \times \rrrr$.

7703: Let $H = \wls(\gamma_1)$. 

7704: But $F$ does not intersect $H$.

7705: If $F$ escapes down as $\Theta(F)$ approaches $x_1$, then

7706: $\Psi(v) = (\gamma_1)_-$. 

7707: But then

7708: 

7709: $$(\gamma_0)_- \ = \ (\gamma_1)_- \ = \ \Psi(v) 

7710: \ = \ (\gamma_0)_+$$

7711: 

7712: \noindent

7713: contradiction.

7714: This implies that $F$ escapes up as $\Theta(F)$ approaches

7715: $x_1$. 

7716: Hence $\Theta(H)$ has a ray in $\partial \Theta(F)$.

7717: %Then $S$ intersects $H$ and for $t$ near $t_0$, so does $S_t$.

7718: %This implies that $\Theta(H)$ has a ray in $\partial \Theta(F)$.

7719: Therefore $\Psi(s^1_t)$ limits to $(\gamma_1)_+$.

7720: This implies that $(\gamma_0)_+ = (\gamma_1)_+$,

7721: where $\gamma_0, \gamma_1$ are distinct orbits

7722: of $\wwp$ in the {\underline {same}} unstable leaf $S$.

7723: This is dealt with by the following theorem proved

7724: in \cite{Fe5}:

7725: 

7726: \begin{theorem}{}{(\cite{Fe5})}

7727: Let $\Phi$ be a quasigeodesic almost pseudo-Anosov flow in $M^3$ with

7728: $\pi_1(M)$ negatively curved. Suppose there is

7729: an unstable leaf $V$ of $\wlu$ and different orbits

7730: $\beta_0, \beta_1$ in $V$ with $(\beta_0)_+ = (\beta_1)_+$.

7731: Then $C_0 = \wls(\beta_0),  \ C_1 = \wls(\beta_1)$ are both periodic,

7732: invariant under a nontrivial covering translation $f$,

7733: and the periodic orbits in $C_0, C_1$ are connected

7734: by an even chain of lozenges all intersecting $V$.

7735: \label{ident}

7736: \end{theorem}

7737: 

7738: \noindent

7739: {\bf {Remark}} $-$

7740: This result is case 2 of theorem 5.7 of \cite{Fe5}.

7741: In that article the proof is done for quasigeodesic

7742: Anosov flows in $M^3$ with $\pi_1(M)$ negatively

7743: curved. The proof goes verbatin to the case of

7744: pseudo-Anosov flows. The singularities make no

7745: difference. By the blow up operation, the same holds

7746: for almost pseudo-Anosov flows.

7747: \vskip .1in

7748: 

7749: 

7750: %Assume first that none of the $r_t$ has limit $p$.

7751: %In the joint figure in $\mi$, (figure \ref{remo}, b) we see

7752: %a continuun of leaves $g_0$ to $r_t$ which are all escaping to $p$.

7753: %There is a last leaf $G'$. Now the other ideal point

7754: %of $g_0$ cannot be $p$ or else it would be in the

7755: %non separated case.

7756: %If some $l_t$ is part of a leaf of $\wlsf$ with 

7757: %2 ideal points identified, then this $l_t$ is inside a strip ${\cal S}$

7758: %which is a bounded distance from a  geodesic ray. 

7759: %Hence the whole

7760: %half of the strip converges to $(H_i)_+$. Consider a parametrization

7761: %$\{ q_t \ | \ t \geq 0 \}$ for the outer ray $x$ of the strip

7762: %and let $x_t$ be the unstable leaves of $\wluf$ through $q_t$.

7763: %If some has identifications, we produce another pair of

7764: %non separated leaves. As seen before this process has to stop.

7765: %For simplicity assume the last boundary leaf is a stable one

7766: %and call it $s_0$ also.

7767: %

7768: %\blankfig{quilt}{1.1}{Escaping and non separated leaves.}

7769: %

7770: %\vskip .1in

7771: %Suppose first that none of the $r_t$'s has endpoints $p$.

7772: %If the $r_t$ escape in $F$ then as seen in the

7773: %claim, we obtain continuity in this side of $p$. If on the

7774: %other hand we obtain $r_t$ converging to $r'$, then again

7775: %we obtain non separated leaves $R'$ and $G'$ in $\wlu$.

7776: %Again this produces a non Hausdorff strip that in $M$

7777: %is asymptotic to a Reeb annulus in a leaf of $\fol$. Since

7778: %any two such strips are a bounded distance apart, the same

7779: %happens to their limits, so we produce non Hausdorff strips in

7780: %the same leaf of $\fol$. 

7781: %

7782: %Since there are only finitely many non Hausdorff leaves of $\wlu$

7783: %up to covering translations, they will project to the same leaf

7784: %in $M$. This produces a ${\bf Z} \oplus {\bf Z}$ in $\pi_1(M)$

7785: %as seen before and this is impossible.

7786: %

7787: %So now assume that some $r_0$ has ideal point $p$. Consider

7788: %$s_t$, the set of leaves of $\wls$ intersecting $r_{\infty}$.

7789: %One of the endopoints is $p$ and the other endpoint is also $p$

7790: %will produce non Hausdorfwless. So assume the other endpoint

7791: %is not $p$.

7792: %

7793: %Let $s_t \subset S_t \in \wls$.

7794: %Notice that $R_b$ intersects $E_0$ and

7795: %

7796: %$$(R_b \cap E_0)_+ \ = \ (E_0)_+ \ = \ (H^*)_-.$$

7797: %

7798: %

7799: %\noindent

7800: %If $\Theta(r_b)$ escapes in $\Theta(R_b)$ then $r_b$ converges

7801: %to $(R_b)_-$. But this is equal to $(R_b \cap E_0)_+$,

7802: %contradiction.

7803: %So $\Theta(r_b)$ limits to a leaf $U_0$

7804: %which is a stable leaf, so $r_b$ escapes going up and

7805: %$r_b$ limits to

7806: %$(R_b \cap E_0)_+$. Then

7807: %

7808: %$$(R_b \cap E_0)_+ \ = \ (R_b \cap V_0)_+$$

7809: %

7810: %\noindent

7811: %This means that 2 distinct orbits in the same unstable leaf $R_b$

7812: %have the same positive ideal point. This situation was analysed

7813: %before.

7814: %

7815: %The region $B$ bounded $l_0$ and $e_0$ with ideal point

7816: %$v$ does not have any interior singularity. If it did

7817: %we would have a line leaf in $B$ and hence a leaf with 

7818: %both endpoints at $v$. By proposition \ref{band},

7819: %that leaf would be

7820: %part of a spike region and the spike region 

7821: %does not have any singularities in its interior.

7822: %

7823: %\begin{proposition}{}{}

7824: %In the situation above, then $E_0, R_0$ are the sides of lozenges

7825: %connected by a finite chain all intersecting $R_b$.

7826: %\end{proposition}

7827: 

7828: The theorem implies that $G, H$ are in the boundary of a chain

7829: of adjacent lozenges all intersecting $S$. The first lozenge,

7830: call it ${\cal C}$ has one stable side contained in $G$

7831: and an unstable side $D_1$ which makes a perfect fit with

7832: $G$. Suppose first $D_1$ is in the side of $S$ opposite

7833: to $E_0$, see fig. \ref{spread}, a. The other unstable side

7834: of ${\cal C}$ is a leaf $D_2$ which intersects $G$ on the

7835: other side of $S$. Hence $G$ is some $S_c$ with $c > t_0$.

7836: Then $S_c \cap F = s_c$ \ is a leaf of $\wluf$ and

7837: $\Psi(s_c)$ has ideal point $\Psi(v)$. Notice that

7838: $\Theta(s_c)$ (which is contained in $\Theta(F)$) escapes

7839: in $\Theta(F)$ $-$ otherwise it would produce stable/unstable

7840: boundary in $\Theta(F)$ before it hits $\Theta(H)$

7841: and $\Theta(F)$ could not limit on $\Theta(H)$, impossible.

7842: Hence $\Psi(s_c)$ limits to $(S_c)_-$ which is equal to $\Psi(v)$.

7843: Then

7844: 

7845: $$(S_c \cap G)_- \ = \ (S_c)_- \ = \ \Psi(v) \ = \ G_+$$

7846: 

7847: \noindent 

7848: which contradicts the orbit $S_c \cap G$ being a

7849: quasigeodesic.

7850: 

7851: 

7852: %Then the forward ideal point is $(E_0)_+$ which is $(S)_-$, etc..

7853: %If the chain of nonseparated leaves is not on the $H_*$ side,

7854: %then there is $R_c$ with $c > b$ so that $R_c$ makes a perfect

7855: %fit with $A$ one side of a lozenge. Then $r_c$ has ideal point

7856: %$(R_c)_- = (R_c \cap E_0)_-$. This is a contradiction

7857: %because the limit point is also $(R_b \cap E_0)_+ = (E_0)_+$.

7858: 

7859: It follows that the perfect fits with $G$ occurs in the $E$ side

7860: of $S$, see fig. \ref{spread}, b.

7861: Here $\Theta(H), \Theta(D_1)$ are

7862: contained in the boundary of $\Theta(F)$. 

7863: %As seen

7864: %in case ???? any points

7865: %in $F$ in the region between $e_0$ and $s_{t_0}$ can only

7866: %limit to $(D_1)_- = H_+$.

7867: We now look at the region $B$ in $F$ bounded by $s_{t_0} = S \cap F$

7868: and $e_0 = E_0 \cap F$. 

7869: 

7870: 

7871: \vskip .1in

7872: \noindent

7873: {\underline {Claim 1}} $-$ 

7874: The image $\Psi(B)$ can only

7875: limit in $\Psi(v)$. 

7876: 

7877: 

7878: The region $\Psi(B)$ is contained in the region ${\cal E}$ of

7879: $\mi$ which is bounded by $E, D_1$ (maybe other unstable leaves non

7880: separated from $D_1$ as well), $H$ and $S$, see fig. \ref{spread}, b.

7881: Notice that $F$ does not intersect $D_1$ or any leaf non separated from

7882: $D_1$ which is beyond $D_1$. Otherwise $b_0 = (D_1 \cap F)$ is

7883: contained in $B$ and non separated from $e_0$, so it would have

7884: both ideal points $v$. Then it would be contained in the

7885: interior of a spike region and could not be non separated

7886: from another leaf $-$ impossible.

7887: On the other hand since $\Theta(H)$ has a line leaf in

7888: the stable boundary of 

7889: $\Theta(F)$, then $\Theta(D_1)$ has a line

7890: leaf in the unstable boundary of $\Theta(F)$.

7891: Hence $F$ escapes down as $\Theta(F)$ approaches 

7892: $\Theta(D_1)$.

7893: 

7894: Let $z_k$ in $B$ escaping in $F$ and hence

7895: converging to $v$ in $\pin F$.

7896: Suppose that $\Psi(z_k)$ does not converge to $\Psi(v)$.

7897: Given that $z_k$ escapes $F$ and the structure of the region

7898: ${\cal E}$, it follows that up to subsequence

7899: either $\wu(z_k)$ converges

7900: to $D_1$ or $\ws(z_k)$ converges to $H$.

7901: Suppose that $\ws(z_k)$ converges to $H$.

7902: In that case $(z_k)_+$ converges to $H_+ = \Psi(v)$.

7903: Then as seen before if $(z_k)_-$ does not converge

7904: to $\Psi(v)$ we can assume up to subsequence $\wwr(z_k)$

7905: converges to $\wwr(z)$. Then $z$ is in a leaf

7906: non separated from $H$ and since $\Psi(z_k)$ has to

7907: be in ${\cal E}$ then $z$ can only be in $H$.

7908: As $F$ escapes up as $\Theta(F)$ approaches $\Theta(H)$ then

7909: $\Psi(z_k)$ converges to $H_+ = \Psi(v)$.

7910: The case $\wu(z_k)$ converges to $D_1$ leads to 

7911: $\wwr(z_k)$ converging to $\wwr(z)$ with $z$ in

7912: unstable leaf non separated from $D_1$. As $F$ escapes

7913: down as $\Theta(F)$ approaches these unstable leaves,

7914: then $\Psi(z_k)$ converges to $(D_1)_- = \Psi(v)$.

7915: Since this works for any subsequence of $z_k$, then

7916: $\Psi(z_k)$ has to converge to $\Psi(v)$ always.

7917: This proves claim 1.

7918: 

7919: 

7920: \begin{figure}

7921: \centeredepsfbox{gm16.eps}

7922: \caption{

7923: Perfect fit

7924: with $G$ in the side opposite to $E_0$,

7925: b. Perfect fit in the $E$ side.}

7926: \label{spread}

7927: \end{figure}

7928: 

7929: 

7930: 

7931: 

7932: 

7933: \vskip .1in

7934: Let $G_0 = G$.  Notice that $G$ is periodic and connected to $H$ 

7935: by an even chain of lozenges. 

7936: We consider the ray $s_{t_0} = S \cap F$ which has

7937: ideal point $v$. Parametrize it as \ $\{ z_t \ | \ t \geq 0 \}$.

7938: Let $y_t$ be the leaf of $\wlsf$ through $z_t$ and $y^1_t$ the

7939: component of $(y_t - z_t)$ in the side opposite to $e_0$.

7940: The ray $s_{t_0}$ has the same behavior as the original ray

7941: $e_0$. Hence we obtain continuity in that side of $s_{t_0}$

7942: unless $y^1_t$ converges to a leaf $\mu$ of $\wlsf$ with

7943: ideal point $v$. Let $G_1$ in $\wls$ with $\mu \subset G_1$.

7944: Then $G_1$ is non separated from $H$, see fig. \ref{spread}, b \

7945: and therefore connected to it by a chain of lozenges.

7946: It follows that $G_1$ is connected to $G_0$ by a chain

7947: of lozenges.

7948: As in the proof of claim 1, the region $B_1$ of $F$

7949: between $e_0$ and $(F \cap G_1)$ has image

7950: $\Psi(B_1)$ which can limit only in $\Psi(v)$.

7951: 

7952: We restart the process with $g_1 = G_1 \cap F$ 

7953: instead of $g$. The leaves of $\wluf$ through

7954: points of $g_1$ already converge to the unstable

7955: leaf $(D_3 \cap F)$ of $\wluf$ \ ($D_3$ is depicted in fig.

7956: \ref{spread}, b).

7957: The leaf $(D_3 \cap F)$ cannot be non separated from any

7958: other leaf of $\wluf$ in that side of $(D_3 \cap F)$.

7959: It follows that 

7960: the unstable leaves intersected by $g_1$ escape

7961: in $F$. The only case to be analysed is that some of

7962: these unstable leaves have ideal point $v$. This brings

7963: the process exactly to the situation of some $s^1_t$

7964: of $\wluf$ having ideal point $v$ as described before

7965: (it was $s^1_{t_0}$). So this would produce 

7966: $H_1$ of $\wls$ with similar properties as $H$.

7967: This process can now be iterated.

7968: As in claim 1 the region of $F$ between $g_i$ and $g_{i+1}$

7969: maps to $\mi$ to a region which can only limit in $\Psi(v)$.

7970: 

7971: We show that this process has to stop. Otherwise produce

7972: $G_i$ leaves of $\wls$ which are all connected to $G_0$ by

7973: a chain of lozenges. The $G_i$ are all non separated from

7974: some other leaf of $\wls$, Hence there are $G_i, G_j$ which

7975: project to the same stable leaf in $M$.

7976: There is a covering translation $h$ taking $G_i$ to $G_j$.

7977: If $f$ is a generator of the isotropy group of $G_0$ leaving

7978: all sectors invariant, then it leaves invariant

7979: all lozenges in any chain starting in $G_0$ so leaves

7980: invariant all the $G_i$. As before this leads to $h^{-1} f h = f^n$

7981: for some $n$ in ${\bf Z}$ and to a ${\bf Z} \oplus {\bf Z}$ in

7982: $\pi_1(M)$. This is disallowed.

7983: Therefore the process finishes after say $j$ steps and we

7984: obtain continuity of $\Psi$ at $v$ in that side of $g_j = G_j \cap F$.

7985: As seen above the region between $s_{t_0}$ and $g_j$ maps by $\Psi$

7986: into a region that can only limit in $\Psi(v)$. 

7987: This proves continuity of $\Psi$ at $v$ in that side of $e_0$.

7988: This finishes the analysis of situation 1.

7989: 

7990: 

7991: 

7992: 

7993: \vskip .15in

7994: \noindent

7995: {\underline {Situation 2}} $-$ There is $l^1_{t_0}$ with ideal

7996: point $v$.

7997: 

7998: Recall the setup before the analysis of situation 1. 

7999: Let $\{ u_t \ | \ t \geq 0 \}$  be the collection

8000: of unstable leaves intersected by the ray $l^1_{t_0}$.

8001: The analysis is 

8002: extremely similar to the analysis of situation 1, which shows all

8003: cases produce continuity in the first step except

8004: when $u_t$ converges to a leaf $u$ of $\wluf$ 

8005: with ideal point $v$.

8006: Then consider the stable leaves intersecting $u$.

8007: The analysis of situation 1 shows continuity unless there 

8008: is stable leaf

8009: with ideal point $v$.

8010: From now on the analysis is exactly the same as in situation

8011: 1, with unstable replaced by stable and vice versa.

8012: 

8013: 

8014: %\vskip .2in

8015: %Now look at the stable leaves intersected by $R_a$ which are the $S_t$.

8016: %One of the ideal points of the $S_t$ is $p$. If the other one is

8017: %$p$ we produce a non Hausdorff band. If the $s_t$ escape in $F$

8018: %we obtain continuity on that side. Finally if the $s_t$ limit

8019: %to a leaf we obtain a leaf $E_1$, non separated from $E_0$.

8020: %Again look at $X_t$ $-$ the leaves intersecting $E_1$ 

8021: %and $x_t = X_t \cap F$. Applying the same procedure

8022: %we either end or continue this chain. Since there

8023: %are non separated leaves in the chain, they are all periodic.

8024: %After a while we produce a ${\bf Z} \oplus {\bf Z}$ in

8025: %$\pi_1(M)$, contradiction. 

8026: %

8027: %What this shows is that the process needs to terminate after

8028: %finitely many steps. 

8029: %

8030: 

8031: \vskip .15in

8032: So far we proved continuity of $\Psi$ at $v$

8033: from the side of $e_0$ opposite to $l$.

8034: The same works for the other side of $l$, producing

8035: $l_0$ with similar properties as $e_0$.

8036: We now must consider the regions between $e_0$ and

8037: $e$, between $e$ and $l$ and between $l$ and $l_0$.

8038: 

8039: First consider the region between $e$ and $e_0$,

8040: which occurs only when they are distinct. This implies

8041: that the ray $e_0$ is a bounded distance from a

8042: geodesic ray in $F$ with ideal point $v$.

8043: Let $\{ \mu_t \ | \ t \geq 0 \}$ be a parametrization

8044: of the stable leaves of $\wlsf$ through $e$.

8045: Let $\mu^1_t$ be the component of $(\mu^1_t - e)$ 

8046: in the side of $e$ we are considering. If some 

8047: $\mu^1_t$ has ideal point $v$ then both ideal

8048: points of $\mu_t$ are $v$ and $\mu_t$ is inside

8049: a spike region. The same is true for $e$ and

8050: so $e$ is a bounded distance from a geodesic

8051: ray in $F$ with ideal point $a$.

8052: %We refer to fig. \ref{vera} a.

8053: Hence the region between $e$ and $e_0$ is a bounded

8054: distance from a geodesic ray and we are finished in this

8055: case.

8056: 

8057: %\blankfig{vera}{1.1}{a. The case $\mu^1_t$ has ideal point $v$,

8058: %b. The remaining case.}

8059: 

8060: The remaining case to be analysed here is that $\mu^1_t$

8061: has no ideal point $v$. Then $\mu^1_t$ does not escape

8062: $F$ as $t \rightarrow \infty$, because $e_0$ is in that

8063: side of $e$.

8064: %, see fig. \ref{vera}, b.

8065: So $\mu^1_t$ converges to a leaf $\mu$ which has ideal

8066: point $v$.

8067: Now consider a parametrization $\{ \nu_t \ | \ t \geq 0 \}$ of

8068: the unstable leaves intersected by $\mu$.

8069: Then $\nu_t$ converges to the leaf $e$.

8070: If it converges to some other leaf, then $e$ is

8071: a bounded distance from a geodesic ray in $F$ and we are done.

8072: Otherwise it must be that some $\nu_t$ has ideal

8073: point $v$. Therefore we exactly in the setup analysed

8074: in situation 1 above. 

8075: 

8076: This shows continuity of $\Psi$ for the region between $e$ and

8077: $e_0$ and similarly for the region between $l$ and $l_0$.

8078: 

8079: Finally we analyse the region $B$ between $e$ and $l$.

8080: First notice there is no singularity in the interior

8081: of $B$. Otherwise there would be a line leaf in

8082: $B$ and hence a leaf with both endpoints $v$. 

8083: It would have to be part of a spike region and the

8084: spike region does not have any singularities in its 

8085: interior.

8086: 

8087: Parametrize the leaves of $\wluf$ through $l$ as

8088: $\{ e_s \ | \ s \geq 0 \}$ and similarly

8089: those of $\wlsf$ through $e$ as $\{ l_t \ | \ t \geq 0 \}$.

8090: Let $L, L_t$ leaves of $\wls$ with $l \subset L, \ l_t

8091: \subset L_t$ and similarly define $E, E_t$.

8092: There are 2 possibilities:

8093: 

8094: 

8095: 

8096: 

8097: 

8098: \vskip .1in

8099: \noindent

8100: 1) Product case $-$ Any $l_t$ intersects every $e_s$ and

8101: vice versa. 

8102: 

8103: Equivalently $\wlsf, \wluf$ define  a product structure 

8104: in the region $B$ bounded by $l_0$ and $e_0$. If the $L_t$ escapes

8105: in $\mi$ as $t \rightarrow \infty$, then there is a stable

8106: product region defined by a segment in $L_0$. 

8107: But then theorem \ref{prod} implies that

8108: $\Phi$ is topologically conjugate to a suspension,

8109: contradiction.

8110: It follows that the 

8111: $L_t$ converge to $H_1 \cup ... H_m$ as $t \rightarrow \infty$.

8112: %As in case ??? for any $t$, $l_t$ converges to

8113: Since the $l_t$ are stable leaves, it follows that

8114: $F$ escapes up as $\Theta(F)$ appraches $\Theta(H_i)$.

8115: This implies that $\Psi(e)$ limits to $(H_i)_+$ which

8116: is then equal to $\Psi(v)$.

8117: Similarly $E_s$ converges to $V_1 \cup ... V_n$ and

8118: $F$ escapes down as $\Theta(F)$ approaches $\Theta(V_j)$. 

8119: Hence 

8120: $\Psi(l)$ limits to $(V_j)_- = \Psi(v)$.

8121: If some $H_i$ intersects some

8122: $V_j$, then 

8123: 

8124: 

8125: $$(V_j \cap H_i)_+ \ = \ (H_i)_+ 

8126: \ = \ \Psi(v) \ = \ 

8127: (V_j)_- \ = \ (V_j \cap H_i)_-,$$

8128: 

8129: \noindent

8130: contradiction.

8131: %Let now $u_i$ in $F$ in the region $B$. 

8132: Let now $\{ z_k \}$ be a sequence in $B$ converging

8133: to $v$. The product structure implies that up to subsequence

8134: we may assume that either $\ws(z_k)$ converges to

8135: $H_i$ or $\wu(z_k)$ converges to $V_j$. 

8136: This is analysed carefully in Claim 1 above, which

8137: shows that $\Psi(z_k)$ must converge to $\Psi(v)$.

8138: This shows continuity of $\Psi$ when restricted to the

8139: region $B$.

8140: 

8141: 

8142: %Suppose $u_i$ converges

8143: %to $q$ not equal $\varphi(p)$. Then $u_i$ are in $\ws(p_{t_i})$ for some

8144: %$t_i > 0$. Suppose first the $t_i$ does not converge to infinity.

8145: %Up to subsequence we may assume that $t_i \rightarrow t_0$.

8146: %\blankfig{regi}{1.1}{a. A region where there is a product structure.}

8147: %Also $(v_i)_- = (q_{t_i})_-$. Here $t_i \rightarrow \infty$ else

8148: %

8149: %$$v_i \  \rightarrow  \ \ws(p_{t_0}) \cap \wu(q_{b_0}) \cap F.$$

8150: %

8151: %\noindent

8152: %So $(v_i)_- = (q_{c_t})_-$ converges to $(V_j)_-$.

8153: %If $v_i$ does not converge to $(V_j)_-$ then up to subsequence

8154: %$\wwr(v_i)$ converges to $\wwr(v)$ and $v$ is in one of the $V_i$,

8155: %say in $V_{i_0}$. But then $F$ escapes down so $v_i$ converges

8156: %to $(V_j)_-$.

8157: %Similarly  if $b_i$ does not converge to infinity. Finally if

8158: %$t_i, b_i$ converge to infinity then

8159: %

8160: %$$(v_i)_+ = (p_{t_i})_+ \rightarrow (H_1)_+, \ \ 

8161: %(v_i)_- = (q_{b_i})_+ \rightarrow (V_j)_-.$$

8162: %

8163: %\noindent

8164: %Hence since the orbits are uniform quasigeodesics we obtain 

8165: %$v_i \rightarrow (H_1)_-$ as required.

8166: %This finishes the analysis in the product situation.

8167: %

8168: 

8169: \vskip .1in

8170: \noindent

8171: 2) Non product case.

8172: 

8173: There are $t, u > 0$ with $l_t \cap e_u = \emptyset$.

8174: Consider one such $u$.

8175: Let $a$ be the infimum of $t$ with $l_t \cap e_u = \emptyset$.

8176: Now let $b$ be the infimum of $u$ with $l_a \cap e_u = \emptyset$.

8177: Then $l_a \cap e_b = \emptyset$, but for any $0 \leq t < a$

8178: and $0 \leq u < b$ one has 

8179: $l_t \cap e_u \not = \emptyset$. 

8180: Since $l_a \cap e_b = \emptyset$, then $L_a \cap E_b = \emptyset$.

8181: It follows that $L_a, E_b$ form a perfect fit,

8182: see fig. \ref{nene}, a.

8183: If $\Theta(l_a)$ does not escape in $\Theta(L_a)$, then

8184: there would be unstable boundary of $\Theta(F)$ in the

8185: limit and that would keep $F$ from intersecting $E_b$,

8186: contradiction. Hence $\Theta(l_a)$ escapes in $\Theta(L_a)$

8187: and $\Theta(e_b)$ escapes in $\Theta(E_b)$.

8188: Hence $\Psi(l_a)$ limits to $(L_a)_+$ and $\Psi(e_b)$ limits

8189: to $(E_b)_-$. 

8190: Also $l_a, e_b$ limit to $v$ in $\pin F$. 

8191: 

8192: Let $p_t = l_t \cap e$. If $\Theta(p_t)$ escapes in $\Theta(E)$,

8193: then $\Psi(e)$ converges to $E_-$. Notice that $\Psi(e)$ converges

8194: to $\Psi(v)$ so:

8195: 

8196: 

8197: $$E_- \ = \ \Psi(v) \ = \ (L_a)_+ \ = \ (L_a \cap E)_+$$

8198: 

8199: \noindent

8200: contradiction.

8201: It follows that $\wwr(p_t)$ converges to

8202: $\wwr(x)$ with $x$ in $E$.

8203: Also $F$ has to escape up as $\Theta(F)$ approaches $\Theta(x)$

8204: $-$ same as in Situation 1 above.

8205: Hence $\Psi(e)$ limits to $x_+$. So

8206: 

8207: $$x_+ \ = \ \Psi(v) \ = \ (L_a)_+ \ = \ (p_a)_+$$

8208: 

8209: \noindent

8210: Let $X = \wws(x)$.

8211: Then $x, p_a$ are in 2 distinct orbits of $E$ with

8212: the same positive ideal point.

8213: Therefore theorem \ref{ident} implies that

8214: $L_a,  X$ are connected by an even chain of lozenges

8215: all intersecting $E$.

8216: Let ${\cal C}$ be the first lozenge. It has a stable

8217: side in $L_a$ and one unstable side, call it $D_1$ which

8218: makes a perfect fit with $L_a$. Suppose first that $D_1$ is

8219: in the component of $\mi - E$ opposite to $E_b$.

8220: Then the other unstable side of ${\cal C}$, call it $D_2$

8221: has to intersect $L_a$ in the other side of $E$.

8222: %, see

8223: %fig. \ref{nene}, a. 

8224: Then $D_2$ must be some $E_t$,

8225: let it be $E_{b'}$,

8226:  see

8227: fig. \ref{nene}, a. 

8228: Then $\Theta(e_{b'})$ has to escape in $\Theta(E_{b'})$ 

8229: or else one produces stable boundary to $\Theta(F)$ and

8230: $\Theta(F)$ cannot limit to $\Theta(x)$ contradiction.

8231: Hence $\Psi(e_{b'})$ converges to $\Psi(v)$ and 

8232: also to $(E_{b'})_-$.

8233: But then 

8234: 

8235: $$(E_{b'} \cap L_a)_- \ = \ 

8236: (E_{b'})_- \ = \ \Psi(v) \ = \ (E_b)_- \ = \ (L_a)_+$$

8237: 

8238: \noindent

8239: again a contradiction.

8240: 

8241: 

8242: \begin{figure}

8243: \centeredepsfbox{gm17.eps}

8244: \caption{

8245: a. Reaching before, b. Reaching at the exact time.}

8246: \label{nene}

8247: \end{figure}

8248: 

8249: 

8250: This implies that $D_1$ is on the side of $E$ containing

8251: $E_b$, see fig. \ref{nene}, b.

8252: 

8253: If there are only 2 lozenges in the chain from $L_1$ to

8254: $X$, then  $D_1$ also makes a perfect fit with

8255: $X$. Otherwise there are $D_2, ..., D_i$ all non separated

8256: from $D_1$ and so that $D_i$ makes a perfect fit 

8257: with $X$ and the $D_j$ are all in the boundary of the

8258: chain of lozenges.

8259: As seen in claim 1 above, $F$ cannot intersect

8260: any $D_j$ ($1 \leq j \leq i$), but all

8261: $\Theta(D_j)$ are contained in the unstable boundary

8262: of $\Theta(F)$. Also $F$ escapes down as $\Theta(F)$

8263: limits to $\Theta(D_j)$.

8264: The set $\Theta(X)$ also has a line leaf which is

8265: a stable boundary of $\Theta(F)$ and $F$ escapes

8266: up when $\Theta(F)$ approaches $\Theta(X)$.

8267: 

8268: The same discussion applies to $L$, so there is 

8269: $y$ in $L$, $Y = \widetilde W^u(y)$ with $\Theta(Y)$

8270: having a line leaf in the unstable boundary

8271: of $\Theta(F)$ and $F$ escapes down accordingly.

8272: There are $C_1, ..., C_n$ leaves in $\wlu$, 

8273: all non separated from each other and in the

8274: boundary of the lozenges in the chain from $E_b$ to

8275: $Y$ so that $C_1$ makes a perfect fit with $E_b$ and

8276: $C_n$ makes a perfect fit with $Y$, see fig. \ref{nene}, b.

8277: Finally $\Theta(C_j)$ has a line leaf in the

8278: stable boundary of $\Theta(F)$ and $F$ escapes up

8279: accordingly.

8280: 

8281: Let ${\cal E}$ be the region in $\mi$ bounded

8282: by

8283: 

8284: $$E, \ L, \ X, \ Y, \ C_1, ..., C_n, \ D_1, ..., D_i$$

8285: 

8286: \noindent

8287: Then ${\cal E} \cap F$ is exactly the region $B$

8288: bounded by the rays $e$ and $l$.

8289: Let $z_k$ in $B$ escaping to $v$.

8290: Then the region ${\cal E}$ shows that

8291: up to subsequence one of the following must occur:

8292: 

8293: \vskip .08in

8294: 1) $\ws(z_k)$ converges to either $X$ or $C_1$. 

8295: The analysis of claim 1 above shows that

8296: $\Psi(z_k)$ converges to either $X_+$ or $(C_1)_+$

8297: both of which are equal to $\Psi(v)$.

8298: 

8299: 2) $\wu(z_k)$ converges to either $Y$ or $D_1$.

8300: Here $\Psi(z_k)$ converges to either $Y_-$ or

8301: $(D_1)_-$ both of which are equal to $\Psi(v)$.

8302: 

8303: In any case this shows continuity of $\Psi$ in the

8304: region $B$.

8305: This finishes the non product case.

8306: 

8307: \vskip .05in

8308: This finishes the proof of theorem \ref{exten},

8309: the continuous extension theorem.

8310: \end{proof}

8311: 

8312: 

8313: 

8314: 

8315: 

8316: 

8317: %$(b_1)_+ = (L_{t_0} \cap E_0)_+$.

8318: %As seen before this implies that $L_{t_0}$ and $A = \ws(b_1)$

8319: %are connected by an even chain of lozenges, all

8320: %intersecting a common stable???  leaf.

8321: 

8322: 

8323: %If the asymptotic sides are in the side not containing $E_u, u > 0$,

8324: %then there is an unstable side $E_{u_1}, u_1 < u_0$, see fig. \ref{nene}.

8325: %But then

8326: %

8327: %$$(e_{u_1})_{\infty} \ = \ (E_{u_1})_- \ = \ (E_{u_1} \cap L_{t_0})_-$$

8328: %

8329: %\noindent

8330: %But 

8331: %

8332: %$$(e_{u_0})_- \ = \ (E_{u_0})_- \ = \ (L_{t_0})_+,

8333: %\ \ {\rm so} \ \ (L_{t_0})_+ \ = \ (E_{u_1} \cap L_{t_0})_-$$

8334: %

8335: %\noindent

8336: %which is impossible.

8337: %Therefore the asymptotic boundary lozenges are on the other side,

8338: %see fig. \ref{nene}, b.

8339: 

8340: 

8341: %So if $v_i$ is in $B$ and between $H_1, L_{t_0}$ and $E_0$ and

8342: %it escapes $F$ then if it escapes $\oo$, it will

8343: %converge to $(H_1)_-$. If it converges to a point

8344: %in the boundary, then it will go to either $H_1$

8345: %or to $V_1 \cup V_2 ...$. In either case either converges

8346: %to $(H_1)_-$ or to $(V_i)_+$ which are equal. Hence we

8347: %obtain continuity at $p$ in the other side of $l_{t_0}$.

8348: %Similarly for the other side of $e_{u_0}$.

8349: %Finally in the region between $l_{t_0}$ and $e_{u_0}$ if

8350: %$\Theta(v_i)$ escapes, we have that 

8351: %

8352: %

8353: %$$\wu(v_i) \rightarrow E_{u_0} \ \ {\rm and} \ \ 

8354: %\ws(v_i) \rightarrow L_{t_0}. \ \ {\rm So} \ \ 

8355: %(v_i)_- \rightarrow (L_{t_0})_+,

8356: %(v_i))- \rightarrow (E_{u_0})_-$$

8357: %

8358: %\noindent

8359: %which are equal. So again we obtain continuity as desired.

8360: %This finishes the proof of the theorem.

8361: %\end{proof}

8362: %

8363: %Conclusion: finite depth implies continuous extension.

8364: 

8365: 

8366: 

8367: 

8368: 

8369: 

8370: 

8371: 

8372: \section{Foliations and Kleinian groups}

8373: 

8374: There are many similarities between foliations in hyperbolic

8375: 3-manifolds and Kleinian groups.

8376: We refer to \cite{Mi,Can,Mar} for basic definitions concerning

8377: degenerate and non degenerate Kleinian groups, in particular

8378: singly and doubly degenerate groups.

8379: 

8380: 

8381: If the foliation is $\rrrr$-covered then the limit set

8382: of any leaf in $\mi$ is the whole sphere \cite{Fe3}. 

8383: This corresponds to doubly degenerate surface

8384: Kleinian groups \cite{Th1,Mi,Can,Mar,Bon}.

8385: There is always a pseudo-Anosov flow which is transverse

8386: to the foliation \cite{Fe7,Cal1}. If the flow is quasigeodesic then

8387: the results of this article imply that the foliation

8388: has the continuous extension property.

8389: 

8390: If the foliation has one sided branching, say branching down,

8391: then limit sets of leaves can only have domain of discontinuity

8392: ``above" \cite{Fe3}. 

8393: Let $F$ in $\fn$ and $\Lambda_F$ its limit set. If $p$ is not

8394: in $\Lambda_F$, the $p$ is said to be {\em above} $F$ if

8395: there is a neighborhood $V$ of $p$ in $\mi \cup \si$, so

8396: that $V \cap \mi$ is on the positive side of $F$.

8397: This corresponds to simply degenerate surface Kleinian

8398: groups \cite{Th1,Mi,Can}. There are examples of foliations with one sided

8399: branching transverse to 

8400: suspension pseudo-Anosov flows

8401: provided by Meigniez \cite{Me}. Suspension flows are always

8402: quasigeodesic flows \cite{Ze}. 

8403: The results of this article show the continuous extension property

8404: for such foliations.

8405: Under these conditions, the limit

8406: sets are locally connected, the continuous extension provides

8407: parametrizations of these limit sets.

8408: 

8409: Finally if there is branching in both directions, 

8410: %for example

8411: %in true Reebless finite depth foliations (not perturbations

8412: %of fibrations). 

8413: then there can be domain of discontinuity above

8414: and below leaves. This corresponds to non degenerate

8415: Kleinian groups \cite{Th1,Mi,Can}. These occur for example

8416: in the case of finite depth foliations, where

8417: the depth 0 leaves are not virtual fibers \cite{Fe6}.

8418: 

8419: There are many interesting questions:

8420: 

8421: \vskip .1in

8422: \noindent

8423: {\bf {Question 1}} $-$ Given a 

8424: foliation $\fol$, is it  $\rrrr$-covered if and only if 

8425: for every $F \in \fn$ then the limit set $\Gamma_F$ is $\si$?

8426: 

8427: The forward direction is true. The backwards direction

8428: is true if there is a compact leaf \cite{Go-Sh}. In addition if there is one

8429: leaf with limit set the whole sphere then all leaves have

8430: limit set the whole sphere \cite{Fe3} $-$ whether $\fol$ is

8431: $\rrrr$-covered or not.

8432: 

8433: \vskip .1in

8434: \noindent

8435: {\bf {Question 2}} $-$ Given $\fol$ an $\rrrr$-covered foliation, is there 

8436: a quasigeodesic transverse pseudo-Anosov flow?

8437: 

8438: This is true in the case of slitherings or uniform foliations

8439: as defined by Thurston \cite{Th5}. Examples are fibrations,

8440: $\rrrr$-covered Anosov flows and many others.

8441: There is always a transverse pseudo-Anosov flow, the

8442: question is whether it is quasigeodesic.

8443: 

8444: \vskip .1in

8445: \noindent

8446: {\bf {Question 3}} $-$ Is there domain of discontinuity of $\Lambda_F$ only

8447: above $F$ if and only if $\fol$ has one sided branching

8448: in the negative direction?

8449: 

8450: This occurs for the examples constructed by Meigniez \cite{Me}.

8451: 

8452: \vskip .1in

8453: \noindent

8454: {\bf {Question 4}} $-$ Are the pseudo-Anosov flows constructed

8455: by Calegari \cite{Cal2} and transverse to one sided branching foliations

8456: quasigeodesic?

8457: 

8458: 

8459: \vskip .1in

8460: \noindent

8461: {\bf {Question 5}} $-$ If $\fol$ has 2 sided branching is there

8462: always domain of discontinuity above and below?

8463: Is there a quasigeodesic pseudo-Anosov flow almost transverse to $\fol$?

8464: 

8465: 

8466: 

8467: 

8468: 

8469: 

8470: 

8471: {\footnotesize

8472: {

8473: \setlength{\baselineskip}{0.01cm}

8474: 

8475: \begin{thebibliography}{Fe-Mo}

8476: 

8477: {

8478: 

8479: %\setlength{\baselineskip}{0pt}

8480: %\baselineskip0pt

8481: 

8482: 

8483: 

8484: \bibitem[Ba1]{Ba1} T. Barbot, {\em Caract\'{e}rization des flots d'Anosov

8485: 	pour les feuilletages}, Erg. Th. Dyn. Sys. {\bf 15} (1995) 247-270.

8486: 

8487: \bibitem[Ba2]{Ba2} T. Barbot, {\em Actions de groupes sur les 

8488: 	$1$-vari\'{e}t\'{e}s non s\'{e}par\'{e}es et feuilletages de

8489: 	codimension un}, Ann. Fac. Sci. Toulose Math. {\bf 7} (1998) 559-597.

8490: 

8491: 

8492: \bibitem[Be-Me]{Be-Me} M. Bestvina and G. Mess, {\em The boundary of

8493: 	negatively curved groups}, Jour. Amer. Math. Soc.

8494: 	{\bf 4} (1991) 469-481.

8495: 

8496: \bibitem[Bl-Ca]{Bl-Ca} S. Bleiler and A. Casson, {\em Automorphims of

8497: 	surfaces after Nielsen and Thurston}, Cambridge Univ. Press, 1988.

8498: 	

8499: \bibitem[Bon]{Bon} F. Bonahon, {\em Bouts des vari\'{e}t\'{e}s 

8500: 	hyperboliques de dimension 3}, Ann. of Math. {\bf 124} (1986) 71-158.

8501: 

8502: \bibitem[Br1]{Br1} M. Brittenham, {\em Essential laminations in, {\em The boundary of

8503: 	negatively curved groups}, Jour. Amer. Math. Soc.

8504: 	{\bf 4} (1991) 469-481.

8505: 	Seifert fibered spaces}, Topology {\bf 32} (1992) 61-85. 

8506: 

8507: 

8508: \bibitem[Cal1]{Cal1} D. Calegari, {\em The geometry of $\rrrr$-covered

8509: 	foliations}, Geom. Topol. {\bf 4} (2000) 457-515 (eletronic).

8510: 

8511: \bibitem[Cal2]{Cal2} D. Calegari, {\em Foliations with one-sided branching},

8512: 	Geom. Ded. {\bf 96} (2003) 1-53.

8513: 

8514: %\bibitem[Cal3]{Cal3} D. Calegari, {\em Universal circles for

8515: %quasigeodesic flows}, eprint math.GT/0406040.

8516: 

8517: \bibitem[Cal3]{Cal3} D. Calegari, {\em Promoting essential laminations},

8518: 	eprint math.GT/0210148.

8519: 

8520: \bibitem[Ca-Du]{Ca-Du} D. Calegari and N. Dunfield, {\em Laminations

8521: 	and groups of homeomorphisms of the circle}, 

8522: 	Inven. Math. {\bf 152} (2003) 149-204.

8523: 

8524: \bibitem[Ca]{Ca} A. Candel, {\em Uniformization of surface laminations},

8525: 	Ann. Sci. Ecole Norm. Sup. {\bf 26} (1993) 489-516.

8526: 

8527: \bibitem[Can]{Can} R. Canary, {\em Ends of hyperbolic $3$-manifolds},

8528: 	Jour. Amer. Math. Soc. {\bf 6} (1993) 1-35.

8529: 

8530: \bibitem[Ca-Th]{Ca-Th} J. Cannon and W. Thurston, {\em Group invariant

8531: 	Peano curves}, preprint, 1987.

8532: 

8533: \bibitem[Ca-Co]{Ca-Co} J. Cantwell and L. Conlon, {\em Smoothability of	

8534: 	proper foliations}, Ann. Inst. Fourier Grenoble 

8535: 	{\bf 38} (1988) 403-453.

8536: 

8537: \bibitem[CDP]{CDP} M. Coornaert, T. Delzant and A. Papadopoulos,

8538: 	{\em G\'{e}om\'{e}trie and th\'{e}orie des groupes, Les groupes

8539: 	hyperboliques de Gromov}, Lec. Notes in Math {\bf 1441} 

8540: 	Springer (1991).

8541: 

8542: 

8543: \bibitem[Di]{Di} P. Dippolito, {\em Codimension one foliations of closed

8544: 	manifolds}, Ann. of Math. {\bf 107} (1978) 403-453.

8545: 

8546: %\bibitem[Fe1]{Fe1} S. Fenley, {\em Asymptotic properties of depth one

8547: %foliations in hyperbolic 3-manifolds}, Jour. Diff. Geom.

8548: %{\bf 36} (1992) 269-313.

8549: 

8550: \bibitem[Fe1]{Fe1} S. Fenley, {\em Quasi-isometric foliations},

8551: 	Topology {\bf 31} (1992) 667-676.

8552: 

8553: \bibitem[Fe2]{Fe2} S. Fenley, {\em Anosov flows in $3$-manifolds},

8554: 	Ann. of Math. {\bf 139} (1994) 79-115. 

8555: 

8556: %\bibitem[Fe4]{Fe4} S. Fenley, {\em Quasigeodesic Anosov flows and

8557: %homotopic properties of flow lines}, Jour. Diff. Geom.

8558: %{\bf 41} (1995) 479-514.

8559: 

8560: \bibitem[Fe3]{Fe3} S. Fenley, {\em Limit sets of foliations},

8561: 	Topology {\bf 37} (1998) 875-894.

8562: 

8563: \bibitem[Fe4]{Fe4} S. Fenley, {\em The structure of branching

8564: 	in Anosov flows of $3$-manifolds}, 

8565: 	Comm. Math. Helv. {\bf 73} (1998) 259-297.

8566: 

8567: \bibitem[Fe5]{Fe5} S. Fenley, {\em Local and global properties of Anosov

8568: 	foliations in 3-manifolds with negatively curved fundamental

8569: 	group}, Trans. A.M.S. {\bf 350} (1998) 3923-3941.

8570: 

8571: \bibitem[Fe6]{Fe6} S. Fenley, {\em Foliations with good geometry},

8572: 	Jour. Amer. Math. Soc. {\bf 12} (1999) 619-676.

8573: 

8574: \bibitem[Fe7]{Fe7} S. Fenley, {\em Foliations, topology and geometry

8575: 	of 3-manifolds: $\rrrr$-covered foliations and transverse

8576: 	pseudo-Anosov flows}, Comm. Math. Helv. {\bf 77} (2002) 415-490.

8577: 

8578: \bibitem[Fe8]{Fe8} S. Fenley, {\em Pseudo-Anosov flows and incompressible

8579: 	tori}, Geom. Ded. {\bf 99} (2003) 61-102.

8580: 

8581: \bibitem[Fe9]{Fe9} S. Fenley, {\em Regulating flows, topology of foliations

8582: 	and rigidity},

8583: 	Trans. A.M.S. {\bf 357} (2005) 4957-5000.

8584: 

8585: \bibitem[Fe10]{Fe10} S. Fenley, in preparation.

8586: 

8587: \bibitem[Fe-Mo]{Fe-Mo} S. Fenley and L. Mosher, {\em Quasigeodesic

8588: 	flows in hyperbolic 3-manifolds}, Topology {\bf 222} (2001) 503-537.

8589: 

8590: \bibitem[Fr]{Fr} D. Fried, {\em Transitive Anosov flows and pseudo-Anosov

8591: 	maps}, Topology {\bf 22} (1983) 299-303.

8592: 

8593: \bibitem[Ga1]{Ga1} D. Gabai, {\em Foliations and the topology of 3-manifolds},

8594: 	J. Diff. Geo. {\bf 18} (1983) 445--503.

8595: 

8596: \bibitem[Ga2]{Ga2} D. Gabai, {\em Foliations and the topology 

8597: 	of 3-manifolds II},

8598: 	J. Diff. Geo. {\bf 26} (1987) 461--478.

8599: 

8600: \bibitem[Ga3]{Ga3} D. Gabai, {\em Foliations and the topology 

8601: 	of 3-manifolds III},

8602: 	J. Diff. Geo. {\bf 26} (1987) 479--536.

8603: 

8604: %\bibitem[Ga4]{Ga4} D. Gabai, {\em $8$ problems in foliations

8605: %and laminations}, in Geometric Topology, W. Kazez, ed.,

8606: %Amer. Math. Soc. 1987, 1-33.

8607: 

8608: %\bibitem[Ga5]{Ga5} D. Gabai, {\em $3$ lectures on foliations

8609: %	 and laminations on $3$-manifolds}, Laminations and foliations

8610: %in dynamics, geometry and topology (Stony Brook, NY 1998),

8611: %	87-109, Contemp. Math., 269, Amer. Math. Soc. Providence, 2001.

8612: 

8613: \bibitem[Ga-Oe]{Ga-Oe} D. Gabai and U. Oertel, {\em Essential

8614: 	laminations and $3$-manifolds}, Ann. of Math.

8615: 	{\bf 130} (1989) 41-73.

8616: 

8617: \bibitem[Ga-Ka]{Ga-Ka} D. Gabai and W. Kazez, {\em Group negative

8618: 	curvature for 3-manifolds with genuine laminations},

8619: 	Geom. Top. {\bf 2} (1998) 65-77.

8620: 

8621: \bibitem[Gh-Ha]{Gh-Ha} E. Ghys and P. De la Harpe, editors,

8622: 	{\em Sur les groupes hyperboliques d'apr\`{e}s

8623: 	Mikhael Gromov}, 

8624: 	Progress in Math. {\bf 83}, Birkh\"{a}user, 1990.

8625: 

8626: \bibitem[Go-Sh]{Go-Sh} S. Goodman and S. Shields, A condition for the

8627: 	stability of R-covered foliations of 3-manifolds, Trans. Amer.

8628: 	Math. Soc. {\bf 352} (2000) 4051-4065.

8629: 

8630: \bibitem[Gr]{Gr} M. Gromov, {\em Hyperbolic groups}, in Essays on

8631: 	Group theory, Springer-Verlag, 1987, pp. 75-263.

8632: 

8633: \bibitem[Han]{Han} M. Handel, {\em Global shadowing of pseudo-Anosov

8634: 	homemorphisms}, Ergod. Th. Dyn. Sys. {\bf 5} (1985) 373-377.

8635: 

8636: %\bibitem[Ha-Th]{Ha-Th} M. Handel and W. Thurston, {\em Anosov flows in new

8637: %3-manifolds}, Inven. Math. {\bf 59} (1980) 95-103.

8638: 

8639: \bibitem[He]{He} J. Hempel, {\em 3-manifolds}, Ann. of Math. Studies 86,

8640: 	Princeton University Press, 1976.

8641: 

8642: \bibitem[Im]{Im} H. Imanishi, {\em On the theorem of Denjoy-Sacsteder

8643: 	for codimension one foliations without holonomy},

8644: 	J. Math. Kyoto Univ. {\bf 14} (1974) 607-634.

8645: 

8646: \bibitem[In-Ma]{In-Ma} T. Inaba and S. Matsumoto, {\em  Non singular 

8647: 	expansive flows on 3-manifolds and foliations with circle

8648: 	prong singularities}, Japan. Jour. Math. {\bf 16} (1990) 329-340.

8649: 

8650: %\bibitem[Ja-Oe]{Ja-Oe} W. Jaco and U. Oertel, {\em An algorithm to 

8651: %	decide if a $3$-manifold is a Haken manifold}, Topology

8652: %	{\bf 23} (1984) 195-209.

8653: 

8654: \bibitem[Le]{Le} G. Levitt, {\em Foliations and laminations

8655: 	in hyperbolic surfaces}, Topology {\bf 22} (1983) 119-135.

8656: 

8657: \bibitem[Man]{Man} B. Mangum, {\em Incompressible surface and

8658: 	pseudo-Anosov flows}, Topol. Appl. {\bf 87} (1998) 29-51.

8659: 

8660: \bibitem[Mar]{Mar} A. Marden, {\em The geometry of finitely generated

8661: 	Kleinian groups}, Ann. of Math. {\bf 99} (1974) 383-462.

8662: 

8663: %\bibitem[Mas]{Mas} B. Maskit, {\em Kleinian groups}, Grundlehren {\bf 287}

8664: %	Springer 1987.

8665: 

8666: \bibitem[Me]{Me} G. Meigniez, {\em Bouts d'un groupe  op\'{e}rant sur la

8667: 	droite: 2. Applications \`{a} la topologie de feuilletages},

8668: 	T\^{o}hoku Math. Jour. {\bf 43} (1991) 473-500.

8669: 

8670: \bibitem[Mor]{Mor} J. Morgan, {\em On Thurston's uniformization 

8671: 	theorem for 3-dimensional manifolds}, in The Smith Conjecture,

8672: 	J. Morgan and H. Bass, eds., Academic Press, New York, 1984,

8673: 	37-125.

8674: 

8675: %\bibitem[Mo1]{Mo1} L. Mosher, {\em Dynamical systems and the homology norm

8676: %of a $3$-manifold I. Efficient interesection of surfaces and flows},

8677: %Duke Math. Jour.  {\bf 65} (1992) 449--500.

8678: 

8679: \bibitem[Mi]{Mi} Y. Minsky, {\em On Thurston's ending lamination conjecture},

8680: 	in Proc. Low-Dimensional topology, May 18-23, 1992, Inter. Press.

8681: 

8682: \bibitem[Mo1]{Mo1} L. Mosher, {\em Equivariant spectral decomposition

8683: 	for flows with a ${\bf Z}$-action}, Erg. Th. Dyn. Sys.

8684: 	{\bf 9} (1989) 329-378.

8685: 

8686: \bibitem[Mo2]{Mo2} L. Mosher, {\em Surfaces and branched surfaces

8687: 	transverse to pseudo-Anosov flows on $3$-manifolds},

8688: 	Jour. Diff. Geom. {\bf 34} (1990) 1-36.

8689: 

8690: \bibitem[Mo3]{Mo3} L. Mosher, {\em Dynamical systems and the homology 

8691:  	norm of a $3$-manifold II},

8692: 	Invent. Math. {\bf 107} (1992) 243--281.

8693: 

8694: \bibitem[Mo4]{Mo4} L. Mosher, {\em Examples of quasigeodesic flows on

8695: 	hyperbolic $3$-manifolds}, in Proceedings of the Ohio State

8696: 	University Research Semester on Low-Dimensional topology,

8697: 	W. de Gruyter, 1992.

8698: 

8699: \bibitem[Mo5]{Mo5} L. Mosher, {\em Laminations and flows transverse

8700: 	to finite depth foliations}, Part I: Branched surfaces and

8701: 	dynamics $-$ available from

8702: 	http://newark.rutgers.edu:80/~mosher/, 

8703: 	Part II in preparation.

8704: 

8705: %\bibitem[Mu]{Mu} J. Munkres, {\em A first course in Topology}, 

8706: %Prentice Hall, 2000.

8707: 

8708: 

8709: \bibitem[No]{No} S. P. Novikov, {\em Topology of foliations}, 

8710: 	Trans. Moscow Math. Soc. {\bf 14} (1963) 268-305.

8711: 

8712: \bibitem[Pa]{Pa} F. Palmeira, {\em Open manifolds foliated by 

8713: 	planes}, Ann. of Math. {\bf 107} (1978) 109-131.

8714: 

8715: \bibitem[Pe1]{Pe1} G. Perelman, {\em The entropy formulat for the 

8716: 	Ricci flow and its geometric applications}, eprint math.DG/0211159.

8717: 

8718: \bibitem[Pe2]{Pe2} G. Perelman, {\em Ricci flow with surgery on

8719: 	3-manifolds}, eprint math.DG/0303109.

8720: 

8721: \bibitem[Pe3]{Pe3} G. Perelman, {\em Finite extinction time for the 

8722: 	solutions to the Ricci flow in certain 3-manifolds},

8723: 	eprint math.DG/0307245.

8724: 

8725: \bibitem[Pl1]{Pl1} J. Plante, {\em Solvable groups acting on the line},

8726: 	Trans. A.M.S. {\bf 278} (1983) 401-414.

8727: 

8728: \bibitem[Pl2]{Pl2} J. Plante, {\em Anosov flows

8729: 	transversely affine foliations and a conjecture of

8730: 	Verjovsky}, Jour. London Math. Soc. {\bf 23} (1981) 359-362.

8731: 

8732: %\bibitem[Re]{Re} G. Reeb, {\em Sur certains propri\'{e}t\'{e}s

8733: %topologiques des vari\'{e}t\'{e}s feuillet\'{e}es},

8734: %Actual. Sci. Ind. {\bf 1183}, Hermann, Paris, 1952.

8735: 

8736: 

8737: 

8738: \bibitem[Ro1]{Ro1} R. Roberts, {\em Constructing taut foliations},

8739: 	Comm. Math. Helv. {\bf 70} (1995) 516-545.

8740: 

8741: \bibitem[Ro2]{Ro2} R. Roberts, {\em Taut foliations in punctured 

8742: 	surface bundles, I}, Proc. London Math. Soc. {\bf 82} (2001) 747-768.

8743: 

8744: \bibitem[Ro3]{Ro3} R. Roberts, {\em Taut foliations in punctured 

8745: 	surface bundles, II}, Proc. London Math. Soc. {\bf 83} (2001) 443-471.

8746: 

8747: \bibitem[Ro-St]{Ro-St} R. Roberts and M. Stein, {\em Group actions

8748: 	on order trees}, Topol. Appl. {\bf 115} (2001) 175-201.

8749: 

8750: \bibitem[Ros]{Ros} Rosenberg, {\em Foliations by planes},

8751: 	Topology {\bf 7} (1968) 131-138.

8752: 

8753: \bibitem[Th1]{Th1} W. Thurston, {\em The geometry and topology of 3-manifolds},

8754: 	Princeton University Lecture Notes, 1982.	

8755: 

8756: \bibitem[Th2]{Th2}  W. Thurston, {\em Three dimensional manifolds, Kleinian

8757: 	groups and hyperbolic geometry}, Bull. A.M.S. {\bf 6} (1982) 357-381.

8758: 

8759: \bibitem[Th3]{Th3}  W. Thurston, {\em Hyperbolic structures on $3$-manifolds II:

8760: 	Surface groups and $3$-manifolds that fiber over the circle}, preprint.

8761: 

8762: \bibitem[Th4]{Th4}  W. Thurston, {\em On the geometry and dynamics of

8763: 	diffeomorphisms of surfaces}, Bull. A.M.S. {19} (1988) 417-431.	

8764: 

8765: 

8766: \bibitem[Th5]{Th5}  W. Thurston, {\em 3-manifolds, foliations and circles I},

8767: 	preprint.

8768: 

8769: \bibitem[Th6]{Th6}  W. Thurston, {\em 3-manifolds, foliations and circles II:

8770: 	Transverse asymptotic geometry of foliations}, preprint.

8771: 

8772: \bibitem[Th7]{Th7}  W. Thurston, {\em The universal circle for foliations

8773: 	and transverse laminations}, lectures in M.S.R.I., 1997.

8774: 

8775: %\bibitem[Th7]{Th7}  W. Thurston, {\em The universal circle for foliations,

8776: %Preprint}

8777: 

8778: 

8779: %\bibitem[Wi]{Wi} R. Wilder, {\em Topology of Manifolds}, A.M.S. 

8780: %Colloquium Publ. {\bf 32} (1949).

8781: 

8782: \bibitem[Ze]{Ze} A. Zeghib, {\em Sur les feuilletages g\'{e}od\'{e}siques

8783: 	continus des vari\'{e}t\'{e}s hyperboliques}, Inven. Math.

8784: 	{\bf 114} (1993) 193-206.

8785: 

8786: 

8787: 

8788: }

8789: \end{thebibliography}

8790: 

8791: 

8792: 

8793: \noindent

8794: Florida State University

8795: 

8796: \noindent

8797: Tallahassee, FL 32306-4510, USA

8798: 

8799: }

8800: }

8801: 

8802: \end{document}

8803: 

8804: