1: % ----------------------------------------------

2: % Dressing preserving the fundametal group

3: %

4: % Josef Dorfmeister

5: % Martin Kilian

6: %

7: % May 08, 2003

8: % ----------------------------------------------

9: 

10: % debug flag

11: \newif\ifdebug

12: \debugfalse

13: 

14: % preliminary flag

15: \newif\ifpreliminary

16: \preliminarytrue

17: 

18: \documentclass[12pt,twoside]{article}

19: %\usepackage{a4Camb}

20: 

21: \usepackage{amsmath}

22: \usepackage{amstext}

23: \usepackage{amsbsy}

24: \usepackage{amssymb}

25: \usepackage{amsfonts}

26: \usepackage{latexsym}

27: 

28: \usepackage{graphicx}

29: %\usepackage[draft]{graphicx}

30: 

31: % gzipped images

32: \DeclareGraphicsRule{.eps.gz}{eps}{.eps.bb}{`gunzip -c #1}

33: 

34: \ifdebug

35:   \usepackage[notref,notcite]{showkeys}

36: \fi

37: \makeatletter

38: \@addtoreset{equation}{subsection}

39: \def\theequation{\thesubsection.\arabic{equation}}

40: \makeatother

41: 

42: \renewcommand{\thefootnote}{\fnsymbol{footnote}}

43: 

44:   \vskip6.94444pt

45: 

46: \newcommand{\separate}{\medskip\noindent}

47: 

48: \def\newsection{ \separate

49:    \refstepcounter{subsection}

50:    {\large\bf \thesubsection\kern.3em}

51: }

52: \def\mytheorem#1{

53:    \separate{\large\bf Theorem#1:\kern.3em}}

54: 

55: \def\mylemma#1{

56:    \separate{\large\bf Lemma#1:\kern.3em}}

57: 

58: \def\mycorollary#1{

59:    \separate{\large\bf Corollary#1:\kern.3em}}

60: 

61: \def\myprop#1{

62:    \separate{\large\bf Proposition#1:\kern.3em}}

63: 

64: \def\myremark#1{

65:    \separate{\large\bf Remark#1:\kern.3em}}

66: 

67: \def\mydefinition#1{

68:    \separate{\large\bf Definition#1:\kern.3em}}

69: 

70: \def\myexample#1{

71:    \separate{\large\bf Example#1:\kern.3em}}

72: 

73: \def\Proof{\separate\underline{Proof:}\kern1em}

74: 

75: \newcommand{\field}[1]{\mathbb{#1}}

76: \newcommand{\C}{\field{C}}

77: \newcommand{\D}{\field{D}}

78: \newcommand{\R}{\field{R}}

79: \newcommand{\Q}{\field{Q}}

80: \newcommand{\N}{\field{N}}

81: \newcommand{\Z}{\field{Z}}

82: \newcommand{\CPE}{{\mathchoice{\C{\rm P}_1}{\C{\rm P}_1}{\C\!\!

83: \mbox{\rm\tiny P}_1}{\C\!\!\mbox{\rm\tiny P}_1}}}

84: \newcommand{\ad}{{\mathchoice{\mbox{\rm ad}}{\mbox{\rm ad}}{%

85: \mbox{\scriptsize\rm ad}}{\mbox{\scriptsize\rm ad}}}}

86: \newcommand{\unity}{{\setlength{\unitlength}{1em}

87:                      \begin{picture}(0.75,1)

88:                      \put(0,0){$1$}

89:                      \put(0.34,0){\line(0,1){0.65}}

90:                      \end{picture}

91:                    }}

92: \newcommand{\X}{\mathbb X}

93: \newcommand{\Xh}{\widehat{\mathbb X}}

94: \newcommand{\T}{\mathbb T}

95: \newcommand{\hash}{\mbox{\tiny{\#}}}

96: \newcommand{\MT}{\mbox{\small{$\widetilde{M}$}}}

97: \newcommand{\s}{\mbox{\boldmath $\sigma$}}

98: \newcommand{\ci}{\mbox{\boldmath $i$}}

99: \newcommand{\DPWLab}{{\tt dpwlab}}

100: \newcommand{\CP}{\C \mathbb{P}^1}

101: \newcommand{\tr}{\mathrm{tr}}

102: \newcommand{\Id}{\mathrm{Id}}

103: \newcommand{\Order}{{\rm O}}

104: \newcommand{\FID}{\mathcal{F}_{\mbox{\tiny{$\Id$}}}

105: (\MT)}

106: 

107: \def\QED{\hfill$\Box$}

108: \def\inv{^{-1}}

109: \def\bref#1{(\ref{#1})}

110: \def\tmatrix#1#2#3#4{

111:     \left(\begin{array}{cc} #1 & #2 \\ #3 & #4 \end{array}\right)}

112: 

113: %MATRICES

114: 

115: \def\SL{{\mbox{\bf SL}(2,\C)}}

116: \def\SO{{\mbox{\bf SO}}}

117: \def\GL{{\mbox{\bf GL}(2,\C)}}

118: \def\SU{{\mbox{\bf SU}(2)}}

119: 

120: \newcommand{\Utwo}{\mbox{\bf U}(2)}

121: \def\U1{{\mbox{\bf U}(1)}}

122: 

123: \def\B{{\mbox{\bf B}}}

124: \def\G{{\mbox{\bf G}}}

125: \def\sl{{\mbox{\bf sl}(2,\C)}}

126: \def\gl{{\mbox{\bf gl}(2,\C)}}

127: \def\su{{\mbox{\bf su}(2)}}

128: 

129: \newcommand{\n}{\noindent}

130: 

131: %LOOPGROUPS

132: 

133: %GL splitting on r-circle

134: 

135: \def\LSL{{\Lambda \SL}}

136: \def\L{{\Lambda_r \GL_\sigma}}

137: \def\LP{{\Lambda_r^+ \GL_\sigma}}

138: 

139: \def\LU{{\Lambda_r \U_\sigma}}

140: \newcommand{\LB}{\Lambda_{r,\T}^{\mbox{\tiny{$+$}}} \GL _\sigma}

141: 

142: % SL loops on two different circles

143: \def\LGC{{\Lambda_r \SL_\sigma}}

144: \def\LGCs{{\Lambda_s \SL_\sigma}}

145: 

146: % Borel factor

147: \newcommand{\LGB}{\Lambda_{r,\T}^{\mbox{\tiny{$+$}}} \SL _\sigma}

148: 

149: % Birkhoff factors 

150: \def\LGP{{\Lambda_r^{\mbox{\tiny{$+$}}} \SL _\sigma}}

151: \def\LGM{{\Lambda_{r}^{\mbox{\tiny{$-$}}} \SL _\sigma}}

152: 

153: % SU loops

154: \def\LGU{{\Lambda_r \SU_\sigma}}

155: \def\LSU{{\Lambda \SU_\sigma}}

156: \def\Lhol{{\Lambda_{\mbox{\tiny{$\C^*$}}}\SU_\sigma}}

157: 

158: %LOOPALGEBRAS

159: \def\LAC{{\Lambda_r \sl _\sigma}}

160: \def\LAI{{\Lambda_r \sl _\sigma}}

161: \def\LAE{{\Lambda_r \sl _\sigma}}

162: \def\Lsu{{\Lambda \su _\sigma}}

163: \def\LGe{{\Lambda_r \su _\sigma}}

164: 

165: \newcommand{\delpot}{\xi^{\mbox{\tiny{\rm{Del}}}}}

166: \newcommand{\cylpot}{\xi_{\mbox{\tiny{\rm{cyl}}}}}

167: 

168: 

169: %DPW-POTENTIALS

170: \newcommand{\pot}{\Lambda \Omega(M)}

171: \newcommand{\pott}{\Lambda \Omega(\MT)}

172: \newcommand{\Lsl}{\Lambda \vspace{-.3mm}\hspace{-.2mm}^{\sigma}_{\mbox{\tiny{$-1,\infty$}}} \Sl}

173: 

174: \newcommand{\gauge}{\mathcal{G}_r (\MT)}

175: \newcommand{\gaugebase}{\mathcal{G}_{r,\T} (\MT)}

176: 

177: \def\u{{\mbox{\bf u}}}

178: \def\b{{\mbox{\bf b}}}

179: \def\g{{\mbox{\bf g}}}

180: \def\k{{\mbox{\bf k}}}

181: \def\p{{\mbox{\bf p}}}

182: \def\s{{\mbox{\bf $\sigma$}}}

183: 

184: \def\l{{\mbox{\footnotesize \sc l}}}

185: 

186: 

187: 

188: %\input epsf

189: 

190: \setlength{\topmargin}{-.6in}

191: \setlength{\textheight}{8.75in}

192: \ifdebug

193:   \setlength{\textwidth}{5in}

194: \else

195:   \setlength{\textwidth}{6in}

196:   \setlength{\oddsidemargin}{3.5ex}

197:   \setlength{\evensidemargin}{0 pt}

198: \fi

199: \setlength{\headsep}{.5in}

200: \setlength{\footskip}{.5in}

201: \setlength{\footskip}{.5in}

202: \setlength{\parskip}{1ex}

203: \setlength{\baselineskip}{1.2\baselineskip}

204: \setlength{\parindent}{0in}

205: 

206: 

207: % uses lower-case Roman numbers in 'enumerate'

208: \renewcommand{\theenumi}{\roman{enumi}}

209: \renewcommand{\labelenumi}{(\theenumi)}

210: \renewcommand{\theenumii}{\roman{enumii}}

211: \renewcommand{\labelenumii}{(\theenumii)}

212: %\renewcommand{p@enumii}{\theenumi--}

213: 

214: 

215: \begin{document}

216: 

217: \begin{center}

218: {\LARGE Dressing preserving the fundamental group}

219: 

220: \vskip1cm

221: \begin{minipage}{6cm}

222: \begin{center}

223: Josef~Dorfmeister \\ 

224: TU M\"{u}nchen\\ 

225: Zentrum Mathematik \\ 

226: Boltzmannstr. 3 \\ 

227: D-85747 Garching, Germany.

228: \end{center}

229: \end{minipage}

230: \begin{minipage}{6cm}

231: \begin{center}

232: Martin~Kilia$\mbox{n}^*$ \\ 

233: Mathematical Sciences \\

234: University of Bath \\ 

235: Claverton Down \\

236: Bath, BA2 7AY, UK.

237: \end{center}

238: \end{minipage}

239: \vspace{0.5cm}

240: 

241: \end{center}

242: \footnotetext[0]{1991 Mathematics Subject 

243: Classification. Primary 53A10; Secondary 53C20.\\

244:   $^*$ Supported by CEC Contract 

245: HPRN-CT-2000-00101 EDGE and EPSRC GR/S28655/01. \date{\today}}

246: 

247: \begin{abstract}

248: In this note we consider the 

249: relationship between the dressing action and 

250: the holonomy representation in the context of 

251: constant mean curvature surfaces. 

252: We characterize dressing elements that preserve 

253: the topology of a surface and discuss dressing 

254: by simple factors as a means of adding bubbles 

255: to a class of non finite type cylinders.

256: \end{abstract}

257: 

258: \refstepcounter{subsection}

259: 

260: \textbf{Introduction.} 

261: The equation for a harmonic map from a Riemann 

262: surface to a Riemannian symmetric space has a 

263: zero-curvature representation, and so 

264: corresponds to a loop of flat connections. 

265: Uhlenbeck discovered in her study \cite{Uhl} of 

266: harmonic maps into a compact Lie group 

267: $G$ that such maps correspond to certain holomorphic 

268: maps into the based loop group of $G$ and used 

269: this to define the \textit{dressing action} of 

270: a certain loop group on the space of harmonic maps. 

271: Dressing, or the \emph{vesture method} \cite{ZakS2}, 

272: was first developed in soliton theory to generate 

273: new solutions from old by solving a matrix Riemann 

274: problem. Unfortunately, the new solution does not 

275: automatically inherit properties of the old, 

276: such as domain, periods or asymptotics nor is 

277: it easy to control such properties when solving 

278: a matrix Riemann problem. 

279: 

280: When the target is the two dimensional 

281: round sphere, harmonic maps are precisely 

282: the Gauss maps of surfaces with constant mean 

283: curvature \cite{RuhV}. In the article \cite{DorPW} 

284: a method was presented by which all conformally 

285: immersed surfaces with constant mean curvature 

286: (CMC) can be obtained. 

287: The construction involves solving a meromorphic 

288: linear differential system with values in 

289: a loop group and then Iwasawa decomposing the 

290: solution to obtain the extended unitary frame of 

291: the surface. Both these steps make it 

292: difficult to keep track of the topology. 

293: Variation of the initial condition 

294: corresponds to the dressing action and is 

295: an integral part of the theory. 

296: Not only can dressing be used to close periods, but 

297: also to generate new CMC surfaces from old ones 

298: without altering the topology. In this work, 

299: we look into the relationship 

300: between dressing and topology in the framework of 

301: meromorphic ODE's and loop group factorizations 

302: in the context of CMC surfaces.

303:  

304: The question, when a conformal CMC immersion 

305: factors through a given surface has been discussed 

306: for tori \cite{Bob:tor}, \cite{DorH:per}, 

307: \cite{PinS}, cylinders \cite{DorH:cyl}, 

308: \cite{Kil:thesis}, 

309: \cite{KilMS} and trinoids \cite{BobPS}, 

310: \cite{DorW:noids}, \cite{KilMS}, \cite{Sch:tri}. 

311: It will be interesting to see more results in 

312: this direction, with particular emphasis on 

313: complete, non-compact surfaces with genus and 

314: compact surfaces of genus $g \geq 2$.

315: 

316: Sterling \& Wente \cite{SteW} discovered that 

317: one can add ``bubbles'' to a round cylinder 

318: while preserving the mean curvature by 

319: Bianchi-- B\"{a}cklund transformations. 

320: In \cite{Kil:thesis} it was shown that these 

321: ``bubbletons'' can be produced by dressing 

322: from the standard round cylinder by the so called 

323: {\em simple factors} of Terng and Uhlenbeck 

324: \cite{TerU}. Subsequently, it was shown in 

325: \cite{Kob} that the simple bubbletons in 

326: \cite{Kil:thesis} coincide with those of Sterling 

327: and Wente \cite{SteW}. Recently, A. Mahler 

328: \cite{Mah} has shown that indeed 

329: Bianchi -- B\"{a}cklund transformations  

330: can be achieved by dressing. Further, it was 

331: recently shown in \cite{KilSS} that it is 

332: also possible to dress CMC surfaces homeomorphic 

333: to the n-punctured sphere with suitably chosen 

334: simple factors without changing the topology.

335:  

336: Every CMC immersion of a Riemann surface $M$ induces

337: a 'monodromy representation' 

338: $\chi = \chi (\gamma,\,\lambda )$ of the 

339: fundamental group $\pi_1 (M)$ of $M$ with values 

340: in a unitary loop group. Moreover, as a function 

341: of the loop parameter $\lambda$, the monodromy 

342: matrices are holomorphic on $\C^*$. 

343: Dressing a given immersion with some arbitrary 

344: matrix will in general destroy the topology and 

345: the dressed surface will have a trivial 

346: fundamental group. Thus the question is: 

347: For which dressing matrices will the dressed 

348: immersion have the same fundamental group?

349: 

350: To explain this in more detail we note that in 

351: our method we associate with a given CMC immersion 

352: $f :M  \rightarrow \R^3$ its associated family 

353: $f_\lambda :\MT \to \R^3$, where $\MT$ 

354: denotes the universal cover of $M$ and 

355: $\lambda \in S^1$. This way we obtain the extended 

356: frame $F = F(z,\, \bar{z},\, \lambda )$ for 

357: $\lambda \in S^1$ and $z \in \MT$.

358: If $\gamma \in \pi_1(M)$ and we also denote the 

359: corresponding deck transformation by $\gamma$ and 

360: write 

361: $\gamma^* F = F(\gamma(z),\,\overline{\gamma(z)},\,\lambda)$  

362: then we have $\gamma^*F = \chi\,F\,k$ for some smooth 

363: $k:\MT \to \U1$. 

364: Since, by assumption, the immersion $f_1 = f$ 

365: descends to $M$, the monodromy matrices satisfy 

366: the two closing conditions 

367: $ \chi (\gamma,\, 1) = \pm \Id$ and 

368: $\left.\partial_{\lambda} \chi \right|_{\lambda = 1}

369:  = 0$ for all $\gamma \in \pi_1(M)$.

370: 

371: If $h = h(\lambda )$ is some dressing element 

372: and $\hat F$ denotes the dressed extended 

373: frame and we write $hF = \hat{F} V_+$ for 

374: the dressing equation, then $\gamma^* \hat{F} = 

375: h\,\chi\,h^{-1}\,\hat{F}\, V_+\,\gamma^* V_+^{-1}$. 

376: Thus $\hat{F}$ has some monodromy 

377: $\hat{\chi}$ if and 

378: only if $\hat{L} \hat{F} = \hat{F} \hat{W}_+$, 

379: where $\hat{W}_+ = 

380: V_+ \, \gamma^* V_+ ^{-1}$ and 

381: $\hat{L} = ( h \chi h^{-1})^{-1}

382: \hat{\chi}$. As a consequence, 

383: $\hat{\chi} = h \chi h^{-1} \hat{L} = 

384: h \chi L h^{-1}$, where $L = h^{-1} \hat{L} h$.

385: 

386: We prove 

387: %in Lemma \ref{th:iso_problem} 

388: that 

389: a dressing matrix $h$ preserves the fundamental 

390: group of an immersion if and only if for 

391: every $\gamma \in \pi_1 (M)$ there exists some 

392: matrix $L = L(\gamma,\,\lambda)$, such that 

393: 

394: (i) $LF = F W_+$,\vspace{-1mm}

395: 

396: (ii) $h \chi h^{-1}$ is holomorphic for 

397:   $\lambda \in \C^*$,\vspace{-1mm} 

398: 

399: (iii) $h \chi L h^{-1}$ is unitary on $S^1$, 

400:   and \vspace{-1mm}

401: 

402: (iv) The closing conditions are satisfied for 

403:   $h\chi L h^{-1}$.

404: 

405: Clearly, the situation simplifies considerably, 

406: if the original immersion has an umbilic point, 

407: since then $L = \pm \Id$ \cite{DorH:dre}. 

408: Let us assume this for now. 

409: Considering in this case condition (ii) we see 

410: that $\chi$ is holomorphic for 

411: $\lambda \in \mathbb{C}^*$ and also 

412: $h \chi h^{-1}$ needs to be holomorphic on 

413: $\mathbb{C}^*$, in spite of the fact

414: that $h$ may only be defined on some circle 

415: of radius $ 0 < r \leq 1$. 

416: We prove 

417: %in Theorem \ref{th:h-split} 

418: that in 

419: the situation discussed here we can assume that 

420: $h = \mathcal{M}\,\mathcal{C}$, where 

421: $[\,\mathcal{C}, \,\chi \,] = 0$ and $\mathcal{M}$ 

422: is meromorphic on some open dense subset 

423: $\mathbb{S} \subset \mathbb{C}^*$, 

424: which contains an open annulus about $S^1$.

425: Moreover, $\mathbb{S}$ is solely defined 

426: by the eigenvalues of $\chi$.

427: In view of condition (iii) we further show 

428: %in Theorem \ref{th:h-split1} 

429: that actually 

430: $\mathcal{M}$ can be chosen so that it is 

431: unitary on $S^1$.

432: 

433: This gives altogether a fairly complete 

434: description for the case, where an umbilic 

435: point exists. Similarly complete is the case, 

436: dealt with in Theorem \ref{th:h-split_cyl}, 

437: where $\chi$ is the monodromy of the standard 

438: round cylinder. For the general umbilic free 

439: case we have, at this point, only some partial 

440: results.

441: 

442: Let us briefly outline the contents of this paper. 

443: The first chapter sets the scene by defining the 

444: relevant loop groups, dressing and gauge actions 

445: and recalls the DPW representation \cite{DorPW} 

446: of CMC surfaces. 

447: In the second chapter we consider how automorphisms 

448: affect maps at the various levels of the DPW 

449: construction as a prerequisite to understanding 

450: monodromy. In chapter 3 we present a factorization 

451: as a necessary condition on the dressing matrix 

452: to ensure that the dressed surface retains the 

453: topology of the original surface. 

454: In the fourth chapter we apply our 

455: methods to the dressing orbit of the vacuum 

456: and give a self contained account of twisted 

457: simple factors. We conclude this work by 

458: discussing how dressing with simple factors is 

459: related to the monodromy and apply these methods 

460: to a class of CMC cylinders with umbilics.

461: 

462: The second author wishes to thank F Burstall for 

463: many useful discussions and the Department of 

464: Mathematical Sciences at the University of Bath 

465: for its hospitality.

466: 

467: \section{Notation and basic results}\label{basics}

468: \message{[basics]} 

469: We begin by collecting some well known results 

470: on loop groups and the dressing action. 

471: We shall use the following notation 

472: for diagonal and off--diagonal matrices:

473: %

474: \begin{equation*}

475: \mathrm{diag}[u,\, v] = 

476: \bigl( \begin{smallmatrix} 

477: u & 0 \\ 0 & v \end{smallmatrix} \bigr),\,

478: \mathrm{off}[u,\, v] = 

479: \bigl( \begin{smallmatrix} 

480: 0 & u \\ v & 0 \end{smallmatrix} \bigr). 

481: \end{equation*}

482: %

483: \newsection{\bf Loops.} \label{loopgroups}

484: For real $r \in (0,1]$, let 

485: $C_r = \{ \lambda \in \C : | \lambda | = r \}$ 

486: and denote the $r$--Loop group of $\SL$ by

487: %

488: \begin{equation*}

489:   \Lambda_r \SL = \left\{ g : C_r \to \SL

490:   \mbox{ smooth} \right\}.

491: \end{equation*}

492: %

493: We have an involution on maps $C_r \to \gl$ 

494: defined by 

495: %

496: \begin{equation} \label{eq:sigma}

497:   \sigma : g(\lambda) \mapsto 

498:   \sigma_3 g(-\lambda) \sigma_3^{-1},\,

499:   \sigma_3 = \mathrm{diag}[1,-1],

500: \end{equation}

501: % 

502: and denote the \textit{twisted} $r$--Loop group 

503: of $\SL$ by

504: %

505: \begin{equation*}

506:   \LGC = \left\{ g \in \Lambda_r \SL: 

507:   \, \sigma g = g \right\}.

508: \end{equation*}

509: %

510: Analogously, one can define the Lie algebras of 

511: these groups, denoted by

512: $\LAC$. To make these loop groups 

513: complex Banach Lie groups, we equip them, 

514: as in \cite{DorPW},

515: with some $H^s$ topology for $s>1/2$ or some 

516: (possibly weighted) Wiener topology. 

517: Elements of these twisted loop

518: groups are matrices whose off--diagonal entries 

519: are odd functions, 

520: while the diagonal entries are even functions of 

521: the parameter $\lambda$. 

522: All entries are in the Banach algebra 

523: ${\cal A}_r$ of

524: $H^s$--smooth functions or of finite 

525: (possibly weighted) Wiener norm on $C_r$.

526: Furthermore, we will use the following subgroups 

527: of $\LGC$: Let 

528: $I_r = \{ \lambda \in \C : |\lambda|<r \}$ 

529: and denote

530: %

531: \begin{equation*}

532:   \LGP = \left\{ g \in \LGC : g 

533:   \text{ extends analytically to } I_r \right\}.

534: \end{equation*}

535: %

536: Let $A_r = \{ \lambda \in \C : 

537: r<|\lambda|<1/r \}$, and by abuse of 

538: notation we denote 

539: %

540: \begin{equation*}

541:   \LGU = \left\{ g \in \LGC :  g 

542:   \text{ extends analytically to $A_r$ 

543:   and}\left. g \right|_{S^1} \in \SU \right\}.

544: \end{equation*}

545: %

546: Let $g:A_r \to \SL_\sigma$ be analytic. 

547: Then $g \in \LGU$ if and only if 

548: $\varrho \, g = g$ for 

549: %

550: \begin{equation} \label{eq:varrho}

551:   (\varrho g)(\lambda) :=

552:   \overline{g(1/\bar{\lambda})}^{t-1}.

553: \end{equation}

554: %

555: or alternatively, $g \in \LGU$ if and only if 

556: %

557: \begin{equation} \label{eq:unitarity_def}

558:   (g^*)(\lambda) := 

559:   \overline{g(1/\bar{\lambda})}^t = 

560:   g(\lambda)^{-1}.

561: \end{equation}    

562: %

563: For $r=1$ we will always omit the subscript '$r$'.

564: All the groups above are Banach Lie subgroups 

565: of $\LGC$. The corresponding Banach Lie 

566: sub algebras of $\LAC$ are defined analogously. 

567: 

568: 

569: %%%%%%%%%%%

570: 

571: \newsection{\bf Iwasawa decomposition.} 

572: Of fundamental importance in our investigation 

573: is a certain Loop group factorization. 

574: We modify a result from \cite{McI} to obtain:

575: 

576: \separate  Multiplication 

577: $\LGU \times \LGP \to \LGC$

578: is a real analytic surjection. 

579: An associated splitting $g = FB$ 

580: of $g \in \LGC$ with $F \in \LGU$ and $B \in \LGP$ 

581: will be called an Iwasawa decomposition of $g$. 

582: Note that

583: %

584: \begin{equation} \label{eq:intersect}

585:   \LGU \cap \LGP = \U1.

586: \end{equation}

587: %

588: If we demand that $B(\lambda=0)$ has positive real 

589: diagonal entries, then the multiplication map 

590: above is a real analytic diffeomorphism onto.

591: We will call this decomposition the "unique" Iwasawa

592: decomposition.

593: 

594: %%%%%%%%%%%

595: 

596: \newsection{\bf Untwisted loops.} 

597: We shall be primarily working with the twisted 

598: loop groups and algebras. On occasion it is 

599: advantageous to switch to the untwisted setting 

600: via the isomorphism between untwisted and twisted 

601: $r$--loops: 

602: let $X_u \in \Lambda_r \SL$ be an 

603: untwisted loop. 

604: Then the corresponding twisted loop is given by 

605: $X_t(\lambda) = D\,X_u(\lambda^2)\,D^{-1}$ 

606: where $D=\mathrm{diag}[\,\sqrt{\lambda},\,

607: 1/\sqrt{\lambda}\,]$. In matrix notation, this 

608: isomorphism is given by

609: %

610: \begin{equation} \label{eq:twist}

611:   \begin{pmatrix} a(\lambda) & b(\lambda) \\

612:   c(\lambda) & d(\lambda) \end{pmatrix} \cong 

613:   \begin{pmatrix} a(\lambda^2) & 

614:   \lambda\, b(\lambda^2) \\

615:   \lambda^{-1} c(\lambda^2) & 

616:   d(\lambda^2)\end{pmatrix}. 

617: \end{equation}

618: % 

619: One technical issue in switching between untwisted 

620: and twisted loops is that under this twisting 

621: isomorphism, an untwisted positive loop is not 

622: automatically positive when twisted, since the 

623: untwisted loop at $\lambda = 0$ need not be 

624: upper triangular. To circumvent this issue, 

625: we must first perform a Gram-Schmidt 

626: orthonormalization at $\lambda = 0$ 

627: before twisting the loop. 

628: 

629: More precisely, consider an untwisted loop 

630: $\Phi_u \in \Lambda_r\SL$ with an $r$-Iwasawa 

631: decomposition  $\Phi_u = F_u\,B_u$ 

632: with untwisted $F_u \in \Lambda_r \SU$ and 

633: $B_u \in \Lambda_r^+ \SL$. Now consider the 

634: twisted loop $\Phi_t(\lambda) = D\,\Phi_u(\lambda^2)\,D^{-1}$ 

635: obtained from $\Phi_u$ via the 

636: isomorphism \eqref{eq:twist} and let 

637: $\Phi_t = F_t\,B_t$ be an $r$-Iwasawa decomposition 

638: with twisted $F_t \in \LGU$ and $B_t \in \LGP$. 

639: If $B_u$ has an expansion 

640: $B_u = B_0 + \lambda B_1 + \ldots$, 

641: then as $B_0 \in \SL$, we may write  

642: $B_0 = Q\,R$ via Gram-Schmidt orthonormalization 

643: with $Q \in \SU$ and $R \in \SL$ upper triangular. 

644: Consequently, $Q^{-1}B_u \in \Lambda_r^+\SL$ and 

645: $\left.Q^{-1}B_u \right|_{\lambda=0} = R$ and thus 

646: $D\,Q^{-1}B_u(\lambda^2)\,D^{-1} \in \LGP$. 

647: Furthermore, we then have that 

648: %

649: \begin{equation}

650:   \Phi_t(\lambda) = \left( D\,F_u(\lambda^2) 

651:   \,Q\,D^{-1} \right) \left( D\,Q^{-1} 

652:   B_u(\lambda^2) \,D^{-1} \right)

653: \end{equation}

654: % 

655: with $D\,F_u(\lambda^2)\,Q\,D^{-1} \in \LGU$ and 

656: $D\,Q^{-1}B_u(\lambda^2)\,D^{-1} \in \LGP$ is an 

657: $r$-Iwasawa decomposition. It is in this sense that 

658: we may carry an Iwasawa decomposition in the 

659: untwisted setting over to the twisted setting.  

660: 

661: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

662: 

663: \newsection{\bf DPW method.} \label{sec:DPW}

664: Let $\Omega(M)$ denote the holomorphic 

665: $1$--forms on a Riemann surface $M$ and define 

666: the 

667: %

668: \begin{equation*}

669:   \pot = \Omega(M) \otimes \left\{ 

670:   \xi : \C^* \to  \sl 

671:   \mbox{ holomorphic}: \xi(\lambda) = 

672:   \sum_{j \geq -1} 

673:   \xi_j \lambda^j,\,\s \xi = \xi \right\}.

674: \end{equation*}

675: %

676: CMC surfaces come in $S^1$ families, 

677: the \textit {associated family}. 

678: The DPW representation \cite{DorPW} constructs 

679: all conformal CMC immersions of the universal 

680: cover $\MT$ in the following three steps: 

681: Let $\xi \in \pott$, $\tilde{z}_0 \in 

682: \MT$ and $\Phi_0 \in \LGC$ for some $r \in (0,1]$. 

683: To avoid totally umbilical surfaces, we assume 

684: $\det \xi_{-1} \not \equiv 0$.

685: 

686: \noindent 1. Solve the initial value problem 

687: %

688: \begin{equation}\label{eq:IVP}

689:   d\Phi = \Phi \xi,\, \Phi(\tilde{z}_0)= \Phi_0

690: \end{equation}

691: %

692: to obtain a unique \textit{holomorphic frame} 

693: $\Phi : \MT \to \LGC$.

694: 

695: \noindent 2. Iwasawa decompose the map 

696: $\Phi : \MT \to \LGC$ point-wise on $\MT$ 

697: %

698: \begin{equation} \label{eq:Iwasawa}

699:   \Phi(z,\,\lambda) = 

700:   F(z,\,\bar{z},\,\lambda) \, 

701:   B(z,\,\bar{z},\,\lambda) 

702: \end{equation}

703: %

704: We will always assume 

705: that the factors $F$ and $B$ in \eqref{eq:Iwasawa} 

706: are real analytic in $z$. 

707: The map $F : \MT \to \LGU$ will be called 

708: \textit{unitary frame}. If we use the unique 

709: Iwasawa decomposition, then the factors $F$ and $B$ 

710: are automatically real analytic.

711: 

712: 

713: \noindent 3. Let $H \in \R^*$ and 

714: $\partial_\lambda = 

715: \tfrac{\partial}{\partial \lambda}$. 

716: Plug $F$ into the Sym-Bobenko formula 

717: %

718: \begin{equation} \label{eq:Sym}

719:   f_\lambda = - \tfrac{1}{2H}\left( 

720:   i\lambda \tfrac{\partial F}{\partial\lambda} 

721:   \,F^{-1} + \tfrac{i}{2} F \sigma_3 F^{-1} \right)

722: \end{equation}

723: %

724: to obtain (possibly branched) conformal 

725: CMC immersions $f_\lambda :\MT \to \Lsu$ with mean 

726: curvature $H$, that is, for each 

727: $\lambda_0 \in S^1$ we have a conformal CMC 

728: immersion $f_{\lambda_0} :\MT \to \su \cong \R^3$. 

729: Note, $f_\lambda (z)$ is branched at $z_0$ if 

730: and only if for $\xi_{-1} = \mathrm{off}[a,\,b]$ we 

731: have $a(z_0)=0$.

732: 

733: \newsection{\bf Frames.} We call $\xi \in \pott$ a 

734: \textit{holomorphic potential}. 

735: If $\Phi_0 = \phi_u \phi_+$ is an Iwasawa 

736: decomposition, then it is not hard to see that 

737: the immersions obtained from 

738: $(\xi,\Phi_0,\tilde{z}_0)$ and 

739: $(\xi,\phi_+,\tilde{z}_0)$ 

740: differ by a $\lambda$-dependent rigid motion. 

741: Thus the choice of initial 

742: condition may be restricted to $\LGP$. 

743: Further, if $\Phi_0 \in \LGP$ 

744: then $F(\tilde{z}_0) \in \U1$ and 

745: $\Phi = F F(\tilde{z}_0)^{-1} F(\tilde{z}_0)B$. 

746: Hence we may assume without loss of generality that

747: %

748: \begin{equation} \label{eq:F_initial}

749:   F(\tilde{z}_0) \equiv \Id \mbox{ for all }

750:   \lambda \in S^1.

751: \end{equation}

752: %

753: Let $\sl = \mathfrak{k} \oplus \mathfrak{p}$ 

754: be the Cartan decomposition induced by the 

755: involution $\s$ defined in \eqref{eq:sigma}. 

756: The specific form of $\xi$ ensures that the 

757: Maurer--Cartan form of $F$ acquires the form

758: %

759: \begin{equation} \label{eq:MC}

760:   F^{-1}dF = 

761:   \lambda^{-1} \alpha_{\mathfrak{p}}^{\prime} + 

762:   \alpha_{\mathfrak{k}} + 

763:   \lambda \, \alpha_{\mathfrak{p}}^{\prime\prime}  

764: \end{equation}

765: %

766: where $\alpha_{\mathfrak{p}}^\prime$ and 

767: $\alpha_{\mathfrak{p}}^{\prime\prime}$ are 

768: $(1,\,0)$ respectively $(0,1)$-forms on $\MT$ 

769: with values in $\mathfrak{p}$ and 

770: $\alpha_{\mathfrak{k}}$ takes values in 

771: $\mathfrak{k}$. 

772: For each $p \in \MT$, 

773: $\lambda \mapsto F(p,\lambda)$ is 

774: holomorphic on $\C^*$. Maps 

775: %

776: \begin{equation}

777:   F:\MT \to \Lhol := 

778:   \bigcap_{r \in (0,1]}\LGU

779: \end{equation}

780: %

781: with the property \eqref{eq:MC} are called 

782: \textit{extended unitary frames} 

783: and denoted by $\mathcal{F}(\MT)$.

784: We call the extended unitary frames 

785: for which \eqref{eq:F_initial} holds the 

786: \textit{normalized extended unitary frames} 

787: and denote these by

788: %

789: \begin{equation}

790:   \FID = \left\{ 

791:   F \in \mathcal{F}(\MT) : 

792:   F(\tilde{z}_0) = \Id \mbox{ for some } 

793:   \tilde{z}_0 \in \MT \right\}.

794: \end{equation}

795: %

796: 

797: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

798: 

799: \newsection{\bf Gauge.} \label{Gauge}

800: The DPW representation $(\xi,\Phi_0,\tilde{z}_0) 

801: \mapsto \mathcal{F}(\MT)$ is a surjective map 

802: \cite{DorPW}. Injectivity fails since the 

803: \textit{Gauge group} 

804: %

805: \begin{equation}

806:   \gauge = \{ G : \MT

807:   \to \LGP \mbox{ holomorphic } \},

808: \end{equation}

809: % 

810: acts by right multiplication on the fibers of 

811: this map. On the level of the potential, this 

812: \textit{gauge action} is given by 

813: %

814: \begin{equation}\label{eq:gauge}

815:   \xi.G = G^{-1} \xi \ G + G^{-1} dG.

816: \end{equation}

817: %

818: A computation shows that if $\Phi$ solves 

819: \eqref{eq:IVP} with triple 

820: $(\xi,\Phi_0 ,\tilde{z}_0 )$ and 

821: $G \in \gauge$ then the triples 

822: $(\xi,\Phi_0 ,\tilde{z}_0 )$ and 

823: $(\xi.G,\Phi_0 G(\tilde{z}_0),\tilde{z}_0 )$ 

824: induce the same CMC immersions, assuming $H \neq 0$.

825: 

826: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

827: 

828: \newsection{\bf Dressing.} \label{Dressing}

829: For $r \in (0,1]$, $h \in \LGC$ and 

830: $F_0 \in \mathcal{F}(\MT)$ let

831: %

832: \begin{equation}\label{eq:dress}

833:   hF_0 = F \, B 

834: \end{equation}

835: %

836: be a point-wise Iwasawa decomposition of $hF_0$ in 

837: $\LGC$, where every factor is analytic in $z$. Then 

838: $F \in \mathcal{F}(\MT)$.

839: 

840: Note that $F:\MT \to \LGU$ and 

841: $B:\MT \to \LGP$ are not uniquely determined, 

842: since for any smooth $U:\MT \to \U1$, we again 

843: have an Iwasawa decomposition given by 

844: $hF_0 = FU \,U^{-1}B$. We shall write 

845: %

846: \begin{equation}

847:   F \in [h\#F_0]

848: \end{equation}

849: %

850: to signify that $F:\MT \to \LGU$ satisfies 

851: \eqref{eq:dress}. We call $[h\#F_0]$ the 

852: \textit{dressing class} of $F_0$ under $h$ 

853: and say that $F \in [h\#F_0]$ was obtained 

854: by \textit{dressing} $F_0 \in \mathcal{F}(\MT)$ 

855: by $h \in \LGC$. 

856: 

857: To preserve the base point condition 

858: \eqref{eq:F_initial}, one needs to restrict 

859: to dressing with elements $h \in \LGP$.

860:  

861: \mylemma{}\label{th:initial} 

862: Let $h \in \LGC$ and $F_0 \in \FID$ be a normalized 

863: extended unitary frame. Then there exists a 

864: normalized extended unitary frame in $[h\#F_0]$ if 

865: and only if $h \in \LGP$.

866: 

867: \Proof A straightforward computation, 

868: see Proposition 2.9 of \cite{BurP:dre}, 

869: shows that the Maurer--Cartan 

870: form of any element in $[h\#F_0]$ 

871: is again of the form \eqref{eq:MC}. 

872: The issue is that for

873: $F_0 \in \FID$ we have 

874: $F_0(\tilde{z}_0) = \Id$, while for 

875: $F \in [h\#F_0]$ we have apriori only 

876: $F(\tilde{z}_0) \in \U1$. 

877: 

878: If $hF_0 = FB$ is an Iwasawa decomposition, 

879: then $F_0(\tilde{z}_0) = F(\tilde{z}_0) = \Id$, 

880: implies $h = B(\tilde{z}_0) \in \LGP$. 

881: 

882: Conversely, if $hF_0 = FB$ is an 

883: Iwasawa decomposition, then so is

884: %

885: \begin{equation}

886:   hF_0 = F F(\tilde{z}_0)^{-1} F(\tilde{z}_0)B.

887: \end{equation}

888: %

889: Hence $F F(\tilde{z}_0)^{-1}\in [h\#F_0]$ and 

890: satisfies \eqref{eq:F_initial}. Thus 

891: $F F(\tilde{z}_0)^{-1} \in \FID$. \QED

892: 

893: We thus have a left action of $\LGP$ 

894: on $\FID$ and shall write 

895: $F = h \# F_0$ to signify that $F \in [h\#F_0]$ 

896: with $F(\tilde{z}_0) = \Id$. Evaluating $hF_0 = 

897: h\#F_0 \, B$ at $\tilde{z}_0$ also gives 

898: $h=B(\tilde{z}_0)$. 

899: %

900: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

901: 

902: \newsection{\bf Isotropies.}

903: We denote the isotropy group of a map

904: $F: \MT \to \LGU$ under dressing by 

905: %

906: \begin{equation}

907:   \mathrm{Iso}_r(F) = \left\{ h \in \LGC : 

908:   F \in [h\# F] \right\}.

909: \end{equation} 

910: %

911: For a holomorphic frame $\Phi:\MT \to \LGC$ 

912: we define the isotropy under dressing by

913: %

914: \begin{equation}

915:   \mathrm{Iso}_r(\Phi) = \{ g \in \LGC : 

916:   g \, \Phi = \Phi \, G \mbox{ for some } G 

917:   \in \mathcal{G}_r(\MT) \}.

918: \end{equation}

919: %

920: If $g \in \mathrm{Iso}_r(\Phi)$ with  

921: $g\Phi = \Phi G$ and $\Phi = F\,B$ 

922: is an Iwasawa decomposition of $\Phi$, 

923: then $gF = FBGB^{-1}$ with $BGB^{-1} \in \LGP$. 

924: Hence $g \in \mathrm{Iso}_r(F)$. Conversely, if 

925: $h \in \mathrm{Iso}_r(F)$, and $hF=FH$ is an Iwasawa 

926: decomposition, then $h\Phi = \Phi B^{-1}HB$ 

927: and thus 

928: %

929: \begin{equation} \label{eq:isotropies}

930:   \mathrm{Iso}_r(\Phi) = \mathrm{Iso}_r(F).

931: \end{equation}

932: %

933: Further, using \eqref{eq:intersect} it is 

934: straightforward to verify that 

935: %

936: \begin{equation} \label{eq:iso_intersect}

937:   \mathrm{Iso}_r(F) \cap \LGU = \U1 \mbox{ for } 

938:   F \in \FID.

939: \end{equation}

940: %

941: Umbilic points occur naturally on compact 

942: CMC surfaces of genus $g \geq 2$ as well 

943: as on complete, open CMC surfaces with more than 

944: two ends. In this context, we quote 

945: 

946: \mytheorem{ \cite{DorH:dre}}

947: \label{th:isotropy}

948: If the associated family of 

949: $F \in \mathcal{F}(\MT)$ has umbilics, 

950: then $\mathrm{Iso}_r(F) = \{ \pm \Id \}$ for all 

951: $r \in (0,\,1]$. \QED

952: 

953: \section{Symmetries}

954: % 

955: The notion of symmetry for CMC immersions has been 

956: discussed in the articles 

957: \cite{DorH:per,DorH:sym2}. 

958: It turns out that symmetry can be defined on 

959: various levels in the context of the DPW 

960: representation.

961: The problem of dealing with symmetries in the 

962: DPW representation stem mostly from the fact 

963: that symmetries on the potential level are, 

964: by and large, defined as coming from symmetries 

965: on the immersion level and have not yet been 

966: defined completely intrinsically on the potential 

967: level. Since symmetries associated with 

968: automorphisms $Aut(\MT)$ of the universal 

969: cover are central to this article, we recall 

970: the basic facts, retaining the notation of 

971: section \ref{sec:DPW}. 

972: 

973: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

974: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

975: 

976: \newsection{\bf Definitions.} \label{symmetries}

977: Let $f : \MT \to \Lsu$ be a CMC $H \neq 0$ immersion 

978: generated by a triple $( \xi, \Phi_0, \tilde{z}_0 )$ 

979: with $\Phi$ its holomorphic frame and $F$ its 

980: unitary frame. In the rest of 

981: this section we will always assume 

982: $\gamma \in Aut(\MT)$ and for

983: a map G with domain $\MT$ we shall write 

984: $\gamma^* G = G \circ \gamma$.  

985: 

986: \n {\bf{Immersion level}}: 

987: A triple 

988: $(\gamma,\,\X,\,T)$ with 

989: $\X \in \LGU$ and $T \in \LGe$ 

990: is called a symmetry of $f$ if and only if 

991: %

992: \begin{equation}\label{eq:sym_immersion}

993:   \gamma^* f = \X \, f \, \X^{-1} 

994:   + T.

995: \end{equation}

996: % 

997: 

998: \n {\bf{Unitary frame level}}: 

999: A pair $(\gamma,\,\X)$ with 

1000: $\X \in \LGU$ 

1001: is called a symmetry of $F$ if and 

1002: only if there exists a smooth map 

1003: $k \in C^\infty(\MT,\U1)$ such that  

1004: %

1005: \begin{equation}\label{eq:sym_frame}

1006:   \gamma^* F = \X \,F \,k.

1007: \end{equation}

1008: %

1009: 

1010: \n {\bf{Holomorphic frame level}}: 

1011: A pair $(\gamma,\,\X)$ with 

1012: $\X \in \LGU$ 

1013: is a symmetry of $\Phi$ if and only if 

1014: there is a map $H \in \gauge$ such that

1015: %

1016: \begin{equation}\label{eq:sym_curve}

1017:   \gamma^* \Phi = \X \,\Phi \,H.

1018: \end{equation}

1019: % 

1020: 

1021: \n {\bf{Potential level}}: 

1022: A pair $(\gamma,\,G)$ with 

1023: $G \in \gauge$ is called a 

1024: symmetry of $\xi$ if and only if 

1025: %

1026: \begin{equation}\label{eq:sym_potential}

1027:   \gamma^* \xi = \xi.G \,.

1028: \end{equation}

1029: % 

1030: where $\xi.G$ denotes the gauge 

1031: transformation of $\xi$ by $G$ as defined in 

1032: \eqref{eq:gauge}.

1033: á

1034: 

1035: \mytheorem{}\label{th:symmetries} 

1036: Let $f : \MT \to \Lsu$ be a CMC $H\neq 0$ immersion 

1037: generated by the triple 

1038: $(\xi,\Phi_0,\tilde{z}_0 )$ with $\Phi$ and 

1039: $F$ the holomorphic respectively unitary frame. 

1040: %

1041: \begin{enumerate}

1042:   \item If $(\gamma,\,\X_1,\,T_1)$ and 

1043:   $(\gamma,\,\X_2,\,T_2)$ are symmetries 

1044:   of $f$, then $\X_2 = \pm \X_1$ and 

1045:   $T_1 = T_2$. 

1046:   \item If $(\gamma,\,\X_1)$ and $(\gamma,\,\X_2)$ 

1047:   are symmetries of $F$ then $\X_2 = \pm \X_1$. 

1048:   \item If $(\gamma,\,\X_1)$ and $(\gamma,\,\X_2)$ 

1049:   are symmetries of $\Phi$ then $\X_2 = \pm \X_1$. 

1050: \end{enumerate}

1051: %

1052: \Proof (i) If $(\gamma,\,\X_1,\,T_1)$ and 

1053: $(\gamma,\,\X_2,\,T_2)$ are symmetries, then 

1054: $\gamma^* f = \X_1 f \X_1^{-1} + T_1 = 

1055: \X_2 f \X_2^{-1} + T_2$ implies 

1056: $\X \, f \, \X^{-1} + T = f$ with 

1057: $\X = \X_2^{-1}\X_1$ and 

1058: $T = \X_2^{-1}(T_1-T_2)\X_2$. Since a CMC 

1059: $H \neq 0$ surface in $\R^3$ is never contained 

1060: in an affine plane, 

1061: $\X \, f \, \X^{-1} + T = f$ implies 

1062: $\X = \pm \Id$ and $T=0$. 

1063: 

1064: (ii) If $(\gamma,\,\X_1)$ and $(\gamma,\,\X_2)$ 

1065: are symmetries of $F$, then 

1066: $\gamma^* F = \X_1 F k_1 = \X_2 F k_2$. 

1067: Plugging $\X_1 F k_1$ and $\X_2 F k_2$ into 

1068: the Sym--Bobenko formula and equating yields

1069: %

1070: \begin{equation}

1071:   \X_1 f \X_1^{-1} - \tfrac{i\lambda}{2H}

1072:   (\partial_{\lambda} \X_1 ) \X_1^{-1} = 

1073:   \X_2 f \X_2^{-1} - \tfrac{i\lambda}{2H}

1074:   (\partial_{\lambda} \X_2 ) \X_2^{-1}.

1075: \end{equation}

1076: %

1077: Now $\X_2^{-1} \X_1 \,f \, \X_1^{-1} \X_2 + 

1078: \tfrac{i\lambda}{2H} \X_2^{-1} 

1079: \left( (\partial_{\lambda} \X_2) \X_2^{-1} - 

1080: (\partial_{\lambda} \X_1 ) \X_1^{-1} \right) \X_2 

1081: = f$ and part (i) implies $\X_2 = \pm \X_1$.

1082: 

1083: (iii) If $(\gamma,\,\X_1)$ and 

1084: $(\gamma,\,\X_2)$ are symmetries of $\Phi$ then 

1085: $\gamma^* \Phi = \X_1 \Phi H = \X_2 \Phi G$ 

1086: implies $\X_2^{-1} \X_1 \Phi = \Phi G H^{-1}$ 

1087: and shows that 

1088: $\X_2^{-1} \X_1 \in \mathrm{Iso}(\Phi)$. 

1089: Let $\Phi = F B$ be an Iwasawa decomposition of 

1090: $\Phi$. Then 

1091: $\X_2^{-1} \X_1 \in \mathrm{Iso}(F)$, since 

1092: $\mathrm{Iso}(\Phi) = \mathrm{Iso}(F)$ by 

1093: \eqref{eq:isotropies}. 

1094: Hence $\X_2^{-1} \X_1 = \pm \Id$ 

1095: by \eqref{eq:iso_intersect}. \QED

1096: 

1097: For symmetries $(\gamma,\,\X_\gamma)$ and 

1098: $(\mu,\,\X_\mu)$ of $F \in \mathcal{F}(\MT)$ and 

1099: corresponding maps 

1100: $k_\gamma,\,k_\mu$,  

1101: we have $\mu^*\,\gamma^*\,F = 

1102: \X_\gamma \, \X_\mu \,F\,k_\mu

1103:   \,\mu^* k_\gamma$ and if we assume that 

1104: $((\gamma \circ \mu),\,\X_{\gamma\mu})$ 

1105: is a symmetry, then part (ii) of Theorem 

1106: \ref{th:symmetries} implies $\X_{\gamma \mu} = 

1107: \pm \X_\gamma \, \X_\mu$ and consequently

1108: %

1109: \begin{equation}

1110:     k_{\gamma \mu} = \pm k_\mu \, \mu^* k_\gamma.

1111: \end{equation}

1112: % 

1113: 

1114: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1115: 

1116: \newsection{\bf Symmetries of $\Phi$.} 

1117: We briefly investigate how symmetries 

1118: of a holomorphic potential descend to symmetries 

1119: of the corresponding holomorphic frame and note a 

1120: simple consequence in case the potential is 

1121: invariant. In view of these results and our 

1122: subsequent inquiry into the relationship with the 

1123: dressing action, we allow symmetries on the level 

1124: of holomorphic frames of the form $(\gamma, X)$ 

1125: with $X \in \LGC$.

1126: 

1127: \myprop{} Let $\gamma \in Aut(\MT)$ and 

1128: $(\gamma,\,G)$ be a symmetry of 

1129: $\xi \in \pott$. Let $\Phi:\MT \to \LGC$ 

1130: be a solution to $d\Phi = \Phi \xi$. 

1131: Then there exists $X \in \LGC$ 

1132: such that $\gamma^* \Phi = X \Phi G$.

1133: 

1134: \Proof Since $\gamma^* \Phi$ and 

1135: $\Phi G$ both solve the differential equation 

1136: $dY = Y\gamma^*\xi$, there exists 

1137: $X \in \LGC$ with $\gamma^*\Phi = X \Phi G$. 

1138: More explicitly, if $\Phi$ solves 

1139: $d\Phi = \Phi \xi,

1140: \, \Phi(\tilde{z}_0) = \Phi_0$, then 

1141: $X = \Phi(\gamma(\tilde{z}_0)) 

1142: G(\tilde{z}_0)^{-1} \Phi_0^{-1}$.

1143: \QED 

1144: 

1145: We shall again call a pair $(\gamma,\,X)$ with 

1146: $\gamma \in Aut(\MT)$ and $X \in \LGC$ a 

1147: symmetry of a map $\Phi:\MT \to \LGC$ 

1148: if there exists a $G \in \gauge$ such that 

1149: $\gamma^* \Phi = X\,\Phi \,G$.

1150: 

1151: \mycorollary{} \label{th:monophi} 

1152: Let $\gamma \in Aut(\MT)$ and 

1153: $\xi\in \pott$ with $\gamma^* \xi = \xi$. 

1154: Let $\Phi$ be a solution to $d\Phi = \Phi \xi$. 

1155: Then there exists $X \in \LGC$ such that 

1156: $\gamma^* \Phi = X \Phi$. \QED

1157: 

1158: 

1159: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1160: 

1161: \newsection{\bf Symmetries of $F$.} 

1162: In view of the previous section, 

1163: we now characterize how symmetries 

1164: of the holomorphic frame descend to symmetries 

1165: of the corresponding extended unitary frame 

1166: and show that for open Riemann surfaces, the 

1167: co-cycle factor in equation \eqref{eq:sym_frame} 

1168: can be gauged away.

1169: 

1170: \mylemma{}\label{th:symF} 

1171: Let $(\gamma,\,X)$ be a symmetry 

1172: of $\Phi:\MT \to \LGC$. Let 

1173: $\Phi = F\,B$ be an Iwasawa decomposition 

1174: of $\Phi$. Then there exists an $\X \in \LGU$ 

1175: such that $(\gamma,\X)$ is a symmetry of 

1176: $F$ if and only if there exists an element 

1177: $L \in \mathrm{Iso}_r(\Phi)$ with $LX^{-1} \in \LGU$.

1178: 

1179: \Proof Let $G \in \gauge$ such that 

1180: $\gamma^* \Phi = X \Phi G$. If $(\gamma,\X)$ 

1181: with $\X \in \LGU$ is a symmetry of $F$, then by 

1182: \eqref{eq:sym_frame} there 

1183: exists a differentiable map $k:\MT \to \U1$ 

1184: such that $\gamma^* F = \X F k$. In combination 

1185: with $F = \Phi B^{-1}$, after rearranging, 

1186: we obtain 

1187: %

1188: \begin{equation}

1189:   \X^{-1} X \, \Phi = 

1190:   \Phi B^{-1}k\gamma^*B \, G^{-1}.

1191: \end{equation}

1192: %

1193: Then $L:= \X^{-1} X \in \mathrm{Iso}_r(\Phi)$ as 

1194: $B^{-1}k\gamma^*B \, G^{-1} \in \gauge$ and 

1195: $L \, X^{-1} = \X^{-1} \in \LGU$.

1196: 

1197: Conversely, let $L \in \mathrm{Iso}_r(\Phi)$ and 

1198: $L \, X^{-1} \in \LGU$. Then there exists 

1199: an element $H \in \gauge$ with 

1200: $L\, \Phi = \Phi H$. By \eqref{eq:isotropies}, 

1201: $L \in \mathrm{Iso}_r(F)$ and $LF=FV$ for 

1202: $V = BHB^{-1} \in \gauge$. Then 

1203: $\gamma^* \Phi = X \Phi G$ yields 

1204: %

1205: \begin{equation}

1206:   \gamma^* F = X \,F\, B\, G\,  \gamma^* B^{-1} = 

1207:   X\, L^{-1}F\, V\, B \,G\, \gamma^*B^{-1}.

1208: \end{equation}

1209: %

1210: Define $\X:=X\,L^{-1}$ and 

1211: $k:=VBG\gamma^*B^{-1}$. 

1212: On the one hand, $k \in \gauge$ 

1213: while on the other hand 

1214: $k= F^{-1}\X^{-1}\gamma^*F$ 

1215: takes values in $\LGU$, since 

1216: $\X \in \LGU$ by assumption. From equation 

1217: \eqref{eq:intersect} we conclude that 

1218: $k:\MT \to \U1$. \QED

1219: %

1220: 

1221: \mytheorem{}

1222: Let $M$ be an open Riemann surface 

1223: with Fuchsian group $\Gamma$ and let 

1224: $f_\lambda: \MT \to \R^3$ be an associated family 

1225: of CMC $H \neq 0$ immersions. 

1226: Then there exists an extended unitary frame 

1227: $F \in \mathcal{F}(\MT)$ for $f_\lambda$ such 

1228: that for every $\gamma \in \Gamma$ there exists 

1229: $\X \in \Lhol$ such that $\gamma^*F = \X \,F$.

1230: 

1231: \Proof First we apply \cite{DorH:sym2}, 

1232: Theorem 2.3 and infer that there 

1233: exists some extended frame 

1234: $\tilde{F} \in \mathcal{F}(\MT)$ for $f_\lambda$ 

1235: such that for every $\gamma \in \Gamma$ we have 

1236: $\gamma^* \tilde F =

1237: \X \,\tilde{F} \,k$, with $\X \in \Lhol$ and 

1238: $k:\MT \to \U1$ as in \eqref{eq:sym_frame}. 

1239: Writing $\tilde{F} = \Phi \,B$ with 

1240: $\Phi:\MT \to \LGC$ holomorphic and 

1241: $B \in \gauge$, we obtain 

1242: $\gamma^* \Phi = \X\,\Phi\,W$ 

1243: with $W \in \gauge$. It is easy to see that 

1244: $W$ satisfies the co-cycle condition. Therefore, 

1245: since the open Riemann surface $\MT$ is Stein, 

1246: by \cite{Bun}, $W$ is a co-boundary and whence 

1247: $W = W\,\gamma^*W^{-1}$. Hence 

1248: $\tilde{\Phi} = \Phi \,W$ satisfies 

1249: $\gamma^* \tilde{\Phi} = \X \, \tilde{\Phi}$. 

1250: Iwasawa splitting $\tilde{\Phi} = F \hat{B}$ with 

1251: $\hat{B}$ normalized such that the $\lambda^0$ 

1252: coefficient in $\hat{B}$ has positive real entries 

1253: we obtain $\gamma^* F = \X\,F$. \QED

1254: 

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

1256: 

1257: \newsection{{\bf Symmetries \& Dressing.} 

1258: \label{sec:sym} 

1259: Next, we characterize how symmetries between two 

1260: dressing equivalent unitary frames are related.

1261: 

1262: \mytheorem{} \label{th:sym1}

1263: Let $F_0 \in \mathcal{F}(\MT)$ have symmetry 

1264: $(\gamma,\X_0)$ and $h \in \LGC$. Then 

1265: $F \in [h \# F_0]$ has symmetry $(\gamma,\X)$ 

1266: if and only if there exists a 

1267: $L \in \mathrm{Iso}_r(F)$ such that 

1268:  $\X = h \X_0 h^{-1} L^{-1}$. 

1269: Furthermore, if both $F_0,\,F \in \FID$, then 

1270: $L \in \LGP$.

1271: 

1272: \Proof

1273: %

1274: Writing $hF_0 = F B$ for an Iwasawa decomposition 

1275: and using $\gamma^* F_0 = \X_0 F_0 k_0$, we obtain 

1276: $h \X_0 h^{-1} F  = 

1277:   \gamma^* F \gamma^* B k_0^{-1} B^{-1}$. 

1278: If $(\gamma,\X)$ is a symmetry of $F$ then 

1279: $\gamma^* F = \X F k$, whence

1280: %

1281: \begin{equation} \label{eq:iso_con}

1282:   \X^{-1} h \X_0 h^{-1} F =  

1283:   F k \gamma^*B k_0^{-1} B^{-1}.

1284: \end{equation}

1285: %

1286: Set $L :=\X^{-1} h \X_0 h^{-1}$ and 

1287: $H := k \gamma^*B k_0^{-1} B^{-1}$. Clearly 

1288: $L \in \LGC$ and $H \in \gauge$, so equation 

1289: \eqref{eq:iso_con} reads $LF=FH$ and is an 

1290: Iwasawa decomposition of $LF$. Hence 

1291: $L \in \mathrm{Iso}_r(F)$, and by construction    

1292: $\X = h \X_0 h^{-1} L^{-1}$.

1293: 

1294: Conversely, assume there exists a map 

1295: $L \in \mathrm{Iso}_r(F)$ 

1296: such that $L F = F H$ for some $H \in \gauge$ 

1297: and $h \X_0 h^{-1} L^{-1} \in \Lhol$. 

1298: Defining $\X = h \X_0 h^{-1} L^{-1}$ and 

1299: $k = H B k_0 \gamma^* B ^{-1}$, a computation 

1300: yields $\gamma^* F = \X F k$. 

1301: A priori, $k:\MT \to \LGP$, but as 

1302: $\gamma^* F,\, \X$ and $F$ take values 

1303: in $\Lhol$ so does $k$. 

1304: Hence $k:\MT \to \U1$ and consequently 

1305: $(\gamma,\X)$ is a symmetry of $F$.

1306: \QED

1307: 

1308: \mycorollary{}

1309: If in addition to the assumptions of Theorem 

1310: \ref{th:sym1}, $\mathrm{Iso}_r(F) = \{ \pm \Id \}$, 

1311: then $(\gamma,\X)$ is a symmetry of 

1312: $F$ if and only if $\X = \pm h\, \X_0 h^{-1}$.

1313: 

1314: \Proof

1315: The assumption $\mathrm{Iso}_r(F) = 

1316: \{ \pm \Id \}$ implies that for the maps $L,\,H$ in the 

1317: proof of Theorem \ref{th:sym1} we have  

1318: $L = \Id$ and $H \equiv \Id$, yielding the claim.

1319: \QED

1320: 

1321: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1322: 

1323: \newsection{\bf Groups of symmetries.}

1324: Extending the results of section \ref{sec:sym} 

1325: to groups $\Gamma \subset Aut(\MT)$ of symmetries 

1326: one obtains:

1327: 

1328: \mytheorem{} \label{th:mainresult}

1329: Let $F_0 \in \mathcal{F}_r(\MT)$ admit symmetries 

1330: $(\gamma,\X_0(\gamma))$ for all 

1331: $\gamma \in \Gamma$ and let $h \in \LGC$. 

1332: Then $F \in [h \# F_0]$ 

1333: has symmetries $(\gamma,\X(\gamma))$ for all 

1334: $\gamma \in \Gamma$ if and only if there exist 

1335: maps $L(\gamma) \in \mathrm{Iso}_r(F)$ such that 

1336: $\X(\gamma)=h\X_0(\gamma) h^{-1} L(\gamma)^{-1}$. 

1337: Moreover, if $\mathrm{Iso}_r(F) = \{ \pm \Id \}$, 

1338: then $F$ admits symmetries $(\gamma,\X(\gamma))$ 

1339: for all $\gamma \in \Gamma$ if and only if 

1340: $\X(\gamma) = h \X_0(\gamma) h^{-1}$ for all 

1341: $\gamma \in \Gamma$. \QED

1342: 

1343: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1344: 

1345: \newsection{\bf Symmetries \& Isotropy} 

1346: Let $(\gamma,\X)$ be a symmetry of 

1347: $F \in \mathcal{F}(\MT)$ and 

1348: $\gamma^* F = \X F k$. 

1349: If $L \in \mathrm{Iso}_r(F)$ and $LF = FB$ for an 

1350: Iwasawa decomposition, then 

1351: $\X^{-1}L \X F = F k\gamma^*Bk^{-1}$. 

1352: As $(k\gamma^*Bk^{-1})$ is $\LGP$--valued, 

1353: $\X^{-1} L \X \in \mathrm{Iso}_r(F)$ 

1354: and every symmetry 

1355: $(\gamma,\X)$ induces an inner automorphism 

1356: $S:\mathrm{Iso}_r(F) \to \mathrm{Iso}_r(F),\,

1357: L \mapsto \X^{-1} L \X$. 

1358: 

1359: Now consider $F_0 \in \FID,\,

1360: h \in \LGP$ and let $F = h \# F_0$. Assume that 

1361: $(\gamma,\,\X_0)$ and $(\gamma,\,\X)$ are 

1362: symmetries of $F_0$ respectively $F$. 

1363: If further $h \in \mathrm{Iso}_r(F)$, then we 

1364: obtain a map $S : \mathrm{Iso}_r(F) \to 

1365: \mathrm{Iso}_r(F),\,h \mapsto 

1366: \X^{-1} h \X_0 h^{-1}$ with the property 

1367: $S(h_1 h_2) = S(h_1) h_1 S(h_2) h_1^{-1}$.

1368: 

1369: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1370: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1371: 

1372: \section{Invariant potentials \& Monodromy}

1373: %

1374: We start our investigation of the relationship 

1375: between monodromy and dressing by defining the 

1376: notion of monodromy on the level of 

1377: holomorphic and unitary frames in the DPW framework. 

1378: 

1379: \mydefinition{} \label{def:monodromies} 

1380: Let $M$ be a connected Riemann surface 

1381: with universal cover $\MT$ and $\Delta$ 

1382: the group of deck transformations. 

1383: Let $\xi \in \pott$ and $\Phi: \MT \to \LGC$ be a 

1384: solution of $d \Phi = \Phi \xi$ and 

1385: $\tau \in \Delta$. A loop 

1386: $\varrho(\tau) \in \LGC$ such that 

1387: $(\tau,\,\varrho(\tau))$ is a symmetry of 

1388: $\Phi$ will be called a \textit{monodromy} of 

1389: $\Phi$ with respect to $\tau$. 

1390: Let $\Phi = F\,B$ be an Iwasawa decomposition. 

1391: A loop $\chi(\tau) \in \LGU$ such that 

1392: $(\tau,\,\chi(\tau))$ is a symmetry of 

1393: $F$ is called a \textit{monodromy} of $F$ 

1394: with respect to $\tau$.

1395: 

1396: \myremark{} A monodromy of a normalized extended 

1397: unitary frame $F \in \FID$ takes values in $\Lhol$.

1398: %

1399: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1400: %

1401: 

1402: \newsection{\bf Existence of monodromies.} 

1403: Apriori, it is not clear that 

1404: for some $\tau \in \Delta$ there exist monodromies 

1405: $\varrho(\tau)$ and 

1406: $\chi(\tau)$ as in the definitions above. 

1407: In case the 

1408: corresponding potentials are invariant under 

1409: $\Delta$ we can evoke Corollary \ref{th:monophi} 

1410: and obtain the following sufficient conditions 

1411: for the existence of monodromies.

1412: 

1413: \myprop{} (i) Let $\xi \in \pott$ and 

1414: $\Phi$ be a solution of $d\Phi = \Phi\,\xi$ with 

1415: initial condition $\Phi_0 \in \LGC$. 

1416: Let $\tau \in \Delta$ with $\tau^*\xi = \xi$. 

1417: Then there exists a monodromy 

1418: $\varrho(\tau) \in \LGC$ such that 

1419: $\tau^* \Phi = \varrho(\tau) \,\Phi$.

1420: 

1421: (ii) Let $F:\MT \to \LGU$ and $\alpha = F^{-1}dF$. 

1422: If $\tau^*\alpha = \alpha$ for $\tau \in \Delta$, 

1423: then there exists a monodromy $\chi(\tau) \in \LGU$ 

1424: such that $\tau^* F = \chi(\tau) \,F$. \QED

1425: 

1426: It is shown in \cite{DorH:cyl} that CMC 

1427: immersions of open Riemann surfaces $M$ 

1428: can always be generated by $\Delta$--invariant 

1429: potentials $\xi \in \pott$. 

1430: If $\Phi = F B$ is the Iwasawa 

1431: decomposition of a holomorphic frame 

1432: $\Phi : \MT \to \LGC$, then we shall need 

1433: to study the monodromy of $F$ in order to control 

1434: the periodicity of the 

1435: resulting immersion \eqref{eq:Sym}. 

1436: Even if $F$ is obtained from a potential that 

1437: is invariant under $\Delta$, we are a priori 

1438: not assured that there exist loops 

1439: $\chi(\tau) \in \LGU$ and maps 

1440: $k(\tau):\MT \to \U1$ such that 

1441: equation \eqref{eq:sym_frame} holds for all 

1442: $\tau \in \Delta$ or even more strongly, 

1443: that for $\alpha = F^{-1}dF$ we have 

1444: $\tau^* \alpha = \alpha$ for all 

1445: $\tau \in \Delta$. In analogy to Theorem 

1446: \ref{th:sym1}, this issue is characterized 

1447: by the following 

1448: 

1449: \mylemma{} \label{th:iso_problem} 

1450: Let $(\xi, \Phi_0, \tilde{z}_0 )$ generate the 

1451: holomorphic frame $\Phi:\MT \to \LGC$ and 

1452: extended unitary frame $F \in \mathcal{F}(\MT)$. 

1453: Let $\tau \in \Delta$ and $\varrho(\tau)$ 

1454: be a monodromy of $\Phi$. 

1455: Then the following are equivalent.

1456: %

1457: \begin{enumerate} 

1458:   \item There exists a monodromy 

1459:   $\chi(\tau) \in \LGU$ of $F$ with 

1460:   respect to $\tau$.

1461: %

1462:   \item There exists an element  

1463:   $L(\tau) \in \mathrm{Iso}_r(\Phi)$ 

1464:   with $L(\tau)\varrho^{-1}(\tau) \in \LGU$. 

1465: \end{enumerate}

1466: 

1467: \Proof We first show that (i) implies (ii). 

1468:   Let $\chi(\tau) \in \LGU$ 

1469:   and $k(\tau):\MT \to \U1$ such that 

1470:   $\tau^* F = \chi(\tau) F\, k(\tau)$. 

1471:   Using $F = \Phi B^{-1}$ and 

1472:   $\tau^* \Phi = \varrho(\tau) \Phi$ and 

1473:   rearranging, we obtain

1474:   $\chi(\tau)^{-1}\varrho(\tau) \, \Phi = 

1475:   \Phi\, B^{-1} k(\tau) \,\tau^* B$.

1476:   Then $L(\tau) :=  \chi(\tau)^{-1}\varrho(\tau) 

1477:   \in \mathrm{Iso}_r(\Phi)$ and 

1478:   $L(\tau)\varrho(\tau)^{-1} \in \LGU$. 

1479: 

1480: Conversely, let $L(\tau) \in \mathrm{Iso}_r(\Phi)$ 

1481: and assume $L(\tau)\varrho(\tau)^{-1} \in \LGU$. 

1482: Then there exists an element 

1483: $G(\tau) \in \gauge$ with 

1484: $L(\tau) \, \Phi = \Phi \, G(\tau)$.

1485: Multiplying this last equation on the right by 

1486: $\tau^* \Phi^{-1}$ and rearranging gives

1487: %

1488: \begin{equation} \label{eq:iso_phi2} 

1489:   F^{-1}L(\tau) \, \varrho(\tau)^{-1} \,\tau^*F 

1490:   = B \, G(\tau) \, \tau^*B^{-1}.

1491: \end{equation}

1492: %

1493: The left respectively right hand side of 

1494: \eqref{eq:iso_phi2} takes values in $\LGU$ 

1495: respectively $\LGP$. 

1496: Hence, by \eqref{eq:intersect}, 

1497: both sides are $\lambda$--independent and 

1498: $\U1$--valued. Set $k(\tau):= F^{-1}L(\tau)\,

1499: \varrho(\tau)^{-1}\tau^*F$ 

1500: and $\chi(\tau) := \varrho(\tau)\,L(\tau)^{-1}$. 

1501: Then $\tau^* F = \chi(\tau)\, F\, k(\tau)$ and 

1502: $\chi(\tau) \in \LGU$. 

1503: \QED

1504: 

1505: \mycorollary{} \label{th:chi=rho} 

1506: Let $(\xi, \Phi_0, \tilde{z}_0 )$ generate the 

1507: holomorphic frame $\Phi:\MT \to \LGC$ and 

1508: extended unitary frame $F \in \mathcal{F}(\MT)$. 

1509: Let $\tau \in \Delta$ and $\varrho(\tau)$ 

1510: be a monodromy of $\Phi$ 

1511: and $\chi(\tau)$ a monodromy of $F$ with 

1512: respect to $\tau \in \Delta$. 

1513: If $\mathrm{Iso}_r(F) = \{ \pm \Id \}$, 

1514: then $\chi(\tau) = \pm \varrho(\tau)$. \QED

1515:  

1516: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1517: 

1518: \newsection{\bf A factorization Theorem.} 

1519: In this section we work exclusively with 

1520: untwisted loops.Consider an analytic map $H:\C^* \to \SL$, 

1521: for which $\tr H$ is not independent of $\lambda$.

1522: Its eigenvalues are 

1523: %

1524: \begin{equation}

1525:   \mu_{\pm} = \frac{1}{2} \left( \tr H \pm 

1526:   \sqrt{(\tr H)^2 - 4}\right).

1527: \end{equation}

1528: %

1529: We will need to have $\mu_{\pm}$ holomorphic on 

1530: some open and dense subset $\mathbb S$ of $\C^*$.

1531: To this end we slightly generalize the procedure 

1532: used in \cite{DorH:cyl}. 

1533: 

1534: \myprop{}

1535: Let $g:\C^* \to \C,\,g \not\equiv 0$ be holomorphic. 

1536: Then there exists some connected, open, 

1537: and dense subset $\mathbb S \subset \C^*$ 

1538: such that there exists a well defined square 

1539: root function $\sqrt{g}$ of $g$ on $\mathbb S$.

1540: 

1541: \Proof

1542: Since $g(\lambda)$ does not vanish 

1543: identically it has at most finitely 

1544: many roots on $S^1$. Moving slightly inside, 

1545: if necessary, we can choose any $0 < r \leq 1$ 

1546: such that $g$ does not vanish on $C_r$.

1547: We form a (scalar) Birkhoff splitting 

1548: of $g$ on $C_r$: $g = g_- \lambda^N g_+$.

1549: It is easy to verify that, since $g$ is 

1550: holomorphic on $\C^*$, also $g_-$ and 

1551: $g_+$ are holomorphic on $\C^*$. 

1552: Now we introduce some cuts: Consider 

1553: the roots of $g_+$. They are all contained 

1554: in the complement of $I_r$.

1555: We cut the Riemann sphere from every odd 

1556: ordered root of $g_+$ to infinity. 

1557: Similarly we cut the Riemann sphere from 

1558: each odd ordered root of $g_-$ to the origin.

1559: Thus $g_-$ and $g_+$ have well defined 

1560: root functions on the cut domain.

1561: If $N$ is odd, then we also need to 

1562: introduce one further cut from the origin to the 

1563: point at infinity. This resulting domain will 

1564: be denoted by $\mathbb S$. 

1565: \QED

1566: 

1567: Applying this to the function 

1568: $g(\lambda) = {\tr H}^2 -4$ we obtain

1569: 

1570: \mycorollary{} \label{th:mu}

1571: The eigenvalue functions $\mu_{\pm}(\lambda)$ 

1572: of a loop $H \in \Lambda_{\C^*}\SL$ are analytic 

1573: on a connected, open and dense set 

1574: $\mathbb S \subset \C^*$, obtained by making 

1575: radial branch cuts from odd ordered roots $\alpha$ 

1576: of ${\tr H}^2 -4$ to the origin 

1577: respectively the point at infinity, depending 

1578: on whether $|\alpha| <r$ respectively 

1579: $|\alpha| >r $. In addition there is 

1580: possibly a cut from $0$ to $\infty$.

1581: \QED

1582: 

1583: For our approach it will be convenient 

1584: to diagonalize $H$

1585: %

1586: \begin{equation} \label{eq:diag}

1587:   Y\, H \, Y^{-1}= \mathrm{diag}[\mu_+,\,\mu_-] 

1588: \end{equation}

1589: %

1590: If the matrix entries $H_{1,2}$ and $H_{2,1}$ 

1591: of $H$ do not vanish identically, 

1592: then one can choose as diagonalizing matrix $Y$ 

1593: ( see  e.g. (3.5.18) in \cite{DorH:cyl})

1594: %

1595: \begin{equation}

1596:   Y = \begin{pmatrix}

1597:   1 &  -\frac{H_{12}}{u-iv-H_{11}}\\

1598:   -\frac{H_{21}}{2iv}& \frac{u + iv-H_{22}}{2iv}

1599:   \end{pmatrix},

1600: \end{equation}

1601: %

1602: where  $H_{ij}$ denotes the $(i,j)-$coefficient 

1603: of $H$ and $ 2u = \tr H$ and $ v^2 = u^2 -1$.

1604: If one of the off-diagonal coefficients vanishes 

1605: identically, then we choose the corresponding 

1606: matrix $Y$ listed in \cite{DorH:cyl}. In this 

1607: case the diagonal entries are $\mu_{\pm}$. Note 

1608: that this works in our setting.(Not only in the 

1609: somewhat more special setting of \cite{DorH:cyl}.)

1610: 

1611: From the specific form of the matrices 

1612: $Y$ mentioned above we infer that $Y$ 

1613: is meromorphic on $\mathbb S $, provided

1614: the trace $\tr H$ of $H$ is not independent 

1615: of $\lambda$.

1616: 

1617: The next result is crucial for our 

1618: characterization of dressing matrices

1619: preserving the fundamental group. As before, 

1620: we denote an open annulus about $S^1$ by 

1621: $A_r = \{ \lambda \in \C : r<|\lambda|<1/r \}$ and 

1622: by $C_r = \{ \lambda \in \C : |\lambda| = r \}$ 

1623: a circle of radius $r$. With the above notations 

1624: we have the following

1625: 

1626: \mytheorem{} \label{th:h-split}

1627: Let $h \in \Lambda_r^+\SL$ for some $0<r\leq1$ and 

1628: $H  \in \Lambda_{\C^*}\SL$ such that the trace 

1629: of $H$ is not constant. 

1630: Let $\mathbb S$ be as in Corollary \ref{th:mu}. 

1631: If $h\, H  \,h^{-1}$ is meromorphic on 

1632: $\mathbb S \cap A_s$ for some 

1633: $0 < s < r$, then $h$ can be factored, on a possibly 

1634: segmented circle $C_l \cap \mathbb S$ for 

1635: $s\leq l \leq r$, into 

1636: %

1637: \begin{equation}\label{eq:h-split}

1638:   h = \mathcal{M}\, \mathcal{C},

1639: \end{equation}

1640: % 

1641: where $\mathcal{C}$ is  $\SL$--valued 

1642: on $\mathbb S\cap C_l$ and 

1643: $[\, \mathcal{C},\, H  \,] = 0$ there. Moreover, 

1644: $\mathcal{M}$ is $\SL$--valued and meromorphic 

1645: on the connected open set $\mathbb S \cap A_s$. 

1646: 

1647: \Proof From the above discussion, 

1648: there exists some $Y$  

1649: such that \eqref{eq:diag} 

1650: holds. Let us write 

1651: %

1652: \begin{equation}

1653:   h Y^{-1} = \begin{pmatrix} 

1654:   a & b \\ 

1655:   c & d \end{pmatrix}.

1656: \end{equation}

1657: %

1658: Computing $h\,H \,h^{-1}$ and evaluating 

1659: the diagonal entries gives that both 

1660: %

1661: \begin{align}

1662:   &ad\mu_- -bc\mu_+ , \label{eq:r22}\\

1663:   &ad\mu_+ - bc\mu_- \label{eq:r11}

1664: \end{align}

1665: %

1666: are meromorphic on $\mathbb S \cap A_s$. 

1667: Multiplying \eqref{eq:r22} by $\mu_+$ and 

1668: \eqref{eq:r11} by $\mu_-$ and subtracting 

1669: the resulting equations, 

1670: implies that $bc(\mu_+^2 - \mu_-^2)$ 

1671: is meromorphic on $\mathbb S \cap A_s$. 

1672: Multiplying \eqref{eq:r22} by $\mu_-$ and 

1673: \eqref{eq:r11} by $\mu_+$ and subtracting the 

1674: resulting equations, implies that 

1675: $ad(\mu_-^2 - \mu_+^2)$ is 

1676: meromorphic on $\mathbb S \cap A_s$. Assuming 

1677: $\mu_-^2 - \mu_+^2 = 0$ would imply 

1678: $\mu_{\pm}^4 = 1$ and contradict the assumption 

1679: that $\tr H$ is not independent of $\lambda$. 

1680: Since $\mu_{\pm}$ are 

1681: holomorphic on $\mathbb S \cap A_s$ both $ad$ 

1682: and $bc$ are meromorphic on $\mathbb S \cap A_s$. 

1683: Similarly, evaluating the off--diagonal terms of 

1684: $h\,H \,h^{-1}$ we obtain that both $ab$ 

1685: and $cd$ are meromorphic on $\mathbb S \cap A_s$.

1686: 

1687: If $d \neq 0$ on some circle 

1688: $C_l$, for $s \leq l \leq r$, then we may write 

1689: $h\,Y^{-1}=\widehat{h}\,\mathrm{diag}[\,1/d,\,d\,]$ 

1690: with

1691: %

1692: \begin{equation} \label{eq:h-split_case1}

1693:   \widehat{h} := \begin{pmatrix} 

1694:   ad & \tfrac{b}{d} \\ cd & 1 \end{pmatrix} 

1695:   \mbox{ meromorphic on $\mathbb S\cap A_s$.} 

1696: \end{equation}

1697: %

1698: We define $\mathcal{C} := 

1699: Y^{-1} \,\mathrm{diag}[\,1/d, d\,]\,Y$ and 

1700: $\mathcal{M} := \widehat{h}\, Y$. 

1701: Then $h = \mathcal{M}\, \mathcal{C}$ 

1702: with $\mathcal{C} = \mathcal{C}(\lambda)$ 

1703: defined on $C_l \cap \mathbb S$ and 

1704: $[\,\mathcal{C},\,H  \,] = 0$ there. Moreover, 

1705: $\mathcal{M}$ is meromorphic on 

1706: $\mathbb S \cap A_s$.

1707:  

1708: If $d \equiv 0$ on an arc $C_l \cap \mathbb S$ for 

1709: some $s \leq l \leq r$, then both 

1710: $cd \equiv 0$ on this arc and consequently 

1711: $cd \equiv 0$ in $\mathbb S \cap I_r$, 

1712: and $ad \equiv 0$ on this arc and consequently 

1713: $ad \equiv 0$ in $\mathbb S \cap I_r$.

1714: Thus $d \equiv 0$ on $C_l \cap \mathbb S$ 

1715: implies that both $b,\,c \neq 0$ on 

1716: $\mathbb S \cap I_r$ and we may write 

1717: $h\,Y^{-1} = \widetilde{h}\,

1718: \mathrm{diag}[\,c,\,1/c\,]$ 

1719: with

1720: %

1721: \begin{equation*} \label{eq:h-split_case2}

1722:   \widetilde{h} = \begin{pmatrix} 

1723:   a/c & bc \\ 1 & cd

1724:   \end{pmatrix} 

1725:   \mbox{ meromorphic on $\mathbb S\cap A_s$.} 

1726: \end{equation*}

1727: %

1728: Set 

1729: $\mathcal{C} = Y^{-1} \mathrm{diag}

1730: [\, c, 1/c \,] Y$ and   

1731: $\mathcal{M} = \widetilde{h} Y$. Then

1732: $h = \mathcal{M}\, \mathcal{C}$ with 

1733: $[\, \mathcal{C},\,\chi \,] =0$ 

1734: and $\mathcal{M}$ meromorphic on 

1735: $\mathbb S\cap A_s$. \QED

1736: 

1737: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1738: 

1739: Given an open dense subset $\mathbb S \subset \C^*$, 

1740: obtained by making branch cuts from the points 

1741: $\{ \alpha_j \} \subset \C^*$ to the origin 

1742: respectively $\infty$, depending on whether 

1743: $|\alpha_j|<r$ respectively $|\alpha_j| > r$, let 

1744: $\hat{\mathbb S}$ be the set obtained by making 

1745: branch cuts according to this rule from the points 

1746: $\{ 1/\bar{\alpha}_j \}$. Then the set 

1747: ${\mathbb S}^* =  \mathbb S \cap \hat{\mathbb S}$ is 

1748: invariant under $\lambda \mapsto 1/\bar{\lambda}$. 

1749: Note though that if there are at least two distinct 

1750: $\alpha_j \in S^1$, then $\mathbb S^*$ is not 

1751: connected. 

1752: 

1753: We now improve on these results by exploiting the 

1754: reality condition of loops in $\Lambda_{\C^*} \SU$. 

1755: 

1756: \mylemma{} \label{th:S}

1757: The eigenvalue functions  $\mu_{\pm}$ of a loop 

1758: $H \in \Lambda_{\C^*}\SU$ 

1759: are of the form $\mu_{\pm} = u \pm iv$, 

1760: where $2u = \tr H$ and 

1761: $v^2 =1 - u^2 = \lambda^N \cdot \hat{v}^2$ with 

1762: $N \in 2\Z$ and $\hat{v}$ holomorphic on 

1763: $\mathbb S$. In particular, 

1764: $S^1 \subset \mathbb S = {\mathbb S}^*$.

1765: 

1766: \Proof A loop $H \in \Lambda_{\C^*}\SU$ 

1767: is of the form 

1768: %

1769: \begin{equation*}

1770:   H(\lambda) = \begin{pmatrix} 

1771:   x(\lambda) & y(\lambda) \\

1772:   -\overline{y(1/\bar{\lambda})} & 

1773:   \overline{x(1/\bar{\lambda})} \end{pmatrix}

1774: \end{equation*}

1775: %

1776: with analytic functions $x,\,y:\C^* \to \C$. Hence

1777: $\left. u \right|_{S^1} \in [-1,\,1]$. 

1778: Consequently, the function $v^2$ is real and 

1779: non-negative on $S^1$ and thus has only even 

1780: roots on $S^1$. If 

1781: $\{ \lambda_1,\,\ldots\,,\lambda_m \}$ are the 

1782: roots of $v^2$ on $S^1$, we may write 

1783: %

1784: \begin{equation*}

1785:   v^2 = \left( \prod_{j=1}^m 

1786:   (\lambda - \lambda_j)^{2K_j} \right) \, s

1787: \end{equation*}

1788: % 

1789: where $s:\C^* \to \C$ is analytic and 

1790: $\left.s\right|_{S^1} \neq 0$. We write 

1791: $s = c_0\,s_- \, \lambda^N \, s_+$ 

1792: for the scalar Birkhoff decomposition of 

1793: $s$ on $S^1$, with $c_0 \in \C^*$ and $N \in \Z$, 

1794: such that $s_-(\infty)=s_+(0)=1$. 

1795: With these normalizations, the reality condition 

1796: $u^* = u$ implies $s_+^* = s_-$ as well as 

1797: %

1798: \begin{equation*}

1799:   N = - \sum_{j=1}^m K_j \mbox{ and } 

1800:   \overline{c_0} = c_0 \, \prod_{j=1}^m 

1801:   \lambda_j^{2K_j}.

1802: \end{equation*}

1803: % 

1804: Whence also $s_+^*(\infty)=1$ and 

1805: $v^2 = \lambda^N \hat{v}^2$ with

1806: % 

1807: \begin{equation}

1808:   \hat{v}^2 =  c_0\,s_-\,s_+ \, \prod_{j=1}^m 

1809:   (\lambda - \lambda_j)^{2K_j}.

1810: \end{equation}

1811: %

1812: Since $s$ is analytic in $\C^*$, both $s_{\pm}$ 

1813: are analytic in $\C^*$, in fact $s_+(\lambda)$ 

1814: is entire, and thus also $\hat{v}^2$ is analytic 

1815: in $\C^*$. There are two cases depending on the 

1816: parity of $N \in \Z$. If $N$ is even, then 

1817: making branch-cuts from the odd ordered 

1818: roots of $s_+$ to the point at infinity and 

1819: branch-cuts from the odd ordered roots of $s_-$ to 

1820: the origin gives an open and dense set 

1821: $\mathbb S \subset \C^*$ on which $\hat{v}^2$ has 

1822: an analytic square root. Since the roots of 

1823: $\hat{v}^2$ on $S^1$ are all even, we have that 

1824: also $v^2$ has an analytic square root on 

1825: $\mathbb S$ and $S^1 \subset \mathbb S$. 

1826: Further, $s_+^* = s_-$ implies 

1827: $\mathbb S^* = \mathbb S$. 

1828: 

1829: Assume that $N \in \Z$ is odd. Then $\mathbb S$ 

1830: is the set obtained as above but with an 

1831: additional branch cut from the origin to 

1832: the point at infinity. Now $u=\sqrt{\tr^2H-4}$ 

1833: has an absolutely convergent power series expansion 

1834: on $S^1$ and is thus well defined, so there can not 

1835: be a branch cut through $S^1$. Hence $N$ is even.

1836: \QED

1837: 

1838: \mycorollary{} \label{co:h-split} 

1839: Let $h \in \Lambda_r^+ \SL$ for some $0<r\leq 1$ 

1840: and $H \in \Lambda_{\C^*}\SU$ with non-constant 

1841: trace function and let $\mathbb S$ be as in Lemma 

1842: \ref{th:S}. If $h\, H  \,h^{-1}$ 

1843: is meromorphic on $\mathbb S \cap A_s$ for some 

1844: $0 < s < r$ and in addition we require that 

1845: $(h\, H  \,h^{-1})^* = (h\, H  \,h^{-1})^{-1}$ 

1846: for all $\lambda \in \mathbb S \cap A_s$, then 

1847: %

1848: \begin{equation}

1849:   h = U\,C 

1850: \end{equation}

1851: %

1852: with meromorphic loops $U,\,C$ on 

1853: a cut but connected annulus respectively 

1854: a possibly segmented circle $C_l$ for some 

1855: $s<l<r$ with $[\,C,\,H\,]=0$ there, and 

1856: $U \in \Lambda_{r'} \,\SU$ for some 

1857: $r' \in (0,\,1)$.

1858: 

1859: \Proof

1860: Note that we have $ S^1 \subset \mathbb S = {\mathbb S}^*$ 

1861: under our assumptions.

1862: We apply Theorem \ref{th:h-split} and decompose  

1863: $h = \mathcal{M} \,\mathcal{C}$ on $C_l$ for some 

1864: $s<l<r$ with $\mathcal{M}$ meromorphic on 

1865: $\mathbb S \cap A_s$ and 

1866: $[\, \mathcal{C} ,\,H  ]=0$.  

1867: Consequently, we have $h\,H \,h^{-1} = 

1868: \mathcal{M}\,H\,\mathcal{M}^{-1}$ on 

1869: $\mathbb S \cap A_s$ and 

1870: $(h\, H  \,h^{-1})^* = 

1871:   (h\, H  \,h^{-1})^{-1}$ 

1872: is equivalent to 

1873: %

1874: \begin{equation}

1875:   [\,\mathcal{M}^* \,\mathcal{M} ,\,

1876:   H  \,]=0

1877: \end{equation}

1878: % 

1879: on $\mathbb S\cap A_s$. Next, we seek a loop $L$, 

1880: meromorphic on a subset of $\mathbb S \cap A_s$, 

1881: of the form  $L = f \, \Id + g \, \mathcal{M}^*\mathcal{M}$ 

1882: with meromorphic functions $f,\,g$ on 

1883: $\mathbb S \cap A_s$, such that 

1884: %

1885: \begin{equation} \label{eq:Luni}

1886:   (\mathcal{M} L^{-1})^* = 

1887:   (\mathcal{M} L^{-1})^{-1}.

1888: \end{equation}

1889: %

1890: Then $h = \mathcal{M} L^{-1} L \mathcal{C}$ 

1891: is the desired factorization. 

1892: Equation \eqref{eq:Luni} is equivalent to 

1893: $\mathcal{M}^* \mathcal{M} = L^* L$ and by setting 

1894: $P := \mathcal{M}^* \mathcal{M}$, yields 

1895: %

1896: \begin{equation} \label{eq:P} \begin{split}

1897:   P &= ff^* \Id + (f^* g + f\,g^*)P + g\,g^* P^2 \\ 

1898:   &= (ff^* - gg^*)\Id + (f^*g + fg^* + gg^* \tr(P))P

1899:   \end{split}

1900: \end{equation}

1901: %

1902: since $P^2 = \tr(P)P - \Id$ by Cayley-Hamilton. 

1903: The Ansatz $f=f^*=g$ reduces \eqref{eq:P} to 

1904: %

1905: \begin{equation}\label{eq:g^2}

1906:    g^2 = \frac{1}{2 + \tr P}.

1907: \end{equation}

1908: % 

1909: If we denote the entries of $\mathcal{M}$ by 

1910: $m_{ij}$, then the function

1911: %

1912: \begin{equation}

1913:   \tr P(\lambda) = \sum_{i,j=1}^2 m_{ij}(\lambda)

1914:   {m_{ij}^*(\lambda)}

1915: \end{equation}

1916: %

1917: is meromorphic on $\mathbb S \cap A_s$ and 

1918: is real and positive on $S^1\subset \mathbb S \cap A_s$ .

1919: With the help of a scalar Birkhoff decomposition of $g^2$ 

1920: it is straightforward to see that

1921: there exists a well defined square root of \eqref{eq:g^2} 

1922: on $S^1$, which extends meromorphically to the set 

1923: $\mathbb S_P \subset \mathbb S$, obtained by making branch-cuts 

1924: from the odd ordered roots 

1925: of $2 + \tr P$ in the usual fashion. 

1926: Note that $S^1 \subset \mathbb S_P$ and 

1927: $\mathbb S_P^* = \mathbb S_P$. 

1928: Now that $g = (2 + \tr P)^{-1/2}$ is 

1929: meromorphic on $\mathbb S_P \cap A_s$, we can set 

1930: $U:=\mathcal{M}L^{-1}$, which is meromorphic on 

1931: $\mathbb S_P \cap A_s$ with $U^* = U^{-1}$, 

1932: and $C : = L\,\mathcal{C}$, which is defined on 

1933: $C_l \cap \mathbb S_P$ with $[\,C,\,H\,]=0$. 

1934: It remains to show that there exists 

1935: an $r' \in (0,\, 1)$ such that 

1936: $U \in \Lambda_{r'} \,\SU$. 

1937: 

1938: We choose $0<r'<1$ such that both of the 

1939: following conditions hold:

1940: 

1941: (i) The only poles of $\mathcal{M}$ in 

1942:     $A_{r'} \cap \mathbb S$ are on $S^1$, 

1943: 

1944: (ii) $A_{r'} \cap \mathbb S_P = A_{r'}$.

1945: 

1946: Denote the pole set of $\mathcal{M}$ in 

1947: $A_{r'}$ by 

1948: $\mathcal{P} = \{ p_1,\dots,p_K \} \subset S^1$. 

1949: Let $V_j \subset A_{r'}$ be an open 

1950: neighborhood of $p_j$ such that 

1951: $V_j \cap \mathcal{P} = \{ p_j \}$. 

1952: Let $V^*_j = V_j \setminus \{ p_j \}$.

1953: Since $U$ is unitary on 

1954: $\mathbb S_P \cap U^*_j \cap S^1$,  

1955: its entries are bounded. 

1956: Hence $U$ extends analytically to $p_j$ 

1957: for all $j=1,\dots,K$. Consequently, 

1958: $U$ is analytic in $A_{r'}$ with $U^* = U^{-1}$. 

1959: Thus $U \in \Lambda_{r'}\,\SU$, concluding this proof. 

1960: \QED

1961: 

1962: In the following sections we will apply the 

1963: results of the present section to the twisted 

1964: loop groups.

1965: 

1966: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1967: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1968: 

1969: \newsection{} 

1970: Let $F \in \FID$ and $h \in \LGP$ and let

1971: $\hat F = h \# F$. Then $\hat F \in \FID$. 

1972: Let $\tau \in \Delta$ and $\chi(\tau)$, 

1973: $\hat\chi(\tau)$ be monodromies of $F$ 

1974: respectively $\hat F$. Writing $hF = \hat FB$ 

1975: for an Iwasawa decomposition and 

1976: $\tau^*F = \chi(\tau)\,F\,k(\tau)$ and 

1977: $\tau^*\hat F=\hat\chi(\tau)\,\hat F\,\hat k(\tau)$, 

1978: we obtain 

1979: %

1980: \begin{equation} \label{eq:mono_dress1}

1981:   \hat\chi(\tau)^{-1} h\,

1982:   \chi(\tau) \, h^{-1} \hat F =

1983:   \hat F\,\hat k(\tau)\,(\tau^* B)\,

1984:   k(\tau)^{-1}B^{-1}.

1985: \end{equation}

1986: %

1987: Since $\hat k(\tau)\,(\tau^* B)\,k(\tau)^{-1}

1988: B^{-1} \in \gauge$, 

1989: this proves that 

1990: %

1991: \begin{equation} \label{eq:mono_dress2} 

1992:   \hat \chi (\tau)^{-1}h\,\chi(\tau)\,h^{-1} 

1993:   \in \mathrm{Iso}_r(\hat F).

1994: \end{equation}

1995: %

1996: Since $F(\tilde{z}_0)= \hat F(\tilde{z}_0)=\Id$, 

1997: evaluation of $hF = \hat FB$

1998: at the base-point $\tilde{z}_0$ yields 

1999: $h=B(\tilde{z}_0)$. Evaluating equation 

2000: \eqref{eq:mono_dress1} at $\tilde{z}_0$ 

2001: and solving for $\hat \chi(\tau)$ gives  

2002: %

2003: \begin{equation} \label{eq:monodromy_dressed1}  

2004:   \begin{split}

2005:   \hat \chi (\tau) &= h\,\chi(\tau)\, 

2006:   k(\tau,\tilde{z}_0)\,

2007:   B\left(\tau(\tilde{z}_0)\right)^{-1}

2008:   \hat k(\tau,\tilde{z}_0)^{-1} \\

2009:   &= h\, \chi(\tau) \, h^{-1} \,L,\,

2010:   L \in \mathrm{Iso}_r(\hat F) \\

2011:   &= h\,\chi(\tau)\,b\, h^{-1},\,

2012:   b := h^{-1} L \, h \in \mathrm{Iso}_r(F). 

2013:   \end{split}

2014: \end{equation}

2015: % 

2016: This leads us to an application of Theorem 

2017: \ref{th:h-split} and shows that if a 

2018: dressing matrix preserves the topology, then it 

2019: factorizes as in \eqref{eq:h-split}.  

2020: 

2021: We now denote by $\mathbb S = {\mathbb S}^*$ 

2022: the set obtained from the eigenvalues of 

2023: a monodromy $\chi$.

2024: 

2025: \mytheorem{} \label{th:h-split1} 

2026: Let $\tau \in \Delta$ and $F \in \FID$ 

2027: with monodromy $\chi = \chi(\tau)$. We 

2028: assume that the trace of $\chi$ is not 

2029: constant. Let $h \in \LGP$ such that 

2030: $h\,\chi\,h^{-1}$ is meromorphic on 

2031: $\mathbb S \cap A_s$ for some $0<s<r$, 

2032: where $\mathbb S$ is as in Lemma \ref{th:S}. 

2033: If $\hat F = h\#F$ has a monodromy 

2034: $\hat{\chi}$ with respect to 

2035: $\tau$, and if $\hat{\mathbb S}$ is defined 

2036: for $\hat{\chi}$ according to Lemma \ref{th:S},

2037: then $h$ can be factored, on a possibly 

2038: segmented circle 

2039: $C_l \cap \mathbb S \cap \hat{\mathbb S}$ for 

2040: $l \in [s,\,r]$, into 

2041: $h = \mathcal{M} \,\mathcal{C}$ 

2042: where $\mathcal{C} = \mathcal{C}(\tau,\,\lambda)$ 

2043: is twisted $\SL$--valued and defined on 

2044: $C_l \cap \mathbb S \cap \hat{\mathbb S}$ and 

2045: $[\, \mathcal{C},\, \chi  \,] = 0$ there, and 

2046: $\mathcal{M} = \mathcal{M}(\tau,\,\lambda)$ 

2047: is twisted $\SU$--valued and meromorphic 

2048: on the connected open set 

2049: $ S^1 \subset \mathbb S \cap 

2050: \hat{\mathbb S} \cap A_s$.

2051: %

2052: 

2053: \Proof By the above \eqref{eq:monodromy_dressed1}, 

2054: for the dressed monodromy we have  

2055: $\hat{\chi} = h\,\chi\,h^{-1}\,\hat{L}$ with 

2056: $\hat{L} \in \mathrm{Iso}_r(\hat F)$. Clearly, 

2057: $\hat{L}$ is defined and meromorphic on 

2058: $\mathbb S \cap A_s$. Moreover, we can write 

2059: $\hat{\chi} = h\,\chi\, L h^{-1}\,$, where 

2060: $L \in \mathrm{Iso}_r(F)$. 

2061: From this we conclude that the eigenvalue 

2062: functions of $\chi\,L$ are the same as those 

2063: of $\hat{\chi}$. At this point we undress 

2064: every occurring matrix and continue, 

2065: until the very end of the proof, to work 

2066: in the undressed setting. We note that 

2067: the domains of definition of the dressed and 

2068: the undressed matrices are in a natural 

2069: correspondence and that unitary matrices 

2070: are mapped to unitary matrices, 

2071: while the "positive matrices" almost correspond. 

2072: For simplicity of notation we will use the 

2073: same notation as in the twisted situation.

2074: First we note that we can apply Theorem 

2075: \ref{th:h-split} to $h\,\chi\,h^{-1}$.

2076: Hence  $h = \mathcal{M} \,\mathcal{C}$, 

2077: where $\mathcal{M}$ has values in $\SL$ 

2078: and is defined on $\mathbb S$.

2079: As a consequence, $\hat{\chi} = 

2080: \mathcal{M} \chi {\mathcal{M}}^{-1} \hat{L} = 

2081: \mathcal{M} \chi \,L'{\mathcal{M}}^{-1}$. 

2082: From this expression we infer that $L'$ is 

2083: defined and meromorphic on $\mathbb S$.

2084: Now we can apply almost verbatim the proof of 

2085: \ref{co:h-split}, if we replace everywhere

2086: $\mathbb S$ by $\mathbb S \cap \hat{ \mathbb S} = 

2087: \tilde{\mathbb S}$ and obtain that we can assume 

2088: that $\mathcal{M}$ is defined and holomorphic 

2089: on some open neighborhood of $S^1$ and unitary 

2090: on $S^1$. Finally, twisting every occurring matrix 

2091: again we obtain the desired result. \QED

2092: %

2093: 

2094: When the associated family of $F \in \FID$ 

2095: possesses umbilic points, then, since dressing 

2096: preserves umbilic points \cite{Wu:dre}, 

2097: $\rm{Iso}_r(\hat F) = \left\{ \pm \Id \right\}$ 

2098: for $\hat F = h \# F$ and \eqref{eq:mono_dress2} 

2099: implies 

2100: %

2101: \begin{equation} \label{eq:monodromy_dressed2}

2102:   \hat \chi (\tau) = \pm h\,\chi(\tau)\,h^{-1} 

2103:   \in \Lhol.

2104: \end{equation}

2105: % 

2106: This allows us to restate Corollary 

2107: \ref{co:h-split} for the trivial isotropy case: 

2108: 

2109: %%%%%%%%%%

2110: 

2111: \mycorollary{} \label{co:h-split1} 

2112: Let $F \in \FID$ with 

2113: $\mathrm{Iso}(F) = \{ \pm \Id \}$. Let 

2114: $\chi = \chi(\tau) \in \Lhol$ be a

2115: monodromy of $F$ with non-constant trace. 

2116: Let $h \in \LGP$ for some $r \in (0,\,1]$. 

2117: If $h\#F$ also has a 

2118: monodromy with respect to $\tau$, then $h$ can be 

2119: factored into $h = U \,C$ 

2120: with $[\, C,\,\chi  \,] = 0$  and 

2121: $U = U(\tau,\,\lambda)$ the meromorphic 

2122: extension of an element of 

2123: $\Lambda_{r^\prime} \SU_\sigma$ 

2124: for some $r^\prime \in [r,\,1]$. \QED

2125: 

2126: We were not able to prove the corresponding result for a 

2127: general $F \in \FID$ without the isotropy 

2128: condition. It turns out, that we do 

2129: have an analogous result for the dressing orbit of 

2130: the vacuum, to which we now turn our attention.

2131: 

2132: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2133: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2134: 

2135: \section{Dressing the vacuum}

2136: %

2137: \newsection{\bf Definitions.} 

2138: The standard round cylinder (=''the vacuum'') 

2139: has the conformal structure of the punctured 

2140: plane $\C^*$. If we identify 

2141: $\C^* \cong \C / q\, \Z$ for some 

2142: $q \in \C^*$, then 

2143: %

2144: \begin{equation}

2145:   \exp:\C \to \C^*,\,w \mapsto z = \exp(q\,w)

2146: \end{equation}

2147: %

2148: is the universal covering map. The group of deck 

2149: transformations $\Delta \cong \Z$ is generated by 

2150: the translation

2151: %

2152: \begin{equation}  \label{eq:translation}

2153:   \tau_q : w \mapsto w + q.

2154: \end{equation}

2155: % 

2156: An extended unitary frame of the associated 

2157: family of the vacuum is given by

2158: %

2159: \begin{equation} \label{eq:F_c}

2160:   F_c(w,\,\lambda) = 

2161:   \exp((w \lambda^{-1} 

2162:   - \overline{w} \lambda)\, A)

2163: \end{equation}

2164: %

2165: where

2166: \begin{equation}

2167:   A = \begin{pmatrix} 

2168:   0 & 1 \\ 1 & 0 \end{pmatrix}.

2169: \end{equation}

2170: %

2171: A monodromy with respect to $\tau_q$ of 

2172: $F_c$ is given by 

2173: % 

2174: \begin{equation} \label{eq:cylinder_monodromy}

2175:   \chi_c (\tau_q) = 

2176:   \exp((q \lambda^{-1} - \bar{q} 

2177:   \lambda )\,A).

2178: \end{equation}

2179: %

2180: The map $F_c:\C \to \Lhol$ is obtained in the 

2181: DPW framework from the triple 

2182: $((\lambda^{-1} + \lambda )A\,dw,\,\Id,\,0 )$. 

2183: Note that we have chosen the base-point 

2184: $w_0 = 0$. The corresponding holomorphic frame 

2185: is 

2186: % 

2187: \begin{equation} \label{eq:Phi_c}

2188:   \Phi_c(w,\,\lambda) = 

2189:   \exp((\lambda^{-1} + \lambda)\,w\, A)

2190: \end{equation}

2191: %

2192: with monodromy with respect to $\tau_q$ given by 

2193: % 

2194: \begin{equation} \label{eq:rho_c}

2195:   \varrho_c(\tau_q) = 

2196:   \exp((\lambda^{-1} + \lambda)\,q\, A).

2197: \end{equation}

2198: %

2199: Note that if $q \in i\R$, then 

2200: $\varrho_c(\tau_q) \in \Lhol$ and 

2201: $\varrho_c(\tau_q) = \chi_c(\tau_q)$. 

2202: 

2203: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2204: %

2205: \newsection{\bf Commuting flows.} 

2206: Denote the abelian sub-algebra of elements of 

2207: $\Lambda_{\mbox{\tiny{$\C^*$}}}\sl_\sigma$ 

2208: that commute with $A$ and have a pole at 

2209: $\lambda = 0$ by 

2210: %

2211: \begin{equation*}

2212:   \mathcal{Z} = \left\{ \phi \,A: 

2213:   \phi (\lambda) = \Sigma

2214:   \phi_j \lambda^j \mbox{ odd },\, j\geq -K,\, 

2215:   K = 2g +1 \in \N \right\}. 

2216: \end{equation*}

2217: %

2218: Let $h \in \LGP,\,\eta \in \mathcal{Z}$ and 

2219: define $h \sharp \eta$ via the $r$--Iwasawa 

2220: decomposition 

2221: %

2222: \begin{equation}

2223:   h \exp(\eta) = U \, h \sharp \eta.

2224: \end{equation}

2225: % 

2226: Further, for $F = h \# F_c$ define 

2227: $F \sharp \eta := (h \sharp \eta) \# F_c$. 

2228: By \cite{BurP:dre}, Proposition 4.1, this is 

2229: an action on the dressing orbit of $F_c$: 

2230: If $\eta_1,\,\eta_2 \in \mathcal{Z}$, then 

2231: $F \sharp (\eta_1 + \eta_2) = 

2232: (F \sharp \eta_2)\sharp\eta_1  = (F \sharp \eta_1)\sharp\eta_2$ and 

2233: if $\eta_1 - \eta_2 \in \Lambda^+_r \sl$, then 

2234: $F \sharp \eta_1 = F \sharp \eta_2$. 

2235: By construction $\zeta = \phi A \in \mathcal{Z}$ 

2236: is analytic for $\lambda \in \C^*$ with a pole at 

2237: $\lambda = 0$. Expanding $\phi(\lambda) = 

2238: \sum_{j\geq -K} \phi_j \lambda^j$, observe that 

2239: $\exp(\zeta) \# F_c = \exp(\eta) \# F_c$ 

2240: where $\eta = \phi^\prime A$ and 

2241: $\phi^\prime = \sum_{j=-K}^0\phi_j \lambda^j$ is 

2242: analytic on $\CP \setminus \{ 0 \}$. Since  

2243: $\eta^* = \phi^* A$ is analytic 

2244: in $\C$, $e^{\eta - \eta^*} \in \Lhol$ and 

2245: $e^{\zeta}\# F_c = e^{\eta-\eta^*}F_c$. 

2246: Hence, when dressing with $e^\zeta$, we restrict 

2247: without loss of generality to 

2248: $\zeta \in \mathcal{Z}$ of the form 

2249: %

2250: \begin{equation} \label{eq:zeta}

2251:   \zeta = \sum_{j=1}^K (\phi_j \lambda^{-j} - 

2252:   \bar{\phi}_j \lambda^j ) A

2253: \end{equation}

2254: %

2255: where $\phi_j \in \C$ and summation is over odd 

2256: indices $j=1,\,3,\ldots,K = 2g+1 \in \N$. Then

2257: %

2258: \begin{equation}

2259:   \exp (\zeta) \, \# \, 

2260:   F_c(w,\,\lambda) = 

2261:   \exp \bigl(\sum_{j=3}^{K} (\phi_j \lambda^{-j} - 

2262:   \bar{\phi}_j \lambda^j )A \bigr) \, 

2263:   F_c(w+\phi_1,\,\lambda).

2264: \end{equation}

2265: %

2266: %and in particular, for $K=1$ we have 

2267: %$\exp((\phi_0\lambda^{-1}-\bar{\phi}_0\lambda)A) 

2268: %\# F_c(z) = F_c(z + \phi_0)$. 

2269: 

2270: For $\zeta \in \mathcal{Z}$ of the form 

2271: \eqref{eq:zeta} we have $e^{t\zeta}\in \Lhol$ 

2272: for all $t \in \R$. 

2273: Hence, for any extended unitary frame 

2274: $F \in \mathcal{F}(\MT)$, we have 

2275: $e^{t\zeta}\#F = e^{t\zeta}F$ which on the level 

2276: of the immersions has the effect 

2277: %

2278: \begin{equation*}

2279:   f \longmapsto \exp(t\,\zeta)\,f\,\exp(-t\,\zeta) - 

2280:   \tfrac{1}{2H} i\lambda t 

2281:   \tfrac{\partial \zeta}{\partial \lambda}.

2282: \end{equation*}

2283: %

2284: Each $\zeta \in \mathcal{Z}$ defines a flow on the 

2285: dressing orbit of the standard round cylinder by 

2286: %

2287: \begin{equation}

2288:   F_t := F \sharp t \zeta = (h\sharp t\zeta)\#F_c 

2289:   = U(\lambda,\,t)^{-1} h\#e^{t\zeta}F_c

2290: \end{equation}

2291: %

2292: where we have defined $U(\lambda,\,t)$ via the 

2293: $r$--Iwasawa decomposition 

2294: $he^{t\zeta}=U(\lambda,\,t)(h\sharp t\zeta)$ and 

2295: $F = h \# F_c$ for some $h \in \LGP$. 

2296: The flow $F_t$ is called 

2297: \textit{trivial} if and only 

2298: if $F_t = V(t) F V(t)^{-1}$ for a 

2299: $\lambda$--independent map $V:\R \to \SU$ 

2300: and $F$ is of \textit{finite type} if and 

2301: only if the subspace 

2302: $\mathcal{Z}^\prime \subset \mathcal{Z}$ of 

2303: trivial flows has finite co-dimension in 

2304: $\mathcal{Z}$.

2305: 

2306: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2307: 

2308: \newsection{\bf Monodromy.} 

2309: Let $h \in \LGP$ and $hF_c = FB$ the 

2310: $r$--Iwasawa decomposition of $hF_c$ such that 

2311: at the  base-point $w_0=0$ we have 

2312: $F_c(0,\,\lambda) = F(0,\,\lambda) = \Id$ and 

2313: $h(\lambda) = B(0,\,\lambda)$. 

2314: Let us assume that $F$ has symmetry 

2315: $(\tau_q,\,\chi(\tau_q))$. 

2316: From \eqref{eq:monodromy_dressed1} we obtain

2317: %

2318: \begin{equation}\label{eq:dress_hol} 

2319:   \chi (\tau_q) = h\, \chi _c(\tau_q)\, B(q)^{-1}.

2320: \end{equation}

2321: % 

2322: For completeness of exposition we prove the 

2323: following result, first obtained in sections 

2324: 3.1--3.4 of \cite{DorH:per}. 

2325: 

2326: \mylemma{}\label{th:f_q}

2327: Let $h \in \LGP$ and $F_c$ 

2328: be as in \eqref{eq:F_c}. 

2329: Assume that $h \# F_c$ has monodromy 

2330: $\chi(\tau_q) \in \Lhol$. 

2331: Then there exists an odd function 

2332: $f_q :C_r \to \C$ with a holomorphic extension 

2333: to $I_r$ such that 

2334: $B(q)^{-1} = \exp(f_q A) \, h^{-1}$. 

2335: Setting 

2336: %

2337: \begin{equation} \label{eq:p_q}

2338:   p_q(\lambda) = q\lambda^{-1}-\bar{q}\lambda 

2339:   + f_q(\lambda),

2340: \end{equation}

2341: % 

2342: we may write the monodromy of $h \# F_c$ as 

2343: %

2344: \begin{equation} \label{eq:dress_monodromy}

2345:   \chi (\tau_q) = h\, \exp(p_q A)\, h^{-1}.

2346: \end{equation}

2347: %

2348: 

2349: \Proof 

2350: Let $\chi_c(\tau_q)$ be the monodromy of 

2351: $F_c$ as in \eqref{eq:cylinder_monodromy}.

2352: Then

2353: %

2354: \begin{equation}

2355:   \chi_c(\tau_q) F_c = \tau_q^* F_c = 

2356:   h^{-1} \chi (\tau_q) h F_c B^{-1} \tau_q^* B

2357: \end{equation}

2358: %

2359: in combination with \eqref{eq:dress_hol} gives 

2360: $B(q)^{-1}B(0)  F_c = F_c \, \tau_q^* B^{-1} B$. 

2361: 

2362: Hence $[B(q)^{-1}B(0), A] = 0$ and for the matrix 

2363: %

2364: \begin{equation}\label{eq:T}

2365:   T = \frac{1}{\sqrt{2}}  \begin{pmatrix} 

2366:   1 & 1 \\ -1 & 1 \end{pmatrix}

2367: \end{equation}

2368: %

2369: we have $TAT^{-1} = \sigma_3$ and 

2370: $[TB(q)^{-1}B(0)T^{-1}, \sigma_3] =0$. 

2371: Thus 

2372: $TB(q)^{-1}B(0)T^{-1} = 

2373: \mathrm{diag}[s_q,\, s_q^{-1}]$ 

2374: for some holomorphic function $s_q:I_r \to \C^*$. 

2375: Defining $s_q = \exp(f_q)$ yields 

2376: $B(q)^{-1}B(0) = \exp(f_q A)$. 

2377: Since $B(0) = h$, this proves the claim.

2378: \QED

2379: 

2380: The point of Lemma \ref{th:f_q} is that 

2381: we can now write a monodromy of an extended frame 

2382: in the dressing orbit of the vacuum as

2383: %

2384: \begin{equation}\label{eq:sinhcosh}

2385:   \cosh(p_q) \,\Id + \sinh(p_q) \,h\,A\,h^{-1}

2386: \end{equation}  

2387: %

2388: and much of the analysis reduces to the study of 

2389: the two scalar functions

2390: %

2391: \begin{equation} \label{eq:alphabeta}

2392:   \alpha=\cosh(p_q)\mbox{ and }\beta = \sinh(p_q).

2393: \end{equation}

2394: %

2395: In analogy to Corollary \ref{co:h-split1} 

2396: we have the following result for the monodromy 

2397: representation of surfaces in the dressing orbit 

2398: of the cylinder.

2399: %

2400: 

2401: \mytheorem{}\label{th:h-split_cyl}

2402: Let $h \in \LGP$ and $F_c$ 

2403: be as in \eqref{eq:F_c}. 

2404: Assume that $h \# F_c$ has symmetry 

2405: $(\tau_q,\,\chi(\tau_q))$. 

2406: Then $h = \mathcal{M}(\tau_q)\,\mathcal{C}(\tau_q)$ 

2407: with $\mathcal{M}(\tau_q) \in \Lambda_{r^\prime} 

2408: \SU_\sigma$ for suitable $r \leq r^\prime < 1$ and 

2409: $[\,\mathcal{C}(\tau_q),\,A\,]=0$.

2410: 

2411: \Proof For $T$ defined in \eqref{eq:T} and by

2412: Lemma \ref{th:f_q}, we may write

2413: %

2414: \begin{equation} \label{eq:chi}

2415:   \chi(\tau_q) = h \,T^{-1} \,

2416:   \mathrm{diag}[\,\exp(-p_q),\, 

2417:   \exp(p_q)\,]\, T\, h^{-1}.

2418: \end{equation} 

2419: %

2420: for $0< \mid \lambda \mid < r$. 

2421: Let $h T^{-1} = \bigl( \begin{smallmatrix} 

2422: a & b \\ c & d \end{smallmatrix} \bigr)$.

2423: The diagonal entries of $\chi(\tau_q)$ 

2424: are analytic functions in $\lambda \in \C^*$ 

2425: and given by

2426: %

2427: \begin{equation} \begin{split} 

2428:   a \, d \, \exp(-p_q) - b\,c\,\exp(p_q),\\

2429:   a \, d \, \exp(p_q) - b\,c\,\exp(-p_q). \end{split}

2430: \end{equation}

2431: % 

2432: Adding these implies that $\alpha = \cosh(p_q)$ 

2433: is an analytic function on $\C^*$. 

2434: More precisely, $\alpha$ has an analytic 

2435: extension from $ 0 <\mid \lambda \mid <r$ to $\C^*$.

2436: Solving for $\exp(p_q)$ we obtain

2437: %Then 

2438: %

2439: %\begin{equation}

2440: %  p_q = \cosh^{-1}(g) = 

2441: %  \log(g+\sqrt{g^2 -1})

2442: %\end{equation} 

2443: %

2444:  $\exp(p_q) = \alpha + \sqrt{\alpha^2-1}$. 

2445: This is analytic on 

2446: the set $\mathbb S \subset \C^*$ obtained by 

2447: making branchcuts from the odd--ordered roots of 

2448: $\alpha^2-1$ in the usual fashion. 

2449: Note that the equation \eqref{eq:chi} above 

2450: implies that $\mu_{\pm} = \exp (\pm p_q)$ 

2451: are the eigenvalues of $\chi (\tau_q)$. 

2452: Hence, by Lemma \ref{th:S} we know that 

2453: $\mathbb S$ is open, connected, and dense 

2454: in $\C^*$. Moreover we have $S^1 \subset \mathbb S$ 

2455: and $\mathbb S = {\mathbb S }^*$.

2456: 

2457: Consequently, $\exp(-p_q)$ is holomorphic on 

2458: $\mathbb S$. Multiplying the $(1,1)$ entry of 

2459: $\chi(\tau_q)$ by $\exp(p_q)$ and the $(2,2)$ 

2460: entry of $\chi(\tau_q)$ by $\exp(-p_q)$ 

2461: and subtracting the resulting equations implies 

2462: that $bc \sinh (p_q)$ is holomorphic on $\mathbb S$. Note that $\beta = \sinh (p_q) \not\equiv 0$. Hence

2463: $bc$ is meromorphic on $\mathbb S$. 

2464: Similarly one shows that $ad$ is meromorphic on 

2465: $\mathbb S$. 

2466: 

2467: The off--diagonal terms of $\chi(\tau_q)$ 

2468: are $2 ab \beta$ and $-2 cd \beta$. 

2469: Hence both $ab$ and $cd$ are meromorphic on 

2470: $\mathbb S$. 

2471: If $d \neq 0$ on the circle $C_r$, then as in 

2472: \eqref{eq:h-split_case1}, we may write 

2473: $hT^{-1} = \hat{h}D_1$ with 

2474: $D_1= \mathrm{diag}[1/d,\,d]$ 

2475: and $\hat{h}$ meromorphic on $\mathbb S$. 

2476: 

2477: Defining $\mathcal{C}(\tau_q) = T^{-1}D_1 T$, 

2478: $\mathcal{M}(\tau_q) = \hat{h} T$ yields 

2479: $h = \mathcal{M}(\tau_q) \mathcal{C}(\tau_q)$ 

2480: with $[\, \mathcal{C}(\tau_q),\,A \,] =0$ and 

2481: $\mathcal{M}(\tau_q)$ meromorphic on $\mathbb S$.

2482: Arguing as in the proof of Theorem \ref{th:h-split}, 

2483: $d \equiv 0$ on an arc $C_s \cap \mathbb S$ 

2484: implies that both $b,\,c \neq 0$ on $\mathbb S$ 

2485: and we may write $hT^{-1} = \widetilde{h}D_2$ with 

2486: $D_2= \mathrm{diag}[c,\,1/c]$ and 

2487: $\widetilde{h}$ meromorphic on $\mathbb S$. 

2488: Defining $\mathcal{C}(\tau_q) = T^{-1}D_2 T$, 

2489: $\mathcal{M}(\tau_q) = \widetilde{h} T$ yields 

2490: $h = \mathcal{M}(\tau_q) \mathcal{C}(\tau_q)$ 

2491: with $[\,\mathcal{C}(\tau_q),\,A\,] =0$ and 

2492: $\mathcal{M}(\tau_q)$ meromorphic on $\mathbb S$ 

2493: in this case. In either case, on $C_r$ we may write

2494: %

2495: \begin{equation} \label{eq:U} 

2496:   \chi(\tau_q) = \mathcal{M}(\tau_q)\, 

2497:   \exp(p_q\,A)\,\mathcal{M}(\tau_q)^{-1}.

2498: \end{equation}

2499: %

2500: Introducing an additional cut from $0$ to $-\infty$ 

2501: we obtain a simply connected domain 

2502: $\hat{\mathbb S} = \hat{\mathbb S}^*$, 

2503: on which we can take the logarithm of 

2504: the holomorphic function $\exp(p_q)$, 

2505: thus defining $p_q$ on $\hat{\mathbb S}$. 

2506: Then on $\hat{\mathbb S}$ we combine 

2507: $\chi(\tau_q)^* = \chi(\tau_q)^{-1}$ 

2508: with \eqref{eq:U} to obtain

2509: %

2510: \begin{equation} \label{eq:mstarm}

2511:   \mathcal{M}^*(\tau_q) \,\mathcal{M}(\tau_q) 

2512:   \,\exp(-p_q\,A) = \exp(p^*_q\,A)\, 

2513:   \mathcal{M}^*(\tau_q) \,\mathcal{M}(\tau_q).

2514: \end{equation}

2515: %

2516: For $T$ defined in \eqref{eq:T}, recall that 

2517: $T\,A\,T^{-1} = \mathrm{diag}[1,\,-1]$. 

2518: Since $T \in \SU$, the matrix  

2519: $X:= T\,\mathcal{M}^*(\tau_q) \,

2520: \mathcal{M}(\tau_q)\,T^{-1}$ is hermitian and 

2521: of the form $X = W\,W^*$. In particular, if we write 

2522: $X = \bigl( \begin{smallmatrix} x & y \\ 

2523: \bar{y} & z \end{smallmatrix} \bigr)$, 

2524: then $x \neq 0$.

2525: We can rewrite \eqref{eq:mstarm} as 

2526: $X\,T\,\exp(-p_q\,A)\,T^{-1} = 

2527: T\,\exp(p^*_q\,A)\,T^{-1}\,X$ and obtain

2528: %

2529: \begin{equation}

2530:   \begin{pmatrix} x & y \\ \bar{y} & z \end{pmatrix}

2531:   \begin{pmatrix} \exp(-p_q) & 0 \\ 0 &  \exp(p_q) 

2532:   \end{pmatrix} = \begin{pmatrix} \exp(p^*_q) & 0 \\ 

2533:   0 & \exp(-p^*_q) \end{pmatrix} 

2534:   \begin{pmatrix} x & y \\ \bar{y} & z \end{pmatrix}.

2535: \end{equation}

2536: %

2537: Since $x \neq 0$, then $\exp(-p_q) = \exp(p^*_q)$ 

2538: on $\hat{\mathbb S}$ and we conclude that 

2539: $\cosh(p^*_q) = \cosh(p_q)$ and 

2540: $\sinh(p^*_q) = - \sinh(p_q)$. Consequently 

2541: $\exp(p^*_q\,A) = \exp(-p_q\,A)$ 

2542: on $\hat{\mathbb S}$ and equation 

2543: \eqref{eq:mstarm} reads 

2544: %

2545: \begin{equation}

2546:   [\,\mathcal{M}^*(\tau_q) \mathcal{M}(\tau_q),\,

2547:   \exp(-p_q\,A)\,] = 0.

2548: \end{equation}

2549: %

2550: From the proof of Corollary \ref{th:S} we 

2551: now obtain some $L(\tau_q)$ 

2552: and $r' \in [r,1)$ such that 

2553: $\mathcal{M}(\tau_q)L(\tau_q) \in 

2554: \Lambda_{r'}\SU_\sigma$ and 

2555: $[\,L^{-1}(\tau_q),\,\mathcal{C}(\tau_q)\,]=0$, and 

2556: concludes the proof. \QED

2557: %

2558: 

2559: 

2560: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2561: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2562: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2563: 

2564: Before going on to evaluate the result above 

2565: to examples we would like to relate the work 

2566: above to \cite{DorH:per}. Of particular 

2567: interest to us is equation (3.6.7) there:

2568: %

2569: \begin{equation} \label{eq:x}

2570:   h_+ = \frac{i}{2 \sqrt{\tilde{b}}}

2571:   \begin{pmatrix}

2572:   (x-x^{-1}) \tilde{b} & (x+x^{-1}) \tilde{b} \\

2573:   (x+x^{-1}) - (x-x^{-1}) \tilde{a} & 

2574:   (x-x^{-1}) - (x+x^{-1}) \tilde{a} \end{pmatrix}

2575: \end{equation}

2576: %

2577: where $\tilde{a}$ and  $\tilde{b}$ 

2578: denote the entries of the matrix $hAh^{-1}$ 

2579: in the $(11)$--position and the $(12)$--position 

2580: respectively.

2581: This is the general form of a dressing matrix 

2582: having on $\chi$ the same effect as $h$.

2583: The question addressed in the Theorem above 

2584: is thus equivalent to the question whether one 

2585: can choose $x$ so that $h_+$ is unitary on $S^1$.

2586: 

2587: \mylemma{} The matrix $h_+$ is unitary on 

2588: $S^1$ if and only if

2589: %

2590: \begin{equation}

2591:   |\,x\,|^2 = \frac{1 + \tilde{a} +  

2592:   \sqrt{|\tilde{b}|^2}}{1-\tilde{a} + 

2593:   \sqrt{|\tilde{b}|^2}}. 

2594: \end{equation}

2595: %

2596: This equation can be solved with some $x$ 

2597: holomorphic at $\lambda =0$ and well defined 

2598: on some sufficiently well cut complex plane.

2599: 

2600: \Proof

2601: The unitarity condition for $h_+$ yields 

2602: two equations. However, a straightforward 

2603: computation shows that

2604: these two equations are equivalent. 

2605: It thus remains to consider

2606: %

2607: \begin{equation}

2608:   (x^* - (x^*)^{-1}) \sqrt{\tilde{b}^*} = 

2609:   \frac{1}{\sqrt{\tilde{b}}} \lbrack (x-x^{-1}) - 

2610:   (x+x^{-1}) \tilde{a} \rbrack.

2611: \end{equation}

2612: %

2613: Splitting this equation into real and imaginary 

2614: part yields two equations. 

2615: However, again, a straightforward computation 

2616: shows that these two equations are equivalent.

2617: It thus suffices to consider

2618: %

2619: \begin{equation}

2620:   (Re(x)^* - (Re(x)^*)^{-1}) \sqrt{|\tilde{b}|^2} =  

2621:   (Re(x)-Re(x)^{-1}) - (Re(x)+ Re(x)^{-1}) \tilde{a}.

2622: \end{equation} 

2623: %

2624: Note that we have used here that $\tilde{a}$ 

2625: is real on $S^1$ by \cite{DorH:per}, Theorem 3.7.

2626: Using $x^{-1} = x/{\mid x \mid}^2$ this 

2627: equation rewrites directly into the claim.

2628: \QED

2629: %

2630: 

2631: \mycorollary{}

2632: $h_+$ can be chosen diagonal if and only 

2633: if $\tilde{a} =0$. In this case the 

2634: diagonal entries of $h_+$ are unitary on $S^1$.

2635: 

2636: \Proof

2637: Using \eqref{eq:x} for general $x$ we see that 

2638: $h_+$ is diagonal only if 

2639: $x + x^{-1}$ vanishes identically, 

2640: since $\tilde{b}$ cannot vanish by b) and c) of 

2641: \cite{DorH:per}, Theorem 3.7.

2642: Inserting this into the second off-diagonal 

2643: expression for $h_+$ we obtain $\tilde{a} =0$. 

2644: The converse is obvious choosing $x = i$.

2645: Assume now that $h_+$ has been chosen as a 

2646: diagonal matrix. Since  $\tilde{a}$ vanishes 

2647: in this case as just shown, c) and e) of 

2648: \cite{DorH:per} imply that   

2649: $\tilde{b}$ is unitary on $S^1$.

2650: Finally, in view of \eqref{eq:x}, we obtain

2651: %

2652: \begin{equation}

2653:   h_+ = \frac{i}{2 \sqrt{\tilde{b}}}(x - x^{-1})

2654:         \mathrm{diag}[b,\,1].

2655: \end{equation}

2656: %

2657: Moreover, $x = -x^{-1}$, whence $x = \pm i$. 

2658: The condition $\det h_+ = 1$ is now satisfied.

2659: Thus necessarily 

2660: $h_+ = \pm \mathrm{diag}[\sqrt{\tilde{b}},\,

2661: 1/\sqrt{\tilde{b}}]$ 

2662: and the diagonal entries of $h_+$ are unitary 

2663: on $S^1$. \QED

2664:  

2665: An immediate consequence of the above discussion leads to 

2666: following observation.

2667: 

2668: \mycorollary{}

2669: Let $h \in \LGP$ such that 

2670: $h\, \exp(p_q A)\, h^{-1} \in \Lhol$ for some 

2671: translation $\tau_q$. If $h$ is diagonal, 

2672: then $h^4$ is rational on $\CP$ and

2673: $h^2$ is analytic in 

2674: $\CP \setminus \mathcal{H}$, where $\mathcal{H}$ 

2675: is the set obtained by making appropriate 

2676: branchcuts from the points lying in the set

2677: %

2678: \begin{equation} \label{eq:branchset}

2679:   \mathcal{H}^\prime = 

2680:   \{ \lambda \in \CP : |\lambda | > r \mbox{ and } 

2681:   p_q(\lambda) \in \pi i \Z \} \cup \{ 

2682:   \mbox{ singularities of } p_q(\lambda) \}

2683: \end{equation}

2684: %

2685: and $h$ is unitary on $S^1$. \QED

2686: 

2687: \myremark{}

2688: Writing $h = \mathrm{diag}[a,\,1/a]$, we have

2689: that $\chi(\tau_q)= \alpha\,\Id + 

2690: \beta\,\mathrm{off}[a^2,\,a^{-2}]$ 

2691: is analytic on $\C^*$. 

2692: Hence $a^2$ and $a^{-2}$ are analytic in 

2693: $\mathcal{H}$ and even in $\lambda$.

2694: Following the procedure for splitting 

2695: a matrix we first untwist $h$. 

2696: The above procedure for 

2697: constructing the factorisation 

2698: $h = \mathcal{MC}$ gives

2699: %

2700: \begin{equation} \label{eq:diagonal_h-split}

2701:   \mathcal{M} = \frac{1}{2} \begin{pmatrix} 

2702:   1+a^2 & 1-a^2 \\

2703:   -1 + a^{-2}  & 1 + a^{-2} \end{pmatrix}, \,

2704:   \mathcal{C} = \frac{1}{2} \begin{pmatrix} 

2705:   a+1/a & a-1/a \\ a-1/a & a+1/a \end{pmatrix},

2706: \end{equation}

2707: %

2708: where a really denotes the untwisted function.

2709: Now we need to twist again. This way we obtain 

2710: the original function $a$ and in addition 

2711: the off-diagonal terms are multiplied by 

2712: $\lambda$ (in the $(12-)$position) and 

2713: by $\lambda^{-1}$ in the $(21)-$position 

2714: respectively. Clearly this leads to a matrix 

2715: $\mathcal{M}$ which is twisted as required. 

2716: %If $a^{*\,2} = \,a^{-2}$ on 

2717: %$\mathcal{H}$, then 

2718: %$\mathcal{M}^* = \mathcal{M}^{-1}$.

2719: 

2720: 

2721: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2722: 

2723: \newsection{\bf Delaunay Surfaces.} 

2724: \label{sec:Delaunayexample} 

2725: As an example, we discuss how to dress the 

2726: vacuum into a Delaunay surface and compute 

2727: the splitting of this dressing matrix according 

2728: to Theorem \ref{th:h-split_cyl}. It is proven in 

2729: \cite{Kil:del} that an associated 

2730: family of the Delaunay surface with neck radius 

2731: $\omega$ is generated by the triple 

2732: $(D\, dz/z,\,\Id,\,0 )$, where

2733: %

2734: \begin{equation} \label{eq:delaunay}

2735:   D = \begin{pmatrix} 0 & \alpha \lambda^{-1} + 

2736:   \beta \lambda \\ \beta \lambda^{-1} + 

2737:   \alpha \lambda & 0 \end{pmatrix},

2738: \end{equation}

2739: %

2740: and $\alpha,\, \beta \in \R$ with 

2741: $\alpha + \beta = 1/2$ and

2742: $\omega = \tfrac{1}{2H}

2743: (1 - \sqrt{1 - 16 \alpha \beta})$. 

2744: The monodromy of the unitary frame with respect to 

2745: the translation $\tau_q$ for $q = 2 \pi i$ is 

2746: given by $\exp ( 2 \pi i D )$. 

2747: Therefore, to dress the 

2748: standard cylinder into a Delaunay surface, 

2749: it suffices to determine a diagonal 

2750: $h = \mathrm{diag}[a,\, 1/a] \in \LGP$

2751: for some suitable $r>0$ such that 

2752: $p_q h A h^{-1} = D$, which is equivalent to 

2753: the two equations

2754: %

2755: \begin{equation*}

2756:   a^{\pm 2}(q\lambda^{-1}- \bar{q}\lambda + f_q )= 

2757:   2 \pi i  (\alpha\lambda^{\mp 1} + 

2758:   \beta \lambda^{\pm 1} ).

2759: \end{equation*}

2760: %

2761: Solving both equations for $a^2$ and equating gives

2762: %

2763: \begin{equation} \label{eq:Del_ex}

2764:   2 \pi i \sqrt{(\alpha + \beta \lambda^2)

2765:   (\beta + \alpha \lambda^2)} =

2766:   q - \bar{q}\lambda^2 + \lambda f_q,

2767: \end{equation}

2768: % 

2769: which in turn yields 

2770: $a^2 = \sqrt{(\alpha + \beta \lambda^2)/

2771: (\beta + \alpha \lambda^2)}$. 

2772: Evaluating \eqref{eq:Del_ex} at $\lambda = 0$ gives 

2773: $q = 2\pi i \sqrt{\alpha\,\beta}$. 

2774: If $\rho:= \min \{ |\sqrt{\alpha/\beta}| ,\, 

2775: | \sqrt{\beta/\alpha}|\}$, 

2776: then $h \in \LGP$ for $0<r<\rho$. 

2777: It is easily verified that 

2778: $p(\lambda) = 2 \pi i \sqrt{- \det D}$ 

2779: vanishes at the singularities of $h$. 

2780: In summary, the matrix that dresses the vacuum 

2781: into a Delaunay surface with 

2782: neck radius $\omega$ is given by 

2783: %

2784: \begin{equation}

2785:   h(\lambda) = \mathrm{diag}[\, 

2786:   \sqrt[4]{\tfrac{\alpha + \beta \lambda^2}

2787:   {\beta + \alpha \lambda^2}},\, 

2788:   \sqrt[4]{\tfrac{\beta + \alpha \lambda^2}

2789:   {\alpha + \beta \lambda^2}}\, ] 

2790: \end{equation} 

2791: % 

2792: Untwisting $h$, factoring 

2793: $h = \mathcal{M}\mathcal{C}$ as 

2794: in \eqref{eq:diagonal_h-split} and twisting back 

2795: gives

2796: %

2797: \begin{equation} 

2798:   \mathcal{M} = \frac{1}{2} \begin{pmatrix} 

2799:   1+ \sqrt{\tfrac{\alpha + \beta \lambda^2}

2800:   {\beta + \alpha \lambda^2}} & 

2801:   \lambda - \lambda \sqrt{\tfrac{\alpha + 

2802:   \beta \lambda^2}{\beta + \alpha \lambda^2}} \\ 

2803:   \lambda^{-1}\sqrt{\tfrac{\beta + 

2804:   \alpha \lambda^2}{\alpha + \beta \lambda^2}} 

2805:   -\lambda^{-1} & 

2806:   1 + \sqrt{\tfrac{\beta + \alpha \lambda^2}

2807:   {\alpha + \beta \lambda^2}} \end{pmatrix}.

2808: \end{equation}

2809: % 

2810: Evidently $\mathcal{M}^* = \mathcal{M}^{-1}$ 

2811: so that this is actually the decomposition 

2812: according to Theorem \ref{th:h-split_cyl}. 

2813: The domain of $\mathcal{M}$ is the genus one 

2814: hyperelliptic curve $\mu^2  + \det D = 0$, 

2815: the \textit{spectral curve} 

2816: of the underlying Delaunay surface. 

2817: 

2818: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2819: 

2820: \newsection{\bf Finite Blaschke Products.} 

2821: The function $f_q$ occurring in Theorem 

2822: \ref{th:f_q} is not always as easy to deal with as 

2823: in the above example nor can it generally be 

2824: explicitly computed. Nonetheless, even the 

2825: trivial case $f_q \equiv 0$ is quite interesting 

2826: and the corresponding dressing matrices are of such 

2827: a simple form that the involved Iwasawa 

2828: decomposition can be explicitly computed.  

2829: 

2830: \mylemma{}\label{th:f_q=0} 

2831: Let $h \in \LGP$ and $F_c$ be the unitary frame 

2832: of the round cylinder with monodromy $\chi_c(\tau_q) 

2833: = \exp((q\lambda^{-1}-\bar{q}\lambda)A)$ for some 

2834: $q \in \C^*$ with 

2835: %

2836: \begin{equation} \label{eq:q_condition}

2837:   \sqrt{\bar{q}} - \sqrt{q} \notin \pi \,\Z.

2838: \end{equation}

2839: % 

2840: Assume $F = h \# F_c$ has a monodromy 

2841: $h\,\chi_c(\tau_q)\,\exp(f_q\,A)\,h^{-1} \in \Lhol$ 

2842: as in equation \eqref{eq:dress_monodromy}. 

2843: Then the following are equivalent 

2844: %

2845: \begin{enumerate}

2846: \item $h \chi_c(\tau_q) h^{-1} \in \Lhol$ 

2847: \item $f_q \equiv 0$

2848: \item $h\,A\,h^{-1}$ is rational on $\CP$ with only 

2849:       simple poles.

2850: \end{enumerate}

2851: 

2852: \Proof

2853: We first show that (i) implies (ii). 

2854: The assumption (i) allows us to write the 

2855: monodromy of $F$ as 

2856: $h\,\chi_c(\tau_q)\,h^{-1} \cdot 

2857: h\exp(f_q\,A)\,h^{-1}$, where the first group 

2858: of factors is unitary and the second group of 

2859: factors is contained in $\LGP$ and has first 

2860: term $I$. Therefore the condition that the whole 

2861: expression is unitary is equivalent to 

2862: $h\exp(f_q\,A)\,h^{-1}$ being unitary. 

2863: But it is also in $\LGP$. Therefore it is in $\U1$.

2864: But since it starts with $I$, it is identically 

2865: equal to $I$. As a consequence $\exp(f_q\,A) = I$.

2866: Hence $f_q \equiv 2 \pi k$ for some $k \in \Z$ 

2867: and $k =0$, since $f_q$ is an odd function by 

2868: Lemma \ref{th:f_q}. This shows that (i) implies (ii).

2869: 

2870: Let us prove that (ii) implies (iii). Recall from 

2871: \eqref{eq:sinhcosh} and \eqref{eq:alphabeta} that 

2872: $h\,\exp(p_q\,A)\,h^{-1} = \alpha\,\Id + 

2873: \beta\,h\,A\,h^{-1}$. Since $f_q =0$ we have 

2874: $p_q = \lambda^{-1}q - \lambda \bar{q}$ and 

2875: therefore 

2876: $\beta = \sinh(\lambda^{-1}q-\lambda\bar{q})$ is 

2877: holomorphic on $\C^*$. 

2878: By assumption, $\beta\,h\,A\,h^{-1}$ 

2879: is holomorphic on $\C^*$ so that 

2880: $h\,A\,h^{-1}$ is meromorphic on $\C^*$. 

2881: Let $\lambda_0 \in \C^*$ be a root of 

2882: $\beta = \sinh(\lambda^{-1}q-\lambda\bar{q})$, 

2883: that is, it lies in the set

2884: %

2885: \begin{equation} \label{eq:spectral_set}

2886:   \mathcal{S}_r^q = \left\{ \lambda \in A_r : 

2887:   \lambda^{-1}q - \lambda\,\bar{q} \in \pi i \Z 

2888:   \right\}. 

2889: \end{equation}

2890: %

2891: Then $\beta'(\lambda_0) = \pm(\lambda_0^{-2}q - 

2892: \bar{q}) \neq 0$ if and only if 

2893: $\lambda_0 \neq \pm i \sqrt{q/\bar{q}}$. Our 

2894: assumption \eqref{eq:q_condition} ensures that 

2895: $\pm i \sqrt{q/\bar{q}} \notin \mathcal{S}_r^q$. 

2896: Hence all roots 

2897: of $\beta$ in $\C^*$ are simple. 

2898: Therefore the entries of $h\,A\,h^{-1}$ 

2899: can only have simple poles in $\C^*$ which 

2900: must lie in $\mathcal{S}_r^q$.

2901: Note also that $h\,A\,h^{-1}$ is holomorphic at 

2902: $\lambda = 0$. Hence $h\,A\,h^{-1}$ is a holomorphic 

2903: germ at the origin with a meromorphic 

2904: extension to $\C^*$. In Section 3.5 of 

2905: \cite{DorH:per} it is shown that the squares of the 

2906: entries of $h\,A\,h^{-1}$ are finite at $\infty$. 

2907: Hence $h\,A\,h^{-1}$ is rational on $\CP$ and 

2908: proves that (ii) implies (iii).

2909: 

2910: Finally we show that (iii) implies (i). 

2911: By equations \eqref{eq:sinhcosh} and 

2912: \eqref{eq:alphabeta} we write 

2913: $\chi = \alpha\,\Id + \beta\,hAh^{-1}$ 

2914: and use the fact that 

2915: $\beta\,h\,A\,h^{-1}$ is holomorphic on $\C^*$. 

2916: With assumption (iii) this implies that $\beta$ 

2917: is meromorphic on $\C^*$.

2918: Since $\alpha$ is holomorphic 

2919: on $\C^*$, so is $\alpha^2$ and consequently 

2920: $\beta^2 = \alpha^2 - 1$ is holomorphic on 

2921: $\C^*$, therefore $\beta$ itself is holomorphic 

2922: on $\C^*$. Further, from 

2923: $\chi^* = \chi^{-1} = \alpha\,\Id - 

2924: \beta\,hAh^{-1}$ we deduce that $\alpha$ is real 

2925: on $S^1$, i.e. $\alpha^* = \alpha$. 

2926: And since $\beta^2$ is real and non-positive 

2927: on $S^1$, and since $\beta$ is holomorphic on 

2928: $\C^*$, we obtain $\beta^* = - \beta$. Consequently, 

2929: $(hAh^{-1})^* = hAh^{-1}$ and thus 

2930: $h\,\chi_c(\tau_q)\,h^{-1} = 

2931: \exp((\lambda^{-1}q - \lambda\,q)h\,A\,h^{-1}) 

2932: \in \Lhol$. This proves that (iii) implies (i) 

2933: and concludes the proof of the lemma. \QED

2934: 

2935: 

2936: \myremark{}

2937: 1. The result above seems to indicate that for 

2938: every "type of $f_q$ " one obtains an associated 

2939: type of dressing matrix.

2940: 

2941: 2. In the case $f_q = 0$ and the setting of 

2942: \cite{DorH:per}, the hyperelliptic surface 

2943: associated with the dressing is simply the 

2944: Riemann sphere $S^2$. 

2945: If this class would contain some torus, then one 

2946: would have an example for the 

2947: "singular tori question" of \cite{Bob:tor}.

2948: 

2949: %3. A complete classification for the case 

2950: %$f_q =0$ should follow from \cite{DorH:per}.

2951: 

2952: 3. The set $\mathcal{S}_r^q$ in 

2953: \eqref{eq:spectral_set} is finite for fixed 

2954: $r$. Further $\mathcal{S}_r^q \subset \R$ if and 

2955: only if $q \in i\R$, in which case $q$ also 

2956: satisfies \eqref{eq:q_condition}. Finally, 

2957: it is easy to see that if 

2958: $\lambda \in \mathcal{S}_r^q$ then also 

2959: $\bar{\lambda},\,1/\lambda,\,1/\bar{\lambda} 

2960: \in \mathcal{S}_r^q$, and in particular, 

2961: $\left(\mathcal{S}_r^q \right)^* = \mathcal{S}_r^q$.

2962: 

2963: We apply Lemma \ref{th:f_q=0} to a class of 

2964: diagonal dressing matrices for which $f_q \equiv 0$.

2965: 

2966: \mycorollary{} 

2967: Let $h = \mathrm{diag}[a,\,1/a] \in \LGP$. 

2968: Then $h$ has an extension to $A_{r'}$ for some 

2969: $r' \in [r,\,1]$ such that $h^* = h^{-1}$ on 

2970: $A_{r'}$ and $h \chi_c(\tau_q) h^{-1} \in \Lhol$ 

2971: if and only if $a^2$ is a finite Blascke product  

2972: %

2973: \begin{equation}\label{eq:a^2}

2974:   a^2 (\lambda) = \prod_{j=1}^N 

2975:   \frac{\alpha_j^2 - \lambda^2}

2976:   {1 - \bar{\alpha}_j^2 \lambda^2},\,\alpha_j 

2977:   \in \mathcal{S}_r^q. 

2978: \end{equation}

2979: %

2980: \Proof Assume $h$ has a unitary branch on $S^1$ 

2981: and $h \chi_c(\tau_q) h^{-1} \in \Lhol$. Then by 

2982: Lemma \ref{th:f_q=0} we conclude that

2983: $h\,A\,h^{-1} = \mathrm{off}[a^2,\,1/a^2]$ 

2984: is rational with simple poles and zeroes located in 

2985: $\mathcal{S}_r^q$. Hence $a^2$ is of the form 

2986: \eqref{eq:a^2}. The converse is proven by direct 

2987: verification.

2988: \QED

2989: 

2990: Dressing matrices characterised in 

2991: the previous proposition are thus of the form

2992: %

2993: \begin{equation} \label{eq:prod_simple}

2994:   h = \prod_{j=1}^N \begin{pmatrix}  

2995:   \sqrt{\tfrac{\alpha_j^2 - \lambda^2}

2996:   {1 - \bar{\alpha}_j^2 \lambda^2}} & 0 \\ 0 & 

2997:   \sqrt{\tfrac{1 - \bar{\alpha}_j^2 \lambda^2}

2998:   {\alpha_j^2 - \lambda^2}} \end{pmatrix},\,

2999:   \alpha_j \in \mathcal{S}_r^q

3000: \end{equation}

3001: %

3002: and are a special instance of an interesting 

3003: class to which we now turn our attention. 

3004: 

3005: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3006: 

3007: \newsection{\bf Simple factors, untwisted case.} 

3008: 

3009: In this section we will deal primarily with 

3010: untwisted loops. The previous discussion has 

3011: naturally lead us to a class of special dressing 

3012: matrices, the so called \textit{simple factors} of 

3013: Uhlenbeck \cite{Uhl} and discussed in similar 

3014: context in \cite{TerU} and \cite{Bur:iso}. 

3015: The main feature is that dressing with 

3016: simple factors is explicit. 

3017: This construction, due to Terng and  

3018: Uhlenbeck \cite{TerU} goes as follows:

3019: 

3020: \par We decompose 

3021: $\C^2 = \mathrm{L} \oplus \mathrm{L}^\perp$ 

3022: for $\mathrm{L} = \C \left( \begin{smallmatrix} 

3023: a \\ b \end{smallmatrix} 

3024: \right)$. Then for all $A \in \GL$ we have 

3025: %

3026: \begin{equation} \label{eq:perp}

3027:   \bar{A}^t \mathrm{L} \perp 

3028:   A^{-1} \mathrm{L}^\perp

3029: \end{equation}

3030: % 

3031: The hermitian projection 

3032: $\pi_\l:\C^2 \to \mathrm{L}$ onto $\mathrm{L}$ is 

3033: given by 

3034: %

3035: \begin{equation*}

3036:  \pi_\l =  \frac{1}{|a|^2+|b|^2} \begin{pmatrix}

3037:   |a|^2 & a\bar{b} \\ \bar{a}b & |b|^2 

3038:   \end{pmatrix}.

3039: \end{equation*}

3040: %

3041: Let $\alpha \in I_1^* = \left\{ \lambda \in \C : 

3042: 0 < |\lambda | < 1 \right\}$. Then the map 

3043: %

3044: \begin{equation*}

3045:   \tau_\alpha(\lambda) = \frac{\alpha - \lambda}

3046:   {1-\bar{\alpha}\lambda}

3047: \end{equation*}

3048: %

3049: is invariant under $\varrho$. For given 

3050: $\mathrm{L}$ and $\alpha \in I_1$, a 

3051: \textit{simple factor} \cite{TerU} is a 

3052: loop of the form 

3053: %

3054: \begin{equation} \label{eq:psi}

3055:   \psi_{\alpha,\l}(\lambda) = \pi_\l + 

3056:   \tau_\alpha(\lambda) \,\pi_\l^\perp.

3057: \end{equation}

3058: % 

3059: By construction, $\psi_{\alpha,\l}: \CP \setminus 

3060: \left\{ \alpha,\,1/\bar{\alpha} \right\} \to \GL$ 

3061: is analytic and since $|\alpha |>0$, 

3062: %

3063: \begin{equation}

3064:   \psi_{\alpha,\l} \in \Lambda_r^+ \GL

3065:   \mbox{ for } r < | \alpha |.

3066: \end{equation}

3067: % 

3068: Clearly, simple factors are not twisted.

3069: Further, since $\psi_{\alpha,\l}^* = 

3070:   \psi_{\alpha,\l}^{-1} = 

3071:   \pi_\l + \tau_\alpha^{-1} \pi_\l^\perp$, 

3072: we have that 

3073: %

3074: \begin{equation}

3075:   \psi_{\alpha,\l} \in \Lambda_r \mathbf{U}(2)

3076:   \mbox{ for } r > | \alpha |.

3077: \end{equation}

3078: %

3079: For later use we also 

3080: note the fact that for any $A \in \Utwo$ 

3081: we have 

3082: %

3083: \begin{equation} \label{eq:change}

3084:   \psi_{\alpha,A\l} = A \,\psi_{\alpha,\l}\,A^{-1}.

3085: \end{equation}

3086: %

3087: 

3088: % 

3089: Let $U \in  \Lambda_{\mbox{\tiny{$\C^*$}}}\SU,\,

3090: \alpha \in I_1^*$ 

3091: and $\mathrm{L} \in \C\mathbb{P}^2$ and  

3092: $\psi_{\alpha,\l}$ be the corresponding simple 

3093: factor. Then it can be shown, see e.g \cite{KilSS}, 

3094: that 

3095: %

3096: \begin{equation}

3097:   \psi_{\alpha,\l}\,U\,

3098:   \psi_{\alpha,\overline{U(\alpha)}^t \l}^{-1} \in 

3099:   \Lambda_{\mbox{\tiny{$\C^*$}}}\SU

3100: \end{equation}

3101: %

3102: Consequently, we have an explicit 

3103: $r$--Iwasawa decomposition of $\psi_{\alpha,\l}\,U$ 

3104: for $r < |\alpha |$, given by 

3105: %

3106: \begin{equation} \label{sdf}

3107:   \psi_{\alpha,\l}\,U = \left( \psi_{\alpha,\l}\,U\,

3108:   \psi_{\alpha,\overline{U(\alpha)}^t \l}^{-1} 

3109:   \right) \psi_{\alpha,\overline{U(\alpha)}^t \l}.

3110: \end{equation}

3111: %

3112: For the geometric applications considered in 

3113: this paper it is of great value to have

3114: a simple and  explicit formula for the dressed 

3115: frame. Simple factors, for which $\mathrm{L}$ 

3116: depends on $\lambda$ and are of the form 

3117: %

3118: \begin{equation}

3119:   \psi_{\alpha,\l(\lambda)} =  \pi_{\l(\lambda)} + 

3120:   \tau_\alpha(\lambda) \,(\Id - \pi_{\l(\lambda)})

3121: \end{equation}

3122: %

3123: will be called \emph{generalised simple factors}. 

3124: We would like to present an analogous result 

3125: to \ref{sdf} for such generalised simple factors: 

3126: 

3127: \mytheorem{}

3128: Let $\mathrm{L}: \C^* \rightarrow \CP$ be holomorphic and 

3129: $\langle \mathrm{L},\,\mathrm{L}^* \rangle \neq 0$ for all 

3130: $\lambda \in \C^*$. 

3131: Let $\alpha \in I_1^*$ and $\psi_{\alpha,\l(\lambda)}$ 

3132: be the corresponding generalised simple factor. 

3133: Then for any $U \in  \Lambda_{\mbox{\tiny{$\C^*$}}}\SU$

3134: %

3135: \begin{equation} \label{gsdf}

3136: \psi_{\alpha,\l(\lambda)}\,\# \,U = 

3137:   \psi_{\alpha,\l(\lambda)}\,U\,

3138:   \psi_{\alpha,\overline{U(\alpha)}^t 

3139:   \l(\alpha)}^{-1}. 

3140: \end{equation}

3141: %

3142: \Proof

3143: Fix $\lambda_0 \in \C^*$ and write 

3144: $\mathrm{L}(\lambda) = [a(\lambda) : b(\lambda)]$. Then for 

3145: $\mathrm{L}_0 = [1 : 0]$ and the holomorphic map

3146: %

3147: \begin{equation}

3148:   W(\lambda) = \begin{pmatrix} 

3149:   a(\lambda) & -\overline{b(1/\bar{\lambda})} \\

3150:   b(\lambda) & \overline{a(1/\bar{\lambda})} \end{pmatrix}

3151: \end{equation}

3152: %

3153: we have $\mathrm{L}(\lambda) = W(\lambda)\,\mathrm{L}_0$ and 

3154: $W(\lambda) \in \Lambda_{\mbox{\tiny{$\C^*$}}} \Utwo$. 

3155: Applying \ref{eq:change} to 

3156: $A = W(\lambda), \,\lambda \in \C^*$, we obtain 

3157: $\psi_{\alpha,\l(\lambda)} = 

3158: W(\lambda)\,\psi_{\alpha,\l_0}\,W(\lambda)^{-1}$ 

3159: and 

3160: %

3161: \begin{align*}

3162:   \psi_{\alpha,\l(\lambda)}\,U(\lambda) &= 

3163:   W(\lambda)\psi_{\alpha,\l_0}W(\lambda)^{-1} 

3164:   U(\lambda) \\ &= \left( W(\lambda)\, 

3165:   \psi_{\alpha,\l_0}\, W(\lambda)^{-1} U(\lambda)\, 

3166:   \psi_{\alpha,\l_0'}^{-1} \right) 

3167:   \psi_{\alpha,\l_0'},

3168: \end{align*}

3169: %

3170: where $\mathrm{L}_0' = \overline{(W(\alpha)^{-1} 

3171: U(\alpha) )}^t\mathrm{L}_0$.

3172: The right side is an Iwasawa decomposition. 

3173: The unitary part can be rewritten in the form

3174: %

3175: \begin{equation*}

3176:   W(\lambda)\,\psi_{\alpha,L_0}\,W(\lambda)^{-1} 

3177:   \,U(\lambda)\,\psi_{\alpha,L_0'}^{-1} =

3178:   \psi_{\alpha,L(\lambda)}\,U(\lambda)\,

3179:   \psi_{\alpha,L_0'}^{-1}. 

3180: \end{equation*}

3181: %

3182: It is straightforward to verify 

3183: $\mathrm{L}_0' = \overline{U(\alpha)}^t 

3184: \mathrm{L}(\alpha)$, thus proving \eqref{gsdf}. \QED

3185: %

3186: 

3187: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3188: 

3189: \newsection{\bf Twisting of simple factors}

3190: 

3191: Since generalised simple factors are in general 

3192: untwisted, we need to

3193: modify the concept so that it can also be used for 

3194: the twisted case. We discuss two approaches to 

3195: deal with simple factors in the twisted case by 

3196: firstly, taking a simple factor and applying the 

3197: 'twisting map' and secondly, looking for products 

3198: of simple factors which happen to be twisted. 

3199: Applying the 'twisting map' \eqref{eq:twist} 

3200: to a simple factor we obtain 

3201: %

3202: \begin{equation*}

3203:   \psi_{\alpha,\l(\lambda)} =D(\lambda)\,

3204:   \psi_{\alpha,\l (\lambda^2)}\,D(\lambda)^{-1} =  

3205:   \psi_{\alpha, D(\lambda) \l (\lambda^2)}.

3206: \end{equation*}

3207: % 

3208: \mylemma{}

3209: a) Twisting a generalised simple factor 

3210: produces again a generalised simple factor.

3211: 

3212: b) Let $\psi$ be a simple factor (L is 

3213: independent of $\lambda$). Then the 

3214: corresponding twisted simple

3215: factor is again independent of 

3216: $\lambda$ if and only if

3217: %

3218: \begin{equation}\label{eq:picase} 

3219:   D\,\pi_\l \,D^{-1} = \pi_\l \mbox{ or } 

3220:   \pi_\l^\perp. 

3221: \end{equation}

3222: %

3223: \QED

3224: 

3225: \myremark{}

3226: Since $D=D(\lambda)$ is diagonal, the only 

3227: projections satisfying the condition 

3228: \eqref{eq:picase} above are 

3229: those onto the canonical basis vectors.

3230: 

3231: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3232: 

3233: \newsection{\bf Twisted products of simple factors}

3234: As mentioned earlier we also consider products of 

3235: two simple factors and determine, when such 

3236: matrices are twisted. Recall the two 

3237: involutions $\varrho,\,\s$ defined in 

3238: \eqref{eq:sigma} respectively \eqref{eq:varrho}. 

3239: 

3240: \myprop{} \label{th:twist}

3241: 

3242: a) Let $g \in \Lambda_r \SL$. 

3243: Then $(\s g) g$ is twisted if and 

3244: only if $[\,\s g,\,g\,]=0$.

3245: 

3246: b) If $g \in \LGU$, then $(\s g)g \in \LGU$.

3247: 

3248: c) Let $\psi_{\alpha,\l}$ and 

3249: $\psi_{\beta,\widehat{\l}}$ be simple factors.

3250: If $\sigma (\psi_{\alpha,\l})\, 

3251: \psi_{\beta,\widehat{\l}}$ is twisted, then 

3252: $\psi_{\alpha,\l} = \psi_{\beta,\widehat{\l}}$.

3253: 

3254: d) If $\sigma (\psi_{\alpha,\l})\psi_{\alpha,\l}$ 

3255: is twisted, then either $\sigma(\pi_\l) = \pi_\l$ 

3256: or $\sigma(\pi_\l) = \pi_\l^\perp$.

3257: 

3258: \Proof a) The loop $(\s g) g$ is twisted if and 

3259: only if $\s \left( (\s g)\,g \right) = g \,\s g = 

3260: (\s g)\,g$, which is equivalent to 

3261: $[\s g,\,g] = 0$. 

3262: 

3263: b) The loop $g \in \LGU$ if and only if 

3264: $\varrho \,g = g$. Then 

3265: $\varrho \left( (\s g)g \right) = 

3266: (\varrho \s g)\,(\varrho g) = 

3267: (\s \,\varrho \,g ) \, (\varrho\, g) = (\s g)\,g$, 

3268: since $[\varrho,\,\sigma ] = 0$. Hence 

3269: $(\s g)\,g \in \LGU$. 

3270: 

3271: c) The product is twisted if and only if 

3272: $\sigma (\psi_{\alpha,L}) \cdot  

3273: \psi_{\beta,\widehat{L}} =  \psi_{\alpha,L} 

3274: \cdot \sigma ( \psi_{\beta,\widehat{L}})$. 

3275: Let us abbreviate $A = \psi_{\alpha,\l}$ and  

3276: $B = \psi_{\beta,\widehat{\l}}$. 

3277: Then the pole of $A$ is at 

3278: $1/\overline{\alpha}$, while the pole of 

3279: $B$ is at  $1/\overline{\beta}$. 

3280: On the other hand, the pole of $\sigma (A)$ 

3281: is at $- 1/\overline{\alpha}$ and the 

3282: analogous result holds for $B$.

3283: Thus comparing the two sides we obtain 

3284: $\alpha = \beta$. Expanding near a pole 

3285: and comparing the factors we derive $\mathrm{L} = 

3286: \widehat{\mathrm{L}}$.

3287: 

3288: (d) If $(\s \psi_{\alpha,\l} )\psi_{\alpha,\l}$ 

3289: is twisted, then by part (a), 

3290: $[\s \psi_{\alpha,\l},\,\psi_{\alpha,\l}]=0$ 

3291: which is equivalent to the eigenspaces of 

3292: $\s \psi_{\alpha,\l}$ and 

3293: $\psi_{\alpha,\l}$ coinciding. Thus there are two 

3294: possibilities:

3295: %

3296: \begin{equation}\label{eq:picase1} 

3297:   \s_3\,\pi_\l \,\s_3^{-1} = \pi_\l \mbox{ or } 

3298:   \pi_\l^\perp. 

3299: \end{equation}

3300: %

3301: This proves (d) and concludes the proof of the 

3302: proposition. \QED

3303: 

3304: We are going to evaluate the two possibilities 

3305: in \eqref{eq:picase1}. 

3306: 

3307: (i) Let us turn to the first case in 

3308: \eqref{eq:picase1}: 

3309: For the line 

3310: $\mathrm{L} = \C \bigl( \begin{smallmatrix} 

3311: 1\\0 \end{smallmatrix} \bigr)$ we have 

3312: $\psi_{\alpha,\l}= \mathrm{diag}[1,\,\tau_\alpha]$. 

3313: Using $\tau_\alpha(\lambda)\tau_\alpha(-\lambda) =  

3314:   \tau_{\alpha^2}(\lambda^2)$ 

3315: we obtain $(\s \psi_{\alpha,\l})\psi_{\alpha,\l} = 

3316: \pi_\l + \tau_{\alpha^2}(\lambda^2)\pi^\perp_\l$. 

3317: Dividing by the square root of the determinant  

3318: we arrive at all diagonal twisted simple factors 

3319: %

3320: \begin{equation} \label{eq:simple1}

3321:  g_{\alpha,\l}(\lambda) = 

3322:   \sqrt{\tau_{\alpha^2}^{-1}

3323:   (\lambda^2)} \, \pi_\l + \sqrt{\tau_{\alpha^2}

3324:   (\lambda^2)}\, \pi^\perp_\l.

3325: \end{equation}

3326: %

3327: Then $g_{\alpha,\l}(\lambda) \in \LGP$ for 

3328: $0< r< |\alpha|$ and in matrix form is given by 

3329: %

3330: \begin{equation}\label{eq:simple1_matrix}

3331:   g_{\alpha,\l}(\lambda) = \begin{pmatrix} 

3332:   \sqrt{\tfrac{1-\bar{\alpha}^2\lambda^2}

3333:   {\alpha^2 - \lambda^2}} & 0 \\ 

3334:   0 & \sqrt{\tfrac{\alpha^2 - \lambda^2}

3335:   {1-\bar{\alpha}^2\lambda^2}}

3336:   \end{pmatrix} 

3337: \end{equation}

3338: %

3339: Notice that $g_{\alpha,\l}(\lambda)^{-1}$ 

3340: corresponds to a factor of the matrix 

3341: given in \eqref{eq:prod_simple}.

3342: 

3343: (ii) Let us turn to the second case in 

3344: \eqref{eq:picase1}: $\s_3 \pi_\l \s_3^{-1} = 

3345: \pi_\l^\perp$. 

3346: For $\mathrm{L} =\C ( a,\,b )^t$ and hermitian 

3347: projection $\pi_\l:\C^2 \to \mathrm{L}$ we have 

3348: $\sigma \pi_\l = \pi_\l^\perp$ if and 

3349: only if $|a|^2=|b|^2 = 1/2$. 

3350: Setting $a=\tfrac{1}{\sqrt{2}}\exp(is)$ and 

3351: $b=\tfrac{1}{\sqrt{2}}\exp(it)$ and 

3352: rescaling, we may assume without loss of 

3353: generality that $\mathrm{L} = 

3354: \C \bigl( e^{i\theta}, \, 1 \bigr)^t$. Then 

3355: $\pi_\l = \tfrac{1}{2}(\Id + \mathrm{off}

3356: [ e^{i\theta},\,

3357: e^{-i\theta}])$ and for 

3358: $\psi_{\alpha,\l} = \pi_\l + \tau_\alpha(\lambda)\, 

3359: \pi_\l^\perp$ 

3360: we have $\sigma \psi_{\alpha,\l} = \pi_\l^\perp 

3361: + \tau_\alpha(-\lambda)\,\pi_\l$ and consequently 

3362: %

3363: \begin{align*}

3364:   (\sigma \psi_{\alpha,\l})\psi_{\alpha,\l} &= 

3365:   \tau_\alpha(\lambda)\,\pi_\l^\perp + 

3366:   \tau_\alpha(-\lambda)\,\pi_\l \\

3367:   &= \tfrac{1}{1-\bar{\alpha}^2\lambda^2} (

3368:   (\alpha - \bar{\alpha}\lambda^2)\,\Id

3369:   + \lambda (1 - |\alpha|^2)\,

3370:   \mathrm{off}[e^{i\theta},\,e^{-i\theta}]).

3371: \end{align*}

3372: %

3373: Normalising, we arrive at the second class of 

3374: twisted simple factors in $\LGP$ for 

3375: $0 < r < |\alpha|$, given by

3376: %

3377: \begin{equation} \label{eq:simple2}

3378:  g_{\alpha,\l}(\lambda) = 

3379:   \sqrt{\tau_\alpha^{-1}(\lambda)

3380:   \,\tau_\alpha (-\lambda)} \, \pi_\l + 

3381:   \sqrt{\tau_\alpha (\lambda) 

3382:   \,\tau_\alpha^{-1} (-\lambda)}\, \pi^\perp_\l.

3383: \end{equation}

3384: %

3385: In matrix form, these simple factors look like

3386: %

3387: \begin{equation} \label{eq:simple2_matrix}

3388:   g_{\alpha,\l}(\lambda) = 

3389:   \sqrt{\frac{1-\bar{\alpha}^2\lambda^2}

3390:   {\alpha^2 - \lambda^2}}

3391:   \begin{pmatrix} \alpha - \bar{\alpha}\lambda^2 & 

3392:   \lambda (1- |\alpha|^2)e^{i\theta} \\ 

3393:   \lambda (1 - |\alpha|^2)e^{-i\theta} & 

3394:   \alpha - \bar{\alpha}\lambda^2 

3395:   \end{pmatrix}.

3396: \end{equation}

3397: %

3398: The simple factors $g_{\alpha,\l}$ derived 

3399: in \eqref{eq:simple1} and \eqref{eq:simple2} 

3400: are uniquely determined by their 'singularity' 

3401: $\alpha \in I_1$ and choice of line 

3402: $\mathrm{L} \in \CP$. Note also that in both cases 

3403: \eqref{eq:simple1} and \eqref{eq:simple2} we have 

3404: that 

3405: %

3406: \begin{equation} \label{eq:simpleunitary}

3407:   g_{\alpha,\l} \in \LGU 

3408:   \mbox{ for } |\alpha| < r \leq 1.

3409: \end{equation}

3410: %

3411: The key aspect is that dressing with simple factors 

3412: is explicit. This idea is due to 

3413: Terng and Uhlenbeck \cite{TerU} and has lead to 

3414: variants as in \cite{Bur:iso} and \cite{KilSS}. 

3415: The version we need is proven by 

3416: twisting Theorem 1.2 in \cite{KilSS} and is as 

3417: follows 

3418: 

3419: \mytheorem{} Let $M$ be a connected Riemann 

3420: surface with universal cover $\MT$ and let 

3421: $F(z,\lambda) \in \mathcal{F}(\MT)$. 

3422: Let $g_{\alpha,\l} \in \LGP$ be a 

3423: simple factor. Then 

3424: %

3425: \begin{equation} \label{eq:simple_dress}

3426:   g_{\alpha,\l} \# F = 

3427:   g_{\alpha,\l} \,F \,g^{-1}_{\alpha,\l^\prime}

3428: \end{equation}

3429: %

3430: where $\mathrm{L}^\prime = 

3431: \overline{F(z,\alpha)}^t \mathrm{L}$ and 

3432: $g^{-1}_{\alpha,\l^\prime}$ is, pointwise in 

3433: $z \in \MT$, a simple factor of the same 

3434: form as $g_{\alpha,\l}$. \QED

3435: 

3436: We use this factorisation theorem to characterise 

3437: when it is possible to dress an extended unitary 

3438: frame with trivial isotropy by a simple factor 

3439: while retaining a symmetry.

3440: 

3441: \mycorollary{} \label{th:invariance} 

3442: Let $\chi(\tau) \in \Lhol$ be a monodromy 

3443: of an extended unitary frame 

3444: $F \in \mathcal{F}(\MT)$. 

3445: Let $g_{\alpha,\l} \in \LGP$ be a simple factor. 

3446: Then 

3447: $g_{\alpha,\l}\,\chi(\tau) \,g_{\alpha,\l}^{-1}$ 

3448: is a monodromy of $g_{\alpha,\l} \# F$ 

3449: with respect to $\tau$ if and only if 

3450: %

3451: \begin{equation} \label{eq:invariance}

3452:   \overline{\chi(\alpha,\tau)}^t 

3453:   \mathrm{L} = \mathrm{L}.

3454: \end{equation}

3455: %

3456: \Proof Let $\hat F = g_{\alpha,\l} \# F$. 

3457: Then $\hat F = g_{\alpha,\l}\,F\,

3458: g^{-1}_{\alpha,\l^\prime}$ with 

3459: $\mathrm{L}^\prime = \overline{F(z,\alpha)}^t 

3460: \mathrm{L}$ by \eqref{eq:simple_dress}. Let 

3461: $\tau^* F = \chi(\tau)\,F\,k$. Then 

3462: %

3463: \begin{align*}

3464:   \tau^* \hat F &= g_{\alpha,\l}\,(\tau^* F)\, 

3465:   g^{-1}_{\alpha,\tau^*\l^\prime} \\ 

3466: 	   &= g_{\alpha,\l}\,\chi(\tau) \,F\, 

3467:   k\,g^{-1}_{\alpha,\l''} \mbox{ with } 

3468:   \mathrm{L}'' = \bar{k}^t\,

3469:   \overline{F(\alpha)}^t \,

3470:   \overline{\chi(\alpha,\tau)}^t \mathrm{L}  = 

3471:   \tau^* \mathrm{L}' \\

3472:   &= g_{\alpha,\l}\,\chi(\tau)\, g_{\alpha,\l}^{-1}

3473:   \,g_{\alpha,\l} \,F \, g^{-1}_{\alpha,\l_1 } 

3474:   k \mbox{ by } \eqref{eq:change}

3475: \end{align*}

3476: %

3477: where $\mathrm{L}_1 = \overline{F(z,\alpha)}^t\,

3478: \overline{\chi(\alpha,\tau)}^t \mathrm{L}$. 

3479: 

3480: If $g_{\alpha,\l}\,\chi(\tau)\, g_{\alpha,\l}^{-1}$ 

3481: is a monodromy of $\hat F$ and we write 

3482: $\tau^* \hat{F} = g_{\alpha,\l}\,\chi(\tau)\, 

3483: g_{\alpha,\l}^{-1} \, \hat{F} \,\hat{k}$ 

3484: and combine with the above, we obtain 

3485: $g_{\alpha,\l_1}^{-1}\,k = 

3486:   g_{\alpha,\l'}^{-1} \,\hat{k}$. 

3487: Hence $g_{\alpha,\l_1}g_{\alpha,\l'}^{-1}$ 

3488: is independent of $\lambda$ and since there exists a 

3489: $\lambda_0 \in \CP$ at which $\tau_\alpha(\lambda_0) = 1$ we obtain 

3490: $g_{\alpha,\l_1}g_{\alpha,\l'}^{-1} = \Id$.  

3491: This implies that $L' = L_1$ which implies \eqref{eq:invariance}. 

3492: The converse is proven by direct verification. 

3493: \QED

3494: 

3495: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3496: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3497: 

3498: \section{Dressing non--finite type surfaces}

3499: 

3500: In this section we discuss a family of CMC cylinders 

3501: found in \cite{KilMS} that arise as perturbations 

3502: of Delaunay surfaces. 

3503: The resulting surfaces may possess an arbitrary 

3504: number of umbilics and are thus not of finite type and  

3505: are CMC cylinders with one Delaunay end \cite{KilKRS}, see Figure 1. 

3506: We can dress this class of cylinders with simple factors, having the 

3507: effect of adding bubbles to the surface, see Figure 2.

3508: 

3509: %

3510: \begin{figure}\label{fig:pert}

3511:   \centering

3512:     \includegraphics[scale=1.5]{pert.eps}

3513:     \includegraphics[scale=1.3]{smythdelbub.eps}\\ 

3514:     \caption{\footnotesize A perturbed round cylinder on the left. A perturbed 

3515:    and dressed round cylinder on the right. 

3516:    Images generated with \emph{CMCLab} \cite{Sch:cmclab}. For more images see 

3517:   \cite{Sch:gallery}.}

3518: \end{figure}

3519: %

3520: %

3521: 

3522: 

3523: \newsection{\bf Dressing cylinders with umbilics.}

3524: We modify a standard result from the theory of 

3525: differential equations with regular singular points. 

3526: Since the eigenvalues are $\lambda$--dependent, 

3527: we can avoid the assumption that two elements in the 

3528: spectrum not differ by an integer by working on an 

3529: appropriate $\lambda$--circle $C_r$.

3530: %

3531: 

3532: \mytheorem{} \label{th:zap}

3533: Let $U_0^* \subset \C$ be an open neighbourhood of 

3534: $z=0$ and $\xi \in \Lambda \Omega(U_0^*)$ with a 

3535: simple pole at $z=0$ and residue

3536: $\mathrm{res}_{\,0} \xi = D$ 

3537: where $D$ has the form \eqref{eq:delaunay}.

3538: Moreover, we assume that $D$ 

3539: satisfies the first closing condition 

3540: \eqref{eq:closing1}. Then there exists 

3541: an $r \in (0,1)$ and a solution of 

3542: $d\Psi = \Psi \xi$ of the form $\Psi = z^D P$ 

3543: with $P:U_0 \to \LGC$ holomorphic. 

3544: Further, $\Psi$ has Delaunay monodromy and is the 

3545: holomorphic frame associated with $(\xi,\,P(1),\,1)$.

3546: %

3547: 

3548: \Proof

3549: Let $U_0 = U_0^* \cup \{ 0 \}$ and write 

3550: $\xi = D \tfrac{dz}{z} + \eta$ with 

3551: $\eta \in \Lambda \Omega_{U_0}$. The differential 

3552: equation for $P$ is 

3553: $dP = P \eta + [ P,\, D \tfrac{dz}{z}]$. 

3554: We will show that this differential equation has a 

3555: holomorphic solution. Expanding 

3556: $P = \sum_{k = 0}^{\infty} P_k z^k$ and 

3557: $\xi = (\tfrac{1}{z}D +\Sigma \eta_k z^k )dz$ 

3558: at $z=0$, the coefficients are recursively given 

3559: by 

3560: %

3561: $$

3562: kP_k + [D , P_k ] = \sum_{r+s=k-1} P_r \eta_s.

3563: $$

3564: % 

3565: By our Ansatz $P_k = 0$ for $k < 0$ and we are free 

3566: to choose $P_0 \in \LGC$ with $[D , P_0 ]=0$. 

3567: (Note this freedom only means 

3568: $P_0 = \alpha \Id + \beta D$, 

3569: since $D$ is a semisimple $2 \times 2-$matrix

3570: with different eigenvalues.)

3571: The eigenvalues of $D$, are of the form $\pm \mu$, 

3572: where $\mu$ is a solution to the equation 

3573: $\mu^2 = - \det D$. Then 

3574: the operators $k \rm{Id} + \rm{ad}_{D}$, 

3575: $k \in \Z$, have spectrum 

3576: $\sigma = \{ k,\, k \pm  2\mu \}$. 

3577: We are only interested in $k \geq 1$, in which case 

3578: $0 \notin \sigma$ if and only if 

3579: $\mu \notin \tfrac{1}{2} \N$. 

3580: 

3581: We need to show that there exists an 

3582: $r \in (0,\,1)$ for which 

3583: $\mu (\lambda) \notin \frac{1}{2}\N$ for every 

3584: $\lambda \in C_r$. 

3585: A priori, $\mu(\lambda) \in \R$ if and only if 

3586: either $\lambda \in S^1$ or $\lambda \in \R$. 

3587: Since we seek $r \in (0,\,1)$ we have 

3588: $\mu(\lambda) \in \R$ if and only if 

3589: $\lambda \in \R$ we need to ensure 

3590: $\mu(\pm r)\notin \tfrac{1}{2}\N$. 

3591: Since $D$ satisfies the first closing 

3592: condition by assumption, we know that 

3593: $\mu(1) = n/2$ for some $n \in \Z$ 

3594: the winding number of the Delaunay surface. 

3595: A direct computation shows 

3596: $\mu(\pm r) \notin \tfrac{1}{2}\Z$ 

3597: if and only if 

3598: $r^{-2} + r^2 \neq \pm (k^2 - n^2)/(4ab) + 2$ 

3599: for all $k \in \N$. 

3600: The sequences $a^{\pm}_k := 

3601: \pm (k^2 - n^2)/(4ab) + 2$ 

3602: are monotonic, hence we can always find an 

3603: $0 < r <1$ for which $r^{-2} + r^2 

3604: \notin \{ a^{\pm}_k \}$. 

3605: 

3606: This proves the existence of a formal power 

3607: series solving the differential equation for $P$. 

3608: The coefficients of this formal power series are 

3609: defined for all $z \in U_0$. The domain of 

3610: analyticity of $P$ is $U_0$ by standard 

3611: ODE arguments, see e.g \cite{Har}. It is shown in 

3612: \cite{DorW:noids} that $P$ has the twisted $\lambda$ 

3613: behaviour and that the coefficients of $P$ as 

3614: functions of $\lambda$ are in $\mathcal{A}_r$. 

3615: To show that $\Psi$ has Delaunay 

3616: monodromy, recall from \ref{sec:Delaunayexample} 

3617: that the Delaunay monodromy is $\exp(2 \pi i D)$. 

3618: For the translation $\tau(z) = z + 2\pi i$ 

3619: the monodromy of $\Psi$ is $\tau^*\Psi \Psi^{-1} = 

3620: \tau^*z^D \tau^*P P^{-1} z^{-D} = \exp(2 \pi i D)$.

3621: \QED

3622: %

3623: 

3624: For our purposes, the specific form of $\Psi$ 

3625: proven above is of great importance to us, 

3626: since it gives easy control over the monodromy 

3627: singled out by the singularity at $D$.

3628: In particular, if we dress the corresponding 

3629: frame  by  some simple factor it is easy to

3630: determine the dressed monodromy, at least if 

3631: the associated surface has umbilical points. 

3632: From a geometric point of view this is 

3633: particularly interesting, since by a result 

3634: of \cite{Mah}, the classical Bianchi--B\"acklund 

3635: transformation does correspond to dressing 

3636: by some simple factor. Actually, in the twisted 

3637: setup, our 'simple factors' are in fact a product 

3638: of two simple factors. These correspond to the 

3639: two step procedure of the Bianchi--B\"acklund 

3640: transformation. As a matter of fact,  

3641: by making the  right choice for the singularity 

3642: of the simple factor, is is possible to control the 

3643: topology of the resulting surface. 

3644: 

3645: \myremark{}

3646: 1) The proof of the Theorem above actually shows 

3647: that there is some $0 < r < 1$ such that for all 

3648: $0 < r \leq r' <1$ the claim holds. 

3649: 

3650: 2) Instead of choosing some $0 < r < 1$, 

3651: one could ask for what potentials $\xi$ 

3652: one can obtain a solution of the

3653: form $ \Psi = z^D \, P$ without 

3654: singularities on $S^1$. Once this is 

3655: achieved, one can conclude as above and 

3656: infer statements about the monodromy and 

3657: thus one can clear the way for some 

3658: understanding for the effect of dressing 

3659: by simple factors. Such a condition has 

3660: been given in \cite{DorW:noids}.

3661: 

3662: In order to dress a CMC surface with non-trivial 

3663: topology into a new CMC surface with the same 

3664: topology, we recall from \cite{DorH:per} when for a 

3665: given triple $(\xi,\Phi_0,\tilde{z}_0)$ one 

3666: specific member of the associated family is 

3667: invariant under $\Delta$, thus characterising 

3668: the period problem in the DPW framework: 

3669: 

3670: Let $f_\lambda$ be an associated family. 

3671: Let $F$ be the extended unitary 

3672: frame of the surface with monodromy 

3673: $\chi : \Delta \to \Lhol$.

3674: Then there exists a $\lambda_0 \in S^1$ such 

3675: that $\tau^* f_{\lambda_0}= f_{\lambda_0}$ for all 

3676: $\tau \in \Delta$ if and only if for all 

3677: $\tau \in \Delta$, $\chi$ satisfies both

3678: %

3679: \begin{align}

3680: \chi (\tau,\,\lambda_0) = 

3681: \pm \rm{Id},&\label{eq:closing1}\\

3682: \left. \tfrac{\partial}{\partial \lambda}

3683: \right|_{\lambda_0} 

3684: \chi(\tau,\,\lambda)

3685:  = 0.& \label{eq:closing2}

3686: \end{align}

3687: %

3688: 

3689: \mycorollary{} 

3690: Let $\xi \in \Lambda \Omega(\C^*)$ as in Theorem 

3691: \ref{th:zap} and $g_{\alpha,\l} \in \LGP$ be a 

3692: simple factor.  Choosing $r \in (0,\,1)$ as above 

3693: and $\alpha \in \C$ with $|\alpha| \in (r,\,1)$ 

3694: such that 

3695: %

3696: \begin{equation} \label{eq:det}

3697:   \det D(\alpha) = \tfrac{n}{4} \mbox{ for some } n \in \Z,

3698: \end{equation}

3699: %  

3700: then $(\xi,\, g_{\alpha,\l}\, P(1), \,1)$ 

3701: generates a CMC cylinder (with winding number $n$).

3702: 

3703: \Proof 

3704: Let $\Phi = z^D\,P$ and $F$ be the holomorphic 

3705: respectively unitary frame generated by 

3706: $(\xi ,\, P(1),\, 1)$. 

3707: Both have Delaunay monodromy 

3708: $\chi (\tau,\,\lambda) = \exp(2\pi iD(\lambda))$ 

3709: with respect to the translation 

3710: $\tau(z) = z + 2\pi i$. 

3711: The condition \eqref{eq:det} ensures that 

3712: $\chi(\tau,\,\alpha) = \pm \Id$ and 

3713: consequently \eqref{eq:invariance} holds 

3714: for any choice of line $\mathrm{L} \in \CP$. 

3715: Hence, by Corollary \ref{th:invariance}, 

3716: $\hat\chi(\tau) = 

3717: g_{\alpha,\l}\,\chi(\tau)\,g_{\alpha,\l}^{-1}$ 

3718: is a monodromy of $g_{\alpha,\l} \# F$. 

3719: The conditions \eqref{eq:closing1} and 

3720: \eqref{eq:closing2} are verified using the facts 

3721: $\chi(\tau,1) = \pm \mathrm{Id}$ and $\left. 

3722: \partial_{\lambda}\chi(\tau)\right|_{\lambda=1}= 0$.

3723: \QED

3724: 

3725: 

3726: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3727: 

3728: \newsection{\bf Dressing $3$-Noids.}

3729: %

3730: Let us briefly outline how the above theory can be 

3731: applied to construct the dressed $3$-Noids of 

3732: \cite{KilSS}. Let 

3733: $\mathcal{T} = \CP \setminus \{0,\,1,\,\infty\}$ 

3734: and $f: \mathcal{T} \to \R^3$ a CMC $3$-Noid with 

3735: Delaunay ends as constructed in \cite{DorW:noids} 

3736: or \cite{Sch:tri}. Let 

3737: $F \in \mathcal{F}(\widetilde{\mathcal{T}})$ 

3738: be an extended frame such that at $\lambda = 1$ the 

3739: immersion $f$ is obtained via the Sym--Bobenko 

3740: formula \eqref{eq:Sym}. 

3741: Then there are three monodromies, 

3742: $\chi_1,\, \chi_2,\,\chi_3$, one for each end, 

3743: which can be computed in terms of 

3744: $\Gamma$--functions \cite{DorW:noids}, 

3745: \cite{KilSS} and satisfy 

3746: $\prod_{j=1}^3 \chi_j = \Id$. 

3747: 

3748: In \cite{KilSS} it was shown that there exist 

3749: values $\alpha \in \C^*,\,|\alpha |<1$ and 

3750: invariant subspaces $\mathrm{L} \in \CP$ such that 

3751: \eqref{eq:invariance} holds for all three 

3752: $\chi_j$'s. Consequently, by Corollary 

3753: \ref{th:invariance} the unitary frame 

3754: obtained by dressing 

3755: with the simple factor $g_{\alpha,\l}$ 

3756: has monodromies $\hat \chi_j = 

3757: g_{\alpha,\l} \, \chi_j \, g_{\alpha,\l}^{-1}$ and 

3758: since the closing conditions \eqref{eq:closing1} and \eqref{eq:closing2} are invariant under conjugation, 

3759: the resulting surface is again a 

3760: CMC $3$--Noid.

3761: %

3762: %.Adding a bubble to the first end can be 

3763: %accomplished by dressing with $h^1C^1$, with $h^1 \in \LGP$ as in the

3764: %example of section ?? and $C \in \LGC$ with $[\, C^1, H^1 \, ]=0$. 

3765: %If we want to dress from trinoids with embedded ends to trinoids 

3766: %with embedded ends,  by Theorem \ref{th:mainresult} we need to have 

3767: %that $h^1 C^1 H_j (h^1C^1)^{-1} \in \LGU$ for all $r>0$. Moreover,

3768: %the closing conditions \eqref{eq:closing1},\eqref{eq:closing2}

3769: %have to be satisfied. For $j=1$ this all holds by \cite{Kil}, 

3770: %Theorem 6.2.2.

3771: 

3772: 

3773: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3774: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3775: 

3776: \providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}

3777: \providecommand{\MR}{\relax\ifhmode\unskip\space\fi MR }

3778: % \MRhref is called by the amsart/book/proc definition of \MR.

3779: \providecommand{\MRhref}[2]{%

3780:   \href{http://www.ams.org/mathscinet-getitem?mr=#1}{#2}

3781: }

3782: \providecommand{\href}[2]{#2}

3783: \begin{thebibliography}{10}

3784: 

3785: \bibitem{Bob:tor}

3786: A.~I. Bobenko, \emph{All constant mean curvature tori in {$\mathbb{R}^3$},

3787:   {$\mathbb{S}^3$}, {$\mathbb{H}^3$} in terms of theta-functions}, Math. Ann.

3788:   \textbf{290} (1991), 209--245.

3789: \bibitem{BobPS}

3790: A.~I. Bobenko, T.~V. Pavlyukevich and 

3791: B. A. Springborn,

3792: \emph{Hyperbolic constant mean curvature one 

3793: surfaces: spinor representation and trinoids in 

3794: hypergeometric functions}, Math.Z \textbf{245}, (2003), 63--91.

3795: 

3796: 

3797: \bibitem{Bun}

3798: L.~Bungart, \emph{On analytic fiber bundles}, Topology \textbf{7} (1968),

3799:   55--68.

3800: 

3801: \bibitem{Bur:iso}

3802: F.~Burstall, \emph{Isothermic surfaces: conformal geometry, {C}lifford algebras

3803:   and integrable systems.}, Integrable systems, Geometry and Topology, vol.~80,

3804:   taiwan, 1995, math.DG/0003096, pp.~353--382.

3805: 

3806: \bibitem{BurP:dre}

3807: F.~Burstall and F.~Pedit, \emph{Dressing orbits of harmonic maps}, Duke Math.

3808:   J. \textbf{80} (1995), 353--382.

3809: 

3810: \bibitem{DorH:per}

3811: J.~Dorfmeister and G.~Haak, \emph{On constant mean curvature surfaces with

3812:   periodic metric}, Pacific J. Math. \textbf{182} (1998), 229--287.

3813: 

3814: \bibitem{DorH:dre}

3815: \bysame, \emph{Investigation and application of the dressing action on surfaces

3816:   of constant mean curvature}, Q. J. Math. \textbf{51} (2000), 57--73.

3817: 

3818: \bibitem{DorH:sym2}

3819: \bysame, \emph{On symmetries of constant mean curvature surfaces. {II}.

3820:   symmetries in a {W}eierstrass-type representation}, Int. J. Math. Game Theory

3821:   and Algebra \textbf{10} (2000), 121--146.

3822: 

3823: \bibitem{DorH:cyl}

3824: \bysame, \emph{Construction of non-simply connected {CMC} surfaces via

3825:   dressing}, J. Math. Soc. Japan \textbf{55} (2003), no.~2, 335--364.

3826: 

3827: \bibitem{DorPW}

3828: J.~Dorfmeister, F.~Pedit, and H.~Wu, \emph{Weierstrass type representation of

3829:   harmonic maps into symmetric spaces}, Comm. Anal. Geom. \textbf{6} (1998),

3830:   no.~4, 633--668.

3831: 

3832: \bibitem{DorW:noids}

3833: J.~Dorfmeister and H.~Wu, \emph{Construction of constant mean curvature

3834:   $n$-noids from holomorphic potentials}, in preparation, 2002.

3835: 

3836: \bibitem{Har}

3837: P.~Hartman, \emph{Ordinary differential equations}, John Wiley and Sons, Inc.,

3838:   New York, 1964.

3839: 

3840: \bibitem{Kil:thesis}

3841: M.~Kilian, \emph{Constant mean curvature cylinders}, Ph.D. thesis, Univ. of

3842:   Massachusetts, Amherst, 2000.

3843: 

3844: \bibitem{Kil:del}

3845: \bysame, \emph{On the associated family of {D}elaunay surfaces},

3846:   Proc. AMS, in press.

3847: 

3848: \bibitem{KilKRS}

3849: M.~Kilian, S.~Kobayashi, W.~Rossman, and N.~Schmitt, \emph{{CMC} surfaces in

3850:   three dimensional space forms}, math.DG/0403366.

3851: 

3852: \bibitem{KilMS}

3853: M.~Kilian, I.~McIntosh, and N.~Schmitt, \emph{New constant mean curvature

3854:   surfaces}, Experiment. Math. \textbf{9} (2000), no.~4, 595--611.

3855: 

3856: \bibitem{KilSS}

3857: M.~Kilian, N.~Schmitt, and I.~Sterling, \emph{Dressing {CMC} n-{N}oids}, 

3858:   Math. Z., \textbf{246}, (2004), 501--519.

3859: 

3860: \bibitem{Kob}

3861: S.~Kobayashi, \emph{Bubbletons in 3-dimensional space forms via the {DPW}

3862:   method}, Master's thesis, Kobe University, 2001.

3863: 

3864: \bibitem{Mah}

3865: A.~Mahler, \emph{{B}ianchi-{B}\"acklund and dressing transformations on

3866:   constant mean curvature surfaces}, Ph.D. thesis, Univ. of Toledo, 2002.

3867: 

3868: \bibitem{McI}

3869: I.~McIntosh, \emph{Global solutions of the elliptic 2d periodic {T}oda

3870:   lattice}, Nonlinearity \textbf{7} (1994), no.~1, 85--108.

3871: 

3872: \bibitem{PinS}

3873: U.~Pinkall and I.~Sterling, 

3874: \emph{On the classification of constant mean 

3875: curvature tori}, Ann. of Math \textbf{130} (1989), 

3876: 407--451.

3877: 

3878: \bibitem{RuhV}

3879: E.~A.~Ruh and J.~Vilms,

3880: \emph{The tension field of the Gauss map}, 

3881: Trans. Amer. Math. Soc. \textbf{149} (1970), 

3882: 569--573.

3883: 

3884: \bibitem{Sch:cmclab}

3885: N.~Schmitt, \emph{{CMCL}ab}, http://www.gang.umass.edu/software.

3886: 

3887: \bibitem{Sch:gallery}

3888: \bysame, \emph{{CMC} {G}allery}, http://www.gang.umass.edu/cmcgallery.

3889: 

3890: \bibitem{Sch:tri}

3891: \bysame, \emph{Constant mean curvature trinoids}, math.DG/0403036.

3892: 

3893: \bibitem{SteW}

3894: I.~Sterling and H.~Wente, \emph{Existence and classification of constant mean

3895:   curvature multibubbletons of finite and infinite type}, Indiana Univ. Math.

3896:   J. \textbf{42} (1993), no.~4, 1239--1266.

3897: 

3898: \bibitem{TerU}

3899: C.~Terng and K.~Uhlenbeck, \emph{B\"{a}cklund transformations and loop group

3900:   actions}, Comm. Pure and Appl. Math \textbf{LIII} (2000), 1--75.

3901: 

3902: \bibitem{Uhl}

3903: K.~Uhlenbeck, \emph{Harmonic maps into Lie groups 

3904:   (Classical solutions of the chiral model)}, 

3905: J. Diff. Geom. \textbf{30} (1989), 1--50.

3906: 

3907: 

3908: \bibitem{Wu:dre}

3909: H.~Wu, \emph{On the dressing action of loop groups on constant mean curvature

3910:   surfaces}, T\^{o}hoku Math. J. \textbf{49} (1997), 599--621.

3911: 

3912: \bibitem{ZakS2} 

3913: V.~E.~Zakharov and A.~S.~Shabat,

3914: \emph{Integration of the nonlinear equations of 

3915: mathematical physics by the method of the inverse 

3916: scattering problem II}, 

3917: Funktsional. Anal. i Prilozhen. \textbf{13} (1978),

3918: 13--22.

3919: 

3920: \end{thebibliography}

3921: 

3922: 

3923: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3924: 

3925: 

3926: \end{document}

3927: 

3928: