1: \documentclass{lms}

2: 

3: %% CARACTERES

4: \usepackage{amssymb,amsmath} 

5: \usepackage[mathscr]{euscript} 

6: \usepackage[all]{xy}

7: 

8: \newcommand{\R}{\ensuremath{\mathbb{R}}}

9: \newcommand{\C}{\ensuremath{\mathbb{C}}}

10: \newcommand{\N}{\ensuremath{\mathbb{N}}}

11: \newcommand{\Z}{\ensuremath{\mathbb{Z}}}

12: \renewcommand{\O}{\mathcal{O}}

13: \renewcommand{\Re}{\mathop{\rm Re}} 

14: \renewcommand{\Im}{\mathop{\rm Im}}

15: \newcommand{\tr}{\mathop{\rm tr}} \newcommand{\Ker}{\mathop{\rm Ker}}

16: \newcommand{\iN}{\hbox{ {\leaders\hrule\hskip.2cm}{\vrule height .22cm} }}

17: \newcommand{\pa}[3][]{ \frac{ \partial^{#1} {#2} }{ \partial {#3}^{#1} } }

18: \renewcommand{\t}[1]{ {}^t \!  #1 } 

19: \newcommand{\at}[1]{\big|_{#1}}

20: \newcommand{\ds}{{\displaystyle } } 

21: \newcommand{\g}{\ensuremath{\mathfrak{g}}}

22: \newcommand{\gC}{\ensuremath{\mathfrak{g}^{\C}}}

23: \renewcommand{\u}{\ensuremath{\mathfrak{u}}}

24: \newcommand{\su}{\ensuremath{\mathfrak{su}}}

25: \renewcommand{\sl}{\ensuremath{\mathfrak{sl}}} \renewcommand{\phi}{\varphi}

26: \newcommand{\Ad}{\mathop{\rm Ad}} \newcommand{\ad}{\mathop{\rm ad}}

27: \renewcommand{\.}{\cdot} \newcommand{\lf}{\hfill\break}

28: 

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

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

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

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

33: 

34: \newnumbered{definition}[theorem]{Definition}

35: \newnumbered{example}[theorem]{Example}

36: \newunnumbered{remark}[theorem]{Remark}

37: 

38: \title{Hamiltonian stationary tori in the complex projective plane}

39: \author{Fr\'ed\'eric H\'elein \and Pascal Romon}

40: 

41: \classno{53C55 (primary), 53C42, 53C25, 58E12 (secondary)}

42: 

43: \begin{document}

44: \maketitle

45: 

46: 

47: 

48: \section{Introduction}

49: Hamiltonian stationary Lagrangian surfaces are Lagrangian surfaces of a given

50: four-dimensional manifold endowed with a symplectic and a Riemannian structure,

51: which are critical points of the area functional with respect to a particular

52: class of infinitesimal variations preserving the Lagrangian constraint: 

53: the compactly supported Hamiltonian vector fields.  The

54: Euler--Lagrange equations of this variational problem are highly simplified when

55: we assume that the ambient manifold ${\cal N}$ is K\"ahler.  In that case we can

56: make sense of a {\em Lagrangian angle function} $\beta$ along any

57: simply-connected Lagrangian submanifold $\Sigma \subset {\cal N}$ (uniquely

58: defined up to the addition of a constant).  And as shown in \cite{ScWo} the mean

59: curvature vector of the submanifold is then $\vec{H}= J\,\nabla \beta$, where

60: $J$ is the complex structure on ${\cal N}$ and $\nabla \beta$ is the gradient of

61: $\beta$ along $\Sigma$. It turns out that $\Sigma$ is Hamiltonian stationary if and

62: only if $\beta$ is a harmonic function on $\Sigma$.\\

63: 

64: \noindent A particular subclass of solutions occurs when $\beta$ is constant:

65: the Lagrangian submanifold is then simply a minimal one.  In the case where

66: ${\cal N}$ is a Calabi--Aubin--Yau manifold, such submanifolds admit an

67: alternative characterization as {\em special Lagrangian}, a notion which has

68: been extensively studied recently because of its connection with string theories

69: and the mirror conjecture, see \cite{SYZ}.\\

70: 

71: \noindent An analytical theory of two-dimensional Hamiltonian stationary

72: Lagrangian submanifolds was constructed by R. Schoen and J. Wolfson \cite{ScWo},

73: proving the existence and the partial regularity of minimizers.  In contrast our

74: results in the present paper rest on the fact that, for particular ambient

75: manifolds ${\cal N}$, Hamiltonian stationary Lagrangian surfaces are solutions

76: of an integrable system.  This was discovered first in the case when ${\cal

77: N}=\C^2$ in \cite{HR1} and \cite{HR2}.  In a subsequent paper \cite{HR3} we

78: proved that the same problem is also completely integrable if we replace $\C^2$

79: by any two-dimensional Hermitian symmetric space.  Among these symmetric spaces

80: one very interesting example is $\C P^2$, because any simply-connected

81: Lagrangian surface in $\C P^2$ can be lifted into a Legendrian surface in $S^5$.

82: Furthermore the cone in $\C^3$ over this Legendrian surface is actually a

83: singular Lagrangian three-dimensional submanifold in $\C^3$; and the cone in

84: $\C^3$ is Hamiltonian stationary if and only if the surface in $\C P^2$ is so.\\

85: 

86: \noindent A similar correspondence has been remarked and used in \cite{J},

87: \cite{McI3} and \cite{Has} in the case of minimal Lagrangian surfaces in $\C

88: P^2$ and allows these Authors to connect results on minimal Lagrangian surfaces

89: in $\C P^2$ \cite{Sh} to minimal Legendrian surfaces in $S^5$ \cite{MM} and

90: special Lagrangian cones in $\C^3$.\\

91: 

92: \noindent Our aim in this paper is the following:

93: \begin{itemize}

94: \item to expound in details the correspondence between Hamiltonian stationary

95: Lagrangian surfaces in $\C P^2$ and Hamiltonian stationary Legendrian surfaces

96: in $S^5$ and a formulation using a family of curvature free connections of this

97: integrable system (theorem \ref{1.6.theobis}).  

98: We revisit here the formulation given in \cite{HR3}, using {\em twisted loop groups}.  

99: Roughly speaking it rests on the identifications 

100: $\C P^2\simeq SU(3)/S(U(2)\times U(1))$ and $(S^5,\hbox{ contact structure}) \simeq

101: \left( U(3)/U(2)\times U(1), A^3_3=0\right)$, where $A^3_3$ is a component of

102: the Maurer--Cartan form.  We also show that this problem has an alternative

103: formulation, analogous to the theory of K. Uhlenbeck \cite{U} for harmonic maps

104: into $U(n)$, using {\em based loop groups}.

105: 

106: \item to define the notion of {\em finite type} Hamiltonian stationary

107: Legendrian surfaces in $S^5$: we give here again two definitions, in terms of

108: twisted loop groups (which is an analogue to the description of finite type

109: harmonic maps into homogeneous manifolds according to \cite{BP}) and in terms of

110: based loop groups (an analogue to the description of finite type harmonic maps

111: into Lie groups according to \cite{BFPP}).  We prove the equivalence between the

112: two definitions because we actually need this result for the following.  We

113: believe that this fact should be well known to some specialists in the harmonic

114: maps theory, but we did not find it in the literature.

115: 

116: \item we prove in theorem \ref{4.MainTheo} that all Hamiltonian stationary 

117: Lagrangian tori in $\C P^2$ (and

118: hence Hamiltonian stationary Legendrian tori in $S^5$) are of finite type.  This

119: is the main result of this paper.  Our proof focuses on the case of Hamiltonian

120: stationary tori which are not minimal, since the minimal case has been studied

121: by many authors (\cite{BFPP}, \cite{Sh}, \cite{MM}, \cite{Has},

122: \cite{McI1},\cite{McI2}, \cite{McI3}, \cite{J}).  The method here is adapted

123: from the similar result for harmonic maps into Lie groups in \cite{BFPP}.

124: However the strategy differs slightly: we use actually the two existing

125: formulations of finite type solutions, using twisted or based loop groups.  One

126: crucial step indeed is the construction of a {\em formal Killing field},

127: starting from a given torus solution.  This step can be slightly simplified here

128: in the twisted loop groups formulation, because the semi-simple element we start

129: with is then just constant.  However proving that the formal Killing field is

130: {\em adapted} requires more work in the twisted loop groups formulation

131: (actually we were not able to do it directly) than in the based loop groups

132: formulation; here we take advantage from the two formulations to avoid the

133: difficulties and to conclude.

134: 

135: \item lastly we give some examples of Hamiltonian stationary Legendrian tori in

136: $S^5$: we construct in theorem \ref{homogeneous} a family of solutions which are equivariant in some sense

137: under the action of the torus, that we call {\em homogeneous} Hamiltonian

138: stationary tori.  These are the simplest examples that one can build.

139: \end{itemize}

140: 

141: \noindent Let us add that the structure of the integrable system studied here

142: fits in a classification of elliptic integrable systems proposed by C.L. Terng

143: \cite{T}, as a 2nd $(U(3),\sigma,\tau)$-system\footnote{We have here exchanged

144: the notations $\sigma$ and $\tau$ with respect to \cite{T} in order to be

145: consistent with our notations in \cite{HR3}.}, where $\sigma$ is an involution

146: of $U(3)$ such that its fixed set is $U(3)^\sigma\simeq U(2)\times U(1)$ and

147: $U(3)/U(2)^\sigma\simeq \C P^2$ and $\tau$ is a 4th order automorphism (actually

148: $\tau^2=\sigma$) which encodes the symplectic structure on $\C P^2$ or the

149: Legendrian structure on $S^5$.\\

150: 

151: \noindent {\em Notations} --- For any matrix $M\in GL(n,\C)$, we denote by

152: $M^\dagger :=\,^t\overline{M}$.

153: 

154: 

155: 

156: \section{Geometrical description of Hamiltonian stationary Lagrangian surfaces

157: in $\mathbb{C}P^n$}

158: 

159: \subsection{The Lagrangian angle}

160: 

161: The complex projective space $\mathbb{C}P^n$ can be identified with the quotient

162: manifold $S^{2n+1}/S^1$. It is a complex manifold with complex structure $J$. We

163: denote by $\pi:S^{2n+1}\longrightarrow \mathbb{C}P^n$

164:  the canonical projection a.k.a. Hopf fibration, and equip $\mathbb{C}P^n$ with the

165: Fubini-Study Hermitian metric, denoted by $\langle \cdot , \cdot \rangle _{\mathbb{C}P^n} =

166: \langle \cdot , \cdot \rangle  - i\omega(\cdot , \cdot )$, where

167: $\langle \cdot , \cdot \rangle$ is a Riemannian metric and $\omega$ is the

168: K\"ahler form\footnote{note that the sign

169: convention may vary in the literature, e.g.\,some Authors use $\langle \cdot ,

170: \cdot \rangle _{\mathbb{C}P^n} = \langle \cdot , \cdot \rangle +

171: i\omega(\cdot , \cdot )$.}. 

172: For each $z\in S^{2n+1}$ we let ${\cal H}_z$ be the complex $n$-subspace 

173: in $T_{z} S^{2 n+1}\subset \mathbb{C}^{n+1}$ which is Hermitian orthogonal to $z$ 

174: (and hence to the fiber of $d\pi_{z}$). By construction of the Fubini-Study metric,

175: $d\pi_z:{\cal H}_z\longrightarrow T_{\pi(z)}\mathbb{C}P^n$ is 

176: an isometry between complex Hermitian spaces. We call the

177: subbundle ${\cal H}:= \cup_{z\in S^{2n+1}}{\cal H}_z$ of $T S^{2n+1}$

178: the {\em horizontal distribution}. It defines in a natural way a connection

179: $\nabla^\mathit{Hopf}\simeq \nabla ^H$ on the Hopf bundle 

180: $\pi:S^{2n+1}\longrightarrow \mathbb{C}P^n$, whose curvature is $2i\omega$. 

181: As a consequence \cite{Re2}:

182: \begin{proposition}\label{2.1.prop}

183: Let $\Omega$ be a {\em simply connected} open subset of $\mathbb{R}^n$ and 

184: $u:\Omega \longrightarrow \mathbb{C}P^n$

185: be a smooth Lagrangian immersion, i.e.\,such that $u^*\omega = 0$. Then there exists a lift

186: \begin{center}

187: \hskip2cm

188: \xymatrix{

189: & S^{2n+1} \ar@{->}[d]^\pi \\

190: \Omega \ar@{->}[ur]^{\widehat{u}} \ar@{->}[r]^u & \mathbb{C}P^n

191: }

192: \end{center}

193: such that $\left(u^*\nabla^H\right) \widehat{u} = 0$ (where $u^*\nabla^H$ is 

194: the pull-back by $u$ of the connection $\nabla^H$). This lift is unique up to multiplication

195: by a unit complex number. Moreover the pull-back by $\widehat{u}$ of the symplectic

196: form  $\omega$ on $\mathbb{C}^{n+1}$ vanishes; we say that $\hat{u}$ is 

197: \emph{Legendrian}.

198: \end{proposition}

199: 

200: Taking $u,\hat{u}$ as above, we define, for any orthonormal framing

201: $(e_{1},\ldots,e_{n})$ of $T\Omega$, the Lagrangian angle $\beta$ by

202: \[

203: e^{i \beta} = dz^1 \wedge \ldots \wedge dz^{n+1} \big( \hat{u}, d\hat{u}(e_{1}),

204: \ldots, d\hat{u}(e_{n}) \big) .

205: \]

206: which makes sense because $( \hat{u}, d\hat{u}(e_{1}),\ldots, d\hat{u}(e_{n}))$ is

207: a Hermitian-orthonormal frame, for any $x\in\Omega$. Furthermore, the result is

208: independent from the choice of the framing, and depends on the choice of the lift

209: $\hat{u}$ only through multiplication by a unit complex constant. Hence $\beta$ is

210: defined up to an additive constant and $d\beta$ is always well-defined along any

211: Lagrangian immersion $u$. Another characteristic property of the Lagrangian angle

212: relates it to the mean curvature vector field $\vec{H}$ along $u$:

213: \begin{equation}\label{2.2.H=JDb}

214: \vec{H} = {1\over n}J\nabla \beta

215: \end{equation}

216: or equivalently $d\beta = - \vec{H} \iN \omega$ (see~\cite{Br,Da} for details). 

217: 

218: \subsection{Hamiltonian stationary Lagrangian submanifolds}

219: %

220: A \emph{Hamiltonian stationary} Lagrangian submanifold $\Sigma$ in

221: $\mathbb{C}P^n$ is a Lagrangian submanifold which is a critical point of the

222: $n$-volume functional ${\cal A}$ under first variations which are {\em

223: Hamiltonian vector fields} with compact support.  This means that for any smooth

224: function with compact support $h\in {\cal

225: C}^\infty_c(\mathbb{C}P^n,\mathbb{R})$, we have

226: \[

227: \delta {\cal A}_{\xi_h}(\Sigma) := \int_\Sigma \left\langle \vec{H},

228: \xi_h\right\rangle_E d\hbox{vol} = 0,

229: \]

230: where $\xi_h$ is the Hamiltonian vector field of $h$, i.e.\,satisfies

231: $\xi_h \iN \omega +dh = 0$ or $\xi_h = J\nabla h$.  We also remark that if

232: $f\in {\cal C}^\infty_c(\Sigma,\mathbb{R})$, then there exist smooth extensions

233: with compact support $h$ of $f$, i.e.\,functions $h\in {\cal

234: C}^\infty_c(\mathbb{C}P^n,\mathbb{R})$ such that $h_{|\Sigma} = f$, and moreover

235: the normal component of $\left(\xi_h\right)_{|\Sigma}$ does not depend on the

236: choice of the extension $h$ (it coincides actually with $J\nabla f$, where

237: $\nabla$ is here the gradient with respect to the induced metric on $\Sigma$).

238: So we deduce from above that $\delta {\cal

239: A}_{\xi_h}(\Sigma) = {1\over n}\int_\Sigma \left\langle \nabla \beta, \nabla

240: f\right\rangle_E d\hbox{vol}$.  This implies the following.

241: \begin{corollary}

242: Any Lagrangian submanifold $\Sigma$ in $\mathbb{C}P^n$ is Hamiltonian stationary

243: if and only if $\beta$ is a harmonic function on $\Sigma$, i.e.

244: \[

245: \Delta_\Sigma \beta = 0.

246: \]

247: \end{corollary}

248: 

249: \noindent This theory extends to non simply connected surfaces $\Sigma$

250: with the following restrictions.  Let $\gamma$ be a homotopically non trivial

251: loop.  The Legendrian lift of $\gamma$ needs not close, so that in general its

252: endpoints $p_1, p_2 \in S^{2n+1}$ are multiples of each other by a factor $e^{i

253: \theta}$.  The same holds for the Lagrangian angle: $\beta ( p_2 ) \equiv \beta

254: ( p_1 ) + (n+1) \theta \bmod 2 \pi$ (since the tangent plane is also shifted by

255: the Decktransformation $z \longmapsto e^{i \theta} z$).  In particular $\beta$

256: is not always globally defined on surfaces in $\mathbb{C} P^n$ with non trivial

257: topology, unless the Legendrian lift is globally defined in $S^{2n+1} /

258: \mathbb{Z}_{n+1}$ (here $\mathbb{Z}_{n+1}$ stands for the $n+1$-st roots of

259: unity in $SU(n+1)$).

260: 

261: 

262: \subsection{Conformal Lagrangian immersions into $\mathbb{C}P^2$}

263: We now set $n= 2$.  We suppose that $\Omega$ is a simply connected open subset

264: of $\mathbb{R}^2\simeq \mathbb{C}$ and consider a conformal Lagrangian immersion

265: $u:\Omega\longrightarrow \mathbb{C}P^2$.  This implies that we can find a

266: function $\rho:\Omega\longrightarrow \mathbb{R}$ and two sections $E_1$ and

267: $E_2$ of $u^*T\mathbb{C}P^2$ such that $\forall (x,y)\in \Omega$,

268: $(E_1(x,y),E_2(x,y))$ is an Euclidean orthogonal basis over $\mathbb{R}$ of

269: $T_{u(x,y)}u(\Omega)$ and

270: \[

271: d u = e^\rho\left( E_1d x + E_2d y\right).

272: \]

273: We observe that, due to the fact that $u$ is Lagrangian, $(E_1,E_2)$ is a also a

274: {\em Hermitian} basis over $\mathbb{C}$ of $T_{u(x,y)}\mathbb{C}P^2$.\\

275: 

276: \noindent Let \xymatrix{ \Omega\ar@{->}[r]^{\widehat{u}}\ar@{->}[dr]^u & S^5

277: \ar@{->}[d]^\pi \\

278:  & \mathbb{C}P^2 } be a parallel lift of $u$ as in Proposition \ref{2.1.prop}

279:  and $(e_1,e_2)$ be the unique section of $\widehat{u}^*{\cal H}\times

280:  \widehat{u}^*{\cal H}$ which lifts\footnote{recall that the condition that

281:  $v\in \left(\widehat{u}^*{\cal H}\right)_{(x,y)}$ means that $v$ is in the

282:  horizontal subspace ${\cal H}_{\widehat{u}(x,y)}$} $(E_1,E_2)$.  Then we have

283: \begin{equation}\label{duchapeau}

284: d\widehat{u} = e^\rho\left( e_1d x + e_2dy\right).

285: \end{equation}

286: Note that $\forall (x,y)\in \Omega$, $(e_1(x,y),e_2(x,y))$ is a Hermitian basis

287: of ${\cal H}_{\widehat{u}(x,y)}$, which is Hermitian orthogonal to

288: $\widehat{u}(x,y)$.  Hence $\forall (x,y)\in \Omega$,

289: $(e_1(x,y),e_2(x,y),\widehat{u}(x,y))$ is a Hermitian basis of $\mathbb{C}^3$.

290: Thus this triplet can be identified with some $\widehat{F}(x,y)\in U(3)$.  We

291: hence get the diagram \xymatrix{ & U(3)\ar@{->}[d]^{(\cdot \cdot *)}\\

292: \Omega \ar@{->}[ur]^{\widehat{F}} \ar@{->}[r]^{\widehat{u}} \ar@{->}[dr]^u &

293: S^5\ar@{->}[d]^{\pi}\\

294: & \mathbb{C}P^2 }, where $(\cdot \cdot *)$ is the mapping

295: $(e_1,e_2,e_3)\longmapsto e_3$.\\

296: 

297: \noindent We define the Maurer--Cartan form $\widehat{A}$ to be the 1-form on

298: $\Omega$ with coefficients in $\u(3)$ such that $d\widehat{F} = \widehat{F}\cdot

299: \widehat{A}$.  Then we remark that the horizontality assumption $\langle

300: d\widehat{u},\widehat{u}\rangle _{\mathbb{C}^3} = 0$ exactly means that

301: \begin{equation}\label{A33=0}

302: \widehat{A}^3_3 = 0.

303: \end{equation}

304: Moreover the Lagrangian angle function $\beta_{\widehat{u}}$ along $\widehat{u}$

305: can be computed by

306: \[

307: e^{i\beta_{\widehat{u}}} = dz^1 \wedge dz^2 \wedge dz^3 (e_1,e_2,\widehat{u}) 

308: = \det \widehat{F}.

309: \]

310: As in \cite{HR1} we consider a larger class of framings of $u$ as follows.

311: \begin{definition}

312: A \emph{Legendrian framing of $u$ along $\widehat{u}$} is a map

313: $F:\Omega\longrightarrow U(3)$ such that

314: \begin{itemize}

315: \item $(\cdot \cdot *)\circ F = \widehat{u}$ \item $\det F =

316: e^{i\beta_{\widehat{u}}}$.

317: \end{itemize}

318: \end{definition}

319: It is easily seen that the first condition is equivalent to the fact that there

320: exists a smooth map $G:\Omega \longrightarrow U(3)$ (a gauge transformation) of

321: the type

322: \[

323: G(x,y) = \left(\begin{array}{cc}g(x,y) & 0 \\ 0 & 1\end{array}\right),\quad

324: \hbox{where }g:\Omega\longrightarrow U(2)

325: \]

326: such that \[F(x,y) = \widehat{F}(x,y)\cdot G^{-1}(x,y).

327: \]

328: And then the second one is equivalent to say that $g$ takes values in $SU(2)$.

329: 

330: \subsection{A splitting of the Maurer--Cartan form of a Legendrian framing}

331: Using (\ref{duchapeau}) and (\ref{A33=0}) one obtains the following

332: decomposition of $\widehat{A}$:

333: \[

334: \widehat{A} = \widehat{A}_{\u(1)} + \widehat{A}_{\su(2)} +

335: \widehat{A}_{\mathbb{C}^2},

336: \]

337: with the notations

338: \[

339: \widehat{A}_{\u(1)} = \left(\begin{array}{cc}\widehat{\alpha}_{\u(1)} & 0\\0 &

340: 0\end{array}\right),\quad \widehat{A}_{\su(2)} =

341: \left(\begin{array}{cc}\widehat{\alpha}_{\su(2)} & 0\\0 & 0\end{array}\right),

342: \]

343: \[

344: \hbox{and}\quad \widehat{A}_{\mathbb{C}^2} = e^\rho\left(\begin{array}{cc} 0 &

345: \epsilon dz + \overline{\epsilon}d\bar{z}\\

346: -\,^t\!\left(\epsilon dz + \overline{\epsilon}d\bar{z}\right) &

347: 0\end{array}\right),

348: \]

349: where $\widehat{\alpha}_{\u(1)}$ is a 1-form on $\Omega$ with coefficients in

350: $\u(1)\simeq \mathbb{R}\left(\begin{array}{cc}i&0\\0&i\end{array}\right)$,

351: $\widehat{\alpha}_{\su(2)}$ a 1-form on $\Omega$ with coefficients in $\su(2)$,

352: $\epsilon:={1\over 2}\left(\begin{array}{c}1\\-i\end{array}\right)$ and

353: $\overline{\epsilon}:={1\over 2}\left(\begin{array}{c}1\\i\end{array}\right)$

354: (so that $\epsilon dz +

355: \overline{\epsilon}d\bar{z}=\left(\begin{array}{c}dx\\dy\end{array}\right)$).

356: Note that $\det F = e^{i\beta_{\widehat{u}}}$ implies $\widehat{\alpha}_{\u(1)}

357: = {d\beta_{\widehat{u}}\over

358: 2}\left(\begin{array}{cc}i&0\\0&i\end{array}\right)$.\\

359: \noindent Now we let $F:\Omega\longrightarrow U(3)$ be a Legendrian framing and

360: $A:= F^{-1}\cdot dF$.  The relation $F = \widehat{F}\cdot G^{-1}$ implies that

361: $A = G\cdot \widehat{A}\cdot G^{-1} - dG\cdot G^{-1}$.  Hence

362: \[

363: A = A_{\u(1)} + A_{\su(2)} + A_{\mathbb{C}^2},

364: \]

365: where, using the fact that $\u(1)$ commutes with $\su(2)$,

366: \[

367: A_{\u(1)} = \left(\begin{array}{cc}\alpha_{\u(1)} & 0\\0 & 0\end{array}\right) =

368: \left(\begin{array}{cc}\widehat{\alpha}_{\u(1)} & 0\\0 & 0\end{array}\right) =

369: \left(\begin{array}{ccc}i{d\beta_{\widehat{u}}\over

370: 2}&0&0\\0&i{d\beta_{\widehat{u}}\over 2}&0\\

371: 0&0&0\end{array}\right),

372: \]

373: \[

374: A_{\su(2)} = \left(\begin{array}{cc}\alpha_{\su(2)} & 0\\0 & 0\end{array}\right)

375: = \left(\begin{array}{cc}g\cdot \widehat{\alpha}_{\su(2)}\cdot g^{-1} -dg\cdot

376: g^{-1} & 0\\0 & 0\end{array}\right)

377: \]

378: and

379: \[

380: A_{\mathbb{C}^2} = e^\rho\left(\begin{array}{cc} 0 & g\cdot\left( \epsilon dz +

381: \overline{\epsilon}d\bar{z}\right)\\

382: -\left(g\cdot \left(\epsilon dz +

383: \overline{\epsilon}d\bar{z}\right)\right)^\dagger & 0\end{array}\right).

384: \]

385: We can further split the last term $A_{\mathbb{C}^2}$ along $dz$ and $d\bar{z}$

386: as $A_{\mathbb{C}^2} = A_{\mathbb{C}^2}' + A_{\mathbb{C}^2}''$ where

387: \[

388: A_{\mathbb{C}^2}':= e^\rho\left(\begin{array}{cc} 0 & g\cdot \epsilon \\

389: -\left(g\cdot \overline{\epsilon}\right)^\dagger & 0\end{array}\right) dz

390: \quad\hbox{and}\quad A_{\mathbb{C}^2}'':= e^\rho\left(\begin{array}{cc} 0 &

391: g\cdot \overline{\epsilon} \\

392: -\left(g\cdot \epsilon\right)^\dagger & 0\end{array}\right) d\bar{z}.

393: \]

394: 

395: \subsection{Interpretation in terms of an automorphism}

396: As expounded in \cite{HR1} and \cite{HR3} the key point in order to exploit the

397: structure of an integrable system is to observe that the splitting $A =

398: A_{\u(1)} + A_{\su(2)} + A_{\mathbb{C}^2}' + A_{\mathbb{C}^2}''$ corresponds to

399: a decomposition along the eigenspaces of the following automorphism in

400: $\u(3)^{\mathbb{C}}$, the complexification\footnote{we can define

401: $\u(3)^{\mathbb{C}}$ as the set $M(3,\mathbb{C})$ with its standard complex

402: structure and with the conjugation mapping $c:M\longmapsto -M^\dagger$; clearly

403: $c$ is a Lie algebra automorphism, an involution and the set of fixed points of

404: $c$ is $\u(3)$.\\

405: Similarly the complexification $U(3)^{\mathbb{C}}$ is the set $GL(3,\mathbb{C})$

406: with its standard complex structure and the conjugation map $C:G\longmapsto

407: \left(G^\dagger\right)^{-1}$.} of $\u(3)$.  We let $J:=

408: \left(\begin{array}{cc}0&-1\\1&0\end{array}\right)$ and

409: \[

410: \begin{array}{cccl}

411: \tau: & \u(3)^{\mathbb{C}} & \longrightarrow & \u(3)^{\mathbb{C}}\\

412:  & M & \longmapsto & - \left( \begin{array}{cc}-J & 0\\0 & 1\end{array}\right)

413:  \cdot \,^t\!M\cdot \left( \begin{array}{cc}J & 0\\0 & 1\end{array}\right).

414: \end{array}

415: \]

416: It is then straightforward that $\tau$ is a Lie algebra automorphism, that

417: $\u(3)$ is stable by $\tau$ and that $\tau^4 = Id$.  Hence we can diagonalize

418: the action of $\tau$ over $\u(3)^{\mathbb{C}}$ and in the following we denote by

419: $\u(3)^{\mathbb{C}}_a$ the eigenspace of $\tau$ for the eigenvalue $i^a$, for $a

420: = -1,0,1,2$.  We first point out that the eigenspaces $\u(3)^{\mathbb{C}}_0$ and

421: $\u(3)^{\mathbb{C}}_2$, with eigenvalues 1 and $-1$ respectively, are the

422: complexifications of $\u(3)_0$ and $\u(3)_2$ respectively, where

423: \[

424: \u(3)_0:= \left\{ \left(\begin{array}{cc}g & 0 \\0 & 0\end{array}\right) /g\in

425: \su(2)\right\} \hbox{ and } \u(3)_2:= \left\{ \left(\begin{array}{ccc}\lambda

426: i&0&0\\0&\lambda i&0\\0&0&\mu i\end{array}\right) / \lambda, \mu\in

427: \mathbb{R}\right\}.

428: \]

429: This can be obtained by first computing that

430: \[

431: \tau \left(\begin{array}{cc}A & X \\-\,^t\!Y & d\end{array}\right) =

432: \left(\begin{array}{cc}J\,^t\!AJ & -JY \\-\,^t\!XJ & -d\end{array}\right), \quad

433: \forall A\in M(2,\mathbb{C}),\forall X,Y\in \mathbb{C}^2,\forall d\in

434: \mathbb{C},

435: \]

436: and by using the fact that $\forall A\in \sl(2,\mathbb{C})$, $J\,^t\!AJ = A$.

437: Similarly the eigenspaces $\u(3)^{\mathbb{C}}_{1}$ and

438: $\u(3)^{\mathbb{C}}_{-1}$, with eigenvalues $i$ and $-i$ respectively, are found

439: to be

440: \[

441: \u(3)^{\mathbb{C}}_{1} = \left\{ \left(\begin{array}{cc}0 & X \\-\,^t\!Y &

442: 0\end{array}\right) /X,Y\in \mathbb{C}^2, JY = -iX\right\}

443: \]

444: and

445: \[

446: \u(3)^{\mathbb{C}}_{-1} = \left\{ \left(\begin{array}{cc}0 & X \\-\,^t\!Y &

447: 0\end{array}\right) /X,Y\in \mathbb{C}^2, JY = iX\right\}.

448: \]

449: Now we have the following

450: \begin{lemma}

451: The eigenspaces $\u(3)^{\mathbb{C}}_{1}$ and $\u(3)^{\mathbb{C}}_{-1}$ can be

452: characterized by

453: \[

454: \u(3)^{\mathbb{C}}_{1} = \left\{\lambda \left(\begin{array}{cc}0 & h\cdot

455: \overline{\epsilon} \\

456: -\left(h\cdot \epsilon\right)^\dagger & 0\end{array}\right) / \lambda\in

457: [0,\infty), h\in SU(2)\right\}

458: \]

459: and

460: \[

461: \u(3)^{\mathbb{C}}_{-1} = \left\{ \lambda\left(\begin{array}{cc}0 & h\cdot

462: \epsilon \\

463: -\left(h\cdot \overline{\epsilon}\right)^\dagger & 0\end{array}\right) /

464: \lambda\in [0,\infty), h\in SU(2)\right\}.

465: \]

466: \end{lemma}

467: \begin{proof} This can be proved either by adapting the argument in section 2.4

468: of \cite{HR1} or by a straightforward computation which exploits the fact that

469: $\forall h\in SU(2)$, $\overline{h}J = Jh$, $J\epsilon = i\epsilon$ and

470: $J\overline{\epsilon} = -i \overline{\epsilon}$.  \end{proof}

471: 

472: \noindent We conclude that if, using $\u(3)^{\mathbb{C}} =

473: \u(3)^{\mathbb{C}}_{-1}\oplus \u(3)^{\mathbb{C}}_0\oplus

474: \u(3)^{\mathbb{C}}_1\oplus \u(3)^{\mathbb{C}}_2$, we decompose $A$ as

475: \[

476: A = A_{-1} + A_0 + A_1 + A_2,

477: \]

478: where each $A_a$ is a 1-form with coefficients in the $\u(3)^{\mathbb{C}}_a$,

479: then we recover the previous splitting by setting $A_0 = A_{\su(2)}$, $A_2 =

480: A_{\u(1)}$, $A_{-1} = A_{\mathbb{C}^2}'$ and $A_1 = A_{\mathbb{C}^2}''$.  Note

481: that the two last conditions actually reflects the conformality of $u$.

482: 

483: \begin{remark}

484: Note that by the automorphism property $[\u(3)_a,\u(3)_b]\subset \u(3)_{a+b\bmod

485: 4}$.

486: \end{remark}

487: 

488: \subsection{Legendrian framings of Hamiltonian stationary Lagrangian immersions}

489: \noindent Given the Legendrian framing $F$ of a conformal Lagrangian immersion

490: $u$ in $\mathbb{C}P^2$, we define the family of deformations $A_\lambda$ of its

491: Maurer--Cartan form $A$, for $\lambda\in S^1\subset \mathbb{C}^*$ by

492: \begin{equation}\label{1.6.Alambda}

493: A_\lambda:= \lambda^{-2}A_2' + \lambda^{-1}A_{-1} + A_0 + \lambda A_1 +

494: \lambda^2A_2'',

495: \end{equation}

496: where $A_2':= A_2(\partial /\partial z)dz$ and $A_2'':= A_2(\partial /\partial

497: \bar{z})d\bar{z}$.  We then have the following:

498: 

499: \begin{theorem}\label{1.6.theo}

500: Given a conformal Lagrangian immersion $u:\Omega\longrightarrow \mathbb{C}P^2$

501: and a Legendrian framing $F$ of $u$, the Maurer--Cartan form of $F$ satisfies

502: \begin{equation}\label{1.6.lag}

503: A_{-1} = A_{-1}'= A_{\mathbb{C}^2}'\hbox{ and }A_1 = A_1''= A_{\mathbb{C}^2}'.

504: \end{equation}

505: Furthermore $u$ is Hamiltonian stationary if and only if, defining $A_\lambda$

506: as in (\ref{1.6.Alambda}),

507: \begin{equation}\label{1.6.curvature}

508: dA_\lambda + A_\lambda\wedge A_\lambda = 0,\quad \forall \lambda\in S^1.

509: \end{equation}

510: \end{theorem}

511: %

512: \begin{remark}

513: For $\lambda = 1$, $A_1 = A$ and equation (\ref{1.6.curvature}) is a consequence

514: of its definition $A:=F^{-1}\cdot dF$.

515: \end{remark}

516: %

517: \begin{proof} See \cite{HR1} and \cite{HR3}.\end{proof}

518: \noindent We remark that all the conditions that have been collected about the

519: components $A_a$ can be encoded by the following twisting condition on

520: $A_\lambda$:

521: \[

522: \forall \lambda\in S^1,\quad \tau(A_\lambda) = A_{i\lambda}.

523: \]

524: Thus we are led to define the following {\em twisted loop algebra}

525: \[

526: \Lambda \u(3)_\tau := \{S^1\ni\lambda \longmapsto \xi_\lambda\in \u(3)/\forall

527: \lambda\in S^1, \tau(\xi_\lambda) = \xi_{i\lambda}\},

528: \]

529: and $A_\lambda$ is a 1-form on $\Omega$ with coefficients in $\Lambda

530: \u(3)_\tau$.\\

531: 

532: \noindent Actually $\Lambda \u(3)_\tau$ is the Lie algebra of the following {\em

533: twisted loop group}

534: \[

535: \Lambda U(3)_\tau := \{S^1\ni\lambda \longmapsto g_\lambda\in U(3)/\forall

536: \lambda\in S^1, \tau(g_\lambda) = g_{i\lambda}\},

537: \]

538: where the Lie algebra automorphism $\tau :\u(3)^{\mathbb{C}}\longrightarrow

539: \u(3)^{\mathbb{C}}$ has been extended to the Lie group automorphism by

540: \[

541: \begin{array}{cccc}

542: \tau: & U(3)^{\mathbb{C}} & \longrightarrow & U(3)^{\mathbb{C}}\\

543:  & M & \longmapsto & \left( \begin{array}{cc}-J & 0\\0 & 1\end{array}\right)

544:  \cdot \,^t\!M^{-1}\cdot \left( \begin{array}{cc}J & 0\\0 & 1\end{array}\right).

545: \end{array}

546: \]

547: Now if we assume that $\Omega$ is simply connected, then relation

548: (\ref{1.6.curvature}) allows us to integrate $A_\lambda$, i.e.\,to find a map

549: $F_\lambda:\Omega\longrightarrow U(3)$ for any $\lambda\in S^1$ such that

550: $dF_\lambda = F_\lambda\cdot A_\lambda$.  Moreover if we choose some base point

551: $z_0\in \Omega$, then by requiring further that $F_\lambda(z_0) = Id$,

552: $F_\lambda$ is unique.  A key observation is then that $\tau(A_\lambda) =

553: A_{i\lambda}$ implies $\tau(F_\lambda) = F_{i\lambda}$, $\forall \lambda\in

554: S^1$.  Hence, a conformal Lagrangian immersion $u:\Omega\longrightarrow

555: \mathbb{C}P^2$ is Hamiltonian stationary if and only if any Legendrian lift $F$

556: of it can be deformed into a map $F_\lambda:\Omega\longrightarrow \Lambda

557: U(3)_\tau$, such that $F_\lambda^{-1}\cdot dF_\lambda$ has the form

558: (\ref{1.6.Alambda}).  Summarizing this result with the observations in the

559: previous section we have:

560: 

561: \begin{theorem}\label{1.6.theobis}

562: Given a simply connected domain $\Omega\subset \mathbb{C}$ and a base point

563: $z_0\in \Omega$, the set of Hamiltonian stationary conformal Lagrangian

564: immersions $u:\Omega\longrightarrow \mathbb{C}P^2$ such that $u(z_0) = [0:0:1]$

565: is in bijection with the set of maps $F_\lambda:\Omega\longrightarrow \Lambda

566: U(3)_\tau$, such that $F_\lambda(z_0)=\mathit{Id}$ and the Fourier decomposition

567: of $A_\lambda:=F_\lambda^{-1}\cdot dF_\lambda$, $A_\lambda = \sum_{k\in

568: \mathbb{Z}}\widehat{A}_k\lambda^k$ satisfies

569: \begin{equation}\label{1.6.-infty}

570: \forall k\in \mathbb{Z},\quad k\leq -3 \Longrightarrow \widehat{A}_k = 0,

571: \end{equation}

572: \begin{equation}\label{1.6.-2}

573: \widehat{A}_{-2}=

574: a(z)dz\left(\begin{array}{ccc}i&0&0\\0&i&0\\0&0&0\end{array}\right), \quad

575: \hbox{where }a\in{\cal C}^\infty(\Omega,\mathbb{C}),

576: \end{equation}

577: \begin{equation}\label{1.6.-1}

578: \widehat{A}_{-1}=\widehat{A}_{-1}(\partial /\partial z)dz,\quad \hbox{i.e. }

579: \widehat{A}_{-1}(\partial /\partial \bar{z}) = 0.

580: \end{equation}

581: \end{theorem}

582: \begin{proof} For any conformal Lagrangian Hamiltonian stationary immersion $u$

583: the existence of $F_\lambda$ and the properties (\ref{1.6.-infty}),

584: (\ref{1.6.-2}) and (\ref{1.6.-1}) are immediate consequences of Theorem

585: \ref{1.6.theo}.  Conversely for any map $F_\lambda$, conditions

586: (\ref{1.6.-infty}), (\ref{1.6.-2}) and (\ref{1.6.-1}) and the reality condition

587: $\overline{A_\lambda} = A_\lambda$ imply that $A_\lambda$ must satisfy

588: (\ref{1.6.Alambda}).  In particular we remark that condition (\ref{1.6.-2}) is a

589: reformulation of (\ref{A33=0}).  Thus by theorem \ref{1.6.theo} we deduce that

590: $F_1$ is the Legendrian lift of some Hamiltonian stationary conformal Lagrangian

591: immersion.\end{proof}

592: %

593: \begin{remark}

594: >From the analysis of the Maurer--Cartan form of a Legendrian lift we know that

595: actually the function $a$ in (\ref{1.6.-2}) is ${1\over 2}\partial

596: \beta/\partial z$, where $\beta$ is the Lagrangian angle function.  In

597: particular since $u$ is Hamiltonian stationary $\beta$ is harmonic and hence $a$

598: is holomorphic.

599: \end{remark}

600: 

601: \subsection{An alternative characterization}

602: We introduce here another construction using based loop groups for

603: characterizing Hamiltonian stationary Lagrangian conformal immersions.  Consider

604: \[

605: E_\lambda:= F_\lambda\cdot F^{-1}.

606: \]

607: We can observe that $E_\lambda$ is a map with values in the {\em based loop

608: group}

609: \[

610: \Omega U(3):= \{S^1\ni\lambda\longmapsto g_\lambda\in U(3)/g_{\lambda=1}=1\},

611: \]

612: since $F_{\lambda=1}=F$.  It is easy to check that $\Omega U(3)$ is a loop

613: group, the Lie algebra of which is

614: \[

615: \Omega \u(3):= \{S^1\ni\lambda\longmapsto \xi_\lambda\in

616: \u(3)/\xi_{\lambda=1}=0\}.

617: \]

618: Note that the (formal) Fourier expansion of an element $\xi_\lambda\in \Omega

619: \u(3)$ can be written $\xi_\lambda = \sum_{k\in\mathbb{Z}\setminus

620: \{0\}}\widehat{\xi}_k(\lambda^k-1)$.\\

621: 

622: \noindent The Maurer--Cartan form of $E_\lambda$ is

623: \[

624: \begin{array}{ccl}

625: \Gamma_\lambda & := & E_\lambda^{-1}\cdot dE_\lambda \\

626: & = & F\cdot\left( F_\lambda^{-1}\cdot dF_\lambda - F\cdot dF\right) \cdot

627: F^{-1} = F\cdot\left(A_\lambda - A\right) \cdot F^{-1}\\

628: & = & (\lambda^{-2}-1)\Gamma_2' + (\lambda^{-1}-1)\Gamma_{-1} +

629: (\lambda-1)\Gamma_1 + (\lambda^{2}-1)\Gamma_2'',

630:  \end{array}

631: \]

632: where $\Gamma_2':= F\cdot A_2'\cdot F^{-1}$, $\Gamma_{-1}:= F\cdot A_{-1}\cdot

633: F^{-1}$, $\Gamma_1:= F\cdot A_1\cdot F^{-1}$ and $\Gamma_2'':= F\cdot A_2''\cdot

634: F^{-1}$.  We can observe in particular that

635: \[

636: \Gamma_2' = ia \pi^\perp dz,\quad \hbox{where }\pi^\perp:=

637: F\cdot\left(\begin{array}{ccc}1&&\\&1&\\&&0\end{array}\right)\cdot F^{-1}.

638: \]

639: Note that $\pi^\perp$ is the Hermitian orthogonal projection in $\C^3$ onto the

640: plane $\widehat{u}^\perp$ (moreover $\pi^\perp$ is actually independent of the

641: lift $\widehat{u}$ chosen for $u$).\\

642: 

643: \noindent Lastly we point out the following equivariance property with respect

644: to the automorphism $\tau_u$ defined\footnote{ We can remark that the definition

645: of $\tau_u$ is independent from the choice of the Legendrian framing $F$ of $u$,

646: and depends only on $u$.  This means that for any pair of Legendrian framings

647: $F$ and $\widehat{F}$ such that $ \widehat{F} = F\cdot G$, where

648: $G=\left(\begin{array}{cc}g&\\&1\end{array}\right)$ and $g:\Omega\longrightarrow

649: SU(2)$, we have $\widehat{F}\cdot \tau(\widehat{F}^{-1}\cdot M\cdot

650: \widehat{F})\cdot \widehat{F}^{-1} = F\cdot \tau (F^{-1}\cdot M\cdot F)\cdot

651: F^{-1}$.  This can be checked by a computation using the fact that

652: $\left(\begin{array}{cc}\pm J&\\&1\end{array}\right)\cdot G =\overline{G}\cdot

653: \left(\begin{array}{cc}\pm J&\\&1\end{array}\right)$.} by

654: \[

655: \tau_u(M) = F\cdot \tau (F^{-1}\cdot M\cdot F)\cdot F^{-1}.

656: \]

657: We have obviously $\tau_u^4=1$.  Moreover, setting

658: \begin{eqnarray*}

659: \gamma_\lambda & := & \lambda^{-2}\Gamma_2' + \lambda^{-1}\Gamma_{-1} +

660: \lambda\Gamma_1+ \lambda^{2}\Gamma_2'' \\

661: & = & F\cdot\left( \lambda^{-2}A_2'+\lambda^{-1}A_{-1} + \lambda A_1+

662: \lambda^{2}A_2''\right)\cdot F^{-1}

663: \end{eqnarray*}

664: and $\gamma:= \gamma_{\lambda=1} = F\cdot\left( A_2'+A_{-1} + A_1+

665: A_2''\right)\cdot F^{-1}$, so that $\Gamma_\lambda = \gamma_\lambda -\gamma$, we

666: have

667: \[

668: \tau_u(\gamma_\lambda) = \gamma_{i\lambda}.

669: \]

670: 

671: 

672: \section{Finite type solutions}

673: \noindent In \cite{HR3} we showed how Theorem \ref{1.6.theobis} allows us to

674: adapt the theory of J. Dorfmeister, F. Pedit and H.Y. Wu \cite{DPW}, in order to

675: build a Weierstrass type representation theory of {\em all} conformal Lagrangian

676: Hamiltonian stationary immersions, i.e.\,using holomorphic data.  Here we want

677: to exploit Theorem \ref{1.6.theobis} in order to construct a particular class of

678: examples of solutions: the {\em finite type} ones.

679: 

680: \subsection{Definitions}

681: \noindent We invite the Reader to consult \cite{BFPP}, \cite{G} or \cite{H} for

682: more details.  We first observe that $U(3)_0:=\{g\in U(3)/ \tau(g)=g\}$, the

683: fixed set of $\tau$, is a subgroup of $U(3)$, the Lie algebra of which is

684: $\u(3)_0$ (same observation about $U(3)_0^{\C}$).  Actually $U(3)_0$ is

685: isomorphic to $SU(2)$ so that we make the identifications $U(3)_0\simeq SU(2)$

686: and $\u(3)_0\simeq \su(2)$.  We will need an Iwasawa decomposition of

687: $SU(2)^{\mathbb{C}}$ for our purpose: it will be a pair $(SU(2),\mathfrak{B})$

688: of {\em subgroups} of $SU(2)^{\mathbb{C}}$, such that $\forall g\in

689: SU(2)^{\mathbb{C}}$, $\exists !  (f,b)\in SU(2)\times \mathfrak{B}$ with

690: $g=f\cdot b$, a property that we summarize by writing $SU(2)^{\mathbb{C}} =

691: SU(2)\cdot \mathfrak{B}$.  Moreover $\mathfrak{B}$ is a solvable Borel subgroup.

692: We can choose for example

693: \[

694: \mathfrak{B}:=\left\{\left(\begin{array}{cc}T^1_1 &

695: 0\\T^2_1&T^2_2\end{array}\right)/ T^1_1,T^2_2\in (0,\infty), T^2_1\in

696: \mathbb{C}, T^1_1T^2_2 = 1\right\}.

697: \]

698: We denote by $\mathfrak{b}$ the Lie algebra of $\mathfrak{B}$.  The Iwasawa

699: decomposition $SU(2)^{\mathbb{C}} = SU(2)\cdot \mathfrak{B}$ immediately implies

700: the vector space decomposition $\su(2)^{\mathbb{C}} = \su(2)\oplus

701: \mathfrak{b}$, which leads to the definition of the two projection mappings

702: $(\cdot )_{\su}:\su(2)^{\mathbb{C}} \longrightarrow \su(2)$ and

703: $(\cdot)_\mathfrak{b}:\su(2)^{\mathbb{C}} \longrightarrow \mathfrak{b}$ such

704: that

705: \[

706: \forall \xi \in \su(2)^{\mathbb{C}}, \quad \xi =

707: (\xi)_{\su}+(\xi)_\mathfrak{b}\quad \hbox{with }(\xi)_{\su}\in \su(2)\hbox{ and

708: }(\xi)_\mathfrak{b} \in \mathfrak{b}.

709: \]

710: 

711: \noindent Then we define the following twisted loop algebras

712: \[

713: \Lambda \u(3)_\tau^{\mathbb{C}} := \{S^1\ni\lambda \longmapsto \xi_\lambda\in

714: \u(3)^{\mathbb{C}}/ \forall \lambda\in S^1, \tau(\xi_\lambda) =

715: \xi_{i\lambda}\},

716: \]

717: \[

718: \Lambda_{\mathfrak{b}}^+ \u(3)_\tau^{\mathbb{C}} := \{[\lambda \longmapsto

719: \xi_\lambda] \in \Lambda \u(3)_\tau^{\mathbb{C}}/\forall k\in \mathbb{Z}, k\leq

720: -1\Longrightarrow \widehat{\xi}_k = 0\hbox{ and }\widehat{\xi}_0\in

721: \mathfrak{b}\},

722: \]

723: where we use the Fourier decomposition $\xi_\lambda =\sum_{k\in

724: \mathbb{Z}}\widehat{\xi}_k\lambda^k$.\\

725: 

726: \noindent The decomposition $\su(2)^{\mathbb{C}} = \su(2)\oplus \mathfrak{b}$

727: can be extended to loop algebras, i.e. to the splitting $\Lambda

728: \u(3)_\tau^{\mathbb{C}} = \Lambda \u(3)_\tau\oplus \Lambda_{\mathfrak{b}}^+

729: \u(3)_\tau^{\mathbb{C}}$.  This can be checked by using the Fourier expansion of

730: an element $\xi_\lambda\in \Lambda \u(3)_\tau^{\mathbb{C}}$:

731: \[

732: \sum_{k\in \mathbb{Z}}\widehat{\xi}_k\lambda^k = \left(

733: \sum_{k<0}\widehat{\xi}_k\lambda^k + (\widehat{\xi}_0)_{\su} -

734: \sum_{k>0}\left(\widehat{\xi}_{-k}\right)^\dagger\lambda^k \right) + \left(

735: (\widehat{\xi}_0)_\mathfrak{b} + \sum_{k>0}\left(\widehat{\xi}_k

736: +\left(\widehat{\xi}_{-k}\right)^\dagger \right)\lambda^k \right).

737: \]

738: We will denote the corresponding projection mappings by $(\cdot

739: )_{\Lambda_{\su}}:\Lambda \u(3)_\tau^{\mathbb{C}}\longrightarrow \Lambda

740: \u(3)_\tau$ and $(\cdot)_{\Lambda_\mathfrak{b}^+}:\Lambda

741: \u(3)_\tau^{\mathbb{C}}\longrightarrow \Lambda_{\mathfrak{b}}^+

742: \u(3)_\tau^{\mathbb{C}}$.\\

743: 

744: \noindent We also introduce the following finite dimensional subspaces of

745: $\Lambda \u(3)_\tau$: for any $p\in \mathbb{N}$ we let

746: \[

747: \Lambda^{2+4p} \u(3)_\tau:= \left\{[\lambda \longmapsto \xi_\lambda]\in \Lambda

748: \u(3)_\tau/ \xi_\lambda =\sum_{k=-2-4p}^{2+4p}\widehat{\xi}_k\lambda^k\right\}.

749: \]

750: We can now define a pair of vector fields $X_1,X_2:\Lambda^{2+4p}

751: \u(3)_\tau\longrightarrow \Lambda \u(3)_\tau$ by

752: \begin{equation}\label{2.1.X}

753: X_1(\xi_\lambda):= [\xi_\lambda,(\lambda^{4p}\xi_\lambda)_{\Lambda_{\su}}],

754: \quad X_2(\xi_\lambda):=

755: [\xi_\lambda,(i\lambda^{4p}\xi_\lambda)_{\Lambda_{\su}}].

756: \end{equation}

757: Note that $\lambda^{4p}\xi_\lambda$ belongs to $\Lambda \u(3)_\tau^{\C}$, so

758: that $(\lambda^{4p}\xi_\lambda)_{\Lambda_{\su}}$ is well defined.

759: \begin{lemma}

760: Let $p\in \mathbb{N}$ and $X_1$ and $X_2$ defined by (\ref{2.1.X}).  Then

761: \begin{itemize}

762: \item $\forall \xi_\lambda\in \Lambda^{2+4p} \u(3)_\tau$, $X_1(\xi_\lambda),

763: X_2(\xi_\lambda)\in T_{\xi_\lambda}\Lambda^{2+4p} \u(3)_\tau \simeq

764: \Lambda^{2+4p} \u(3)_\tau$, so that $X_1$ and $X_2$ are tangent vector fields to

765: $\Lambda^{2+4p} \u(3)_\tau$.  \item $|\xi_\lambda|^2$ is preserved by $X_1$ and

766: $X_2$.  Hence the flow of these vector fields are defined for all time \item The

767: Lie bracket of $X_1$ and $X_2$ vanishes:

768: \begin{equation}\label{2.1.[,]}

769: [X_1,X_2] = 0.

770: \end{equation}

771: \end{itemize}

772: \end{lemma}

773: \begin{proof} This result follows by a straightforward adaptation of the

774: analogous results for harmonic maps in \cite{BFPP} (see e.g.\,\cite{G} and

775: \cite{H}).  Note that the proof of (\ref{2.1.[,]}) rests upon the crucial

776: property that $\Lambda \u(3)_\tau$ and $\Lambda_{\mathfrak{b}}^+

777: \u(3)_\tau^{\mathbb{C}}$ are Lie algebras (see e.g.\,\cite{BP},

778: \cite{H}).\end{proof}

779: 

780: \noindent This result allows us to integrate simultaneously $X_1$ and $X_2$.  So

781: for any $\xi_\lambda^0\in \Lambda^{2+4p} \u(3)_\tau$ there exists a unique map

782: $\xi_\lambda:\mathbb{R}^2\longrightarrow \Lambda^{2+4p} \u(3)_\tau$ such that

783: $\xi_\lambda(z_0) = \xi_\lambda^0$ and

784: \begin{equation}\label{2.1.flot}

785: {\partial \xi_\lambda\over \partial x}(x,y) = X_1\left(\xi_\lambda(x,y)\right)

786: \quad \hbox{and} \quad {\partial \xi_\lambda\over \partial y}(x,y) =

787: X_2\left(\xi_\lambda(x,y)\right).

788: \end{equation}

789: Denoting by $z=x+iy\in \C$, the system (\ref{2.1.flot}) can be rewritten

790: \[

791: \begin{array}{ccl}

792: d\xi_\lambda & = & \left[\xi_\lambda,

793: \left(\lambda^{4p}\xi_\lambda\right)_{\Lambda_{\su}}dx +

794: \left(i\lambda^{4p}\xi_\lambda\right)_{\Lambda_{\su}}dy\right]\\

795: & = & \left[\xi_\lambda, \left(\lambda^{4p}\xi_\lambda

796: dz\right)_{\Lambda_{\su}}\right].

797: \end{array}

798: \]

799: Let us denote by $A_\lambda:= (\lambda^{4p}\xi_\lambda dz)_{\Lambda_{\su}}$.

800: Since the system (\ref{2.1.flot}) is overdetermined, $A_\lambda$ should satisfy

801: a compatibility condition.  Indeed one can check that

802: \begin{equation}\label{2.1.curv}

803: dA_\lambda + A_\lambda\wedge A_\lambda = 0.

804: \end{equation}

805: This relation can be proved by a method similar to the proof of (\ref{2.1.[,]})

806: (see \cite{H}).  It implies that there exists a map $F_\lambda:\C\longrightarrow

807: \Lambda U(3)_\tau$ such that

808: \begin{equation}\label{2.1.dF}

809: dF_\lambda = F_\lambda\cdot A_\lambda.

810: \end{equation}

811: Now observe that $\lambda^{4p}\xi_\lambda =

812: \sum_{k=-2}^{8p+2}\hat{\xi}_{k-4p}\lambda^k$ implies

813: \[

814: A_\lambda = \lambda^{-2}\hat{\xi}_{-4p-2}dz + \lambda^{-1}\hat{\xi}_{-4p-1}dz +

815: \left(\hat{\xi}_{-4p}dz\right)_{\su} -

816: \lambda\left(\hat{\xi}_{-4p-1}\right)^\dagger d\bar{z} -

817: \lambda^{2}\left(\hat{\xi}_{-4p-2}\right)^\dagger d\bar{z}.

818: \]

819: We recall that $\hat{\xi}_{-4p-2}\in \u(3)_2^{\C}$ and so has the form

820: $\mathrm{diag}(ia,ia,ib)$.  Moreover we have the following result.

821: \begin{lemma}

822: If $\xi_\lambda\longrightarrow \Lambda^{2+4p} \u(3)_\tau$ and $A_\lambda:=

823: (\lambda^{4p}\xi_\lambda dz)_{\Lambda_{\su}}$ are solutions of $d\xi_\lambda =

824: [\xi_\lambda,A_\lambda]$, then $\hat{\xi}_{-4p-2}$ is constant.

825: \end{lemma}

826: \begin{proof} The relevant term in the Fourier expansion of $d\xi_\lambda =

827: [\xi_\lambda,A_\lambda]$ gives

828: \[

829: \begin{array}{ccl}

830: \displaystyle d\hat{\xi}_{-4p-2} & = & \displaystyle \left[ \hat{\xi}_{-4p-2},

831: \left(\hat{\xi}_{-4p}dz\right)_{\su}\right] + \left[ \hat{\xi}_{-4p-1},

832: \hat{\xi}_{-4p-1}\right]dz + \left[ \hat{\xi}_{-4p},

833: \hat{\xi}_{-4p-2}\right]dz\\

834:  & = & \displaystyle

835:  \left[\left(\hat{\xi}_{-4p}dz\right)_\mathfrak{b},\hat{\xi}_{-4p-2}\right].

836: \end{array}

837: \]

838: But since the coefficients of $\left(\hat{\xi}_{-4p}dz\right)_\mathfrak{b}$ are

839: in $\u(3)_0^{\C}$ and $\hat{\xi}_{-4p-2}$ takes values in $\u(3)_2$ we deduce

840: that $d\hat{\xi}_{-4p-2} = 0$, because $\u(3)_0^{\C}$ and $\u(3)_2^{\C}$

841: commute.  \end{proof}

842: 

843: We deduce from this result that if we choose the initial value $\xi_\lambda^0$

844: of $\xi_\lambda$ to be such that $\hat{\xi}^0_{-4p-2} = \mathrm{diag}(ia,ia,0)$

845: then $\hat{\xi}_{-4p-2}$ is equal to that value for all $(x,y)$.  So in this

846: case the map $F_\lambda$ obtained by integrating $A_\lambda$ satisfies all the

847: requirements of Theorem \ref{1.6.theobis}.  It implies that $F_\lambda$

848: represents a (conjugate family) of Hamiltonian stationary conformal Lagrangian

849: immersion(s).  The category of such $F_\lambda$'s are exactly characterized by

850: the following definition.

851: 

852: \begin{definition}\label{2.1.def-ft}

853: Let $F_\lambda$ be a family of Hamiltonian stationary conformal Lagrangian

854: immersions and let $A_\lambda:= F_\lambda^{-1}\cdot dF_\lambda$.  Then

855: $F_\lambda$ is called a family of {\em finite type solutions} if and only if

856: there exists $p\in \mathbb{N}$ and a map $\xi_\lambda:\mathbb{C}\longrightarrow

857: \Lambda^{2+4p} \u(3)_\tau$ such that $\hat{\xi}_{-4p-2} =

858: \mathrm{diag}(ia,ia,0)$, for some constant $a\in \C$, and

859: \begin{equation}\label{2.1.laxfn}

860: d\xi_\lambda = [\xi_\lambda,A_\lambda]

861: \end{equation}

862: \begin{equation}\label{2.1.laxfn2}

863: (\lambda^{4p}\xi_\lambda dz)_{\Lambda_{\su}} = A_\lambda.

864: \end{equation}

865: \end{definition}

866: We also need the following definition in which we introduce an a priori weaker

867: notion of finite type solution.

868: 

869: \begin{definition}

870: Let $F_\lambda$ be a family of Hamiltonian stationary conformal Lagrangian

871: immersions and let $A_\lambda:= F_\lambda^{-1}\cdot dF_\lambda$.  Then

872: $F_\lambda$ is called a family of {\em quasi-finite type solutions} if and only

873: if it satisfies the same requirements as in definition \ref{2.1.def-ft} excepted

874: that condition (\ref{2.1.laxfn2}) is replaced by

875: \begin{equation}\label{2.1.laxqfn2}

876: \exists B\in \Omega^1\otimes \u(3)^{\C}_0,\quad (\lambda^{4p}\xi_\lambda

877: dz)_{\Lambda_{\su}} = A_\lambda + B.

878: \end{equation}

879: \end{definition}

880: We shall see in Section 3.3 that both definitions are actually equivalent.

881: 

882: \subsection{An alternative description of quasi-finite type solutions}

883: We may as well characterize such finite type solutions in terms of $E_\lambda =

884: F_\lambda\cdot F^{-1}$.  For that purpose we need to introduce the untwisted

885: loop Lie algebra

886: \[

887: \Lambda^+\u(3)^{\C}:= \{S^1\ni\lambda\longmapsto \xi_\lambda\in \u(3)^{\C}/

888: \xi_\lambda = \sum_{k=0}^\infty \widehat{\xi}_k\lambda^k\}

889: \]

890: and observe that any $\xi_\lambda=\sum_{k=-\infty}^\infty

891: \widehat{\xi}_k\lambda^k\in \Lambda \u(3)^{\C}$ can be split as

892: \[

893: \xi_\lambda = \left(\sum_{k=-\infty}^{-1}\widehat{\xi}_k(\lambda^k-1) -

894: (\widehat{\xi}_k)^\dagger (\lambda^{-k}-1)\right) + \left(\sum_{k=0}^\infty

895: \widehat{\xi}_k\lambda^k + \sum_{k=1}^\infty \widehat{\xi}_{-k} +

896: (\widehat{\xi}_{-k})^\dagger (\lambda^{k}-1)\right)

897: \]

898: and hence $\Lambda \u(3)^{\C} = \Omega \u(3)\oplus \Lambda^+\u(3)^{\C}$.  This

899: defines a pair of projection mappings $(\cdot)_{\Omega}:\Lambda \u(3)^{\C}

900: \longrightarrow \Omega \u(3)$ and $(\cdot)_{\Lambda^+}:\Lambda \u(3)^{\C}

901: \longrightarrow \Lambda^+\u(3)^{\C}$.

902: 

903: Now consider a family $F_\lambda$ of quasi-finite type, let $A_\lambda:=

904: F_\lambda^{-1}\cdot dF_\lambda$, $A:= F^{-1}\cdot dF$ (where $F=F_{\lambda=1}$)

905: and $\xi_\lambda$ be a solution of (\ref{2.1.laxfn}).  We let

906: \[

907: \eta_\lambda:= F\cdot \xi_\lambda\cdot F^{-1} = \sum_{k=-2-4p}^{2+4p}F\cdot

908: \widehat{\xi}_k\cdot F^{-1}\lambda^k .

909: \]

910: Then (\ref{2.1.laxfn}) implies by a straightforward computation that

911: \[

912: \begin{array}{ccl}

913: d\eta_\lambda & = & F\cdot (d\xi_\lambda + [A,\xi_\lambda])\cdot F^{-1}\\

914: & = & F\cdot ([\xi_\lambda,A_\lambda] - [\xi_\lambda,A])\cdot F^{-1} =

915: [\eta_\lambda, \Gamma_\lambda],

916: \end{array}

917: \]

918: where $\Gamma_\lambda:= E_\lambda^{-1}\cdot dE_\lambda$.  Now setting

919: $R_\lambda:= \sum_{k=-4p}^{2+4p}F\cdot \widehat{\xi}_k\cdot F^{-1}\lambda^k$, we

920: have

921: \begin{eqnarray*}

922: \displaystyle \left(\lambda^{4p}\eta_\lambda dz\right)_\Omega & = &

923: \displaystyle \left( \lambda^{-2}F\cdot \widehat{\xi}_{-2-4p}\cdot F^{-1}dz +

924: \lambda^{-1}F\cdot \widehat{\xi}_{-1-4p}\cdot F^{-1}dz + \lambda^{4p}R_\lambda

925: dz\right)_\Omega\\

926: & = & \displaystyle (\lambda^{-2}-1)F\cdot \widehat{\xi}_{-2-4p}\cdot F^{-1}dz +

927: (\lambda^{-1}-1)F\cdot \widehat{\xi}_{-1-4p}\cdot F^{-1}dz\\

928: && \displaystyle - (\lambda-1)\left(F\cdot \widehat{\xi}_{-1-4p}\cdot

929: F^{-1}\right)^\dagger d\bar{z} - (\lambda^{2}-1)\left(F\cdot

930: \widehat{\xi}_{-2-4p}\cdot F^{-1}\right)^\dagger d\bar{z}.

931: \end{eqnarray*}

932: But relation (\ref{2.1.laxqfn2}) implies in particular that

933: $\widehat{\xi}_{-2-4p} = A_2'(\partial /\partial z)$ and $\widehat{\xi}_{-1-4p}

934: = A_{-1}(\partial /\partial z)$.  So we deduce that

935: \[

936: \displaystyle \left(\lambda^{4p}\eta_\lambda dz\right)_\Omega = F\cdot

937: \left(A_\lambda - A\right) \cdot F^{-1} = \Gamma_\lambda.

938: \]

939: Hence $E_\lambda$ can be constructed by solving a system analogous to

940: (\ref{2.1.laxfn}), (\ref{2.1.laxfn2}), i.e.

941: \begin{equation}\label{2.2.laxfn}

942: d\eta_\lambda + \left[\Gamma_\lambda,\eta_\lambda\right] = 0 \quad \hbox{and}

943: \quad \Gamma_\lambda = \left(\lambda^{4p}\eta_\lambda dz\right)_\Omega.

944: \end{equation}

945: Conversely a similar computation shows that a solution of (\ref{2.2.laxfn})

946: gives rise to a quasi-finite type family of solutions by an inverse

947: transformation, but we shall prove more in the next section.\\

948: 

949: \noindent Note that system (\ref{2.2.laxfn}) can also be interpreted as a pair

950: of commuting ordinary differential equations in the finite dimensional space

951: $\Lambda^{2+4p}\u(3):= \{S^1\ni\lambda\longmapsto\eta_\lambda \in \u(3)/

952: \eta_\lambda =\sum_{k=-2-4p}^{2+4p}\widehat{\eta}_k\lambda^k\}$.  It is the

953: analogue of the definition of a finite type solution according to \cite{BFPP}.\\

954: 

955: 

956: \subsection{Quasi-finite type solutions are actually finite type}

957: 

958: \noindent We show here the following

959: \begin{theorem}\label{3.3.qtf=tf}

960: For any family $F_\lambda$ of Hamiltonian stationary Lagrangian conformal

961: immersions of quasi-finite type, i.e.\,such that there exists

962: $\xi_\lambda:\Omega\longrightarrow \Lambda^{2+4p}\u(3)_\tau$ which satisfies

963: (\ref{2.1.laxfn}) and (\ref{2.1.laxqfn2}), there exists a gauge transformation

964: $F_\lambda\longmapsto F_\lambda^G:=F_\lambda\cdot G$, where $G\in {\cal

965: C}^\infty(\Omega,U(3)_0)$, such that $F_\lambda^G$ is of finite type.  More

966: precisely, denoting by $A_\lambda^G:=G^{-1}\cdot A_\lambda\cdot G + G^{-1}\cdot

967: dG$ and $\xi_\lambda^G:=G^{-1}\cdot \xi_\lambda\cdot G$, then $d\xi_\lambda^G

968: +\left[A_\lambda^G,\xi_\lambda^G\right] = G^{-1}\cdot \left(d\xi_\lambda

969: +[A_\lambda,\xi_\lambda]\right)\cdot G = 0$ and $(\lambda^{4p}\xi_\lambda^G

970: dz)_{\Lambda_{\su}} =A_\lambda^G$.

971: \end{theorem}

972: \begin{proof} We set $E_\lambda:=F_\lambda\cdot F^{-1}$,

973: $\Gamma_\lambda:=E_\lambda^{-1}\cdot dE_\lambda$ and $\eta_\lambda:= F\cdot

974: \xi_\lambda\cdot F^{-1}$ and will use the results of the previous section.\\

975: 

976: {\em A constant in $\Lambda^{2+4p}\u(3)_\tau$ associated to the quasi-finite

977: type family} --- First (\ref{2.2.laxfn}), which is a reformulation of

978: (\ref{2.1.laxfn}), implies

979: \[

980: d\left(E_\lambda\cdot \eta_\lambda\cdot E_\lambda^{-1}\right) = E_\lambda\cdot

981: \left( d\eta_\lambda +[\Gamma_\lambda,\eta_\lambda]\right)\cdot E_\lambda^{-1}

982: =0.

983: \]

984: Hence

985: \[

986: \eta_\lambda^0:= E_\lambda\cdot \eta_\lambda\cdot E_\lambda^{-1}

987: \]

988: is a constant in $\Lambda \u(3)$.  Moreover

989: \[

990: \eta_\lambda^0 = E_\lambda(z_0)\cdot \eta_\lambda(z_0)\cdot E_\lambda^{-1}(z_0)

991: = \eta_\lambda(z_0) = F(z_0)\cdot \xi_\lambda(z_0)\cdot F^{-1}(z_0)

992: =\xi_\lambda(z_0),

993: \]

994: which proves that $\eta_\lambda^0\in \Lambda^{2+4p}\u(3)_\tau$.\\

995: 

996: {\em An auxiliary map into $\Lambda^+U(3)^{\C}$} --- We let

997: \[

998: \Theta_\lambda:= \left(\lambda^{4p}\eta_\lambda dz\right)_{\Lambda^+} =

999: \lambda^{4p}\eta_\lambda dz - \left(\lambda^{4p}\eta_\lambda dz\right)_\Omega.

1000: \]

1001: Then using (\ref{2.2.laxfn}) we have $\Theta_\lambda = \lambda^{4p}\eta_\lambda

1002: dz - \Gamma_\lambda$ and so

1003: \[

1004: d\Gamma_\lambda + \Gamma_\lambda\wedge \Gamma_\lambda + d\Theta_\lambda -

1005: \Theta_\lambda\wedge \Theta_\lambda = -\lambda^{4p} \left(d\eta_\lambda

1006: +[\Gamma_\lambda,\eta_\lambda]\right) \left({\partial \over \partial

1007: \bar{z}}\right)dz\wedge d\bar{z} =0.

1008: \]

1009: But since $d\Gamma_\lambda + \Gamma_\lambda\wedge \Gamma_\lambda = 0$ this

1010: implies that $d\Theta_\lambda - \Theta_\lambda\wedge \Theta_\lambda = 0$.  Hence

1011: $\exists!  V_\lambda:\Omega\longrightarrow \Lambda^+U(3)^{\C}$ such that

1012: \[

1013: dV_\lambda = \Theta_\lambda\cdot V_\lambda \quad \hbox{and}\quad V_\lambda (z_0)

1014: = 1.

1015: \]

1016: Now, starting from $\lambda^{4p}\eta_\lambda dz =

1017: \Gamma_\lambda+\Theta_\lambda$, we deduce that

1018: \[

1019: \begin{array}{ccl}

1020: \displaystyle \lambda^{4p}\eta_\lambda^0dz & = & \displaystyle E_\lambda\cdot

1021: \Gamma_\lambda\cdot E_\lambda^{-1} + E_\lambda\cdot \Theta_\lambda\cdot

1022: E_\lambda^{-1}\\

1023: & = & \displaystyle dE_\lambda\cdot E_\lambda^{-1} + E_\lambda\cdot

1024: dV_\lambda\cdot V_\lambda^{-1}\cdot E_\lambda^{-1}\\

1025: & = & \displaystyle d\left(E_\lambda\cdot V_\lambda\right)\left(E_\lambda\cdot

1026: V_\lambda\right)^{-1},

1027: \end{array}

1028: \]

1029: which can be integrated into the relation

1030: \[

1031: E_\lambda\cdot V_\lambda = e^{\lambda^{4p}(z-z_0)\eta_\lambda^0}.

1032: \]

1033: 

1034: {\em An Iwasawa decomposition of $e^{\lambda^{4p}(z-z_0)\eta_\lambda^0}$} ---

1035: The latter implies

1036: \[

1037: e^{\lambda^{4p}(z-z_0)\eta_\lambda^0} = F_\lambda\cdot F^{-1}\cdot V_\lambda.

1038: \]

1039: >From this relation and the fact that $\eta_\lambda^0$ and $F_\lambda$ are

1040: twisted we deduce that $W_\lambda:= F^{-1}\cdot V_\lambda$ is twisted.  It is

1041: also a map with values in $\Lambda^+U(3)^{\C}_\tau$.  However it may not be not

1042: in $\Lambda^+_{\mathfrak{B}}U(3)^{\C}_\tau$ in general, because in the

1043: development

1044: \[

1045: F^{-1}\cdot V_\lambda = \widehat{W}_0 + \sum_{k=1}^\infty

1046: \widehat{W}_k\lambda^k,

1047: \]

1048: we are not sure that $\widehat{W}_0$ takes values in $\mathfrak{B}$.  But it

1049: takes values in $U(3)^{\C}_0$, so by using the Iwasawa decomposition

1050: $U(3)^{\C}_0 = U(3)_0\cdot \mathfrak{B}$ we know that $\exists !  G\in U(3)_0$,

1051: $\exists !\widehat{B}_0\in \mathfrak{B}$, $\widehat{W}_0 = G\cdot

1052: \widehat{B}_0$.  Hence

1053: \[

1054: G^{-1}\cdot F^{-1}\cdot V_\lambda = \widehat{B}_0 + \sum_{k=1}^\infty

1055: G^{-1}\cdot \widehat{W}_k\lambda^k

1056: \]

1057: takes values in $\Lambda^+_{\mathfrak{B}}U(3)^{\C}_\tau$.  So the splitting

1058: \[

1059: e^{\lambda^{4p}(z-z_0)\eta_\lambda^0} = \left(F_\lambda\cdot G\right)

1060: \left(G^{-1}\cdot F^{-1}\cdot V_\lambda\right)

1061: \]

1062: exactly reproduces the Iwasawa decomposition $\Lambda U(3)_\tau^{\C} = \Lambda

1063: U(3)_\tau \cdot \Lambda^+_{\mathfrak{B}}U(3)^{\C}_\tau$ proved in \cite{DPW}.\\

1064: 

1065: \noindent {\em Conclusion} --- Let us denote by $F_\lambda^G:= F_\lambda\cdot

1066: G$, $A_\lambda^G:= \left(F_\lambda^G\right)^{-1}\cdot dF_\lambda^G = G^{-1}\cdot

1067: A_\lambda\cdot G + G^{-1}\cdot dG$ and $B_\lambda^G:= G^{-1}\cdot F^{-1}\cdot

1068: V_\lambda$ and let us introduce

1069: \[

1070: \xi_\lambda^G:= \left(F_\lambda^G\right)^{-1}\cdot \eta_\lambda^0\cdot

1071: F_\lambda^G.

1072: \]

1073: (These definitions imply immediately $d\xi_\lambda^G +

1074: [A_\lambda^G,\xi_\lambda^G] = 0$.)  The first main observation is that the

1075: relation $\eta_\lambda^0 = E_\lambda\cdot \eta_\lambda\cdot E_\lambda^{-1} =

1076: F_\lambda\cdot F^{-1}\cdot \eta_\lambda\cdot F\cdot F_\lambda^{-1} =

1077: F_\lambda\cdot \xi_\lambda\cdot F_\lambda^{-1}$ implies

1078: \begin{equation}\label{2.3.tf1}

1079: \xi_\lambda^G = G^{-1}\cdot F_\lambda^{-1}\cdot \eta_\lambda^0\cdot

1080: F_\lambda\cdot G = G^{-1}\cdot \xi_\lambda\cdot G.

1081: \end{equation}

1082: Second, from the relation

1083: \[

1084: \begin{array}{ccl}

1085: \displaystyle \lambda^{4p}\eta^0_\lambda dz & = & \displaystyle

1086: d\left(e^{\lambda^{4p}(z-z_0)\eta_\lambda^0}\right) \cdot

1087: e^{-\lambda^{4p}(z-z_0)\eta_\lambda^0}\\

1088:  & = & \displaystyle d\left(F_\lambda^G\cdot B_\lambda^G\right)\cdot

1089:  \left(F_\lambda^G\cdot B_\lambda^G\right)^{-1}\\

1090: & = & \displaystyle dF_\lambda^G\cdot \left(F_\lambda^G\right)^{-1} +

1091: F_\lambda^G\cdot dB_\lambda^G\cdot \left(B_\lambda^G\right)^{-1}\cdot

1092: \left(F_\lambda^G\right)^{-1},

1093: \end{array}

1094: \]

1095: we deduce that

1096: \[

1097: \lambda^{4p}\xi_\lambda^G dz = \left(F_\lambda^G\right)^{-1} \cdot \left(

1098: \lambda^{4p}\eta_\lambda^0 dz\right)\cdot F_\lambda^G =

1099: \left(F_\lambda^G\right)^{-1}\cdot dF_\lambda^G + dB_\lambda^G\cdot

1100: \left(B_\lambda^G\right)^{-1}.

1101: \]

1102: Hence, since $B_\lambda^G$ takes values in

1103: $\Lambda^+_{\mathfrak{B}}U(3)^{\C}_\tau$,

1104: \begin{equation}\label{2.3.tf2}

1105: A_\lambda^G = \left(F_\lambda^G\right)^{-1}\cdot dF_\lambda^G = \left(

1106: \lambda^{4p}\xi_\lambda^Gdz\right)_{\Lambda_{\su}}.

1107: \end{equation}

1108: And relations (\ref{2.3.tf1}) and (\ref{2.3.tf2}) lead to the conclusion.

1109: \end{proof}

1110: 

1111: 

1112: 

1113: \section{All Hamiltonian stationary Lagrangian tori are of finite type}

1114: The subject of this section is to prove the following:

1115: \begin{theorem}\label{4.MainTheo}

1116: Let $u:\mathbb{C}\longrightarrow \mathbb{C}P^2$ be a doubly periodic Hamiltonian

1117: stationary Lagrangian conformal immersion.  Then $u$ is of finite type.

1118: \end{theorem}

1119: We will actually prove a slightly more general result, since we can replace the

1120: doubly periodicity assumption by the hypothesis that the Maurer--Cartan form of

1121: any {\em Legendrian framing} of $u$ is doubly periodic.  This result of course

1122: implies immediately that Hamiltonian stationary Lagrangian tori are of finite

1123: type, since they always can be covered conformally by the plane.\\

1124: 

1125: \noindent Note also that the study of Hamiltonian stationary Lagrangian tori

1126: splits into exactly two subcases: the {\em minimal} Lagrangian tori and the {\em

1127: non minimal} Hamiltonian stationary Lagrangian ones.  The first case occurs when

1128: the Lagrangian angle function along any Legendrian lift is locally constant, the

1129: second one when this function is harmonic and non constant.  In the case of

1130: minimal Lagrangian surfaces, Theorem \ref{4.MainTheo} is a special case of the

1131: result in \cite{BFPP}, since in this case $u$ is a harmonic map into

1132: $\mathbb{C}P^2$, as discussed in \cite{McI1}, \cite{McI2}, \cite{McI3} and

1133: \cite{J}.  The non minimal case however is not covered by the theory in

1134: \cite{BFPP} and is the subject of this section.  \\

1135: 

1136: \noindent Let $F:\mathbb{C}\longrightarrow U(3)$ be a Legendrian framing of $u$,

1137: $A:=F^{-1}\cdot dF$ its Maurer--Cartan form and $A_\lambda$ the family of

1138: deformations of $A$ as defined by (\ref{1.6.Alambda}).  The first basic

1139: observation is that $A_2\left({\partial \over \partial z}\right)$ is holomorphic

1140: and doubly periodic on $\C$, hence constant.  Thus two cases occur: either

1141: $A_2\left({\partial \over \partial z}\right)=0$, which corresponds to the

1142: minimal case that we exclude here, or $A_2\left({\partial \over \partial

1143: z}\right)$ is a constant different from 0, the case that we consider next.\\

1144: 

1145: \noindent In order to show Theorem \ref{4.MainTheo} we need to prove that there

1146: exists some $p\in \mathbb{N}$ and a map $\xi_\lambda:\mathbb{C}\longrightarrow

1147: \Lambda^{2+4p}\u(3)_\tau$ such that $d\xi_\lambda = [\xi_\lambda,A_\lambda]$ and

1148: $A_\lambda = (\lambda^{4p}\xi_\lambda dz)_{\Lambda_{\su}}$.  But thanks to

1149: Theorem \ref{3.3.qtf=tf} it will enough to prove that $A_\lambda -

1150: (\lambda^{4p}\xi_\lambda dz)_{\Lambda_{\su}}$ is a 1-form with coefficients in

1151: $\u(3)_0$.  Our proof here follows a strategy inspired from \cite{BFPP}: a first

1152: step consists in building a formal series $Y_\lambda = \sum_{k=-2}^\infty

1153: \widehat{Y}_k\lambda^k$ which is a solution of $dY_\lambda =

1154: [Y_\lambda,A_\lambda]$.  Such a series is called a {\em formal Killing field}.

1155: We will also require $Y_\lambda$ to be {\em quasi-adapted}, i.e.\,is such that

1156: \begin{equation}\label{4.adapted}

1157: (Y_\lambda dz)_{\Lambda_{\su}} = A_\lambda + B,\quad \hbox{where the

1158: coefficients of }B\hbox{ are in } \u(3)_0.

1159: \end{equation}

1160: This is achieved through a recursion procedure.\\

1161: 

1162: \noindent In a second step we will show that the coefficients of $Y_\lambda$

1163: form a countable collection of doubly periodic functions satisfying an elliptic

1164: PDE and hence, by using a compactness argument, we conclude that they are

1165: contained in a finite dimensional space.  Then we deduce the existence of

1166: $\xi_\lambda$ using linear algebra.

1167: 

1168: 

1169: 

1170: 

1171: \subsection{Construction of an adapted formal Killing field}

1172: 

1173: We first introduce some notations.  We denote by

1174: \[

1175: \pi_0^\perp := \left(\begin{array}{ccc} 1 &&\\ & 1& \\ && 0 \end{array}\right)

1176: \]

1177: and $a := {1\over 2}{\partial \beta\over \partial z}$ (here a constant different

1178: from 0).  Then $A_2' = ia\pi_0^\perp dz$.  We will also set

1179: $X:=A_{-1}\left({\partial \over \partial z}\right)$ and $C:=A_0\left({\partial

1180: \over \partial z}\right)$, so that

1181: \[

1182: A_\lambda = \lambda^{-2}ia\pi_0^\perp dz + \lambda^{-1}Xdz + Cdz -C^\dagger

1183: d\bar{z} -\lambda X^\dagger d\bar{z} + \lambda^2i\overline{a}\pi_0^\perp

1184: d\bar{z}.

1185: \]

1186: We also introduce the linear map $\ad \pi_0^\perp : \u(3)^{\C}\longrightarrow

1187: \u(3)^{\C}$, acting by $\xi\longmapsto [\pi_0^\perp ,\xi]$\footnote{Actually the

1188: map $\xi\longmapsto i[\pi_0^\perp ,\xi]$ corresponds to the complex structure on

1189: the Legendrian distribution.}.  We observe that $\pi_0^\perp $ commutes with the

1190: elements in $\u(3)_0^{\C}$ and $\u(3)_2^{\C}$.  Moreover

1191: \[

1192: \forall a,b\in \C, \; \left[\pi_0^\perp ,\left(\begin{array}{ccc}&& a \\ && b \\

1193: \mp ib &\pm ia \end{array}\right)\right] = \left(\begin{array}{ccc} && a \\ && b

1194: \\ \pm ib & \mp ia& \end{array}\right),

1195: \]

1196: that is $\ad \pi_0^\perp $ maps $\u(3)_{\mp 1}^{\C}$ to $\u(3)_{\pm 1}^{\C}$.

1197: >From that we deduce that $V:= \Ker \ad \pi_0^\perp $ coincides with

1198: $\u(3)_0^{\C}\oplus \u(3)_2^{\C}$ and $V^\perp:= \Im \ad \pi_0^\perp $ coincides

1199: with $\u(3)_{-1}^{\C}\oplus \u(3)_1^{\C}$ (note that $V^\perp$ is actually the

1200: orthogonal subspace to $V$ in $\u(3)^{\C}$).  In our construction we will use

1201: extensively the following properties:

1202: \begin{itemize}

1203: \item the map $\ad \pi_0^\perp \at{V^\perp \to V^\perp}$ is a vector space

1204: isomorphism (it is actually a involution on $V^\perp$), \item the inclusions $VV

1205: \subset V$, $VV^\perp \subset V^\perp$, $V^\perp V \subset V^\perp$ and $V^\perp

1206: V^\perp \subset V$.  These properties can be checked by a direct computation

1207: using the fact that matrices in $V$ are diagonal by blocks and the matrices in

1208: $V^\perp$ are off-diagonal by blocks.  (The three first properties can also be

1209: deduced from the definition of $V$ and $V^\perp$ and the fact that $\ad$ is a

1210: derivation).

1211: \end{itemize}

1212: We look for a formal Killing field $Y_\lambda$, i.e.\,a solution of the equation

1213: \begin{equation} \label{eq:champ_formel}

1214: dY_\lambda = [Y_\lambda,A_\lambda],

1215: \end{equation}

1216: of the form $Y_\lambda = (1+W_\lambda)^{-1}\lambda^{-2}ia\pi_0^\perp

1217: (1+W_\lambda)$, where $W_\lambda = \sum_{k=0}^\infty \widehat{W}_k\lambda^k$ as

1218: in \cite{BFPP}.  In order to have a well-posed problem (and in particular to

1219: guarantee the existence of an unique solution of this type) we assume that

1220: $W_\lambda$ takes values in $V^\perp$.  We start by evaluating

1221: (\ref{eq:champ_formel}) along $\partial/\partial z$.  It gives, after

1222: conjugation by $1+W_\lambda$:

1223: \begin{multline} \label{eq:W}

1224: \lambda^{-2}\frac{\partial a}{\partial z} \pi_0^\perp + \lambda^{-2}

1225: a\big[\pi_0^\perp , \\

1226:  \frac{\partial W_\lambda}{\partial z}(1+W_\lambda)^{-1}

1227:  -(1+W_\lambda)(\lambda^{-2}ia\pi_0^\perp

1228:  +\lambda^{-1}X+C)(1+W_\lambda)^{-1}\big] =0

1229: \end{multline}

1230: Here the fact that $a$ is a constant leads to an immediate simplification,

1231: namely that the bracket in the left hand side of (\ref{eq:W}) is 0.  Thus

1232: equation (\ref{eq:W}) implies that $\frac{\partial W_\lambda}{\partial z}

1233: (1+W_\lambda)^{-1}-(1+W_\lambda)(\lambda^{-2}ia\pi_0^\perp +\lambda^{-1}X+C)

1234: (1+W_\lambda)^{-1}$ lies in $V$, hence there exists a map

1235: $\varphi_\lambda:\C\longrightarrow V$ such that

1236: \[

1237: \pa{W_\lambda}{z}(1+W_\lambda)^{-1} -(1+W_\lambda)(\lambda^{-2}ia\pi_0^\perp

1238: +\lambda^{-1}X+C)(1+W_\lambda)^{-1} = \varphi_\lambda

1239: \]

1240: or

1241: \[

1242: \pa{W_\lambda}{z}-(1+W_\lambda)(\lambda^{-2}ia\pi_0^\perp +\lambda^{-1}X+C) =

1243: \varphi_\lambda(1+W_\lambda),

1244: \]

1245: which can be projected according to the splitting $V \oplus V^\perp$ as

1246: \[

1247: \left\{ \begin{array}{l}\displaystyle \lambda^{-2}ia\pi_0^\perp

1248: +\lambda^{-1}W_\lambda X+C = -\varphi_\lambda \in V

1249: \\

1250: \displaystyle \pa{W_\lambda}{z} - \lambda^{-2}ia W_\lambda \pi_0^\perp -

1251: \lambda^{-1}X - W_\lambda C = \varphi_\lambda W_\lambda \in V^\perp.

1252: \end{array} \right.

1253: \]

1254: Substituting $\varphi_\lambda$,

1255: \[

1256: \pa{W_\lambda}{z} - \lambda^{-2}ia W_\lambda \pi_0^\perp - \lambda^{-1}X-

1257: W_\lambda C + \lambda^{-2}ia\pi_0^\perp W_\lambda+\lambda^{-1}W_\lambda X

1258: W_\lambda +C W_\lambda = 0

1259: \]

1260: or

1261: \[

1262: [ia\pi_0^\perp ,W_\lambda] + \lambda(W_\lambda X W_\lambda-X) +

1263: \lambda^2[C,W_\lambda] + \lambda^2 \pa{W_\lambda}{z} = 0

1264: \]

1265: or

1266: \begin{multline*}

1267: ia\sum_{n\geq 0} [\pi_0^\perp ,\widehat{W}_n]\lambda^n + \sum_{n\geq

1268: 1}\left(\sum_{k=0}^{n-1} \widehat{W}_k X \widehat{W}_{n-1-k}\right) \lambda^n -

1269: \lambda X \\

1270: + \sum_{n\geq 2} \left([C,\widehat{W}_{n-2}] + \pa{\widehat{W}_{n-2}}{z} \right)

1271: \lambda^n = 0.

1272: \end{multline*}

1273: Hence

1274: \[

1275: \left\{ \begin{array}{ll} \displaystyle n=0, & \displaystyle ia[\pi_0^\perp

1276: ,\widehat{W}_0] = 0

1277: \\

1278: \displaystyle n=1, & \displaystyle ia[\pi_0^\perp ,\widehat{W}_1]+\widehat{W}_0

1279: X \widehat{W}_0 - X = 0

1280: \\

1281: \displaystyle n \geq 2, & \displaystyle ia[\pi_0^\perp

1282: ,\widehat{W}_n]+\sum_{k=0}^{n-1} \widehat{W}_k X \widehat{W}_{n-1-k} +

1283: [C,\widehat{W}_{n-2}] + \pa{\widehat{W}_{n-2}}{z} =0 \end{array} \right.

1284: \]

1285: and thus

1286: \[

1287: \left\{ \begin{array}{l} \widehat{W}_0 = 0

1288: \\

1289: \widehat{W}_1 = -ia^{-1}[\pi_0^\perp ,X]

1290: \\

1291: \displaystyle \widehat{W}_n = ia^{-1}\left[\pi_0^\perp ,\sum_{k=0}^{n-1}

1292: \widehat{W}_k X \widehat{W}_{n-1-k} + [C,\widehat{W}_{n-2}] +

1293: \pa{\widehat{W}_{n-2}}{z}\right] \end{array} \right.

1294: \]

1295: We observe that the formal Killing field is {\em quasi-adapted} in the sense

1296: that the two first coefficients are the right ones:

1297: \begin{eqnarray*}

1298: Y_\lambda &=& ia\lambda^{-2} (1+W_\lambda)^{-1}\pi_0^\perp (1+W_\lambda) = ia

1299: \lambda^{-2} \left( \pi_0^\perp - \lambda [\widehat{W}_1,\pi_0^\perp ] +

1300: \O(\lambda^2) \right) \\

1301: &=& \lambda^{-2} i a \pi_0^\perp + \lambda^{-1} X + \O(1).

1302: \end{eqnarray*}

1303: Another pleasant property is that this formal field is automatically twisted (as

1304: in the case of $\C^2$, see \cite{HR1}).  Indeed using the fact that $\tau$ is an

1305: automorphism for the product of matrices as well as for the Lie bracket (and so

1306: $[\u(3)^{\C}_a,\u(3)^{\C}_b] \subset \u(3)^{\C}_{a+b}$ and $\u(3)^{\C}_a

1307: \u(3)^{\C}_b \subset \u(3)^{\C}_{a+b}$), we obtain that

1308: \[

1309: \tau(Y_\lambda) = \tau(1+W_\lambda)^{-1} (-\lambda^2)ia\pi_0^\perp

1310: \tau(1+W_\lambda).

1311: \]

1312: Thus it is enough to show that $1+W_\lambda$ is twisted, i.e.\,$\tau(W_\lambda)

1313: = W_{i\lambda}$.  In terms of the Fourier decomposition of $W_\lambda$ this is

1314: equivalent to proving that $\widehat{W}_n$ belongs to $\u(3)^{\C}_n$.  Let us

1315: prove it by recursion.  We already know that $\widehat{W}_1 =

1316: -ia^{-1}[\pi_0^\perp ,X]$ is in $\u(3)^{\C}_1$.  Assume that the result is true

1317: up to $n-1$,then

1318: \[

1319: \sum_{p=0}^{n-1} \widehat{W}_p X \widehat{W}_{n-1-p} + [C,\widehat{W}_{n-2}] +

1320: \pa{\widehat{W}_{n-2}}{z}

1321: \]

1322: belongs to $\u(3)^{\C}_{n-2}$.  And since $\pi_0^\perp \in \u(3)^{\C}_2$,

1323: $\widehat{W}_n$ is in $\gC_n$.

1324: 

1325: We now prove that (\ref{eq:champ_formel}) is also true along $\partial/\partial

1326: \bar{z}$.  We follow here the same kind of arguments as in \cite{BFPP} slightly

1327: simplified\footnote{essentially the simplifications occur because the

1328: semi-simple term $B$ of \cite{BFPP} is here $ia\pi_0^\perp $ which is constant

1329: for $d$, so we do not need to introduce a flat connection.}.  We want to show

1330: that

1331: \[

1332: \frac{\partial Y_\lambda}{\partial \bar{z}} +

1333: \left[A_\lambda\left(\frac{\partial }{\partial \bar{z}}\right), Y_\lambda\right]

1334: = 0

1335: \]

1336: and for that purpose we rather consider the conjugate of the left hand side

1337: $\zeta_\lambda = (1+W_\lambda) (\frac{\partial Y_\lambda}{\partial \bar{z}} +

1338: [A_\lambda\left(\frac{\partial }{\partial \bar{z}}\right),

1339: Y_\lambda])(1+W_\lambda)^{-1}$.  We then prove two facts

1340: \begin{itemize}

1341: \item $\zeta_\lambda$ takes its values in $V^\perp$: this follows from the

1342: identity

1343: \[

1344: \zeta_\lambda = \left[\lambda^{-2}ia\pi_0^\perp , \frac{\partial

1345: W_\lambda}{\partial \bar{z}} (1+W_\lambda)^{-1} -

1346: (1+W_\lambda)A_\lambda\left(\frac{\partial }{\partial \bar{z}}\right)

1347: (1+W_\lambda)^{-1}\right].

1348: \]

1349: Note that since $\zeta_\lambda$ is twisted the fact that $\zeta_\lambda\in

1350: V^\perp$ implies also that $\zeta_\lambda$ is an odd function of $\lambda$

1351: and so that

1352: \begin{equation}\label{3.1.oddzeta}

1353: \zeta_\lambda = \sum_{k=0}^\infty \widehat{\zeta}_{2k-1}\lambda^{2k-1}.

1354: \end{equation}

1355: 

1356: \item the relation

1357: \begin{equation}\label{3.1.dzeta}

1358: {\partial \zeta_\lambda\over \partial z} = [\varphi_\lambda,\zeta_\lambda].

1359: \end{equation}

1360: Indeed $d+\ad A_\lambda$ is a flat connection and in particular

1361: $\frac{\partial }{\partial z} +\ad A_\lambda(\frac{\partial }{\partial z})$

1362: commutes with $\frac{\partial }{\partial \bar{z}} +A_\lambda(\frac{\partial

1363: }{\partial \bar{z}})$.  Hence

1364: \[

1365: \left( \frac{\partial }{\partial z}+\ad A_\lambda\left(\frac{\partial

1366: }{\partial z}\right) \right) \left( \frac{\partial }{\partial \bar{z}} +\ad

1367: A_\lambda\left(\frac{\partial }{\partial \bar{z}}\right) \right) Y_\lambda

1368: = 0

1369: \]

1370: i.e.

1371: \[

1372: \left( \frac{\partial }{\partial z}+\ad A_\lambda\left(\frac{\partial

1373: }{\partial z}\right) \right)

1374: \left((1+W_\lambda)^{-1}\zeta_\lambda(1+W_\lambda)\right) = 0

1375: \]

1376: Thus (\ref{3.1.dzeta}) follows from a computation which uses

1377: $\varphi_\lambda = {\partial W_\lambda\over\partial z}(1+W_\lambda)^{-1} -

1378: (1+W_\lambda) A_\lambda\left(\frac{\partial }{\partial

1379: z}\right)(1+W_\lambda)^{-1}$.

1380: 

1381: \end{itemize}

1382: Now assume by contradiction that $\zeta_\lambda\neq 0$: in view of

1383: (\ref{3.1.oddzeta}) there exists an integer $k\in \mathbb{N}$ such that

1384: $\widehat{\zeta}_{2k-1}\neq 0$ and $\widehat{\zeta}_{2k-3} = 0$.  By

1385: substituting the Fourier decompositions in (\ref{3.1.dzeta}) and observing that

1386: the Fourier series expansion of $\varphi_\lambda$ starts by $\lambda^{-2}

1387: ia\pi_0^\perp$, we deduce that $0 = \partial \widehat{\zeta}_{2k-3}/\partial z =

1388: [ia\pi_0^\perp , \widehat{\zeta}_{2k-1}]$; but $\ad \pi_0^\perp $ is invertible

1389: on $V^\perp$ and hence $\widehat{\zeta}_{2k-1} = 0$.  So we get a contradiction.

1390: 

1391: \subsection{Polynomial Killing fields}

1392: We now deduce the existence of a non-trivial polynomial Killing field.

1393: 

1394: A first easy consequence of the results of the previous section is that, for all

1395: $n\in \N$ and for all polynomial of the form $P(\lambda)=a_n\lambda ^{-4n} +

1396: a_{n-1}\lambda ^{-4(n-1)} + \cdots + a_0$, where $a_0,a_1,\cdots ,a_n\in \C$ and

1397: $a_n\neq 0$, then $Z_\lambda:=P(\lambda)Y_\lambda$ is again formal Killing

1398: field.  Moreover it is quasi-adapted (modulo the multiplicative factor

1399: $a_n\lambda^{-4n}$), i.e.\,the lower degree terms are $a_n\lambda^{-4n}\left(ia

1400: \lambda^{-2} \pi_0^\perp + \lambda^{-1} X + O(\lambda ^0)\right)$).  Let us

1401: consider

1402: \[

1403: Z_\leq := \sum_{k=-2-4n}^0\widehat{Z}_k\lambda^k,\quad \hbox{and}\quad Z_>:=

1404: \sum_{k=1}^\infty \widehat{Z}_k\lambda^k,

1405: \]

1406: so that $Z_\lambda= Z_\leq + Z_>$.  We study

1407: \begin{equation}\label{dZ-}

1408: R_\lambda:= dZ_\leq +[A_\lambda,Z_\leq].

1409: \end{equation}

1410: We first remark that $R_\lambda$ is necessarily of the form

1411: $R_\lambda=\sum_{k=-4-4n}^2\widehat{R}_k\lambda^k$.  But because of $dZ_\lambda

1412: + [A_\lambda,Z_\lambda]= 0$, we also have

1413: \begin{equation}\label{dZ+}

1414: R_\lambda = -dZ_> - [A_\lambda,Z_>],

1415: \end{equation}

1416: which implies $R_\lambda=\sum_{k=-1}^\infty \widehat{R}_k\lambda^k$.  Hence

1417: finally

1418: \[

1419: R_\lambda = \lambda^{-1}\widehat{R}_{-1} + \widehat{R}_{0} +

1420: \lambda^{1}\widehat{R}_{1} + \lambda^{2}\widehat{R}_{2}.

1421: \]

1422: Each term $\widehat{R}_{k}$ can be evaluated through two different ways: by

1423: using (\ref{dZ-}) or (\ref{dZ+}).  From (\ref{dZ-}) we obtain

1424: \begin{equation}\label{dZ-z}

1425: \left\{ \begin{array}{l} \widehat{R}_{-1}(\partial _z) = \partial _z

1426: \widehat{Z}_{-1} + [A_{-1}(\partial _z), \widehat{Z}_0] + [A_{0}(\partial _z),

1427: \widehat{Z}_{-1}] \\

1428: \widehat{R}_{0}(\partial _z) = \partial _z \widehat{Z}_{0} + [A_{0}(\partial

1429: _z), \widehat{Z}_0] \\

1430: \widehat{R}_{1}(\partial _z) = 0\\

1431: \widehat{R}_{2}(\partial _z) = 0 \end{array}\right.

1432: \end{equation}

1433: and

1434: \begin{equation}\label{dZ-zbar}

1435: \left\{ \begin{array}{l} \widehat{R}_{-1}(\partial _{\bar{z}}) = \partial

1436: _{\bar{z}} \widehat{Z}_{-1} + [A_{0}(\partial_{\bar{z}}), \widehat{Z}_{-1}] +

1437: [A_{1}(\partial_{\bar{z}}), \widehat{Z}_{-2}] + [A_{2}''(\partial _{\bar{z}}),

1438: \widehat{Z}_{-3}] \\

1439: \widehat{R}_{0}(\partial _{\bar{z}}) = \partial_{\bar{z}} \widehat{Z}_{0} +

1440: [A_{0}(\partial _{\bar{z}}), \widehat{Z}_{0}] + [A_{1}(\partial _{\bar{z}}),

1441: \widehat{Z}_{-1}] + [A_{2}''(\partial _{\bar{z}}), \widehat{Z}_{-2}] \\

1442: \widehat{R}_{1}(\partial _{\bar{z}}) = [A_{1}(\partial _{\bar{z}}),

1443: \widehat{Z}_{0}] + [A_{2}''(\partial _{\bar{z}}), \widehat{Z}_{-1}] \\

1444: \widehat{R}_{2}(\partial _{\bar{z}}) = [A_{2}''(\partial _{\bar{z}}),

1445: \widehat{Z}_{0}].  \end{array}\right.

1446: \end{equation}

1447: >From (\ref{dZ+}) we get

1448: \begin{equation}\label{dZ+z}

1449: \left\{ \begin{array}{l} \widehat{R}_{-1}(\partial _z) = - [A_{2}'(\partial _z),

1450: \widehat{Z}_{1}] \\

1451: \widehat{R}_{0}(\partial _z) = - [A_{2}'(\partial _z), \widehat{Z}_{2}] -

1452: [A_{-1}(\partial _z), \widehat{Z}_{1}] \\

1453: \widehat{R}_{1}(\partial _z) = - \partial _z\widehat{Z}_{1} - [A_{2}'(\partial

1454: _z), \widehat{Z}_{3}] - [A_{-1}(\partial _z), \widehat{Z}_{2}] - [A_{0}(\partial

1455: _z), \widehat{Z}_{1}] \\

1456: \widehat{R}_{2}(\partial _z) = - \partial _z\widehat{Z}_{2} - [A_{2}'(\partial

1457: _z), \widehat{Z}_{4}] - [A_{-1}(\partial _z), \widehat{Z}_{3}] - [A_{0}(\partial

1458: _z), \widehat{Z}_{2}] \end{array}\right.

1459: \end{equation}

1460: and

1461: \begin{equation}\label{dZ+zbar}

1462: \left\{ \begin{array}{l} \widehat{R}_{-1}(\partial _{\bar{z}}) = 0 \\

1463: \widehat{R}_{0}(\partial _{\bar{z}}) = 0 \\

1464: \widehat{R}_{1}(\partial _{\bar{z}}) = - \partial _{\bar{z}} \widehat{Z}_{1} -

1465: [A_{0}(\partial _{\bar{z}}), \widehat{Z}_{1}] \\

1466: \widehat{R}_{2}(\partial _{\bar{z}}) = - \partial _{\bar{z}} \widehat{Z}_{2} -

1467: [A_{0}(\partial _{\bar{z}}), \widehat{Z}_{2}] - [A_{1}(\partial _{\bar{z}}),

1468: \widehat{Z}_{1}] .  \end{array}\right.

1469: \end{equation}

1470: Thus in order to obtain an expression of $R_\lambda$ which does depend only on

1471: $\widehat{Z}_{-1}$ and $\widehat{Z}_{0}$, we exploit (\ref{dZ-z}) and the two

1472: last equations in (\ref{dZ-zbar}).  But instead of using the two first equations

1473: of (\ref{dZ-zbar}) we take the two first ones of (\ref{dZ+zbar}).  This gives us

1474: \begin{equation}\label{Rz}

1475: R_\lambda(\partial _z) = \lambda^{-1}\left( \partial _z\widehat{Z}_{-1} +

1476: [A_{-1}(\partial _z),\widehat{Z}_0] + [A_{0}(\partial

1477: _z),\widehat{Z}_{-1}]\right) + \left( \partial _z\widehat{Z}_{0} +

1478: [A_{0}(\partial _z),\widehat{Z}_0] \right),

1479: \end{equation}

1480: \begin{equation}\label{Rzbar}

1481: R_\lambda(\partial _{\bar{z}}) = \lambda \left( [A_{1}(\partial

1482: _{\bar{z}}),\widehat{Z}_0] + [A_{2}''(\partial

1483: _{\bar{z}}),\widehat{Z}_{-1}]\right) + \lambda^2[A_{2}''(\partial

1484: _{\bar{z}}),\widehat{Z}_{0}].

1485: \end{equation}

1486: These relations will imply that $\widehat{Z}_{-1}$ and $\widehat{Z}_{0}$ satisfy

1487: a second order elliptic equation.  In order to prove that we need to establish

1488: another relation between $R_\lambda(\partial _z)$ and $R_\lambda(\partial

1489: _{\bar{z}})$.  For that purpose recall that $dA_\lambda + A_\lambda\wedge

1490: A_\lambda = 0$, which means that the connection $d+\ad A_\lambda$ has a

1491: vanishing curvature.  In particular

1492: \[

1493: 0= \left( d+\ad A_\lambda\right) \circ \left( d+\ad A_\lambda\right)Z_\leq =

1494: dR_\lambda + [A_\lambda \wedge R_\lambda].

1495: \]

1496: This implies

1497: \begin{equation}\label{fermeture}

1498: {\partial R_\lambda(\partial _z)\over \partial \bar{z}} - {\partial

1499: R_\lambda(\partial _{\bar{z}}) \over \partial z} = [A_\lambda(\partial

1500: _z),R_\lambda(\partial _{\bar{z}})] - [A_\lambda(\partial

1501: _{\bar{z}}),R_\lambda(\partial _z)].

1502: \end{equation}

1503: A substitution of (\ref{Rz}) and (\ref{Rzbar}) in (\ref{fermeture}) gives a

1504: system of linear elliptic equations on $\widehat{Z}_{-1}$ and $\widehat{Z}_{0}$.

1505: Since the space of solutions to this system which are periodic is finite

1506: dimensional, it turns out that $\widehat{Z}_{-1}$ and $\widehat{Z}_{0}$ belong

1507: to a finite dimensional vector space.  Hence relations (\ref{Rz}) and

1508: (\ref{Rzbar}) force $R_\lambda(\partial _z)$ and $R_\lambda(\partial

1509: _{\bar{z}})$ to stay in a finite dimensional vector space.

1510: 

1511: We can conclude: let us consider

1512: \[

1513: {\cal R}:= \{ R_\lambda / R_\lambda(\partial _z), R_\lambda(\partial

1514: _{\bar{z}})\hbox{ are given by (\ref{Rz}) and (\ref{Rzbar}) and satisfy

1515: (\ref{fermeture})}\}.

1516: \]

1517: It is a complex finite dimensional vector space.  Let us also denote by ${\cal

1518: P}_n:=\{P(\lambda)=a_n\lambda ^{-4n} + a_{n-1}\lambda ^{-4(n-1)} + \cdots + a_0/

1519: (a_0,\cdots , a_n)\in \C ^{n+1}\}$ and ${\cal P}_\infty :=\cup_{n\in \N}{\cal

1520: P}_n$.

1521: 

1522: The linear map ${\cal P}_\infty \ni P(\lambda)\longmapsto dZ_\leq

1523: +[A_\lambda,Z_\leq]$, where $Z_\leq = \left( P(\lambda)Y_\lambda\right)_\leq$

1524: takes values in ${\cal R}$ and so has a finite rank, say $n$.  Then since

1525: $\hbox{dim}_\mathbb{C} {\cal P}_n=n+1$, the map ${\cal P}_n \ni

1526: P(\lambda)\longmapsto dZ_\leq +[A_\lambda,Z_\leq]$ has a non trivial kernel: let

1527: $P(\lambda)=\sum_{k=0}^na_k\lambda^{-4k}$ be a non trivial polynomial in this

1528: kernel.  Let $4p$ be the degree of $P$ in $\lambda^{-1}$, i.e.\,such that

1529: $P(\lambda)=\sum_{k=0}^pa_k\lambda^{-4k}$ and $a_p\neq 0$.  Without loss of

1530: generality we can assume that $a_p = 1$.  Then $\xi_\lambda:=

1531: \left(P(\lambda)Y_\lambda\right)_\leq -

1532: \left(P(\lambda)Y_\lambda\right)_\leq^\dagger$ is a solution of

1533: (\ref{2.1.laxfn}) and (\ref{2.1.laxqfn2}).

1534: 

1535: 

1536: \section{Homogeneous tori in $\mathbb{C} P^2$}

1537: 

1538: We describe here the simplest examples of Hamiltonian stationary Lagrangian tori

1539: in $\mathbb{C}P^2$: the {{\em homogeneous}} Hamiltonian stationary Lagrangian

1540: tori, i.e. immersions $u$ of $S^1 \times S^1$ into $\mathbb{C} P^2$ such that $u

1541: ( x + t, y ) = e^{t A} u ( x, y )$ and $u ( x, y + t ) = e^{t B} u ( x, y )$ for

1542: some skew-Hermitian matrices $A$ and $B$.  Notice that $A$ and $B$ are only

1543: defined up to addition with a multiple of $i \mathit{Id}$.  The simplest example

1544: is the Clifford torus, namely the image by the Hopf map $\pi$ of the product

1545: torus $\{ z = ( z^1, z^2, z^3 ) ; | z^1 | = | z^2 | = | z^3 | = 1 / \sqrt{3}

1546: \}$.  This torus is minimal.  The main result states that all homogeneous

1547: Hamiltonian stationary Lagrangian tori are similar to the Clifford torus.

1548: 

1549: \begin{theorem} \label{homogeneous}

1550: Any homogeneous Hamiltonian stationary Lagrangian torus in $\mathbb{C} P^2$ is

1551: the image by the Hopf map of some Cartesian product $r_1 S^1 \times r_2 S^1

1552: \times r_3 S^1 = \{ z = ( z^1, z^2, z^3 ) ; | z^1 | = r_1, | z^2 | = r_2, | z^3

1553: | = r_3 \}$ where $r_1^2 + r_2^2 + r_3^2 = 1$, up to $U(3)$ congruence.

1554: Moreover, the torus is special Lagrangian if and only if $r_1 = r_2 = r_3 =

1555: \sqrt{3}$.

1556: \end{theorem}

1557: %

1558: \begin{proof}

1559: Let us first see why $\pi ( T )$ is a Hamiltonian stationary Lagrangian torus in

1560: $\mathbb{C} P^2$, where $T = r_1 S^1 \times r_2 S^1 \times r_3 S^1$.  Indeed it

1561: suffices to show that $\pi ( T )$ admits a Legendrian preimage.  Let \[ f ( x, y

1562: ) : = \left( r_1 e^{i (( 1 - r_1^2 ) x - r_2^2 y )}, r_2 e^{i ( - r_1^2 x + ( 1

1563: - r_2^2 ) y )}, r_3 e^{i ( - r_1^2 x - r_2^2 y )} \right) .

1564: \]

1565: Then the orbit under the Hopf action of the image of $f$ is exactly the 3-torus

1566: $T$ above and $\pi \circ f$ is doubly periodic with periods $(2\pi,0)$ and

1567: $(0,2\pi)$.  Note that this immersion is not conformal but there exists an

1568: orthonormal Hermitian moving frame $(e_1,e_2)$ such that ${\partial f\over

1569: \partial x} = r_1\sqrt{1-r_1^2}\, e_1$ and ${\partial f\over \partial y} =

1570: \frac{r_2}{\sqrt{1-r_1^2}} (r_3 e_2 - r_1r_2 e_1)$.  And it is easy to check

1571: that $f$ is Legendrian (and flat).  Its Lagrangian angle function is

1572: \[

1573: \beta ( x, y ) = x ( 1 - 3 r_1^2 ) + y ( 1 - 3 r_2^2 ) + \pi

1574: \]

1575: and since the metric is flat, $\beta$ is clearly harmonic, and constant if and

1576: only if $r_1 = r_2 = r_3 = 1 / \sqrt{3}$.  Notice that many of these tori do not

1577: lift up to $S^5$ as Legendrian tori (they do not close up).  Indeed the Maslov

1578: class is not always an integer: for the implicit homology basis, $t \longmapsto

1579: ( 2 \pi t, 0 )$ and $t \longmapsto ( 0, 2 \pi t )$, it is $( 1 - 3 r_1^2, 1 - 3

1580: r_2^2 )$.  However, if all $r_i^2$ are rational, the torus in $\mathbb{C} P^2$

1581: possesses a Legendrian toric multiple cover.

1582: 

1583: Suppose now that $u : S^1 \times S^1 \longrightarrow \mathbb{C} P^2$ is a

1584: homogeneous Lagrangian immersion.  According to our definition $u$ has a lift

1585: $\hat{u}$ such that $\pi ( \hat{u} ( x + t, y )) = \pi ( e^{t A} \hat{u} ( x, y

1586: ))$ and $\pi ( \hat{u} ( x, y + t )) = \pi ( e^{t B} \hat{u} ( x, y ))$.  In

1587: particular $\pi ( e^{x A} e^{y B} p ) = \pi ( e^{y B} e^{x A} p )$, for any $p

1588: \in S^5$ in the image.  However the image is never contained in a complex

1589: subspace of $\mathbb{C}^3$, hence $[ A, B ] \in i\mathbb{R} \mathit{Id}$.  Since

1590: $[ A, B ]$ is traceless, $A$ and $B$ commute.

1591: 

1592: The obvious (non Legendrian) lift in $S^5$ is $( x, y ) \longmapsto e^{x A} e^{y

1593: B} p$ where now $p = ( p_1, p_2, p_3 )$ is a fixed point mapped by the Hopf map

1594: $\pi$ to $u ( 0, 0 )$.  A Legendrian lift $\hat{u}$ takes the following form:

1595: $\hat{u} ( x, y ) = e^{i \theta ( x, y )} e^{x A} e^{y B} p$ for some function

1596: $\theta$.  The horizontality condition implies $\langle ( i\frac{\partial

1597: \theta}{\partial x} \mathit{Id} + A ) p, p \rangle_{\mathbb{C}^3} = 0$ so that

1598: $\frac{\partial \theta}{\partial x} = i \frac{\langle A p, p

1599: \rangle_{\mathbb{C}^3}}{| p |^2}$ is a constant.  The same holds in the $y$

1600: direction so we can define the lift $\hat{u} ( x, y ) = e^{x \hat{A} + y

1601: \hat{B}} p$ where $\hat{A} = A + i\frac{\partial \theta}{\partial x}

1602: \mathit{Id}$ and $\hat{B} = B + i\frac{\partial \theta}{\partial y} \mathit{Id}$

1603: are two commuting skew-symmetric matrices.  (Notice that $\hat{u}$ is only

1604: defined on the universal cover $\mathbb{R}^2$.)  The base point $p$ depends of

1605: course on the choice of origin and is only defined up to multiplication by a

1606: complex unit number.  Nevertheless it plays an important role.

1607: 

1608: Consider now the metric induced by $\hat{u}$.  Due to homogeneity, it is a

1609: constant metric on the $( x, y )$-plane.  By doing a simple change in variables,

1610: we may as well assume that the metric is the standard plane metric, in other

1611: words the immersion is isometric.  (Of course that will change the matrices

1612: $\hat{A}$ and $\hat{B}$, but since they are replaced by some real linear

1613: combination of themselves, the properties mentioned above still hold.)

1614: Henceforth we suppose that $\hat{u}$ is an isometric homogeneous Legendrian

1615: immersion of the plane.

1616: 

1617: Up to a unitary rotation in $\mathbb{C}^3$ we may suppose that $\hat{A}$ is

1618: diagonal, and write $\hat{A} = i \,\hbox{diag} ( a_1, a_2, a_3 )$ with real

1619: coefficients $a_1, a_2, a_3$.  We will now consider three cases and show that

1620: only case (i) is possible.

1621: \begin{enumerate}

1622: \item Suppose $B = i \hbox{diag} ( b_1, b_2, b_3 )$ is diagonal.  Then the

1623: surface lies inside the three torus $T = | p_1 | S^1 \times | p_2 | S^1

1624: \times | p_3 | S^1$.  Necessarily it lifts $\pi ( T )$.  Isometry will

1625: constrain the coefficients to be as above.

1626: 

1627: \item One and only one of the off-diagonal coefficients of $B$ is non zero.

1628: We can assume it is $b_{12}$ up to permutation of the coordinates.

1629: Commutation of $\hat{A}$ and $\hat{B}$ forces $a_1 = a_2$, while $a_3 \neq

1630: a_1$, otherwise we would get a contradiction: $A$ cannot be a multiple of

1631: $i\mathit{Id}$.  Let us first look at equations involving $A$.  The

1632: immersion being isometric in $S^5$, $| p | = | A p | = 1$

1633: \[

1634: 1 = | p_1 |^2 + | p_2 |^2 + | p_3 |^2 = a_1^2 ( | p_1 |^2 + | p_2 |^2 ) +

1635: a_3^2 | p_3 |^2 \hspace{0.75em}

1636: \]

1637: but it is also Legendrian, so \[ \omega ( A p, p ) = \langle i A p, p

1638: \rangle_{\mathbb{C}^3} = a_1 ( | p_1 |^2 + | p_2 |^2 ) + a_3 | p_3 |^2 =

1639: 0 .

1640: \]

1641: Hence \[ | p_1 |^2 + | p_2 |^2 = \frac{a_3}{a_3 - a_1} \hspace{0.25em},

1642: \hspace{0.75em} | p_3 |^2 = - \frac{a_1}{a_3 - a_1} \hbox{ and } a_1 a_3

1643: = - 1

1644: \]

1645: excluding thus $a_1 = 0$, and finally \[ | p_1 |^2 + | p_2 |^2 =

1646: \frac{1}{1 + a_1^2} \hspace{0.75em}, \hspace{0.75em} | p_3 |^2 =

1647: \frac{a_1^2}{1 + a_1^2} \hspace{0.75em} .

1648: \]

1649: Take now into account the Legendrian constraints on $B$: \[ B =

1650: \left(\begin{array}{ccc} i b_1 & b_{12} & 0\\

1651: - \overline{b_{12}} & i b_2 & 0\\

1652: 0 & 0 & b_3 \end{array}\right)

1653: \]

1654: \[ 0 = \langle B p, p \rangle_{\mathbb{C}^3} = i \left( \sum_1^3 b_j |

1655: p_j |^2 + 2 \mathrm{Im} ( b_{12} \overline{p_1} p_2 ) \right) \] \[ 0 =

1656: \langle B p, A p \rangle_{\mathbb{C}^3} = \sum_1^3 a_j b_j | p_j |^2 + 2

1657: a_1 \mathrm{Im} ( b_{12} \overline{p_1} p_2 ).

1658: \]

1659: Uniting both, we deduce \[ a_1 b_1 | p_1 |^2 + a_1 b_2 | p_2 |^2 + a_1 b_3

1660: | p_3 |^2 = a_1 b_1 | p_1 |^2 + a_1 b_2 | p_2 |^2 + a_3 b_3 | p_3 |^2.

1661: \]

1662: Since $| p_3 | \neq 0$ and $a_1 \neq a_3$, $b_3$ vanishes.  The remaining

1663: equations are: \[ | p_1 |^2 + | p_2 |^2 = \frac{1}{1 + a_1^2}

1664: \]

1665: \begin{equation}

1666: \label{bmatrix1} b_1 | p_1 |^2 + b_2 | p_2 |^2 + 2 \mathrm{Im} ( b_{12}

1667: \overline{p_1} p_2 ) = 0

1668: \end{equation}

1669: \begin{equation}

1670: \label{bmatrix2} ( b_1^2 + | b_{12} |^2 ) | p_1 |^2 + ( b_2^2 + |

1671: b_{12}|^2 ) | p_2 |^2 + 2 ( b_1 + b_2 ) \mathrm{Im} ( b_{12}

1672: \overline{p_1} p_2 ) = 1.

1673: \end{equation}

1674: Finally we infer a contradiction: indeed equation (\ref{bmatrix1}) amounts to

1675: the existence of an isotropic vector $( p_1, p_2 )$ for the skew-hermitian

1676: matrix $\left( \begin{array}{cc} i b_1 & b_{12}\\

1677: - \overline{b_{12}} & i b_2 \end{array} \right)$, and that requires its

1678: determinant $|b_{12}|^2 - b_1 b_2$ to vanish.  Plugging this into

1679: (\ref{bmatrix2}), we obtain \[ ( b_1 + b_2 ) \big( b_1 | p_1 |^2 + b_2 |

1680: p_2 |^2 + 2 \mathrm{Im} ( b_{12} \overline{p_1} p_2 ) \big) = 1,

1681: \]

1682: obviously contradicting (\ref{bmatrix1}).

1683: 

1684: \item If at least two off-diagonal coefficients of $B$ are non-zero, then $A =

1685: ia_1 \mathit{Id}$.  But that contradicts $( A p | p ) = 0$.  So that case is

1686: also excluded.

1687: \end{enumerate}

1688: \medskip

1689: 

1690: Notice that in the language of integrable systems, homogeneous tori correspond

1691: to vacuum solutions and are of finite type for $p=0$.

1692: \end{proof}

1693: 

1694: \begin{thebibliography}{99}

1695: %

1696: \bibitem{Br}

1697: {\bibname R. Bryant}, `Minimal lagrangian submanifolds of K\"ahler-Einstein

1698: manifolds', Lecture Notes in Math. Vol. 1255, Springer, New York, 1987, 1--12.

1699: 

1700: \bibitem{BP}

1701: {\bibname F. E. Burstall \and F. Pedit}, `Harmonic maps via

1702: Adler--Kostant--Symes theory', \emph{Harmonic maps and integrable systems}, (eds

1703: A. Fordy and J. C. Wood), Aspects of Mathematics E23 (Vieweg 1994); also in

1704: http://www.amsta.leeds.ac.uk/Pure/staff/wood/FordyWood/contents.html~.

1705: 

1706: \bibitem{BFPP}

1707: {\bibname F. Burstall, D. Ferus, F. Pedit \and U. Pinkall}, `Harmonic tori in

1708: symmetric spaces and commuting Hamiltonian systems on loop algebras', \emph{Ann.

1709: of Maths.} (II) 138 (1993), 173--212.

1710: 

1711: %\bibitem{CaUr}

1712: %{\bibname I. Castro \and F. Urbano}, `Examples of unstable Hamiltonian-minimal

1713: %Lagrangian tori in $\C^2$', \emph{Compositio Mathematica} 111 (1998), 1-14.

1714: 

1715: %\bibitem{CM}

1716: %{\bibname B.-Y. Chen \and J.-M. Morvan}, `Deformations of isotropic submanifolds

1717: %in K\"ahler manifolds', \emph{J. of Geometry and Physics} 13 (1994), 79-104.

1718: 

1719: \bibitem{Da}

1720: {\bibname P. Dazard}, `Sur la g\'eom\'etrie des sous-fibr\'es et des feuilletages

1721: lagrangiens', \emph{Ann. Sci. Ecole Norm. Sup}. (4), 14 (1981) 465--480.

1722: 

1723: \bibitem{DPW}

1724: {\bibname J. Dorfmeister, F. Pedit \and Hong-You Wu}, `Weierstrass type

1725: representation of harmonic maps into symmetric spaces', \emph{Comm.  in Analysis

1726: and Geometry} Vol 6, Number 4 (1998), 633-668.

1727: 

1728: \bibitem{G}

1729: {\bibname M. Guest}, {\em Harmonic maps, loop groups and integrable systems},

1730: (Cambridge University Press, Cambridge, 1997).

1731: 

1732: \bibitem{Has}

1733: {\bibname M. Haskins}, `Special Lagrangian Cones', arXiv:math.DG/0005164.

1734: 

1735: \bibitem{H}

1736: {\bibname F. H\'elein}, {\em Harmonic maps, constant mean curvature surfaces and

1737: integrable systems}, Lecture in Mathematics, ETH Z\"urich, (Birkh\"auser, 2001).

1738: 

1739: \bibitem{HR2}

1740: {\bibname F. H\'elein \and P. Romon}, `Weierstrass representation of Lagrangian

1741: surfaces in four-dimensional space using spinors and quaternions',

1742: \emph{Commentarii Mathematici Helvetici} 75 (2000), 668--680.

1743: 

1744: \bibitem{HR1}

1745: {\bibname F. H\'elein \and P. Romon}, `Hamiltonian stationary Lagrangian

1746: surfaces in $\C^2$', \emph{Communications in Analysis Geometry}, Vol.  10, N. 1

1747: (2002), 79--126; arXiv:math.DG/0009202.

1748: 

1749: \bibitem{HR3}

1750: {\bibname F. H\'elein \and P. Romon}, `Hamiltonian stationary Lagrangian

1751: surfaces in Hermitian symmetric spaces', \emph{Differential geometry and

1752: integrable systems (Tokyo, 2000)}, (eds Martin Guest, Reiko Miyaoka, and

1753: Yoshihiro Ohnita), \emph{Contemp.  Math.} 308, (Amer.  Math.  Soc., Providence,

1754: RI, 2002), pp.  161--178.

1755: 

1756: \bibitem{J}

1757: {\bibname D. Joyce}, `Special Lagrangian 3-folds and integrable systems', Proc.

1758: of the 9th Math.  Soc.  Japan Internat.  Res.  Inst.; arXiv:math.DG/0101249.

1759: 

1760: \bibitem{MM}

1761: {\bibname H. Ma \and Y. Ma}, `Totally real minimal tori in $\C P^2$',

1762: arXiv:math.DG/0106141.

1763: 

1764: \bibitem{McI1}

1765: {\bibname I. McIntosh}, `The construction of all non-isotropic harmonic tori in

1766: complex projective space', \emph{Internat.  J. Math.} 6 (1995), 831--879.

1767: 

1768: \bibitem{McI2}

1769: {\bibname I. McIntosh}, 

1770: `Two remarks on the construction of harmonic tori in $\C P^2$', 

1771: \emph{Internat.  J. Math.} 7 (1996), 515--521.

1772: 

1773: \bibitem{McI3}

1774: {\bibname I. McIntosh}, 

1775: `Special Lagrangian cones in $\C^3$ and primitive harmonic maps', 

1776: \emph{J. London Math. Soc., Second Series} 67 (2003), number 3, pp. 769--789;

1777: arXiv:math.DG/0201157.

1778: 

1779: %\bibitem{PrS}

1780: %{\bibname A. Pressley \and G. Segal}, {\em Loop groups}, Oxford Mathematical

1781: %Monographs, (Clarendon Press, Oxford, 1986).

1782: 

1783: \bibitem{Re1}

1784: {\bibname H. Reckziegel}, `Horizontal lifts of isometric immersions into the

1785: bundle space of a pseudo-Riemannian submersion', \emph{Global differential

1786: geometry and global analysis (1984)}, Lecture Notes in Mathematics 1156,

1787: (Springer, 1985), 264-279.

1788: 

1789: \bibitem{Re2}

1790: {\bibname H. Reckziegel}, `A correspondence between horizontal submanifolds of

1791: Sasakian manifolds and totally real submanifolds of K\"ahlerian submanifolds',

1792: \emph{Pap.  Colloq., Hajduszoboszlo/Hung.  1984, Vol.  2}, Colloq.  Math.  Soc.

1793: Janos Bolyai 46, 1063-1081 (1988).

1794: 

1795: \bibitem{ScWo}

1796: {\bibname R. Schoen \and J. Wolfson}, `Minimizing volume among Lagrangian

1797: submanifolds', \emph{Proc.  Sympos.  Pure Math., 65}, Amer.  Math.  Soc.

1798: (1999).

1799: 

1800: \bibitem{Sh}

1801: {\bibname R. Sharipov}, `Minimal tori in the five dimensional sphere in $\C^3$',

1802: \emph{Theor.  and Math.  Phys.} 87 (1991), 363--369.

1803: 

1804: \bibitem{SYZ}

1805: {\bibname A. Strominger, S.-T. Yau \and E. Zaslov}, `Mirror symmetry is

1806: T-duality', \emph{Nucl.  Phys.  B} 479 (1996), 243--259.

1807: 

1808: \bibitem{T}

1809: {\bibname C. L. Terng}, `Geometries and symmetries of soliton equations and

1810: integrable elliptic equations', 

1811: to appear in \emph{Surveys on Geometry and Integrable Systems}, 

1812: Advanced Studies in Pure Mathematics, Mathematical Society of Japan; 

1813: arXiv:math.DG/0212372.

1814: 

1815: \bibitem{U}

1816: {\bibname K. Uhlenbeck}, `Harmonic maps into Lie groups', \emph{Journal of

1817: Differential Geometry} 30 (1989), 1-50.

1818: 

1819: \end{thebibliography}

1820: 

1821: \affiliationone{ Fr\'ed\'eric H\'elein\\

1822: Universit\'e Denis Diderot (Paris 7)\\

1823: Institut de Math\'ematiques de Jussieu -- UMR 7586, Case 7012\\

1824: 2, place Jussieu\\

1825: 75251 Paris Cedex 05\\

1826: France \email{helein@math.jussieu.fr}}

1827: % Important: Do not put any empty line here.

1828: \affiliationtwo{ Pascal Romon\\

1829: Universit\'e de Marne-la-Vall\'ee\\

1830: 5, bd Descartes, Champs-sur-Marne\\

1831: 77454 Marne-la-Vall\'ee Cedex 2 \\

1832: France \email{romon@univ-mlv.fr}}

1833: 

1834: \end{document}

1835: