Dea

Maximal surfaces with conelike
singularities in
complete flat 3-spacetimes
2
Contents
1 Introduction 7
2 Preliminaries 13
2.1 Notation and main definitions . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Maximal immersions with singularities . . . . . . . . . . . . . . . . . . . . 15
3 Maximal surfaces of finite type in L3 23

3.1 Existence and Uniqueness of CMF graphs . . . . . . . . . . . . . . . . . . 24
3.2 The space of CMF maximal surfaces . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 Identifying Mn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.2 Differentiable structure of the spinorial bundle Sn . . . . . . . . . . 34
3.2.3 The map s2 is open . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.4 Proof of technical results . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3 Convergence of CMF graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Maximal surfaces of finite type in L3 /G 53

4.1 Complete flat 3-spacetimes and CMS surfaces . . . . . . . . . . . . . . . . 54
4.2 Complete flat space-times admitting CMF embeddings. . . . . . . . . . . . 56
4.3 The Global Geometry of singly and doubly periodic CMS surfaces . . . . . 63
4.4 Examples of CMF surfaces in quotients of L3 . . . . . . . . . . . . . . . . . 67
5 Parabolicity criteria for CMS surfaces 71
3
4
List of Figures
1.1 The Lorentzian catenoid and a Riemann type example. . . . . . . . . . . 8
2.1 An upward pointing conelike singularity. . . . . . . . . . . . . . . . . . . . 19

2.2 The surface S and the catenoids Ct and Cts . . . . . . . . . . . . . . . . . . 21
3.1 A catenoidal and a planar end. . . . . . . . . . . . . . . . . . . . . . . . . 24

3.2 Two graphs of catenoidal and planar type with two and three singularities,
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Ω(v), N (v) and Jv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 The canonical homology basis B(v) . . . . . . . . . . . . . . . . . . . . . . 30
4.1 Example of singly periodic maximal surface of Scherk type. . . . . . . . . . 68

4.2 Doubly and singly maximal surfaces . . . . . . . . . . . . . . . . . . . . . . 69
5
6
Chapter 1
Introduction
A maximal hypersurface in a Lorentzian manifold is a spacelike hypersurface with zero

mean curvature. They appear as critical points (local maxima) for the area functional
associated to variations by spacelike surfaces. Besides of their mathematical interest these
hypersurfaces and more generally those having constant mean curvature have a signifi-
cant importance in classical Relativity. More information on this aspect can be found for
instance in [3].
When the ambient space is the Minkowski space Ln+1 , one of the most important results
is the proof of a Bernstein-type theorem for maximal hypersurfaces in Ln+1 :
Theorem[Calabi [12], Cheng and Yau [13]] A complete maximal hyper-

surface in Ln+1 is necessarily a spacelike hyperplane.
It is therefore meaningless to consider global problems on maximal and everywhere

regular hypersurfaces in Ln+1 . Problems of interest should deal with hypersurfaces having
non empty boundary or having a certain type of singularities. For instance, Bartnik and
Simon [4], have obtained results on the existence and regularity of (spacelike) solutions
to the boundary value problem for the mean curvature operator in Ln+1 . Klyachin and
Miklyukov [14] have given results on the existence of solutions, with a finite number of iso-
lated singularities, to the maximal hypersurface equation in Ln+1 with prescribed boundary
conditions.
More recently Klyachin [29] has studied the existence, uniqueness and asymptotic be-
havior of entire maximal graphs in Ln+1 with prescribed flux vector at infinity and compact
singular set.
From a different point of view, Umehara and Yamada [6] have proved some results on the
global behavior of maximal immersed surfaces in L3 with analytical curves of singularities.
7
8 Introduction
Maximal surfaces in L3 share some properties with minimal surfaces in the Euclidean
space R3 . Both families arise as solutions of variational problems: local maxima (minima)
for the area functional in the Lorentzian (Euclidean) case. Like minimal surfaces in R3 ,
the Gauss map is conformal and they admit a Weierstrass representation in terms of
meromorphic data (cf.[22]).
For several reasons, isolated singularities of maximal graphs in L3 are specially inter-
esting. They are called conelike singularities, and correspond to points where the Gauss
curvature blows up, the Gauss map has no well defined limit and the surface is asymptotic
to a half light cone (we refer to Klyachin and Miklyukov [5] for a good setting).
Figure 1.1: The Lorentzian catenoid and a Riemann type example.
The simplest example of this phenomenon is the Lorentzian half catenoid which is the
only entire maximal graph over spacelike planes with an isolated singularity ( cf. [7], and see
[22] for a previous characterization). New examples of Riemann’s type were discovered more
recently [8]. They form a one parameter family of entire maximal graphs over spacelike
planes asymptotic to a plane at infinity and having two isolated singularities. Moreover,
they are characterized by the property of being foliated by complete circles, straight lines
or singular points in parallel spacelike planes, besides the catenoid. In Section 3.1 we
construct explicit examples of entire maximal graphs with an arbitrary number of singular
points.
A maximal surface in L3 reflects about conelike points to its mirror, and its Weierstrass
data extend by Schwartz reflection to the double. This fact establishes an important
connection between the theory of complete embedded maximal surfaces in L3 with a finite
number of singularities and classical algebraic geometry. Therefore, moduli problems arise
in a natural way.
The first part of this work is devoted to exploit this idea. We first observe that a
complete embedded maximal surface with a finite number of singularities is an entire max-
imal graph over any spacelike plane with conelike singularities, and it is asymptotic to a
catenoid or a spacelike plane. The corresponding moduli space has a structure of a finite
dimensional real analytic manifold and we compute its dimension. The underlying topo-
9
logical structure is the natural one, that is to say, the induced by the uniform convergence
on compact subsets.
Theorem I: The space Gn of all the complete maximal graphs having n + 1

singularities has a structure of analytic manifold of dimension 3n + 4. The
coordinates are the position of its singular points in L3 and its logarithmic
growth.
Moreover, the underlying topology is the one of the uniform convergence of
graphs on compact subsets.
In the second part of this paper we deal with complete maximal surfaces with conelike
singularities in any complete flat space-time.
A 3-dimensional complete flat spacetime N is a connected 3-manifold with a geodesi-
cally complete flat Lorentz (2, 1)-metric. In terms of geometric structures, it can be iden-
tified with a quotient L3 /G, where L3 is the Lorentz Minkowski space and G ⊂ Iso(L3 )
is a discrete group of isometries acting properly and freely on L3 . Therefore, the prob-
lem of classifying complete flat spacetimes can be rewritten in terms of discrete groups
of isometries in the Lorentz-Minkowski space. By definition, a spacetime N is said to be
orthochronous if any element of G preserve the future direction.
A basic consequence of Calabi’s theorem is the systematic arranging of groups G ⊂
Iso(L3 ) for which L3 /G contains a complete maximal surface S. For topological consid-
erations, it is natural to assume that the fundamental group of the surface is isomorphic
to G. Under these conditions, G contains a maximal free abelian normal subgroup G↑+
of spacelike translations of rank one or two, and the quotient G/G↑+ is either trivial, Z2
or Z2 × Z2 (a detailed description of these groups can be found in Theorem II.A ahead).
Moreover, S must be a flat cylinder, torus, Möbius strip or Klein bottle.
The former result and the one we obtain for complete maximal surfaces of finite type
in L3 suggest that many open questions still remain about complete embedded maximal
surfaces with singularities. We emphasize two of them:
(A) The classification of spacetimes L3 /G containing complete embedded maximal surfaces

of finite type (that is, with a finite number of singularities), and the geometrical
description of these surfaces.
(B) The global geometry of embedded maximal surfaces in L3 with an infinite closed
discrete set of singular points.
10 Introduction
Both problems are naturally interlaced because embedded maximal surfaces of finite type
in a spacetime L3 /G lead, by passing to the universal covering, to maximal graphs in L3
invariant under G and with an infinite closed discrete set of singular points.
The aim of the second part of this work is to approach the above two problems.
Regarding problem (A), we first observe that the fundamental group of any complete
embedded spacelike surface of finite type S in N = L3 /G is injected by the immersion in
Π1 (N ) ≡ G, i.e., S is incompressible. For this reason, and up to embedding S in a suitable
covering of N , it is always supposed that Π1 (S) ∼ = G.
On the other hand, the classification of discrete groups G acting freely and properly in
3
L is far from being trivial (we refer to the surveys [17], [15] for a good setting). Among
other relevant results, Margulis [11] proved that the fundamental group G of a spacetime
L3 /G can be non abelian free, and Mess [24] that G can not be isomorphic to the funda-
mental group of a closed surface with negative Euler characteristic. Mess result has played
a fundamental role in the proof of Theorem A below, which classifies complete flat space-
times L3 /G restraining a complete embedded maximal surface of finite type. We prove that
this family consists of Calabi’s type spacetimes (i.e., those containing a complete maximal
everywhere regular, and so flat, surface). In particular, none of them can be, for instance,
of Margulis type.
Theorem II.A Let N := L3 /G, G 6= {Id}, be a complete flat spacetime, and

suppose N contains a complete embedded maximal surface S of finite type.
Then, up to conjugacy in Iso(L3 ), only the following possibilities can hold:
(a) N orientable and orthochronous. In this case, either G is generated by a

spacelike translation and S is a cylinder of finite conformal type, or G is
a group of spacelike translations with two generators and S is a torus.
(b) N orientable and non orthochronous. In this case, either G =< R0 > and
S is a Möbius strip of finite conformal type, or G =< R0 , T0 > and S
is a Klein bottle, where R0 ((x1 , x2 , x3 )) := (x1 + ν, −x2 , −x3 ), ν 6= 0, and
T0 (x) = x + (0, λ, 0), λ 6= 0.
(c) N non orientable and orthochronous. In this case, either G =< R1 >
and S is a Möbius strip of finite conformal type, or G =< R1 , T1 > and
S is a Klein bottle, where R1 ((x1 , x2 , x3 )) = (x1 + δ, −x2 , x3 ), δ 6= 0, and
T1 (x) = x + (0, λ, 0), λ 6= 0.
(d) N non orientable and non orthochronous. In this case, G =< R2 >
and S is a cylinder of finite conformal type, G =< R2 , T2 > and S is a
11
torus, or G =< R0 , R2 > and S is a Klein bottle, where R2 ((x1 , x2 , x3 )) =

(x1 , x2 + δ, −x3 ), δ 6= 0, T2 (x) = (λ, µ, 0), λ 6= 0 and R0 is defined as
above.
Moreover, in the non compact case, the ends of S are asymptotic to either a
flat cylinder (if S is a cylinder) or a flat Möbius strip (if S is a Möbius strip).
We also study the asymptotic behavior and analytical representation of periodic com-
plete maximal graphs with singularities. By definition, a complete embedded maximal
surface in L3 with a closed discrete set of singularities is said to be singly periodic (respec-
tively, doubly periodic) if it is invariant under the action of a discrete free group of spacelike
translations of rank one (respectively, two), and the subsequent quotient surface is of finite
type (the last restriction is irrelevant in the doubly periodic case). As a consequence of
Theorem II.A, the universal covering S̃ of a complete embedded maximal surface S of finite
type in N ≡ L3 /G is either singly periodic or doubly periodic. Indeed, the surface S̃/G↑+
is of finite type and embedded in L3 /G↑+ .
As regard to the problem (B), we deal with the underlying conformal structure of com-
plete embedded maximal surfaces in L3 with an infinite closed discrete set of singularities.
In the periodic case, we also study the Weierstrass representation of this kind of surfaces.
This analysis leads to the family of entire maximal graphs with a closed discrete set of
conelike singularities. We prove that all these graphs must be parabolic in the sense that
bounded harmonic functions on the surface are uniquely determined by their singular values
(i.e., its values at the singularities). The notion of parabolicity can be easily extended to
the family of non compact maximal graphs with boundary and a closed discrete set of
interior singular points. In this case, a graph is said to be parabolic if bounded harmonic
functions are uniquely determined by their boundary and singular values. This concept
has been helpful in the modern theory of minimal surfaces in the Euclidean space R3 (see
[9], [16] for a good setting). In Theorem II.B we establish a parabolicity criterium for a
wider family of graphs which includes the minimal ones in R3 :
Theorem II.B Let S be a maximal surface with a closed discrete set of interior
singularities, and suppose S is a graph over a closed starlike region in the plane
{x3 = 0}.
Then S is parabolic.
The same result holds for everywhere regular minimal graphs over closed starlike
planar regions in the 3-dimensional Euclidean space R3 .
12 Introduction
The paper is lead out as follows:

In Section 2.1 we establish the basic definitions and the notation will be used in the
development of this work. In Section 2.2 we state some preliminary results concerning
spacelike and maximal surfaces with singularities. The proof of Theorem I can be found in
Chapter 3, while the one of Theorems II.A and II.B lie in Chapters 4 and 5 respectively.
Moreover, in section 3.1 we construct examples of complete maximal surfaces of fi-
nite type in L3 and Section 4.3 is devoted to the study of the asymptotic behavior and
Weierstrass representation of singly and doubly periodic maximal surfaces, as well as the
construction of the examples exhibited in Figures 4.1 and 4.2.
Chapter 2
Preliminaries
2.1 Notation and main definitions

Throughout this paper we use the following notation and conventions:
• C is the extended complex plane C ∪ {∞}.
• D is the unit disc, D = {z ∈ C : |z| < 1}.
• Given N a Riemann surface and k a positive integer, we call
Divk (N ) = {D : D is an integral multiplicative divisor on N of degree k}.
• M will denote a differentiable surface. For convenience we allow ∂(M) 6= ∅, and in

this case we assume that ∂(M) is smooth enough (for instance, at least C 1 ).
• L3 is the three dimensional Lorentz-Minkowski space (R3 , ds2 = dx2 + dy 2 − dz 2 ) and

h, i denotes its associated scalar product.
• N will denote a flat 3-dimensional Lorentzian manifold (i.e., a 3-dimensional differ-

ential manifold endowed with a flat metric of index one).
• H2 is the hyperbolic sphere in L3 of constant intrinsic curvature −1, H2 = {(x1 , x2 , x3 ) ∈

R3 : x21 + x22 − x23 = −1}.
• H2+ := H2 ∩ {x3 ≥ 1} and H2− := H2 ∩ {x3 ≤ −1} are the two connected components
of H2 .
13
14 Preliminaries
• The stereographic projection σ for H2 is defined as follows:

2Im(z) 2Re(z) |z|2 + 1

2
σ : C − {|z| = 1} −→ H ; z → , , ,
|z|2 − 1 |z|2 − 1 |z|2 − 1
where σ(∞) = (0, 0, 1).
• Cp is the light cone at a point p ∈ L3 , ie. Cp = {q ∈ L3 : ||q − p|| = 0}.

Likewise, Int(Cp ) = {q ∈ L3 : ||q − p|| < 0} is the time cone at p, and we denote
Ext(Cp ) = {q ∈ L3 : ||q − p|| > 0}.
• We also call Int(Cp )+ = Int(Cp ) ∩ {x3 > x3 (p)} and Int(Cp )− = Int(Cp ) ∩ {x3 < x3 (p)}
the future cone and the past cone at p.
• Iso(L3 ) is the group of affine isometries of L3 .
• Iso+ (L3 ) and Iso↑ (L3 ) are the subgroup of positive affine isometries and orthochronous
(that is, preserving H2+ ) affine isometries respectively.
• Iso↑+ (L3 ) = Iso+ (L3 ) ∩ Iso↑ (L3 ).
• For any G ⊂ Iso(L3 ) we denote by G↑+ = G ∩ Iso↑+ (L3 ), G+ = G ∩ Iso+ (L3 ) and
G↑ := G ∩ Iso↑ (L3 ).
Remark 2.1.1 A vector v ∈ R3 − {0} is said to be spacelike, timelike or lightlike if

kvk2 := hv, vi is positive, negative or zero, respectively. When v is spacelike, kvk is chosen
non negative. The vector 0 is spacelike by definition. A plane in L3 is spacelike, timelike or
lightlike if the induced metric is Riemannian, non degenerate and indefinite or degenerate,
respectively.
Remark 2.1.2 Since the universal isometric covering of N is L3 (see for example [25],[27]),
N can be regarded as the quotient of L3 by a discrete group G ⊂ Iso(L3 ) acting freely and
properly on L3 .
Remark 2.1.3 We know that G↑+ (resp., G+ , G↑ ) is a subgroup of index k+ ↑

≤ 4 (resp.,
↑ ↑
k+ ≤ 2, k ≤ 2) in G, and the spacetime N+ := L /G+ (resp., N+ := L /G+ , N ↑ :=
↑ 3 3
↑
L3 /G↑ ) is a k+ -sheeted (resp., k+ -sheeted, k ↑ -sheeted) covering of N := L3 /G. By defi-
nition, N+↑ (resp., N+ , N ↑ ) is said to be the orientable orthochronous (resp., orientable,
orthochronous) covering of N . Note also that G↑+ is a subgroup of index ≤ 2 of G+ , (resp.,
G↑ ) and N+↑ is the orthochronous (resp., orientable) covering of N+ (resp., N ↑ ).
2.2 Maximal immersions with singularities 15
Definition 2.1.1 An immersion X : M −→ N is spacelike if the tangent plane at any

point is spacelike, that is to say, the induced metric on M is Riemannian. In this case,
S = X(M) is said to be a spacelike surface in N .
If N = L3 /G, where G is a (possibly trivial) group of translations acting freely and

properly on L3 , the locally well defined Gauss map N0 of X assigns to each point of M a
point of H2 . A connection argument gives that N0 is globally well defined and N0 (M) lies,
up to a Lorentzian isometry, in H2− . This means that M is orientable.
Definition 2.1.2 Let M be a differential surface. A maximal immersion X : M −→ N

is a spacelike immersion with null mean curvature. In this case, S = X(M) is said to be
a maximal surface in N .
Using isothermal parameters, M can be endowed with a conformal structure. In the

orientable case, M becomes a Riemann surface.
2.2 Maximal immersions with singularities

Let M and F ⊂ M be a differentiable surface with (possibly empty) boundary ∂(M) and
a discrete closed subset of M−∂(M), and let ds2 be a Riemannian metric in M−F. Take a
point q ∈ F, an open disk D(q) in M such that D(q)∩F = {q} and a conformal parameter z
for ds2 on D(q)−{q}. Then write ds2 = h(z)|dz|2 , where h(z) > 0 for any z ∈ z(D(q)−{q}).
By definition, the Riemannian metric ds2 is singular at q if for any disk D(q) and any
parameter z as above, the limit limp→q h(z(p)) vanishes (as a matter of fact, it suffices to
check this condition just for a disc and a conformal parameter).
The metric ds2 is said to be singular in F if it is singular at any point of F. In this case,
ds2 is called a (Riemannian) metric with isolated singularities, (M, ds2 ) a Riemannian
surface with isolated singularities and F the singular set of (M, ds2 ).
One can extend the notion of singularity to boundary points, but for our purposes it
is not necessary. Therefore, throughout this paper we always assume that singular points
are interior.
Definition 2.2.1 Let N be a flat 3-dimensional spacetime, and let X : M → N be a

continuous map. Suppose there is F ⊂ Int(M) discrete and closed in M such that X|M−F
16 Preliminaries
is a spacelike immersion and (M, ds2 ) is a Riemannian surface with isolated singularities
in F, where ds2 is the metric induced by X.
Then, X is said to be an immersion with (isolated) singularities at F, and X(M) a
spacelike surface with singularities at X(F ).
If in addition F is finite, X is said to be a spacelike immersion of finite type, and X(M )
a spacelike surface of finite type in N .
We have assumed above that the set F of (interior) isolated singularities is closed (and
so the set of regular points is always open). This restriction is quite natural for the global
results we will approach later. Of course, this does not prevent the existence of spacelike
surfaces with a non closed set of isolated singularities, although this theory will not be
treated in this paper.
As in the case of minimal surfaces in R3 , one of the most powerful tools in the study
of these surfaces follows from the fact that its Gauss map is conformal. This allows to
recover the immersion in terms of meromorphic data on the underlying Riemann surface.
The following theorem describes this representation, called Weierstrass representation. Its
proof can be found in [22].
Theorem 2.2.1 (Weierstrass Representation) If X : M −→ L3 is a maximal im-

mersion and N0 denotes its Gauss map, then
def
• the map g = σ −1 ◦ N0 is meromorphic,
• there exists a holomorphic 1-form φ3 on M such that

i 1 1 1
φ1 = ( − g)φ3 φ2 = − ( + g)φ3 (2.1)
2 g 2 g
are holomorphic 1-forms on M, Φ = (φ1 , φ2 , φ3 ) never vanishes on M and they have
no real periods, that is
Z

Re Φ = 0 for any γ closed curve in M
γ
R
• Up to a translation, the immersion is given by X = Re P0
(φ1 , φ2 , φ3 ), where P0 ∈ M
is an arbitrary point.
Conversely, if M, g and φ3 are a Riemann surface, a holomorphic map on M and an
holomorphic 1-form on M, respectively, such that |g| < 1 in M, Φ is holomorphic, without
R
zeroes and without real periods, then the conformal immersion X := Re Φ : M → L3
is maximal and its Gauss map is N0 = σ ◦ g.
Definition 2.2.2 We call (M, φ1 , φ2 , φ3 ) (or simply (M, g, φ3 )) the Weierstrass represen-
tation of X.
In case ∂(M) 6= ∅, the data (g, φ3 ) are continuous in M and holomorphic in Int(M).
For simplicity and by definition, (g, φ3 ) are said to be holomorphic in M.
Remark 2.2.1 Weierstrass data are also well defined for maximal immersions in L3 /G,
where G is a translational group. In this case, Φ has no necessary vanishing real periods,
but the group generated by them is contained in G.
The induced Riemannian metric ds2 on M is given by

2
2 2 2 2 |φ3 | 1
ds = |φ1 | + |φ2 | − |φ3 | = ( − |g|) .
2 |g|
Since M is spacelike, then |g| 6= 1 on M, and up to a Lorentzian isometry, we always

assume |g| < 1.
The remainder of this section is devoted to the statement of the main properties of
spacelike and maximal surfaces with isolated singularities.
The following lemma shows that isolated singularities of spacelike surfaces correspond
to topological branch points.
Lemma 2.2.1 (Local structure of singularities) Let X : D → L3 be a spacelike im-

mersion defined on a closed disk D, and suppose that X has a unique singularity at an
interior point q ∈ D. Let Π be a spacelike plane containing P0 = X(q) and label π : L3 → Π
as the Lorentzian orthogonal projection.
Then, there exists a small closed disc U in Π centered at P0 such that (π◦X) : V −{q} →
U −{P0 } is a finite covering, where V is the connected component of (π◦X)−1 (U ) containing
{q}. As a consequence, X|V is an embedding if and only if (π ◦ X) : V − {q} → U − {P0 }
is a homeomorphism, that is to say, X(V ) is a graph over any spacelike plane.
Proof : Up to an ambient isometry we can assume that Π = {x3 = 0} and P0 coincides

with the origin. In the sequel, we put D = X(D).
Claim: There exists a small closed disc U in {x3 = 0} centered at the origin
such that the connected component V of (π ◦ X)−1 (U ) containing q satisfies
X(∂V − {q}) ⊂ π −1 (∂U ).
18 Preliminaries
Indeed, assume by contradiction that there exists a positive sequence {sh } → 0 such
that, for each h, X(∂Vh ) is not contained in π −1 (∂Uh ), where Uh is the disc of radius sh
centered at the origin in {x3 = 0} and Vh is the connected component of X −1 (D ∩ π −1 (Uh ))
containing q.
Then, Vh is a connected domain containing q and a piece of ∂D in its boundary. Then,
the set X −1 (D ∩ {(0, 0, x) : x ∈ R}) contains a curve joining the two components of
∂(D − {q}). This contradicts that X is spacelike in D − {q} and proves the claim.
Let us now show that π ◦ X : V − (π ◦ X)−1 (0) → U − {0} is a finite covering. As X
is spacelike on D − {q}, then π ◦ X is a local diffeomorphism on this set. Moreover, since
(π ◦ X)(∂V ) is contained in ∂U, it follows that π ◦ X : V − (π ◦ X)−1 (0) → U − {0} is a local
diffeomorphism. As (π ◦ X)−1 (0) ∩ V is compact, it is clear that π ◦ X : V − (π ◦ X)−1 (0) →
U − {0} is proper and so we infer that π ◦ X : V − (π ◦ X)−1 (0) → U − {0} is a finite
covering. Since U − {0} is a cylinder, we deduce that V − (π ◦ X)−1 (0) is a cylinder too
and (π ◦ X)−1 (0) ∩ V = {q}.
Now, note that X : V −(π ◦X)−1 (0) → L3 is an embedding if and only if the continuous
map x3 separates the fibers of the covering π ◦ X : V − (π ◦ X)−1 (0) → U − {0}. This is
equivalent to saying that this covering is a homeomorphism, that is to say, D is a graph
over {x3 = 0} locally around the singular point.
2
We are interested in embedded singularities, specially in the maximal case. We say

that a singularity is embedded if the surface is locally embedded around the singularity.
The geometry of embedded singularities for maximal surfaces in L3 is well known, and it
is enclosed in the following lemma:
Lemma 2.2.2 (Conelike singularities, [23] ) Let D be an open disc and let X : D →
L3 be a maximal immersion having an embedded singularity at q ∈ D.
Then, D − {q} is conformally equivalent to {z ∈ C : 0 < r < |z| < 1}, and X extends
to Ar := {0 < r < |z| ≤ 1} with P0 := X(q) = X({|z| = 1}).
Moreover, if X : Ar → L3 is the associated conformal reparameterization of X and
(g, φ3 ) its Weierstrass data, then |g| = 1 and g is injective on {|z| = 1}, and φ3 (z) 6= 0,
|z| = 1. In particular, X reflects analytically about {|z| = 1} to the mirror surface
A∗r := {z ∈ C : 1 ≤ |z| < 1/r}, and so g(J(z)) = 1/g(z) and J ∗ (φ3 ) = −φ3 , where
J(z) = 1/z is the mirror involution.
In addition, X({0 < r0 < |z| ≤ 1}) is a graph over any spacelike plane Π asymptotic
to the top or bottom component of the light cone with vertex at P0 , and P0 is said to be a
downward or upward pointing conelike singularity, respectively.
Figure 2.1: An upward pointing conelike singularity.
Obviously, the local behaviour of isolated singularities detailed in Lemmae 2.2.1 and
2.2.2 is valid for immersions in flat Lorentzian 3-manifolds as well, because these manifolds
are locally isometric to L3 .
Any spacelike immersion X : M → N with isolated singularities induces a distance
function on M in the standard way: the distance between two points is defined as the
infimum length of curves joining them (note that this length is well defined even for curves
passing thorough singular points). It is well known that M is a complete metric space
if and only if every divergent path in M has infinite length, and in this case, X (resp.,
X(M)) is said to be a complete immersion (resp., surface).
Of course, this definition is also valid for surfaces with non empty boundary. Anyhow,
in the context of complete surfaces we always suppose ∂(M) = ∅.
The basic global properties of spacelike surfaces with isolated embedded singularities
in L3 are contained in the following lemma.
Lemma 2.2.3 (Global behaviour of spacelike embedded singularities) Let X : M →

L3 be a complete spacelike immersion with isolated embedded singularities.
Then X(M) is a graph over any spacelike plane, that is to say, the Lorentzian orthogo-
nal projection of X(M) over any spacelike plane in L3 is a homeomorphism. In particular,
X is an embedding.
Proof : Let Π = {x3 = 0} and π the canonical projection π : L3 → Π. Then, it is enough

to see that π ◦ X : M → Π = {x3 = 0} is a covering (and therefore, a homeomorphism).
Since the immersion is spacelike out of the singular points, π ◦ X is a local homeo-
morphism on M minus the singular set. On the other hand, lemma 2.2.1 guarantees that
this map is also a local homeomorphism around the singularities, and so π ◦ X : M → Π
20 Preliminaries
is a local homeomorphism. To finish, it suffices to check that π ◦ X has the path-lifting

property.
Let α = α(t) : [0, 1] → Π be a continuous curve. Since π ◦ X is a local homeomorphism,
we can lift α locally around t = 0. Reasoning by contradiction, take α̃ : [0, ε[→ M an
inextensible lift of α and suppose that ε < 1, that is to say, α̃ is a divergent curve. The
completeness of M gives that α̃ has infinite length. If we denote by ds20 and ds2 the induced
metrics on Π and M respectively, it is straightforward that (π ◦ X)∗ (ds20 ) ≥ ds2 , and so
α|[0,ε[ has also infinite length, which is impossible. 2
Corollary 2.2.1 Complete maximal surfaces with conelike singularities in L3 are graphs
over spacelike planes.
Likewise, CMS (resp. CMF) immersion will represent a complete maximal immersion
with a closed discrete subset of conelike singularities (resp. complete maximal immersion
of finite type).
Remark 2.2.2 There are entire spacelike graphs in L3 which are not complete. However,
the answer to the following question remains open:
Is any entire maximal graph with a closed discrete set of singularities complete?
When the number of singularities is finite, the answer is yes [23].
An interesting point is to determine the asymptotic behavior of complete embedded

maximal surfaces with singularities in L3 . The following result deals with this question. It
has been motivated by the Strong Half Space Theorem for minimal surfaces [20].
Theorem 2.2.2 Let S ⊂ L3 be a proper maximal graph over {x3 = 0} with possibly non
empty boundary and with a (possibly empty) closed discrete set of interior upward pointing
singularities.
Assume that S is contained in a horizontal slab {0 < x3 ≤ k} and ∂(S) ⊂ {x3 = k}.
Then S is a planar domain in {x3 = k}.
As a consequence, a complete embedded maximal surface S in {x3 ≥ 0} (∂(S) = ∅),
with a closed discrete set of singularities lie in the smallest closed horizontal half space
{x3 ≥ k0 } containing F− , where F− is the set of downward pointing singularities of S. In
particular, F− 6= ∅.
Proof : Reason by contradiction, and suppose that S is non flat.

Let P0 be a point of S − {x3 = k}, and let C0 ⊂ {x3 ≤ 0} denote a vertical Lorentzian
catenoid with upward singular point at π(P0 ), where π is the orthogonal projection over
{x3 = 0}. Denote by Ct := C0 + (0, 0, t), t ∈ R.
From our hypothesis, Ct ∩ S is compact, for any t, and there is no contact point at
infinity. Moreover, defining A = {t ≥ 0 : Ct ∩ S = ∅}, we infer that A is a non void open
set bounded above. Therefore t0 := Sup A < +∞ and S ∩ Ct0 6= ∅.
Figure 2.2: The surface S and the catenoids Ct and Cts .
Let us see that Ct0 ∩ S is the singular point of Ct0 . Indeed, reason by contradiction and
suppose that Ct0 ∩ S contains a regular point of Ct0 . Note that Ct0 ∩ S consists of either
regular points of S − ∂(S) or upward pointing singularities of S. The second possibility can
not occur because Ct0 and S contact for the first time (here we are taking into account
the asymptotic behaviour around upward pointing singularities given in Lemma 2.2.2).
Therefore, Ct0 and S contact at regular points, which contradicts the maximum principle
for maximal graphs.
Since S is a graph, we deduce that S ∩ Ct0 = {P0 } and P0 is the singular point of Ct0
(in particular, t0 = x3 (P0 )).
This simply means that S lies above the any Lorentzian catenoid with upward singular
point in S −{x3 = k}. Therefore, denoting by Cts the homothetical shrinking s·C0 +(0, 0, t)
and taking into account that lims→0 Cts = {x3 = t} uniformly on compact subsets, it is
not hard to deduce that S ⊂ {x3 ≥ t0 }. But S touches {x3 = t0 } at P0 , which is not a
downward pointing singularity of S (in fact, it can be chosen as a regular point from the
beginning). This contradicts again the maximum principle and proves the first part of the
theorem.
For the second part, observe that by Lemma 2.2.3, complete embedded maximal surfaces
with singularities are entire spacelike graphs over {x3 = 0}. Then use the first part of the
22 Preliminaries
theorem for assuring that S − {x3 ≥ k0 } = ∅. 2

Chapter 3
Maximal surfaces of finite type in

L3
The next lemma describes the asymptotic behaviour of a CMF graph (recall that a CMF
graph is a complete maximal graph with a finite set of conelike singularities, as it was
defined in Section 2.2). The proof follows from some well known classic results by Jorge
and Meeks [30] and R. Schoen [28] for embedded minimal surfaces. A different approach
in the Lorentzian setting can be found in [29]. We omit the proof.
Lemma 3.0.4 Let X : M → L3 be a maximal embedding with a finite set F of singularities
and vertical limit normal vector at the end, and label (g, φ3 ) its Weierstrass representation.
Then, Max{O∞ (φk ), k = 1, 2, 3} = 2, where O∞ (φk ) is the pole order of φk at the end,
k = 1, 2, 3.
Moreover, if we write G := X(M ) = {u(x1 , x2 ) : (x1 , x2 ) ∈ R2 } as an entire graph
over {x3 = 0}, then:
a1 x1 + a2 x2
u(x1 , x2 ) = c log |(x1 , x2 )| + b + 2
+ O(|(x1 , x2 )|−2 ),
|(x1 , x2 )|
for suitable constants c, b, a1 and a2 .
Take an arbitrary complete embedded maximal surface with a finite set of singularities,
and up to a Lorentzian isometry, suppose that the limit normal vector at its unique end is
vertical. From Lemma 3.0.4, the surface is an entire graph u(x1 , x2 ) over {x3 = 0} which
is asymptotic either to a vertical half catenoid (catenoidal end) or to a horizontal plane
(planar end), see Figure 3.1. The asymptotic behaviour of the surface is controlled by the
real constant c appearing in Lemma 3.0.4, called the logarithmic growth of the end (or of
the surface). Note that c = 0 (respectively, c 6= 0) if and only if the end is of planar type
(respectively, of catenoidal type).
23
24 Maximal surfaces of finite type in L3
Figure 3.1: A catenoidal and a planar end.
3.1 Existence and Uniqueness of CMF graphs

Although we have stated it only in the three dimensional case, the following existence and
uniqueness result is also valid for maximal graphs in Ln+1 .
Theorem 3.1.1 (Klyachin [29]) Let h(x) be a function defined on a compact set K ⊂
{x3 = 0}, Int(K) = ∅, and satisfying the inequality |h(x) − h(y)| < |x − y| for all x, y ∈ K,
x 6= y. Then, for every timelike vector v there exists a unique solution u ∈ C 2 (R2 − K) ∩
C(R2 ) to the maximal graph equation such that u|K = h and F∞ = v, where F∞ is the flux
at the end.
If G = {(x, u(x)) : x ∈ R2 } is an entire maximal graph with a finite number of

singularities, it is not hard to check that
|u(x) − u(y)| < |x − y|, for all x, y ∈ R2 , x 6= y.
Therefore, the previous theorem implies that any CMF graph G with vertical limit normal
vector at infinity is uniquely determined by the position of its singular points and its
logarithmic growth at infinity.
In particular, the group of ambient isometries preserving an embedded CMF surface
coincides with:
• the group of ambient isometries leaving the set of its singularities invariant and
preserving a halfspace containing the surface in case the surface has a catenoidal
end,
• the group of ambient isometries leaving the set of its singularities invariant in case
the surface has a planar end.
The following theorem shows that CMF graphs and meromorphic data on compact
Riemann surfaces admitting a mirror involution are closely related. This fact will be
3.1 Existence and Uniqueness of CMF graphs 25
crucial for understanding the moduli space of this kind of surfaces. Moreover, it can be
used to produce explicit examples of CMF graphs with interesting geometrical properties.
We start introducing the following notation.
By definition, an open domain Ω ⊂ C is said to be a circular domain if its boundary
consists of a finite number of circles.
Theorem 3.1.2 Let N be a compact genus n Riemann surface, and let J : N → N be

an antiholomorphic involution in N. Assume that the fixed point set of J consists of n + 1
pairwise disjoint analytic Jordan curves aj , j = 0, 1, . . . , n, and that N − nj=0 aj = Ω∪J(Ω),
S
where Ω is topologically equivalent (and so conformally) to an open planar circular domain.

Let (g, φ3 ) be Weierstrass data on N such that:
(1) g is a meromorphic function on N of degree n + 1, |g| < 1 on Ω and g ◦ J = 1/g on N,
(2) φ3 is a holomorphic 1-form on N − {∞, J(∞)}, ∞ ∈ Ω, with poles of order at most

two at ∞ and J(∞), and satisfying J ∗ (φ3 ) = −φ3 ,
(3) the zeros of φ3 in N − {∞, J(∞)} coincide (with the same multiplicity) with the zeros
and poles of g.
Rz
Then, the maximal immersion X : Ω − {∞} → L3 , X(z) := Real φ1 , φ2 , φ3 , where
φ1 = 2i ( g1 − g)φ3 and φ2 = −1 ( 1 + g)φ3 is well defined and G = X(Ω − {∞}) is an entire
2 g
maximal graph with conelike singularities corresponding to the points qj := X(aj ), j = 0,
1, . . . , n.
Proof : Let us see that the map X is well defined. First, notice that the curves aj , j =
0, 1, . . . , n generate the first homology group on Ω − {∞}. Moreover, J ∗ (φk ) = −φk , k =
1, 2, 3, and J fixes pointwise the curves aj for any j. Hence we deduce that
Z Z Z Z
∗
φk = φk = J (φk ) = − φk ,
ai J(ai ) ai ai
which means that φk has imaginary periods on Ω − {∞}, and so X is well defined.
On the other hand, |g| = 1 on ∂Ω and (3) imply that φ3 does not vanish on aj , j =
0, 1, . . . , n. Moreover, deg(g) = n + 1 gives in addition that g|aj is injective, j = 0, 1, . . . , n,
and so, by Lemma 2.2.1, all the singularities are of conelike type.
Let us prove the completeness of the metric ds2 induced by X. Suppose that the van-
ishing order of φ3 at ∞ and J(∞) is k ≥ −2 (when k < 0, k = 0 or k > 0 this simply
means that φ3 has a pole of order −k, is regular or has a zero of order k, respectively). The
classical theory of compact Riemann surfaces implies that the number of zeros minus the
number of poles of φ3 in N is 2n − 2 (counting multiplicities). Then the number of zeros
of φ3 in N − {∞, J(∞)} is 2n − 2 − 2k and (3) implies that this is the number of poles
and zeros of g in N − {∞, J(∞)}. Since deg(g) = n + 1, g has 2(n + 1) zeros and poles
in N, and so we infer that g has a zero of order k + 2 at ∞ and a pole of order k + 2 at
J(∞) (take into account that |g| < 1 in Ω, |g| > 1 in J(Ω) and |g| = 1 in ∂Ω). Therefore
2
|φ3 | 1
2
the metric ds = 2 ( |g| − |g|) is complete. 2
Corollary 3.1.1 Let N be the compact genus n Riemann surface:
(z − 1) nj=1 (z − cj )
Q
2 2
N = {(z, w) ∈ C : w = },
(z + 1) nj=1 (z − bj )
Q
where cj , bj are pairwise distinct real numbers in R−{−1, 1}, and define on N the following
meromorphic data
w−1 1
g= , φ3 = ( − w)dz.
w+1 w
Let J : N → N be the antiholomorphic involution given by J(z, w) = (z, −w). Call
{∞1 , ∞2 } the two points in z −1 (∞), and let Ω denote the closure of the connected compo-
nent of N − F, where F is the fixed point set of J, containing ∞1 .
Rz
Then X : Ω − {∞1 } → L3 given by X(z) := Real φ1 , φ2 , φ3 , where φ1 and φ2 are
defined as usually, defines a CMF graph with n + 1 conelike singularities.
Proof : First, observe that the fixed point set of J is not empty and consists of the n + 1
analytic circles aj := z −1 (Lj ), j = 0, . . . , n, where Lj are the pairwise disjoint compact real
intervals determined by the points 1, c1 , . . . , cn , −1, b1 , . . . , bn in R. By Koebe’s Uniformiza-
tion theorem, Ω is biholomorphic to a circular domain with n + 1 boundary components.
It is straightforward to check that (g, φ3 ) satisfies the hypothesis of the Theorem 3.1.2, and
so X defines a CMF graph with n + 1 conelike singularities. 2
The Riemann type graph in Figure 1.1 and the surfaces in Figure 3.2 lie in the family
described in the previous corollary for n = 2 and n = 3.
3.2 The space of CMF maximal surfaces

We know that any CMF surface is, up to an ambient isometry, a graph over the plane
{x3 = 0} with vertical limit normal vector at the end. In the sequel we denote by Gn the
space of CMF graphs over the plane {x3 = 0} with vertical limit normal vector at the end
3.2 The space of CMF maximal surfaces 27
Figure 3.2: Two graphs of catenoidal and planar type with two and three singularities,
respectively.
and having n + 1 singularities, n ≥ 1. As a consequence of Corollary 3.1.1, the space Gn

is not empty.
Let G ∈ Gn and label F as its set of singularities. By definition, a mark in G is an
ordering m = (q0 , q1 , . . . , qn ) of the points in F, and we say that (G, m) is a marked graph.
We denote by Mn the space of marked graphs and call s1 : Mn → Gn and s2 : Mn → R3n+4
the maps given by s1 (G, m) = G and s2 (G, m) = (m, c), where c ∈ R is the logarithmic
growth of G at the end.
By Theorem 3.1.1, the map s2 is injective. One can use this map to endow Mn with
an analytic structure, provided s2 (Mn ) is an open subset of R3n+4 . To prove this fact will
be the main goal of this section.
Label Pn+1 as the symmetric group of permutations of order n + 1 and denote by

µ : Pn+1 × Mn → Mn , the natural actionµ(τ, (G, m)) := (G, τ (m)).
Then, our aim in this section is to prove the following theorem.
Main theorem: The set s2 (Mn ) ⊂ R3n+4 is open and hence the one to one
map s2 : Mn → s2 (Mn ) provides a global system of analytic coordinates on
Mn .
Moreover, the action µ is discontinuous and hence the orbit space, naturally
identified to Gn , has a unique analytic structure making s1 an analytic covering
of (n + 1)! sheets.
Sketch of the proof:

In subsection 3.2.1 we identify the space of marked graphs Mn with the space Sn ×R3 ×
S1 ×R∗ , where Sn is given in terms of the divisors of the Weierstrass data. The definition of
Sn involves some elements of classic theory of Riemann surfaces, like the Jacobian bundle
and the Abel-Jacobi map, which will be explained in this subsection.
In subsection 3.2.2 we prove that this space Sn has a natural structure of differentiable
(3n−1)−manifold and thus we use the previous identification to endow Mn with a structure
of differentiable manifold of dimension 3n + 4.
Finally, in subsection 3.2.3 we prove that s2 is open by showing that it is smooth when
we consider the previous differentiable structure on Mn .
Subsection 3.2.4 is devoted to the proof of the technical results involved in the different
steps of the proof.
3.2.1 Identifying Mn
Let Y = (G, m) ∈ Mn , where m = (q0 , q1 , . . . , qn ) ∈ R3n+3 , and orient G with downward
vertical limit normal vector at the end. By Koebe’s uniformization theorem, G − F with
the prescribed orientation is biholomorphic to an once punctured circular domain, where
the puncture corresponds to the end of the surface.
Remark 3.2.1 In what follows we adopt the following notation: for a sequence of centers
and radii v = (c1 , . . . , cn , r1 , . . . , rn ) ∈]1, +∞[×Cn−1 × (R+ )n such that the closed discs
bounded by the circles aj (v) = {|z − cj | = rj }, j = 0, . . . , n with the convention c0 = 0,
r0 = 1, are pairwise disjoint we label Ω(v) ⊂ C as the circular domain bounded by aj (v),
j = 0, . . . , n.
We say that Ω(v) is a marked circular domain. Two marked circular domains Ω(v1 )
and Ω(v2 ) are considered equals if v1 = v2 .
We label Tn ⊂ R3n−1 the open subset consisting of all v ∈]1, +∞[×Cn−1 × (R+ )n such
that the closed discs bounded by the circles aj (v), j = 0, . . . , n are pairwise disjoint.
The mirror surface of Ω(v) will be denoted by Ω(v)∗ and N (v) represent the double
surface, that is N (v) = Ω(v) ∪ Ω(v)∗ identifying ∂Ω(v)∗ ≡ ∂Ω(v). Observe that N (v) is a
compact Riemann surface of genus n.
Finally, we label Jv : N (v) → N (v) as the mirror involution, mapping a point p ∈ Ω(v)
on its mirror image in Ω(v)∗ . Observe that the fixed point set of Jv coincides with ∂Ω(v) =
∪nj=0 aj (v).
Hence, given Y = (G, m) ∈ Mn it is not hard to see that there are unique v ∈ Tn and
conformal maximal immersion X : Ω(v) − {∞} → L3 such that:
• G − F is biholomorphic to Ω(v) − {∞} (in the sequel, they will be identified),
• G = X(Ω(v) − {∞}),
Figure 3.3: Ω(v), N (v) and Jv .
• qj = X(aj (v)), j = 0, . . . , n.
Lemma 2.2.2 gives that the Weierstrass data of X, (g, φ3 ), satisfy the symmetries g◦Jv =
1/g and Jv∗ (φ3 ) = −φ3 , and that g has exactly n + 1 zeros ∞, w1 , . . . , wn ∈ Ω(v) counted
with multiplicity.
Putting D = w1 · . . . · wn ∈ Divn (Ω(v)), it is easy to see that divisors for the Weierstrass
data must be as follows:
D·∞ D · Jv (D)
(g) = and (φ3 ) = (3.1)
Jv (D · ∞) ∞ · Jv (∞)
The divisor D plays an important role in the identification of the space Mn . More
explicitly, the classic theory of Riemann surfaces gives that D determines uniquely the
data (g, φ3 ) up to a multiplicative constant. Then, D encloses the conformal information
about the surface. The Abel-Jacobi map (defined below) allows to characterize those
divisors for which there exist meromorphic function g and 1-form φ3 satisfying Equation
3.1.
In what follows, for any k ∈ N, we denote by
[
Divk = Divk (Ω(v)) = {(v, D) : v ∈ Tn , D ∈ Divk (Ω(v))}
v∈Tn
and we refer to it as the bundle of k-divisors.

It is well known that, for any Riemann surface N, Divk (N ) is the quotient of N k under
the action of the group of permutations of order k, and we denote by pk : N k → Divk (N )
the canonical projection. We endow Divk (N ) with the natural analytic structure induced
by pk . Then, Divk can be endowed with an analytic structure in a natural way (see Remark
3.2.2 for more details).
Let J (v) be the Jacobian variety of the compact Riemann surface N (v) associated to
the following canonical homology basis:
We identify the homology classes of the boundary circles aj (v) in Ω(v) with their repre-
senting curves j = 0, 1, . . . , n. Note first that Jv fixes aj (v) pointwise, and so, Jv (aj (v)) =
aj (v). Take a curve γj ⊂ Ω(v) joining a0 (v) to aj (v), in such a way that the curve
bj (v) obtained by joining γj and Jv (γj ) satisfies that the intersection numbers (bj , bh )
vanish, and (aj (v), bh ) = δjh , where δjh refers to the Kronecker symbol. Observe that
Jv (bj (v)) = −bj (v) in the homological sense, and its homology class does not depend on
the choice of γj . In other words, the identity Jv (bj (v)) = −bj (v) characterizes B(v) =
{a1 (v), . . . , an (v), b1 (v), . . . , bn (v)} as canonical homology basis of N (v) (see Figure 3.4).
Figure 3.4: The canonical homology basis B(v)
Call {η1 (v), . . . , ηn (v)} the dual basis of B(v) for the space of holomorphic 1-forms on
R
N (v), that is to say, the unique basis satisfying ak (v) ηj (v) = δjk , j, k = 1, . . . , n, and put
R
Π(v) = (πj,k (v))j,k=1,...,n for the associated matrix of periods, πj,k (v) = bj (v) ηk (v).
Then the Jacobian variety of N (v) is J (v) = Cn /L(v), where L(v) is the lattice over
j
Z generated by {e1 , . . . , en , π 1 (v), . . . , π n (v)}, where ej = T
(0, . . . , 1, . . . , 0) and π j (v) =
T
(π1,j (v), . . . , πn,j (v)).
The Jacobian bundle is defined as

[
Jn = J (v)
v∈Tn
Jn has a natural structure of analytic manifold (see Remark 3.2.3 for more details).
For any v ∈ Tn , we call ϕv : N (v) → J (v) the Abel-Jacobi embedding defined by

Z z
T
ϕv (z) = pv (η1 (v), . . . , ηn (v))
1
where pv : Cn → J (v) is the canonical projection (recall that 1 ∈ Ω(v) ⊂ N (v) uniformly
on v).
We extend ϕv with the same name to the Abel-Jacobi map ϕv : Divk (N (v)) → J (v)
given by ϕv (P1 · . . . · Pk ) = kj=1 ϕv (Pj ), k ≥ 1.
P
Also, we define ϕ : Divk → Jn by ϕ(v, D) = (v, ϕv (D))
For any v ∈ Tn , label T (v) ∈ J (v) as the image by ϕv of the divisor associated to any
meromorphic 1-form on N (v). By Abel’s theorem, T (v) is independent of the choice of the
meromorphic 1-form (see [31]).
Lemma 3.2.1 The following two maps are smooth:

ϕ : Divk → Jn , ϕ(v, D) = (v, ϕv (D)), and T̂ : Tn → Jn , T̂ (v) = (v, T (v)).
The proof of this lemma is in subsection 3.2.4. However, the necessary notation to
understand the proof is given in subsection 3.2.3. This notation has not been introduced
in this subsection for the sake of clearness. It is because of this that this lemma appears
in subsection 3.2.4 after the proof of Lemma 3.2.4.
Summarizing, we know that given Y = (G, m) ∈ Mn its associated Weierstrass data

(defined on N (v) for some v ∈ Tn ) satisfies Equation 3.1. Abel’s Theorem gives
ϕv (D · ∞) − ϕv (Jv (D · ∞)) = 0 and ϕv (D · Jv (D)) − ϕv (∞ · Jv (∞)) = T (v)
These two equations leads to
2(ϕv (D) − ϕv (Jv (∞))) = T (v).
In what follows, for any v ∈ Tn , we denote by
Sn (v) = {D ∈ Divn (Ω(v)) : 2(ϕv (D) − ϕv (Jv (∞))) = T (v)}
and
Sn = {(v, D) : v ∈ Tn , D ∈ Sn (v)}
We refer to Sn as the spinorial bundle.
Definition 3.2.1 With the previous notation, we call E the map given by
E : Mn → Sn × R3 × S1 × R∗

E(G, m) = (v, D), q0 , g(1), h3 (1)
where φ3 (z) = h3 (z) dz

z
on the planar domain U (v) := Ω(v) − {∞} ∪ {|z| = 1} ∪
∗ ∗

Ω(v) − {Jv (∞)} ⊂ N (v), z = Id|U (v) , (observe that h3 (z) ∈ R on |z| = 1).
Note that the fixed orientation in the graphs is fundamental for the definition of E. The
first coordinate of E encloses the information about the conformal structure and Weierstrass
data of the marked graph, while the last three ones are simply translational, rotational and
homothetical factors respectively.
The map E will provide the desired identification for the space Mn . To prove this, the
following notation and lemma is required.
Consider the holomorphic 1-form Jv∗ (ηj (v)). Taking into account that Jv fixes aj (v)
pointwise, we infer that ak (v) Jv∗ (ηj (v)) = δjk , and so, Jv∗ (ηj (v)) = ηj (v). Moreover, since
R
R
Jv (bj (v)) = −bj (v), then πj,k (v) = bk ηj (v) is an imaginary number, for any j and k.
It follows that there exists a unique analytic mirror involution Iv : J (v) → J (v)
satisfying Iv (pv (w)) = pv (w), for any w ∈ Cn . Since Jv (1) = 1, I satisfies ϕv ◦ Jv = Iv ◦ ϕv .
Define I : Jn → Jn by I(v, pv (w)) = (v, Iv (pv (w))). We refer to it as the mirror invo-
lution of the Jacobian bundle. It is easy to check that I is analytic.
The next result is crucial to recover Weierstrass data from an element in the spinorial
bundle.
Lemma 3.2.2 Let v ∈ Tn and label Kj (v), j = 1, . . . , 22n , the 22n distinct solutions of the
equation 2σ = T (v), σ ∈ J (v).
Then Iv (T (v)) = T (v) and Iv (Kj (v)) = Kj (v), j = 1, . . . , 22n .
As a consequence, given an integral divisor D ∈ Sn (v) the following equalities hold
• ϕv (D) − ϕv (Jv (∞)) − ϕv (Jv (D)) + ϕv (∞) = 0
• ϕv (D · Jv (D)) − ϕv (Jv (∞) · ∞) = T (v)
Proof : Taking for instance ω = ηj (v), which satisfies Jv∗ (ηj (v)) = ηj (v), v ∈ Tn , we can
check that the divisor (w) is invariant under Jv , and hence I(T̂ (v)) = T̂ (v).
As a consequence of the smoothness of T̂ , we can choose the maps v 7→ Kj (v) to be
smooth for j = 1, . . . , 22n . To prove the invariance of Kj (v) note that Iv (Kj (v)) = Kj (v) +
pv ( 12 nh=1 (mh (v)eh + nh (v)π h (v))), where mh (v), nh (v) ∈ Z are continuous functions of
P
v. Using that Tn is connected we get that mh (v), nh (v) are constant. Hence, the set
Aj := {v ∈ Tn : Iv (Kj (v)) = Kj (v)} is either empty or the whole of Tn . On the other
hand, Kj (v) = K1 (v) + qj (v), where 2qj (v) = 0, and so, Iv (qj (v)) = qj (v). Therefore
A1 = Tn if and only if Aj = Tn for any j. Finally, consider the compact genus n Riemann
surface N = {(z, w) ∈ C : w2 = 2n+2

Q
i=1 (z − ci )}, where ci ∈ R and c1 < c2 < . . . < c2n+2 .
Then define the antiholomorphic involution J(z, w) = (z, −w) and the holomorphic 1-form
Qn−1
ω = i=1 (z − ci ) dz
w
. The function w has a well defined branch w+ on the planar domain
n
Σ = C − ∪i=0 [c2i+1 , c2i+2 ], and the domain {(z, w+ (z)) : z ∈ Σ} ⊂ N is biholomorphic
to a circular domain Ω(v0 ), v0 ∈ Tn . Furthermore, up to this biholomorphism, N = N (v0 )
and J = Jv0 .
Observe that the canonical divisor (ω) is given by c21 · . . . · c2n−1 , where, up to the above
identification (ci , 0) ≡ ci ∈ N (v0 ). Since Jv0 (ci ) = ci , then k0 := n−1
P
i=1 ϕv0 (ci ) ∈ J(v0 ) is
invariant under Iv0 and 2k0 = T (v0 ). Up to relabelling, we can suppose that k0 = K1 (v0 )
and hence A1 = Tn , which finishes the first part of proof.
The second part of the lemma is obvious. 2
Proposition 3.2.1 The map E : Mn → Sn × R3 × S1 × R∗ is bijective.
Proof : Take x ∈ Sn , x = (v, D). By Lemma 3.2.2, ϕv (D) − ϕv (Jv (∞)) − ϕv (Jv (D)) +
ϕv (∞) = 0 and ϕv (D · Jv (D)) − ϕv (Jv (∞) · ∞) = T (v). Abel’s theorem gives that there
exists a meromorphic function f of degree n + 1 on N (v) whose principal divisor is equal
D·Jv (D)
to JvD·∞
(D·∞)
and a meromorphic 1-form ν with canonical divisor ∞·J v (∞)
.
Define
1
gx := f
f (1)
and observe that gx ◦ Jv = 1/gx and gx (1) = 1. Moreover, the last equality and the fact
(gx ) = JvD·∞
(D·∞)
characterize gx as meromorphic function on N (v).
On the other hand, observe that Jv∗ (ν) = λ ν, where |λ| = 1 (recall that Jv is an
involution). Then, the 1-form φ = √iλ ν satisfies Jv∗ (φ) = −φ. If we put φ(z) = h(z) dz
z
,
∗
z ∈ U (v), we infer that h(z) ∈ R , |z| = 1. Then, define
1
φ3 (x) := φ
h(1)
D·Jv (D)
and observe that the equations (φ3 (x)) = ∞·J v (∞)
and h3 (1) = 1 characterize φ3 (x) as
meromorphic 1-form on N (v).
With this notation, given x, q0 , θ, r ∈ Sn × R3 × S1 × R∗ , define

gx (θ) = θgx , φ3 (x, r) = r φ3 (x) (3.2)
and define φ1 (x, θ, r) and φ2 (x, θ, r) in the obvious way. As gx (θ) is holomorphic on Ω(v)
and |g| = 1 on ∂Ω(v), the maximum principle implies that |g| < 1 on Ω(v), and so by
Theorem 3.1.2 the map Xx (q0 , θ, r) : Ω(v) − {∞} → L3 ,

Z z

Xx (q0 , θ, r)(z) := q0 + Real φ1 (x, θ, r), φ2 (x, θ, r), φ3 (x, r)
1

provides a CMF graph Gx (q0 , θ, r) := Xx (q0 , θ, r) Ω(v) − {∞} ∈ Gn .
Defining the mark mx (q0 , θ, r) by qj = Xx (q0 , θ, r)(aj (v)), j = 0, . . . , n, it is now clear
that E −1 (x, q0 , θ, r) = {(Gx (q0 , θ, r), mx (q0 , θ, r))}, and so, E is bijective.
2
3.2.2 Differentiable structure of the spinorial bundle Sn

In the previous subsection we have identified Mn with the space Sn × R3 × S1 × R∗ , our
aim now is to show that Sn has a natural structure of differentiable manifold. In order to
do this we need some preliminary notation.
Remark 3.2.2 (Charts for the bundle of k-divisors Divk ) A coordinate chart for Divk
can be constructed as follows:
Take D0 = P1m1 . . . Psms ∈ Divk (Ω(v)) and consider U = U1m1 × . . . × Usms ⊂ Ω(v)k ,
m
where Uj j = Uj × .m.j. ×Uj , (Uj , zj ) is a conformal chart around Pj in Ω(v), zj (Pj ) = 0, and
Uj1 ∩ Uj2 = ∅, j1 6= j2 . The map ξ : pk (U ) → Ck defined by ξ( sj=1 Q1,mj · . . . · Qmj ,mj ) =
Q
Pmj h
((t1,mj , . . . , tmj ,mj )j=1,...,s ), where th,mj = l=1 (zj (Ql,mj )) , h = 1, . . . , mj , j = 1, . . . , s
defines an analytic parameterization around D0 . For more details, see [31].
Then, for (v0 , D0 ) ∈ Divk , take > 0 small, and label V () as the Euclidean ball of
radius in Tn centered at v0 . Write D0 = z1m1 . . . zsms , zl 6= zh if l 6= h, sh=1 mh = k. If
P
is small enough, the set W = ∩v∈V () Ω(v) contains an open disc Uj of radius around zj ,
j = 1, . . . , s, and we can also take the discs {Uj : j = 1, . . . , s} to be pairwise disjoint.
Consider the conformal charts (Uj , wj := z − zj ). Put U = U1m1 × . . . × Usms , and observe
that U can be viewed as a subset of Ω(v)k , for any v ∈ V (). Likewise, pk (U ) ⊂ Divk (Ω(v)),
for any v ∈ V (). The natural chart ξ : pk (Uv ) → Ck is uniformly defined as before for any
v ∈ Tn .
By definition V() := {(v, D) : D ∈ pk (U ) ⊂ Divk (Ω(v)), v ∈ V ()} is a neighborhood

of (v0 , D0 ) in Divk , and we say that V () is its associate open ball in Tn . Moreover, the
map Ψ : V() → V () × ξ(pk (U )) given by:
Ψ(v, D) = (v, ξ(D)) (3.3)
defines a local (analytic) parameterization around (v0 , D0 ) in Divk . We call v : Divk → Tn ,
v(v, D) = v, the natural projection.
Remark 3.2.3 (Charts for the Jacobian bundle Jn ) Label pv : Cn → J (v) = Cn /L(v)
the natural projection and define p : Tn × Cn → Jn by p(v, z) = (v, pv (z)).
Let (v0 , q0 ) ∈ Jn , and consider w0 ∈ Cn such that pv0 (w0 ) = q0 . Let W () be an
open ball of radius in Cn centered at w0 such that pv0 |W () : W () → pv0 (W ()) is a
conformal diffeomorphism. Since L(v) depends smoothly on v (see Lemma 3.2.4), pv |W () :
W () → pv (W ()) is also a conformal diffeomorphism for any v in the open ball V ()
of radius centered at v0 in Tn , provided that is small enough. By definition, the set
W() := p(U () × W ()) is a neighborhood of (v0 , q0 ) in Jn , and the map
Υ : U () × W () → W(), Υ = p|U ()×W () (3.4)
is a local analytic parameterization.
Theorem 3.2.1 (Structure) The space Sn is a smooth real (3n − 1)-dimensional sub-
manifold of Divn and the map v : Sn → Tn , v(v, D) = v, is a finite covering.
Proof : First, observe that Sn 6= ∅ as a consequence of Corollary 3.1.1 and Proposition

3.2.1. Consider the map H : Divn → Jn given by
H(v, D) = (v, 2ϕv (D) − 2ϕv (Jv (∞)) − T (v))
Then Sn = H −1 (0), where 0 = {(v, 0) : v ∈ Tn } is the null section in Jn .

In order to prove that Sn is a differentiable submanifold of Divn , it suffices to check
that dHq is bijective at any point of Sn .
Let q0 := (v0 , D0 ) be an arbitrary point of Sn .
We are going to write the expression of H in local coordinates Ψ around q0 ∈ Divn and
−1
Υ around H(q0 ) = (v0 , 0) ∈ Jn (see equations (3.3) and (3.4)).
To do this, take V (), V (0 ) balls in Tn centered at v0 of radius and 0 respectively
and W (0 ) ∈ Cn the ball of radius 0 centered at the origin. Write D0 = z1m1 · . . . · zsms ∈
Divn (Ω(v0 )), and denote by (Uj , wj := z − zj ) the conformal parameter in Ω(v0 ), where Uj
m
is the open disc of radius centered at Pj , j = 1, . . . , s. Put U = sj=1 Uj j . The following
Q
computations make sense for suitable small enough real numbers 0 > > 0.
Write ηi (v)(wj ) = fv,i,j (wj )dwj on Wj := wj (Uj ) for i = 1, . . . , n, j = 1, . . . s and
v ∈ V ()
The local expression Ĥ of H around q0 , Ĥ := Υ−1 ◦ H ◦ Ψ−1 , is given by
Ĥ : V () × ξ(pn (U )) → V (0 ) × W (0 )

mj Z
s X
!
X wh,mj
Ĥ(v, t) = v, 2 fv,j (wj )dwj + C(v) − T(v)
j=1 h=1 0
T
Pmj l
where fv,j = (fv,1,j , . . . , fv,n,j ), wh,mj ≡ wj , tl,mj = h=1 wh,mj , h = 1, . . . , mj , t =
(t1,mj , . . . , tmj ,mj )j=1,...,s , and C(v), T(v) ∈ Cn such that pv (C(v)) = 2ϕv (D0 )−2ϕv (Jv (∞)),
pv (T(v)) = T (v) and T(v0 ) = C(v0 ).
In order to prove that dĤ(v0 ,0) is bijective, it suffices to see that
(JacH0 )|t=0 6= 0
where H0 is the holomorphic map defined by (v0 , H0 (t)) + C(v) − T(v) = Ĥ(v0 , t), that is
to say
s Xmj Z w
X h,mj
H0 (t) = 2 fv0 ,j (wj )dwj
j=1 h=1 0
In the sequel we denote fj = fv0 ,j .

Put fj (wj ) = ∞ l n
P
l=0 bj,l wj , bj,l ∈ C , j = 1 . . . s. Then the Taylor series for the holo-
R wh,mj R wh,m
fj (wj )dwj is 0 j fj (wj )dwj = ∞ l
P
morphic map wh,mj 7→ 0 l=1 aj,l wh,mj , where aj,l =
P Pm
1
b
l j,l−1
, l ≥ 1, j = 1, . . . , s. It is not hard to check that H0 (t) = 2 sj=1 l=1j aj,l tl,mj + R(t),
where the first derivatives of R with respect to tl,mj vanish at t = 0, and so the column
vectors of the Jacobian matrix of H0 are {2 al,j , l = 1, . . . , mj , j = 1 . . . , s}.
Reasoning by contradiction, suppose that the rows of that matrix are linearly depen-
dent, which is equivalent to saying that there exists a holomorphic 1-form ω0 in N (v0 )
having a zero at zj ∈ Ω(v0 ) ⊂ N (v0 ) of order at least mj , j = 1, . . . , s. A direct application
of Riemann-Roch theorem gives the existence of a non-constant meromorphic function f on
N (v0 ) having poles at z1 , . . . , zs with order at most m1 , . . . , ms , respectively. In particular,
f has degree less than or equal to n. As Jv0 (∞) is not a pole of f, up to adding a constant
we can suppose that f (Jv0 (∞)) = 0.
On the other hand, since D0 ∈ Sn (v0 ), Lemma 3.2.2 gives that ϕv0 (D0 ·∞)−ϕv0 (Jv0 (D0 ·
∞)) = 0. Therefore, a direct application of Abel’s theorem gives the existence of a meromor-
0 ·∞
phic function g of degree n + 1 on N (v0 ) whose principal divisor coincides with Jv D(D 0 ·∞)
.
0
As Jv0 is an antiholomorphic involution with fixed points, it is not hard to check that
g ◦ Jv0 = r/g, r > 0. Hence, up to multiplying g by the factor r−1/2 , we can suppose that
g ◦ Jv0 = 1/g.
Note that g ∈ Fv0 , where for any v ∈ Tn , Fv denotes the family of meromorphic
functions h of degree n + 1 in N (v) with zeroes in Ω(v) and satisfying h ◦ Jv = 1/h.
1+λf
Claim: Let fλ denote the meromorphic function 1+λ(f ◦Jv0 )
, λ ∈ C. Then, fλ is
∗
not constant, for any λ ∈ C . Moreover, gλ := gfλ ∈ Fv0 for any λ ∈ C.
Assume fλ = c, where c, λ ∈ C∗ . Then, we infer that 1 + λf = c(1 + λ(f ◦ Jv0 )) and

so the polar divisor of f, which is contained in D0 , is invariant under Jv0 . This is absurd
because D0 ∈ Divn (Ω(v0 )) and Ω(v0 ) ∩ Jv0 (Ω(v0 )) = ∅.
For the second part of the claim, first note that the principal divisor of gλ is (gλ ) =
Dλ ·∞
Jv0 (Dλ )·Jv0 (∞)
, where Dλ is an integral divisor of degree ≤ n and so the degree of gλ is
≤ n + 1, λ ∈ C. Moreover, gλ is not constant for any λ (otherwise, Jv0 (∞) would be a zero
of 1 + λf, contradicting f (Jv0 (∞)) = 0).
Let A be the set {λ ∈ C : gλ ∈ Fv0 }, and observe that 0 ∈ A. It suffices to see that A
is open and closed.
The openness of A is an elementary consequence of Hurwitz theorem (we are using the
fact that the degree of gλ is at most n + 1). Finally, let us prove that A is closed. Let
λ0 ∈ A, and take {λn }n∈N → λ0 , where {λn : n ∈ N} ⊂ A. The sequence {gn := gλn }n∈N
converges to g0 := gλ0 uniformly on N (v0 ). We know that gn ◦ Jv0 = 1/gn and so the zeros
of gn lie in Ω(v0 ), therefore, gn is holomorphic on Ω(v0 ), n ∈ N and so the same holds
for g0 . Moreover, since |g0 | = 1 on ∂Ω(v0 ) and it is non constant, the maximum principle
implies that |g0 | < 1 on Ω(v0 ) and we infer that g0 has no critical points on ∂Ωv0 . As ∂Ωv0
consists of n + 1 disjoint circles, this means that g0 takes on any complex number θ ∈ S1
at least n + 1 times. Hence the degree of g0 must be n + 1 and g0 ∈ Fv0 . This concludes
the proof of the claim.
To get the desired contradiction take P ∈ ∂Ω(v0 ) such that f (P ) 6= 0, ∞, and choose
λ = f−1
0
(P )
. Since Jv0 (P ) = P, the meromorphic function gλ0 has degree less than n + 1, and
0
so, λ ∈/ A = C, which is absurd.
Summarizing, we have proved that H|Sn : Sn → 0, H(v, D) = (v, 0), is a local diffeo-
morphism, where 0 = {(v, 0) : v ∈ Tn } ⊂ Jn is the null section in the Jacobian bundle.
Consequently, the projection v : Sn → Tn , v(v, D) = v, is a local diffeomorphism too.
To finish, it suffices to check that v is also proper. Indeed, take a sequence {(vk , Dk )}k∈N ⊂
Sn such that {vk }k∈N converges to a point v∞ ∈ Tn . Lemma 3.2.2 and Abel’s Theorem lead
k ·∞
to the existence of a meromorphic function gk ∈ Fvk with canonical divisor Jv D(D k ·∞)
.
k
Let us see that {gk }k∈N → g∞ ∈ Fv∞ . Reflecting about all the components of ∂Ω(vk ),
we can meromorphically extend gk to a planar open neighborhood Wk of Ω(vk ), k ∈ N.
By continuity and for k0 large enough, the set W = ∩k≥k0 Wk is a planar neighborhood
of Ω(v∞ ). Classical normality criteria show that, up to taking a subsequence, {gk }k∈N
converges uniformly on Ω(v∞ ) to a function g∞ which is meromorphic beyond Ω(v∞ ). It is
clear that |g∞ | = 1 on ∂Ω(v∞ ), |g∞ | < 1 on Ω(v∞ ) and g∞ (∞) = 0. This proves that g∞ is
non constant and can be extended to N (v∞ ) by the Schwartz reflection g∞ ◦ J∞ = 1/g∞ .
Since deg(gk ) = n + 1, then Hurwitz theorem implies that deg(g∞ ) ≤ n + 1. On the other
hand, |g∞ | = 1 only on ∂Ω(v∞ ), and so g∞ is injective on every boundary component of
Ω(v∞ ). Therefore, the degree of g∞ must be exactly n + 1 and g∞ ∈ Fv∞ .
∞ ·∞
Finally, note that Jv D(D ∞ ·∞)
where D∞ ∈ Divn and use Hurwitz theorem to infer that
∞
{Dk }k∈N → D∞ ∈ Divn . Since Sn is a closed subset of Divn , we get D∞ ∈ Sn , which proves
the properness of v : Sn → Tn and so the theorem. 2
3.2.3 The map s2 is open

By Proposition 3.2.1 and Theorem 3.2.1 the space of marked maximal graphs Mn can be
endowed with the differentiable structure making E a diffeomorphism.
In order to prove the smoothness of s2 : M : n → R3n+4 we need first the following
lemma, proved in subsection 3.2.4.
Lemma 3.2.3 Given v ∈ Tn , there exists a holomorphic 1-form ω0 in N (v) having 2n − 2

distinct zeroes, none of them contained in ∂Ω(v), and satisfying Jv∗ (ω0 ) = ω0 .
In Subsection 3.2.1, the smoothness of T̂ was derived from the one of the Riemann’s
constants vector. However, this fact can be also deduced from the previous lemma, at
least for the first order derivatives. Indeed, let v0 ∈ Tn and consider a holomorphic 1-
form w(v0 ) on N (v0 ) with simple zeroes (for instance, the one constructed in the lemma).
Put w(v0 ) = nj=1 λj (0)ηj (v0 ) and note that λj (0) ∈ R. Label λ(0) = (λ1 (0), . . . , λn (0)),
P
and without loss of generality, suppose λ1 (0) = 1. Define wλ (v) = nj=1 λj ηj (v), v ∈ Tn ,
P
λ = (1, λ2 , . . . , λn ) ∈ {1} × Rn−1 and observe that Jv∗ (ωλ (v)) = ωλ (v). A direct application
of Hurwitz theorem implies that wλ (v) has also simple zeros which are not contained in
∂Ω(v), for v and λ close enough to v0 and λ(0), respectively. Moreover, the Implicit
Function theorem and Lemma 3.2.4 give that the zeros of wλ (v) depend at least C 1 on
(v, λ) (in fact smoothly, see the Remark below). Taking into account the smoothness of
ϕ : Divn−1 → Jn and I : Jn → Jn , we infer that T̂ : Tn → Jn is at least C 1 .
Remark 3.2.4 Since T̂ and ϕ are smooth, then the zeroes of wλ (v), for (λ, v) in a neigh-
borhood of (λ(0), v0 ), locally define the following (4n − 2)-dimensional smooth submanifold
of Divn−1 :
{(v, D) ∈ Divn−1 : ϕ(v, D) + ϕ(v, Jv (D)) = T̂ (v)}.
Given v ∈ Tn and k1 , k2 ∈ N we denote by Divk1 ,k2 (v) the product manifold, Divk1 ,k2 (v) =
Divk1 (v)×Divk2 (v), and by Divk1 ,k2 its associated bundle over Tn , Divk1 ,k2 = ∪v∈calT n Divk1 ,k2 (v).
Like in the case of Divk , Divk1 ,k2 has a natural structure of analytical manifold. We use
the convention Divk,0 = Divk and Div0,0 = Tn .
For any v ∈ Tn , call C(v) the family of meromorphic functions on N (v). The corre-
sponding bundle over Tn is denoted by Cn = ∪v∈Tn C(v).
Likewise, we call H(v) the space of meromorphic 1-forms on N (v) and denote by
Hn = ∪v∈Tn H(v) the associated bundle over Tn .
We need to prove that Weierstrass data depend smoothly on ((v, D), q0 , θ, r) ∈ Sn ×

R × S1 × R∗ . Therefore, we need to introduce a convenient concept of differentiability
3
for maps from Divk1 ,k2 into Cn or Hn preserving the fibers. We start with the following
definitions:
Definition 3.2.2 Let Mj be a real manifold of dimension mj , j = 1, 2, 3, and let f :

M1 × M2 → M3 be a C k map. The map f is said to be differentiable (or smooth) with k-
regularity in M1 if, for any charts (U1 ×U2 , x ≡ (x1 , . . . , xm1 ), y ≡ (y1 , . . . , ym2 )) in M1 ×M2
and (U3 , z ≡ (z1 , . . . , zm3 )) in M3 , the local expression of f, f (x, y) : x(U1 )×y(U2 ) → z(U3 ),
satisfies that f (·, y) is smooth in x(U1 ) for any y ∈ y(U2 ), and all the partial derivatives of
f (x, y) with respect to variables in x are C k in x(U1 ) × y(U2 ).
Definition 3.2.3 Let v0 ∈ Tn and > 0 small enough. Denote by V () the Euclidean
ball of radius in Tn centered at v0 . Since V () is simply connected, standard homotopy
arguments in differential topology show the existence of a family of diffeomorphisms {Fv :
N (v0 ) → N (v) : v ∈ V ()} such that Fv0 = Id, Fv (∞) = ∞, Jv ◦ Fv ◦ Jv0 = Fv , for any
v ∈ V (), and F : V () × Ω(v0 ) → C, F (v, z) := Fv (z), is smooth.
By definition, we say that {Fv : N (v0 ) → N (v) : v ∈ V ()} is a smooth deformation
of N (v0 ). Moreover note that, for small enough, ∂F∂z
6= 0 in V () × Ω(v0 ).
Let W ⊂ Divk1 ,k2 be a submanifold, and let h : W → Cn be a map preserving the

fibers, that is to say, hv,D1 ,D2 := h(v, D1 , D2 ) ∈ C(v) for any (v, D1 , D2 ) ∈ W. We are going
to define the notion of differentiability with k-regularity of h. Take V() any coordinate
neighborhood in Divk1 ,k2 defined as above and meeting W. Denote by V () the open ball in
Tn associated to V(), and call v0 ∈ V () its center. Take a smooth deformation of N (v0 ),
{Fv : N (v0 ) → N (v) : v ∈ V ()}. We say that h is differentiable with k-regularity in
V()∩W if the map ĥ : (V()∩W )×N (v0 ) → C, given by ĥ((v, D1 , D2 ), x) = hv,D1 ,D2 (Fv (x))
is smooth with k-regularity in V() ∩ W . The map h is said to be differentiable with k-
regularity on W if it does in V() ∩ W, for any coordinate neighborhood V() meeting W. It
is easy to check that this definition does not depend on choice of the smooth deformation
of N (v0 ).
Let ω : W → Hn be a map preserving the fibers, that is to say, ωv,D1 ,D2 := ω(v, D1 , D2 ) ∈
H(v) for any (v, D1 , D2 ) ∈ W. Take V , V (), v0 and {Fv : N (v0 ) → N (v) : v ∈ V ()}, as
(1,0)
above and define ω̂ : V() ∩ W → H(v0 ) by ω̂(v, D1 , D2 ) = Fv∗ (ωv,D1 ,D2 ) , where the
(1,0)
superscript (1, 0) means the (1, 0) part of the 1-form (by definition (f dz +g dz) = f dz).
We say that ω is differentiable in V() ∩ W with k-regularity if for any local chart (U, z) in
N (v0 ), the map fˆ : (V() ∩ W ) × U → C, given by fˆ((v, D1 , D2 ), z) = ω̂(v, D1 , D2 )(z)/dz
is smooth with k-regularity in V() ∩ W. The global concept of differentiability with k-
regularity in W is defined in the obvious way.
m
Given D = sj=1 wj j ∈ Divk (Ω(v)), we denote by τD (v) the unique meromorphic 1-
Q
form on N (v) having simple poles at wj and Jv (wj ), j = 1, . . . , s, and no other poles, and
R
satisfying Residuewj τD (v) = −ResidueJv (wj ) τD (v) = −mj , ai (v) τD (v) = 0, for any j, i.
m
Likewise, take D1 = sj=1 wj,1j , D2 = rh=1 wh,2 nh
Q Q
∈ Divk (Ω(v)) and define κD1 ,D2 (v) as
the unique meromorphic 1-form on N (v) having simple poles at wj,1 , wh,2 and Jv (wj,1 ), Jv (wh,2 (v),
j = 1, . . . , s, h = 1, . . . , r, and no other poles, and satisfying

Residuewj,1 κD1 ,D2 (v) = ResidueJv (wj,1 ) κD1 ,D2 (v) = −mj ,

Residuewh,2 κD1 ,D2 (v) = ResidueJv (wh,2 ) κD1 ,D2 (v) = nh
R
and ai (v) κD1 ,D2 (v) = 0, for any j, h, i.
Our aim is to show that ηj (v), τD (v) and κD1 ,D2 (v) depend smoothly with 1-regularity
on v, (v, D) and (v, D1 , D2 ), respectively. This fact is enclosed in the following technical
lemma whose proof can be found in subsection 3.2.4.
Lemma 3.2.4 The maps ηj : Tn → Hn , v 7→ ηj (v), τ : Divk → Hn , (v, D) 7→ τD (v), and

κ : Divk,k → Hn , (v, D1 , D2 ) 7→ κD1 ,D2 (v) are differentiable with 1-regularity.
R
As a consequence, the functions πj,k (v) := bj (v) ηk (v), are differentiable on Tn .
As in the proof of Proposition 3.2.1, for any x ∈ Sn we label (gx , ϕ3 (x)) the Weierstrass
data of the CMF marked graph E −1 (x, 0, 1, 1) ∈ Mn .
Proposition 3.2.2 Endowing Mn with the unique differentiable structure making E a dif-
feomorphism, the maps Sn → Cn , x 7→ gx , and Sn → Hn , x 7→ φ3 (x), are smooth with
2-regularity and 1-regularity, respectively.
Proof : Indeed, take x0 = (v0 , D0 ) ∈ Sn . From Theorem 3.2.1, there exists an open ball V ()
in Tn centered at v0 of radius > 0 and a local diffeomorphism V () → Sn , v 7→ (v, D(v)),
where D(v0 ) = D0 . For simplicity, we write x(v) := (v, D(v)), v ∈ V ().
Therefore, the map V () → Divn+1 , v → (v, ∞·D(v)) is smooth, and since τ : Divn+1 →
Hn is also smooth with 1-regularity (Lemma 3.2.4), the same holds for the map V () → Hn ,
v 7→ τv := τ∞·D(v) (v).
Take a smooth deformation of N (v0 ), {Fv : N (v0 ) → N (v) : v ∈ V ()}. Let B(v0 ) =
{a1 (v0 ), . . . , an (v0 ), b1 (v0 ), . . . , bn (v0 )} be the canonical homology basis on N (v0 ) defined
as in subsection 3.2.1. In what follows, we deal with any representative curves aj (v0 ),
bj (v0 ), j = 1, . . . , n, of these homology classes for which N (v0 ) − ∪nj=1 (aj (v0 ) ∪ bj (v0 ))
is simply connected and contains the points in ∞ · D0 . For small enough , the curves
aj (v) := Fv (aj (v0 )), bj (v) := Fv (bj (v0 )) do not pass also through the points in ∞ · D(v),
v ∈ V (), j = 1, . . . , n.
By Abel’s theorem, and for z ∈ N (v) − ∪nj=1 (aj (v) ∪ bj (v)) :
Z z n
X
gx(v) (z) = Exp (τv + mj (v)ηj (v)) .
1 j=1

In this expression, the integration paths lie in N (v) − ∪nj=1 (aj (v) ∪ bj (v)) ∪ {1}, and
mj (v) ∈ Z are integer numbers determined by the equation:
n
X
fv (∞ · D(v)) − ϕ
ϕ fv (Jv (∞) · Jv (D(v))) = mj (v)π j (v),
j=1
where ϕfv is the branch of ϕv on N (v) − ∪nj=1 (aj (v) ∪ bj (v)) vanishing at 1.
Since mj (v) depend continuously on v, then mj (v) = mj ∈ Z and so, by Lemma 3.2.4,
gx(v) depends smoothly on v with 2-regularity.
We have to obtain the analogous result for the map V() → Hn , v → φ3 (x(v)). Take the
holomorphic 1-form ω0 on N (v0 ) given in Lemma 3.2.3, write ν(v0 ) := ω0 = nj=1 λj ηj (v0 ),
P
where λj ∈ R, and define ν(v) := nj=1 λj ηj (v). Since the map v 7→ ν(v) is smooth with
P
1-regularity (see Lemma 3.2.4) it suffices to prove that v 7→ φ3ν(v) (x(v))

is smooth with 2-
regularity. By Hurwitz’s Theorem and the implicit function theorem, ν(v) satisfies also
the thesis in Lemma 3.2.3, for small enough . Moreover, as explained during the proof of
Lemma 3.2.3, the map V () → Div2n−2 , v 7→ (v, (ν(v))) is at least C 1 (in fact smooth by
Remark 3.2.4), where as usually (ν(v)) is the canonical divisor associated to ν(v). Hence,
writing (ν(v)) = E(v) · Jv (E(v)), the map V () → Divn−1 , v 7→ E(v) is also smooth,
and therefore, the same holds for V () → Divn,n , v 7→ (v, ∞ · E(v), D(v)). We infer
from Lemma 3.2.4 that the map V () → Hn , v 7→ κv := κ∞·E(v),D(v) (v) is smooth with
1-regularity. Reasoning as above, the map
Z z n
X
fx(v) (z) = Exp (κv + nj ηj (v)) ,
1 j=1
is a well defined meromorphic function on N (v), for suitable integer numbers nj not de-
pending on v and V() → Cn , v 7→ fx(v) , is smooth with 2-regularity. The principal
D(v)·Jv (D(v))
divisor associated to fx(v) is given by (fx(v) ) = ∞·E(v)·J v (∞)·Jv (E(v))
. Therefore, if we write
ν(v) = hv (z) dz
z
on U (v), we infer that φ3ν(v)
(x(v))
= 1
f ,
hv (1) x(v)
and so v 7→ φ3 (x(v)) is smooth
with 1-regularity. This concludes the proof.
2
Remark 3.2.5 Equation (3.2) gives that the maximal immersion Xx (q0 , θ, r) determin-
ing the marked graph (Gx (q0 , θ, r), mx (q0 , θ, r)) depends smoothly on (x, q0 , θ, r) with 2-
regularity (the notion of differentiability with k-regularity for immersions is defined in the
obvious way).
Theorem 3.2.2 (Main theorem) The set s2 (Mn ) ⊂ R3n+4 is open and hence the one
to one map s2 : Mn → s2 (Mn ) provides a global system of analytic coordinates on Mn .
Moreover, the action µ is discontinuous and hence the orbit space, naturally identified to
Gn , has a unique analytic structure making s1 an analytic covering of (n + 1)! sheets.
Proof : Let x ∈ Sn and denote by (Gx , mx ) = E −1 (x, 0, 1, 1) ∈ Gn . Call Xx the associated

maximal immersion and label as (gx , φ3 (x)) its Weierstrass data.
Observe that E −1 (x × R3 × S1 ×]0, +∞[) consists of all the marked graphs which differ
from (Gx , mx ) by translations, rotations about a vertical axis (i.e., parallel to the x3 -axis)
and homotheties.
Since Xx (q0 , θ, r) depends smoothly on (x, q0 , θ, r) with 2-regularity, the map s2 : Mn →
3n+4
R is smooth. By the injectivity of s2 (see Theorem 3.1.1) and the domain invariance
theorem, s2 (Mn ) is open in R3n+4 and hence it is an analytic manifold of dimension 3n + 4.
We can then endow Mn with the unique analytic structure making s2 : Mn → s2 (Mn ) an
analytic diffeomorphism.
To conclude, it remains to check that the action µ is discontinuous. Indeed, let τ :
Mn → Mn denote the diffeomorphism given by τ (G, m) = (G, τ (m)), τ ∈ Pn+1 . Let
(G0 , m0 ) ∈ Mn and write m0 = (q0 , q1 , . . . , qn ) ∈ R3n+3 . Take a neighborhood Uj of qj in
R3 , j = 0, 1, . . . , n, such that Ui ∩Uj = ∅, i 6= j, and call U = nj=0 Uj . Then, it is clear that
Q
τ (s−1 −1
2 (U × R)) ∩ s2 (U × R) = ∅, for any τ ∈ Pn+1 − {Id}, which proves the discontinuity
of µ and concludes the proof.
2
3.2.4 Proof of technical results

Proof of Lemma 3.2.3:
Let ω be a holomorphic non-zero 1-form on N (v) satisfying Jv∗ (ω) = ω. In addition, we
can choose ω such that it does not vanish on ∂Ω(v). Indeed, let h be the unique harmonic
function on Ω(v) satisfying h|aj (v) = δj,0 , j = 0, 1, . . . , n, where δh,k is the Kronecker symbol.
By the maximum principle, h−1 (0) = ∪nj=1 aj (v) and h−1 (1) = a0 (v). Since these level sets
are regular, h has no critical point on ∂Ω(v). The 1-form ω = ∂z h can be extended to N (v)
by Schwartz reflection, and satisfies the desired properties.
Let D = P1n1 · . . . · Psns · Jv (P1 )n1 · . . . · Jv (Ps )ns be the canonical divisor associated to ω.
Since the degree of D is 2n − 2, it is not hard to deduce from Riemann-Roch theorem
that the complex linear space M(D) of meromorphic functions on N (v) having poles only
at the points Pi , Jv (Pi ) with order at most ni , i = 1, . . . , s, has dimension n.
Note that there is a function f ∈ M(D) of degree 2n − 2. Indeed, otherwise the
maximum degree d0 of functions in M(D) would be less than 2n − 2, and so, there would
exist an entire divisor D0 of degree d0 such that D ≥ D0 and M(D0 ) = M(D), where M(D0 )
is defined in a similar way. Given Q ∈ N (v) such that D/D0 ≥ Q, then, with obvious
notations, M(D0 ) ⊂ M(D/Q) ⊂ M(D). We infer that M(D0 ) = M(D/Q) = M(D)
and so M(D/Q) has dimension n. It then follows from Riemann-Roch theorem that the
complex linear space of holomorphic 1-forms ω 0 satisfying (ω 0 ) ≥ D/Q has dimension 2.
Take a holomorphic 1-form ω 0 ∈ H linearly independent from ω. The quotient ω/ω 0 then
defines a meromorphic function on N (v) of degree 1, which is absurd.
Hence, we can take f ∈ M(D) with deg f = 2n − 2. Since the polar divisor of f is D
and Jv (D) = D, we can find λ and µ ∈ R such that f0 := λ(f + f ◦ Jv ) + iµ(f − f ◦ Jv ) has
the same polar divisor as f, moreover, one has f0 ◦ Jv = f0 . Since poles of f0 do not lie in
∂Ω(v), f0 |∂Ω(v) is bounded. Therefore, there exists a large enough real number r such that
f0 + r has only simple zeroes, and these zeroes are not contained in ∂Ω(v). The 1-form
ω0 := (f0 + r)ω satisfies the properties in the statement of the lemma.
Before to prove Lemmae 3.2.1 and 3.2.4 we need the following technical result:
Lemma 3.2.5 Let Ω be an open bounded domain in R2 with smooth boundary and B an
open Euclidean ball in Rm . Let H(t, x), φ(t, x) : B × Ω → R be two C 1 functions, where
φt := φ(t, ·) ∈ C 2,α (Ω), α ∈]0, 1], for all t = (t1 , . . . , tk ), and the map t 7→ φt is C 1 in the C 2,α
norm on Ω. Consider a smooth one parameter family of metrics ds2t on Ω, t ∈ B and denote
by ∆t the associated family of Laplacians. Let ut ∈ C 0 (Ω) ∩ C 2 (Ω) be the solution of the
boundary value problem ∆t ut = Ht , ut |∂Ω = φt |∂Ω , where Ht (x) := H(t, x), (t, x) ∈ B × Ω.
If we define u : B × Ω → R by u(t, x) = ut (x), then:
1. ut ∈ C 2,α (Ω), and
2. the map t 7→ ut is C 1 in the C 2,α norm on Ω,
3. if H and φ are smooth, then for any q, m1 , . . . , mp ∈ N, m1 +. . .+mp = q, i1 , . . . , ip ∈

qu
{1, . . . , m}, the function ∂tm∂1 ...∂t
t
mp ∈ C
2,α
(Ω),
i1 ip
4. if H and φ are smooth, then the map t 7→ ut is C ∞ in the C 2,α norm on Ω. As a

consequence, for any q, m1 , . . . , mp ∈ N, m1 + . . . + mp = q, i1 , . . . , ip ∈ {1, . . . , m},
qu
the function ∂tm1∂...∂tmp ∈ C
2,α
((B × Ω) ∪ T ), where T is any differentiable portion of
i1 ip
∂(B × Ω).
Proof : It is enough to consider the case m = 2 (the general case is similar).

First, note that (1) is a straightforward consequence of global regularity theorem [32]
p.106.
The maximum principle and the classical Schauder estimates ([32] p.35 and p. 93), show
that the family {ut,s : (t, s) ∈ B} is bounded in the C 2,α norm on Ω. Fix (t0 , s0 ), then
for each (t, s), the function ut,s − ut0 ,s0 satisfies: ∆t0 ,s0 (ut,s − ut0 ,s0 ) = (∆t0 ,s0 − ∆t,s )ut,s +
Ht,s − Ht0 ,s0 on Ω and ut,s − ut0 ,s0 = φt,s − φt0 ,s0 on ∂Ω. The maximum principle and
Schauder’s estimates then show that the map (t, s) 7→ ut,s is continuous at (t0 , s0 ) with
respect to the C 2,α norm on Ω. Let us show that the functions wt,s0 = (ut,s0 − ut0 ,s0 )/(t −
t0 ) converge in the C 2,α norm on Ω, as t → t0 , to the solution yt0 ,s0 of the problem:
∆t0 ,s0 yt0 ,s0 = − ∂∆ ∂t
t
(t0 )(ut0 ,s0 ) + ∂H∂tt,s (t0 , s0 ) on Ω and yt0 ,s0 = ∂φ (t , s ) on ∂Ω. Indeed, put
∂t 0 0
L(t, s0 ) = (∆t,s0 − ∆t0 ,s0 )/(t − t0 ) for t 6= t0 . Then the functions wt,s0 are solutions of the
H −H
problem: ∆t0 ,s0 wt,s0 = −L(t, s0 )(ut,s0 ) + t,s0t−t0t0 ,s0 on Ω, wt,s0 |∂Ω = (φt,s0 − φt0 ,s0 )/(t − t0 ).
On the other hand, the function wt,s0 −yt0 ,s0 satisfies: ∆t0 ,s0 (wt,s0 −yt0 ,s0 ) = −L(t, s0 )ut,s0 +
∂∆t,s H 0 −Ht0 ,s0
∂t
(t0 , s0 )ut0 ,s0 + t,s(t−t 0)
− ∂H∂tt,s (t0 , s0 ), (wt,s0 − yt0 ,s0 )|∂Ω = (φt,s0 − φt0 ,s0 )/(t − t0 ) −
∂φ
(t , s ). Therefore, as before, the maximum principle and Schauder’s estimates imply
∂t 0 0
that wt,s0 converges to yt0 ,s0 in the C 2,α norm on Ω. Likewise, the family {yt,s : (t, s) ∈ B}
is bounded with respect to the C 2,α norm on Ω. Furthermore, the function yt1 ,s1 − yt2 ,s2
satisfies the equation ∆t1 ,s1 (yt1 ,s1 −yt2 ,s2 ) = − ∂∆ ∂t
(t1 , s1 )ut1 ,s1 + ∂∆
∂t
(t2 , s2 )ut2 ,s2 + ∂H∂tt,s (t1 , s1 )−
∂Ht,s
∂t
− (∆t1 ,s1 − ∆t2 ,s2 )yt2 ,s2 , (yt1 ,s1 − yt2 ,s2 )|∂Ω = ∂φ
(t2 , s2 ) (t , s ) − ∂φ
∂t 1 1
(t , s ), and hence,
∂t 2 2
using again the maximum principle and Schauder’s estimates, we obtain the continuity of
yt,s in (t, s) in the C 2,α norm on Ω. The same holds for ∂u ∂s
, and proves (2).
Suppose now that H and φ are C ∞ . The above argument can be applied to ∂u∂tt,s and
∂ut,s
∂s
, and so (1) and (2) also hold for these functions. An iterative argument proves that
the map (t, s) 7→ ut,s is C ∞ in the C 2,α norm on Ω, which proves (3) and the first part
qu
of (4). For the second part of (4), let f (t, s, x) denote any partial derivative of ∂tm∂1 ∂s m2 ,
m1 + m2 = q, q ∈ N, of order two with respect to variables in B × Ω. It is enough to check

that ||f ||0,α (where || · ||0,α is the C 0,α norm in B × Ω) is bounded. This follows from the
inequality:

∂f ∂f
||f ||0,α ≤ C Max{||f (t, s, ·)||0,α : (t, s) ∈ B} + || ||0 + || ||0 ,
∂t ∂s
where C is a positive constant and || · ||0 is the C 0 norm on B × Ω. The C 2,α regularity of u
and all its partial derivatives in (t, s) on the smooth portion of ∂(B × Ω) follows also from
the regularity theorem. 2
Proof of Lemma 3.2.4:

Let v0 ∈ Tn , (v0 , D0 ) ∈ Divk and (v0 , D1,0 , D2,0 ) ∈ Divk,k , and take V (), V0 () and
V() the previously defined open neighborhoods of v0 , (v0 , D0 ) and (v0 , D1,0 , D2,0 ) in Tn ,
Divk and Divk,k , respectively.
m m m
Write D = lj=1 wj j , D1 = lj=1 wj,1j,1 and D2 = lh=1
Q Q1 Q2
wh,2h,2 , and denote by Av,D =
R P Pl1 mj,1
Pl mj l2 mh,2
j=1 log |z − wj | , Av,D1 ,D2 = Im h=1 z−wh,2 − j=1 z−wj,1 dz . Observe that Av,D
is well defined on Ω(v) − {∞, w1 , . . . , wl }, and Av,D1 ,D2 is well defined in a small enough
neighborhood of ∂Ω(v) consisting of the union of n+1 small annuli, up to adding constants.
Let hj,v be the unique harmonic function on Ω(v) satisfying hj,v |ak (v) = δjk . Call also
hv,D (resp. hv,D1 ,D2 ) the unique harmonic function on Ω(v) such that hv,D |∂Ω(v) = Av,D
(resp. hv,D1 ,D2 |∂Ω(v) = Av,D1 ,D2 ).

Let η̂j (v), τ̂D (v), κ̂D1 ,D2 (v) denote the 1-forms ∂z hj,v , 2∂z hv,D −Av,D and 2i∂z hv,D1 ,D2 −

Av,D1 ,D2 , which are well defined as meromorphic 1-forms on Ω(v). They are extended by
Schwartz reflection and with the same name to N (v). Moreover, {η̂j (v) : j = 1, . . . , n} is
a basis of the complex linear space of holomorphic 1-forms on N (v), and τ̂D (v), κ̂D1 ,D2 (v)
are meromorphic 1-forms having the same poles (with the same residues) as τD (v) and
κD1 ,D2 (v), respectively.
Claim: The maps η̂j : V () → Hn , v 7→ η̂j (v), τ̂ : V0 () → Hn , (v, D) 7→ τ̂D (v),
and κ̂ : V() → Hn , (v, D1 , D2 ) 7→ κ̂D1 ,D2 (v), are smooth with 1-regularity.
Take a smooth deformation {Fv : N (v0 ) → N (v), v ∈ V ()} of N (v0 ). Note that
Fv (aj (v0 )) = aj (v), Fv (bj (v0 )) = bj (v), in the homological sense, for any v ∈ V () and any
j.
In the sequel, and for the sake of simplicity, we will put hj,v = hv .
Let Γ : B → W be a parameterization in Tn , Divk or Divk,k , where W is an open neigh-
borhood contained in either V (), V0 () or V(), and write Γ(t) = v(t), Γ(t) = (v(t), D(t))
or Γ(t) = (v(t), D1 (t), D2 (t)) in each case, t ∈ B. Call Ft := Fv(t) : Ω(v0 ) → Ω(v(t)) and
ht := hΓ(t) : Ω(v(t)) → R. Then, it suffices to check that the map û : B × Ω(v0 ) → R,
(t, x) 7→ ht (Ft (x)), is smooth with 2-regularity in B.
To do this, observe first that Ω(v0 ) ⊂ C is conformally equivalent to the bounded
domain Ω0 = {1/x : x ∈ Ω(v0 )} ⊂ R2 , where the biholomorphism is given by T : Ω0 →
Ω(v0 ), T (x) = 1/x. If we put u : B × Ω0 → R, u(t, x) = û(t, T (x)), it is clear that û is
smooth with 2-regularity in B if and only if u does.
Consider now the metric ds2t on Ω0 making Ft ◦T : Ω0 → Ω(v(t)) an isometry, and denote
by ∆t the associated family of Laplacians, t ∈ B. Then ut := ht (Ft (T (x))) is the solution
of the boundary value problem ∆t ut = 0, ut |∂Ω0 = φt |∂Ω0 , t ∈ B, where φt (x) = φ(t, x) and
φ is a suitable smooth function in B × Ω0 . From Lemma 3.2.5, u and its partial derivatives
till the second order are smooth with 2-regularity in B, which proves the claim.
From the previous claim, we infer that the maps η̂j : Tn → Hn , v 7→ η̂j (v), τ̂ : Divk →
Hn , (v, D) 7→ τ̂D (v), and κ̂ : Divk,k → Hn , (v, D1 , D2 ) 7→ κ̂D1 ,D2 (v) are smooth with
R
1-regularity. Hence the period functions v 7→ ah (v) η̂j (v) are smooth on Tn , and since
η̂j (v) = nh=1 ah (v) η̂j (v) ηh (v), for all j, we easily check that η1 , . . . , ηn are smooth with
P R
1-regularity in Tn .
Moreover, τD (v) = τ̂D (v)− nh=1 ah (v) τ̂D (v) ηh (v), and likewise κD1 ,D2 (v) = κ̂D1 ,D2 (v)−
P R
Pn R
h=1 ah (v)
κ̂D1 ,D2 (v) ηh (v). Reasoning as before, τ and κ are smooth with 1-regularity
in Divk and Divk,k , respectively, which concludes the proof.
Proof of Lemma 3.2.1

First, the differentiability of the map ϕ : Divk → Jn , follows from the differentiability
of the maps ηj , j = 1, . . . , n showed in Lemma 3.2.4.
To prove T̂ : Tn → Jn is differentiable observe that T (v) = −K(v) where K(v) is the
vector of Riemann constants (see [31] p.298, 1980). It is well known ([31] p.290, 1980)
3.3 Convergence of CMF graphs 47
)that
n Z
X πjj (v) j
K(v) = pv ( e − ϕ̃v ηj (v))
j=1
2 aj (v)
where ϕ̃v |aj (v) is any lift to Cn with respect to pv of ϕv |aj (v) .
Since the map v 7→ K(v) is smooth, the same holds for T̂ .
3.3 Convergence of CMF graphs

During the proof of Theorem 3.2.2 we have shown that the topological structures induced
by E and s2 on Mn are the same.
In this section we prove that this topology coincides with the one of the uniform con-
vergence of graphs on compact subsets of {x3 = 0}. We also study the structure of the
quotients of Mn and Gn when we identify two graph that differ in a Lorentzian similarity
(that is, a composition of translations, rotations or homotethies).
Theorem 3.3.1 Let {Gk }k∈N be a sequence in Gn , and G0 ∈ Gn .

Then {Gk )}k∈N → G0 in the topology of Gn if and only if {Gk }k∈N converges to G0
uniformly on compact subsets of {x3 = 0}.
Proof : Suppose {Gk }k∈N → G0 ∈ Mn in the topology of Mn , and choose marks in such a
way that {(Gk , mk )}k∈N converges to (G0 , m0 ) in Mn .
Write E −1 (Gk , mk ) = (xk , q0 (k), θk , rk ), Xk = Xxk (q0 (k), θk , rk ) and xk = (vk , Dk ) ∈ Sn ,
k ∈ N ∪ {0}.
Let W be any compact domain in R2 ≡ {x3 = 0} containing the singularities in m0
as interior points, and let Wk denote the compact set Xk−1 (Wk × R) ⊂ Ω(vk ) − {∞},
k ∈ N ∪ {0}.
Since {(xk , q0 (k), θk , rk )} → (x0 , q0 (0), θ0 , r0 ) and Xx (q0 , θ, r) depends smoothly on
(x, q0 , θ, r) with 2-regularity, it is not hard to check that limz→∞ ||Xk (z)|| = +∞ uni-
formly in k. In addition, the domains Wk are uniformly bounded in C, {Wk }k∈N → W0 in
the Hausdorff distance and Xk converges uniformly on W0 to X0 . In the last statement we
have used that Xk can be reflect analytically about the circles in ∂Ω(vk ), and so all the
immersions Xk , k large enough, are well defined in a universal neighborhood of W0 in C. It
is then obvious that the function uk : R2 → R defining the graph Gk converges uniformly
on W to the function u0 : R2 → R defining G0 (furthermore, {vk }k∈N → v0 implies that
{mk }k∈N → m0 ). Since W can be as larger as we want, {uk }k∈N → u0 uniformly on compact
subsets of R2 .
Assume now that {Gk }k∈N converges to G0 uniformly on compact subsets of {x3 = 0},
and as above, denote by uk : R2 → R the function defining the graph Gk , k ∈ N ∪ {0}.
Let us show that singular points of G0 are limits of sequences of singular points of
graphs Gk , k ∈ N. Indeed, let p0 = (y0 , u0 (y0 )) ∈ G0 be a singular point, and without
loss of generality, suppose that p0 is a downward pointing conelike singularity. By Lemma
2.2.1, there exists > 0 small enough such that u−1 0 ({x3 ≤ u0 (y0 ) + }) contains a compact
component C0 () with regular boundary and containing y0 as the unique (interior) singular
point. Since {uk }k∈N → u0 uniformly on compact subsets, u−1 k ({x3 ≤ u0 (y0 ) + }) must
contain a compact component Ck () containing y0 as well, k large enough. Furthermore,
{Ck ()} → C0 () in the Hausdorff sense, and by the maximum principle Ck () must contain
at least an interior singular point yk of uk , k large enough. Since C0 () converges to {y0 }
as → 0, we deduce that {pk := (yk , uk (yk ))}k→∞ → p0 .
As a consequence, there exist marked graphs (Gk , mk ) ∈ Mn , k ∈ N ∪ {0}, such that
{mk }k∈N → m0 .
Call ck the logarithmic growth of Gk , k ∈ N∪{0}, and let us see that {ck } → c0 . Indeed,
let γ be a circle in R2 containing all the singular points of u0 in its interior, and let A denote
a closed tubular neighborhood of γ in R2 not containing any singular point of u0 . It is well
known that the function uk − u0 is solution of a uniformly elliptic linear equation over A,
k large enough. Since {uk }k∈N → u0 uniformly on A, the classical Schauder estimates ([32]
p. 93) imply that {uk }k∈N → u0 in the C 2 norm on A. In particular,
Z Z
{ νk (sk )dsk }k∈N → ν0 (s0 )ds0 ,
γ γ
where νk and sk are the conormal vector and the arc-length parameter along γ in Gk ,
R R
respectively, for any k ∈ N ∪ {0}. Since γ νk (sk )dsk = 2π(0, 0, ck ) and γ ν0 (s0 )ds0 =
2π(0, 0, c0 ), we infer that {ck } → c0 .
Since s2 : Mn → s2 (Mn ) ⊂ R3n+4 is an homeomorphism, {(Gk , mk )}k∈N → (G0 , m0 ) in
the manifold Mn , and so, {Gk }k∈N → G0 in the manifold Gn . This concludes the proof. 2
In the sequel, and up to an explicit mention of the contrary, the underlying structure
in Mn will be the analytic one induced by s2 .
There is a natural connection between the n-spinorial bundle Sn and the quotient of
Mn under the action of a suitable subgroup of Lorentzian similarities. Next lemma and
corollary will be devoted to give a detailed explanation of this fact. We first fix the following
notations.
Let R denote the group of Lorentzian similarities generated by translations, rotations

about a vertical axis, symmetries with respect to a horizontal plane and homotheties in
L3 .
Two marked graphs (G1 , m1 ), (G2 , m2 ) ∈ Mn are said to be congruent if there exists a
Lorentzian similarity R̂ ∈ R such that R̂(G1 ) = G2 , R̂(m1 ) = m2 . Call M̂n the quotient
space of Mn under the congruence relation and denote by proj2 : Mn → M̂n the natural
projection.
Likewise G1 , G2 ∈ Gn are said to be congruent if there exists a Lorentzian similarity
R̂ ∈ R such that R̂(G1 ) = G2 . We call Ĝn the quotient space of Gn under the congruence
relation, and label proj3 : Gn → Ĝn the natural projection.
Denote by proj0 : Sn × R3 × S1 × R∗ → Sn the natural projection proj0 (x, q, θ, r) = x.

Let π : R3 → {x3 = 0} denote the orthogonal projection, and call ∆3n+4 the open
subset of R3n+3 × R ≡ R3n+4 given by
∆3n+4 = {((q0 , q1 , . . . , qn ), c) ∈ R3n+4 : π(qi ) 6= π(qj ), i 6= j}.
Let Cn be the quotient of ∆3n+4 under the congruence relation: ((q0 , q1 , . . . , qn ), c) ∼

((q00 , q10 , . . . , qn0 ), c0 ) if there exists a Lorentzian similarity R̂ ∈ R such that R̂(q0 , q1 , . . . , qn ) =
(q00 , q10 , . . . , qn0 ) and R(0,~ 0, c) = (0, 0, c0 ), where R~ is the linear similarity associated to R̂.
3n+4
Label proj1 : ∆3n+4 → ∆ ∼ as the natural projection, and denote by
Hi = {((0, q1 , . . . , qn ), c) : x2 (q1 ) = 0, |q1 | = 1, x1 (q1 ), x3 (qi ) > 0}, i = 1, . . . , n,
Hn+1 = {((0, q1 , . . . , qn ), c) : x2 (q1 ) = 0, |q1 | = 1, x1 (q1 ) > 0, c > 0},

H0 = {((0, q1 , . . . , qn ), 0) : x2 (q1 ) = 0, |q1 | = 1, x1 (q1 ) > 0, x3 (qi ) = 0, i = 1, . . . , n},
where |q1 | here refers to the Euclidean norm of q1 .
∆3n+4
It is clear that ∪n+1
j=1 proj1 (Hj ) = ∼
− proj1 (H0 ). The map proj1 |Hi : Hi → proj1 (Hi )
3n+4
is bijective and provides analytic coordinates on the open set proj1 (Hi ) ⊂ ∆ ∼ . Hence,
∆3n+4
Cn := − proj1 (H0 )
∼
is an analytic manifold.
Moreover, it is easy to check that s2 (Mn ) ⊂ ∆3n+4 − H e 0 = proj−1
e 0 , where H 1 (proj1 (H0 )).
Otherwise, there would be (G, m) ∈ Mn such that s2 ((G, m)) ∈ H0 , and so, G would be a
graph asymptotic to a horizontal plane and with singularities in {x3 = 0}. Theorem 3.1.1
would imply that G = {x3 = 0}, which is absurd.
As a consequence, proj1 (s2 (Mn )) ⊂ Cn .
The connection between the above projections is given in the following lemma.
Lemma 3.3.1 There exists unique maps sˆ2 : M̂n → Cn , sˆ1 : M̂n → Ĝn and Ê : M̂n → Sn
making commutative the following diagrams:
s E
R3n+4 − H
e 0 ←−2−− Mn −−−→ Sn × R3 × S1 × R∗
  
proj proj proj
y 1 y 2 y 0
ŝ Ê
Cn ←−2−− M̂n −−−→ Sn
1 s
Mn −−−→ Gn
 

proj2 y
proj
y 3
1ŝ
M̂n −−−→ Ĝn
Moreover, ŝ2 is injective and Ê is bijective.
Proof : It is natural to define
Ê : M̂n → Sn , Ê(proj2 (G, m)) = proj0 (E(G, m)).
To prove that it is well defined and bijective, it suffices to check that, given (G1 , m1 ),
(G2 , m2 ) ∈ Mn , proj0 (E((G1 , m1 ))) = proj0 (E((G2 , m2 ))) if and only if there exists R̂ ∈ R
such that R̂(G1 ) = G2 , R̂(m1 ) = m2 .
Indeed, write E((Gj , mj )) = (xj , q0,j , θj , rj ), where xj = (vj , Dj ) ∈ Sn , j = 1, 2. For
simplicity, put Xj = Xxj (q0,j , θj , rj ), j = 1, 2, and identify Gj ≡ Ω(vj ) − {∞}, j = 1, 2.
Assume there exists R̂ ∈ R such that R̂(G1 ) = G2 , R̂(m1 ) = m2 , and write R : Ω(v1 ) →
Ω(v2 ) the natural conformal transformation induced by R̂. If we orient G1 and G2 with
downward limit normal vector at the end, any translation, rotation about a vertical axis,
homothety or symmetry with respect to a horizontal plane preserves the orientation of the
graphs, and so, the same holds for R̂. Therefore, R is a biholomorphism (that is to say, a
Möbius transformation). Since R̂ also preserves the end and the singular points, we infer
that R(∞) = ∞, R({|z| = 1}) = {|z| = 1}, and R(∂Brj (v1 ) (cj (v1 ))) = ∂Brj (v2 ) (cj (v2 )), for
any j. This implies that v1 = v2 and R is the identity map on Ω(v1 ).
On the other hand, gx2 (θ2 ) ◦ R = L ◦ gx1 (θ1 ), where L : C → C is the conformal
transformation induced by R̂. In addition, L fixes the origin, because R preserves the end
and the limit normal vector at the ends points downward.Therefore gx2 (θ2 ) = gx2 (θ2 ) ◦ R =
θgx1 (θ1 ), where |θ| = 1, and hence D1 = D2 . Now it is straightforward to check that
x1 = x2 .
For the converse, take (G1 , m1 ) ∈ Mn , call E(G1 , m1 ) = (x1 , q1 , θ1 , r1 ), and observe
that the set (proj0 ◦ E)−1 (x1 ) = { Xx1 (q, θ, r)((Ω(v1 ) − {∞}), mx1 (q, θ, r)) : (q, θ, r) ∈

R3 × S1 × R∗ }. But this set consists of marked graphs in Mn differing from (G1 , m1 ) by

ambient similarities R̂ ∈ R preserving the mark.
Finally, observe that sˆ2 : M̂n → Cn , sˆ2 ([(G, m)]) := [(m, c)] is well defined, and use
Theorem 3.1.1 to show that it is injective. The map sˆ1 : M̂n → Ĝn is given by sˆ1 ([(G, m)]) =
[G], and the commutativity of the diagrams is obvious. 2
From Lemma 3.3.1, M̂n can be endowed with the differentiable structure making Ê,
proj2 and ŝ2 a diffeomorphism, a submersion and a smooth map, respectively (provided
that Mn is endowed with the differentiable structure induced by E). However following
the aim of Theorem 3.2.2 we are interested in viewing M̂n as an analytic manifold. This
is the main goal of the following corollary.
Corollary 3.3.1 The set ŝ2 (M̂n ) ⊂ Cn is open, and the bijective map ŝ2 : M̂n → ŝ2 (M̂n )
provides a unique analytic structure in M̂n making proj2 : Mn → M̂n an analytic submer-
sion.
Moreover, if Σ̂1 , Σ̂2 are two distinct connected components of M̂n , then either sˆ1 (Σ̂1 ) ∩
sˆ1 (Σ̂2 ) = ∅ or sˆ1 (Σ̂1 ) = sˆ1 (Σ̂2 ). In the second case, Σ̂1 and Σ̂2 are analytically diffeomor-
phic.
Proof : To prove that ŝ2 (M̂n ) is open, take into account that s2 (Mn ) is open in ∆3n+4 − H
e0
and that proj1 : ∆3n+4 − H e 0 → Cn is an open submersion. Induce in M̂n the unique
analytic structure making ŝ2 an analytic embedding. Since ŝ2 ◦ proj2 = proj1 ◦ s2 and
proj1 : s2 (Mn ) → ŝ2 (M̂n ) is an analytic submersion, we infer that proj2 is an analytic
submersion.
For the second part of the corollary, let Σ̂1 and Σ̂2 be two connected components in M̂n ,
and suppose that ŝ1 (Σ̂1 ) ∩ ŝ1 (Σ̂2 ) 6= ∅. Take (G, m1 ) ∈ proj−1 −1
2 (Σ̂1 ), (G, m2 ) ∈ proj2 (Σ̂2 ).
Denote by Σ1 the connected component of Mn containing (G, m1 ), and analogously define
Σ2 . Since s1 (G, m1 ) = s1 (G, m2 ), Theorem 3.2.2 shows the existence of τ ∈ Pn+1 such
that τ (Σ1 ) = Σ2 , and so s1 (Σ1 ) = s1 (Σ2 ). Moreover, taking into account that proj3 ◦ s1 =
ŝ1 ◦ proj2 and proj2 (Σj ) = Σ̂j , j = 1, 2, we get ŝ1 (Σ̂1 ) = ŝ2 (Σ̂2 ). In addition, τ : Σ1 → Σ2
induces in a natural way an analytic diffeomorphism τ̂ : Σ̂1 → Σ̂2 , which concludes the
proof. 2
Following Theorem 3.2.2, it is natural to ask whether ŝ1 : M̂n → Ĝn is an analytic covering.
However the class of a marked graph admitting symmetries has non-trivial isotropy group
for the natural action µ̂ : Pn+1 × M̂n → M̂n . Anyway, we can endow Ĝn − proj3 (Sym(Gn ))
(where Sym(Gn ) consists of the family of graphs with non-trivial symmetry group) with
the analytic structure making ŝ1 : M̂n − ŝ−1

1 proj3 (Sym(Gn )) → Ĝn − proj3 (Sym(Gn )) an
analytic covering of (n + 1)! sheets.
Chapter 4
Maximal surfaces of finite type in

L3/G
In this chapter, we will focus our attention on the geometry of complete maximal embed-
dings of finite type in complete flat 3-dimensional spacetimes. One of the main goals of
this paper is the classification of complete flat spacetimes admitting this kind of surfaces.
Next lemma is crucial for some non-existence results of spacelike surfaces. It will also
play a fundamental role in the study of the conformal structure of maximal graphs in
Section 5.
Lemma 4.0.2 Let Ω ⊂ C be a closed starlike region centered at the origin with C 1 boundary
and S̃ = {(z, u(z)) : z ∈ Ω}, u(0) = 0, a spacelike graph in L3 with isolated singularities.
Call F ⊂ Int(S̃) the singular set.
Then, the following statements hold:
(i) S̃ ⊂ Ext(C0 ).

(ii) For each θ ∈ [0, 2π], the function fθ (t) := dist (teiθ , u(teiθ )), C0 ) is increasing on
[0, tθ [, where tθ := Sup{t ∈ R : teiθ ∈ Ω} ∈]0, +∞] (here dist means Euclidean
distance).
(iii) The Lorentzian norm n : S̃ → R, n(p) = kpk is proper.
(iv) If Ω = C ≡ R2 and S̃ does not meet a lightlike line `, then S̃ is disjoint from the
plane containing ` and orthogonal (in the Lorentzian sense) to `.
Proof : For each θ ∈ [0, 2π], label Πθ as the vertical plane in L3 ≡ C × R generated by the
vectors (eiθ , 0) and (0, 1),and denote by Π+ iθ
θ the half plane {(xe , y) : x, y ∈ R, x ≥ 0, }.
53
54 Maximal surfaces of finite type in L3 /G
Then
Int(S̃) ∩ Π+ iθ
θ = {ρθ (t) := (te , uθ (t)) : t ∈ [0, tθ [}, where uθ (t) := u(teiθ )
The simple curve ρθ (t) is differentiable outside the singularities and spacelike (i.e.,
kρ0θ (t)k
> 0), which simply means that |u0θ (t)| < 1. Integrating from t = 0, we get |uθ (t)| < t
and so S̃ ∩ Π+θ ⊂ Ext(C0 ), θ ∈ [0, 2π]. This proves (i).
For (ii), notice that
1
fθ (t) = dist(ρθ (t), C0 ) = √ Min{|t − uθ (t)|, |t + uθ (t)|}, t ∈ [0, tθ [.
2
Since kρθ (t)k, kρ0θ (t)k > 0, it is easy to check that fθ (t) is increasing on [0, tθ [.
To see (iii), observe that the Lorentzian spheres Hs := {x ∈ L3 : kxk = s}, s > 0, are
asymptotic to the light cone C0 in the following Euclidean sense: limk→±∞ (rk,s − rk ) = 0,
√
where rk,s := k 2 + s2 and rk := |k| are the radii of the concentric Euclidean circles
Hs ∩ {x3 = k} and C0 ∩ {x3 = k} respectively. Then it suffices to check that

dist S̃ − D, C0 > > 0, (4.1)
where dist means Euclidean distance and D ⊂ S̃ is a compact domain which orthogonally
projects on {(x, y, 0) ∈ L3 : x2 + y 2 ≤ δ}, δ > 0 small enough. Indeed, take a point
p ∈ S̃ − D, and write p = ρθ (t), for suitable θ ∈ [0, 2π] and t ∈]δ, tθ [. Hence, (ii) gives
dist (p, C0 ) = fθ (t) > fθ (δ) > := dist (∂(D), C0 ) > 0,
which proves (4.1) and (iii).

To prove (iv), let ` be a lightlike line such that S̃ ∩ ` = ∅. Without loss of generality, we
can suppose that ` is above the entire graph S̃. Take a point q ∈ ` and consider the unique
point q 0 ∈ S̃ contained in the same vertical line as q. Up to a translation taking q 0 into the
origin, (i) gives S̃ ⊂ Ext(Cq0 ), and since q is above q 0 , we also obtain that S̃ ∩ Int(Cq )+ = ∅.
Thus, S̃ omit the set ∪q∈` Int(Cq )+ , that is to say, one of the halfspaces determined by the
plane containing ` and orthogonal in the Lorentzian sense to `. 2
4.1 Complete flat 3-spacetimes and CMS surfaces

By definition, a surface S ⊂ L3 is said to be periodic if S is invariant by a group G ⊂ Iso(L3 )
acting properly and freely on L3 . The notions of periodic spacelike surface and periodic
maximal surface with singularities are defined analogously.
4.1 Complete flat 3-spacetimes and CMS surfaces 55
Note that Lemma 2.2.3 yields that any CSS surface in L3 is a graph over spacelike
planes, and in particular is embedded (recall that a CSS surface was defined in Section
2.2).
There is a natural connection between periodic CSS surfaces in L3 and CSS surfaces in
complete flat Lorentzian three manifolds. This fact lies behind the following well known
lemma, whose proof has been omitted.
2
Lemma 4.1.1 Let (M, ds2 ) and (M̃, ds ˜ ) be Riemannian surfaces with isolated singular-
ities, and suppose there is a regular locally isometric covering π : M̃ → M.
˜ 2 ) is complete.
Then,(M, ds2 ) is complete if and only if (M̃, ds
The following proposition explains the relation between CSS immersions in a complete
flat Lorentzian 3-manifold N and periodic CSS embeddings in L3 . The corresponding
version for CSS maximal immersions works as well.
Proposition 4.1.1 Let X : M → N be a CSS immersion, where N is a complete flat

Lorentzian three manifold.
Then X is incompressible, i.e. X∗ : Π1 (M) → Π1 (N ) is injective. Moreover, there
exists a complete flat Lorentzian three manifold N 0 and an isometric covering p : N 0 → N
such that:
• X can be lifted to a CSS embedding X 0 : M → N 0 such that X = p ◦ X 0 and

X∗0 : Π1 (M) → Π1 (N 0 ) is an isomorphism.
• If π : M̃ → M and π0 : L3 → N 0 are the universal coverings of M and N 0 ,

respectively, there exists a periodic CSS embedding X̃ : M̃ → L3 such that π0 ◦ X̃ =
X 0 ◦ π.
As a consequence, M can be identified with the quotient M̃/G, where G ∼ = Π1 (N 0 ) is the

group of affine isometries of L3 given by the deck transformations of π0 .
Conversely, given a periodic CSS embedding X̃ : M̃ → L3 , and labelling G as the
group of affine isometries of L3 acting freely and properly and leaving X̃(M̃) invariant,
the natural immersion X : M := M̃/G → N 0 := L3 /G is a CSS embedding and X∗ :
Π1 (M) → Π1 (N 0 ) ≡ G is an isomorphism.
Proof : Take N as in the statement of the Proposition and put N ≡ L3 /G0 , where G0 is a
group of isometries of L3 which acts freely and properly on it. Let X : M → N be a CSS
immersion and consider the universal covering π : M̃ → M of M. The injectivity of X∗
follows from the fact that the (unique) lift of X to the universal coverings of M and N ,
X̂ : M̂ → L3 , is an embedding by Lemmae 4.1.1 and 2.2.3.
Let G be the subgroup of G0 of all the isometries leaving X̃(M̃) invariant, i.e., up to
natural identifications, G ≡ X∗ (Π1 (M)). Define N 0 := L3 /G and p : N 0 ≡ L3 /G → N =
L3 /G0 the natural covering. It is straightforward to check that X can be lifted to CSS
embedding X 0 : M → N 0 satisfying p ◦ X 0 = X. Clearly X∗0 : Π1 (M) → Π1 (N 0 ) ≡ G is an
isomorphism and π0 ◦ X̃ = X 0 ◦ π, where π0 : L3 → N 0 ≡ L3 /G is the natural projection.
The converse stated in the lemma is obvious. 2
Remark 4.1.1 The last proposition shows that, up to lifting the immersion to a suitable
covering of the 3-dimensional complete flat spacetime N , a CSS immersion X : M → N
is an embedding and X∗ : Π1 (M ) → Π1 (N ) is an isomorphism. For this reason, in the
sequel we always assume that CSS immersions are embeddings inducing an isomorphism
between the fundamental groups of the surface and the spacetime.
4.2 Complete flat space-times admitting CMF em-

beddings.
The aim in this section is to describe the subgroups G of Iso(L3 ) for which N = L3 /G
admits CMF surfaces. By Proposition 4.1.1, this problem is completely equivalent to the
one of classifying complete flat Lorentzian 3-manifolds admitting a CMF embedding. To
sum up, what we are going to show is that any such G must contain subgroup of finite index
generated by either one or two spacelike translations. This result controls the topology
and geometry of both the spacetime and the maximal surface.
Definition 4.2.1 A rotation in L3 is an element of Iso+ (L3 ) whose fixed points are just the
points on a line `, called the axis of the rotation. The rotation is called elliptic, hyperbolic
or parabolic provided ` is a timelike, spacelike or lightlike line, respectively.
An element of Iso+ (L3 ) is said to be a screw motion in L3 if it is the composition
of a rotation followed by a translation and is not a rotation. A screw motion is said to
be elliptic, hyperbolic or parabolic provided its associated rotation is elliptic, hyperbolic or
parabolic, respectively.
It is well known that a positive isometry of L3 is either a translation, a rotation or a

screw motion. Hence, the only positive isometries in L3 without fixed points are transla-
tions of non zero vector and screw motions. Rotations and screw motions in L3 have been
carefully described in the following remark.
4.2 Complete flat space-times admitting CMF embeddings. 57
Remark 4.2.1 Given a positive isometry R ∈ Iso+ (L3 ), it is possible to find a suitable
coordinate system where:
    
cos t sin t 0 x 0
(1) R(x, y, z) = − sin t cos t 0 y  +  0  , λ ∈ R, t ∈]0, 2π[, if R is elliptic.
    
0 0 1 z λ
    
1 0 0 x λ
(2) R(x, y, z) = 0 cosh t sinh t  y  +  0  , λ ∈ R, = ±1, t ∈ R (t 6= 0 if =
    
0 sinh t cosh t z 0
1), if R is hyperbolic.
    
1 −t t x 0
(3) R(x, y, z) =  t 1 − t2 /2 t2 /2  y  +  0  , λ ∈ R, t 6= 0, if R is parabolic.
     
t −t2 /2 1 + t2 /2 z λ
The parameter t is called the angle of R. Moreover, R is a screw motion if and only if
λ 6= 0.
An elliptic or hyperbolic screw motion leaves globally (not pointwise) invariant a

straight line parallel to the axis of its associated rotation, which is called the axis of the
screw motion. This property does not hold for parabolic screw motions. Observe also that
any elliptic or parabolic screw motion is orthochronous, while a hyperbolic screw motion
is orthochronous if and only if = 1 in the above representation.
Lemma 4.2.1 There are no CSS surfaces S̃ in L3 invariant under elliptic or parabolic
screw motions.
As a consequence, if S̃ is invariant under a negative isometry R ∈ Iso− (L3 ) without
fixed points, then in a suitable coordinate system the isometry is given by:
(i) Orthochronous case: R((x1 , x2 , x3 )) = (x1 , −x2 , x3 ) + (δ, 0, 0), δ 6= 0, or
(ii) Non orthochronous case: R((x1 , x2 , x3 )) = (x1 , x2 , −x3 ) + (0, δ, 0), δ 6= 0.
Proof : Let S̃ be a CSS surface in L3 invariant by a elliptic screw motion R with axis `.
Up to an Lorentzian isometry, we can suppose that ` is the x3 −axis and the non zero
translation vector v of R is vertical (see remark 4.2.1). Since S̃ is a graph over the plane
{x3 = 0} (see Lemma 2.2.3), S̃ ∩ ` = {p}. However, R(p) = p + v ∈ S̃ ∩ ` too, which is
absurd.
Assume now that R is a parabolic screw motion. As above, and up to an isometry, we

can suppose that the axis of R ~ is ` = {(0, s, s) : s ∈ R}, and its translation vector is
v = (0, 0, λ), λ 6= 0.
Assume that S̃ ∩ ` 6= ∅. Then, for any p ∈ S̃ ∩ `, R(p) and p would be points in S̃ lying
on the same vertical line, which contradicts that S̃ is a graph over {x3 = 0}.
Suppose now that S̃ ∩ ` = ∅. By Lemma 4.0.2, S̃ is contained in one the halfspaces
determined by {(x, y, z) ∈ L3 : x3 = x2 }. Without loss of generality, we put S̃ ⊂ H =
{(x1 , x2 , x3 ) : x2 ≥ x3 }. For any c ∈ R label Hc = {(x1 , x2 , x3 ) ∈ L3 : x2 ≥ x3 + c}. Then
S̃ = Rk (S̃) ⊂ Rk (H) = H−kλ for any k ∈ Z, which is impossible.
For the final part of the lemma, note that if S̃ is invariant under a negative isometry R,
then R2 must be an orthochronous positive isometry different from an elliptic or parabolic
screw motion. By Remark 4.2.1, −R is either a hyperbolic positive isometry, an elliptic
positive isometry of angle π or a translation. The last possibility can not hold because
R has no fixed points. If −R is elliptic of angle π, R ~ is the Lorentzian symmetry with
respect to a spacelike plane. Hence R2 is a spacelike translation (recall that R2 (S̃) = S̃
and S̃ is spacelike), which in a suitable coordinate system leads to (ii). Finally, if −R is
hyperbolic, it must be non orthochronous and of angle 0 (otherwise, R would have fixed
points). Reasoning as above R2 is a spacelike translation, which in a suitable coordinate
system corresponds to (i). 2
Lemma 4.2.2 Let S̃ be a CSS surface invariant under a discrete subgroup G ⊂ Iso↑+ (L3 )
acting freely and properly on L3 . Assume that G contains a non trivial translation.
Then G is a group of spacelike translations of rank at most two.
Proof : Take T any non zero translation in G and write T (x) = x + w, w 6= 0. Since S̃
is a spacelike graph with singularities, the vector determined by two points of S̃ must be
spacelike, and so w = T (p) − p, p ∈ S̃, is spacelike.
Reasoning by contradiction, suppose G contains a Lorentzian screw motion R. By
Lemma 4.2.1, R(x) = R(x)~ + v, where R ~ is a linear orthochronous hyperbolic rotation of
~
non zero angle, and up to a Lorentzian isometry, v lies in the axis ` of R.
Define T 0 = R−1 ◦ T ◦ R ∈ G and observe that T 0 (x) = x + w0 , where w0 = R ~ −1 (w).
Let us see that the vectors w and w0 are linearly independent. Otherwise, R(w)~ = λw,
~ ~
λ ∈ R. We infer that w is spacelike eigenvector of R, that is to say, w lies in the axis of R.
Therefore λ = 1, T = T 0 and w = µv, µ ∈ R. Since G0 =< T, R > acts freely and properly
as a group of translations on `, it must be cyclic. Hence there are n, m ∈ Z − {0} such
that T m ◦ Rn fixes the origin, which contradicts that G acts freely and properly on L3 .
Denote by Π the spacelike plane generated by w and w0 , and call G1 =< T, T 0 > . As S̃
is a graph over Π, then S̃/G1 is a topological torus and the same holds for S/G since it is
covered by S̃/G1 . This proves that G is a rank two abelian group. Since G1 is a finite index
subgroup of G, we infer that Rn ∈ G1 , for a suitable n > 0, and so Rn is a translation.
This is obviously absurd, and proves that G is a translational group.
If G contains two independent translations, we repeat the last argument to get that
S̃/G is a topological torus again, and so G has rank two. This concludes the proof. 2
The following remarkable result by Mess will play a crucial role in the proof of Theorem
4.2.2:
Theorem 4.2.1 (Mess [24], [10]) If G ⊂ Iso(L3 ) acts freely and properly on L3 , then
G cannot be isomorphic to the fundamental group of a closed surface of negative Euler
characteristic.
As a consequence of Lemmae 4.2.1, 4.2.2 and Theorem 4.2.1, we can prove the following
theorem.
Theorem 4.2.2 If N = L3 /G, G 6= {Id}, contains a CSF (embedded) maximal surface

S, then the subgroup of positive and orthochronous part of G, G↑+ , is a group of spacelike
translations of rank one or two.
Moreover, only the following possibilities can hold:
(a) N+↑ = N :
(i) G is generated by a spacelike translation. In this case S is a cylinder with ends

conformally equivalent to an once punctured disc.
(ii) G is generated by two independent spacelike translations. In this case S is a
torus.
(b) N+↑ 6= N+ = N :
(i) G =< R0 >, where in a suitable coordinate system R0 ((x1 , x2 , x3 )) := (x1 +

ν, −x2 , −x3 ), ν 6= 0. In this case, S is a Möbius strip with end conformally
equivalent to an once punctured disc.
(ii) G =< R0 , T0 >, where in a suitable coordinate system R0 is as above and T0 (x) =
x + (0, λ, 0), λ 6= 0. In this case, S is a Klein bottle.
(c) N+↑ = N+ 6= N :
(i) G =< R1 >, where in a suitable coordinate system R1 ((x1 , x2 , x3 )) = (x1 +

δ, −x2 , x3 ), δ 6= 0. In this case S is a Möbius strip with end conformally equivalent
to an once punctured disc.
x + (0, λ, 0), λ 6= 0. In this case, S is a Klein bottle.
(d) N+↑ 6= N+ 6= N and N+↑ 6= N ↑ 6= N :
(i) G =< R2 >, where in a suitable coordinate system R2 ((x1 , x2 , x3 )) = (x1 , x2 +

δ, −x3 ), δ 6= 0. In this case S is a cylinder with ends conformally equivalent to
an once punctured disc.
(λ, µ, 0), λ 6= 0. In this case, S is a torus.
(iii) G =< R0 , R2 >, where R0 and R2 are as above. In this case S is a Klein bottle.
Proof : Call S̃ ⊂ L3 the universal covering of S. Since Π1 (S) ∼ = G (see Remark 4.1.1) is
the surface S̃ is a (periodic) CSS maximal surface invariant under G, and S+↑ := S̃/G↑+ is
a finite covering of S ≡ S̃/G. Moreover, S+↑ is naturally embedded as CMF surface in the
spacetime N+↑ = L3 /G↑+ .
Let us show that G↑+ is a translational group of rank at most 2. Reasoning by contra-
diction, suppose that G↑+ contains a orthochronous screw motion. From Lemmae 4.2.1 and
4.2.2, we infer that G↑+ consists of orthochronous hyperbolic screw motions, besides the
identity.
Let D ⊂ S+↑ be a closed disc containing the singularities as interior points. As maximal
surfaces have non-negative Gauss curvature, Huber Theorem [21] can be applied. Therefore
the surface S+↑ − Int(D) is conformally equivalent to a compact Riemann surface M with
boundary minus a finite set of interior points {P1 , . . . , Pk }, which are called the ends of
the surface. Let us show that in fact S+↑ − Int(D) has no ends. Suppose not and take a
once punctured conformal disc A0 ⊂ S+↑ − Int(D) centered at an end Pj , j ∈ {1, . . . , k}.
Let γ0 be a loop in A0 winding once around the puncture Pj and call R0 ∈ G↑+ − {Id} the
transformation such that
π −1 (A0 )/ < R0 >≡ A0 ,
where π : S̃ → S+↑ is the covering projection. Let N0 denote the Gauss map N0 : π −1 (A0 ) →
H2− . Since N0 ◦ R0 = R~ 0 ◦ N0 , we can naturally induce a holomorphic map B : A0 → H2 / <
−
~ 0 > . It is clear that the homomorphism between fundamental groups
R
~ 0 >) ≡< R
B∗ : Π1 (A0 ) ≡< R0 >−→ Π1 (H2− / < R ~0 >
is an isomorphism. On the other hand, since R ~ 0 is an hyperbolic transformation, H2− / <

R~ 0 > endowed with the hyperbolic metric is conformally equivalent to an annulus A1 :=
{z ∈ C : 1 < |z| < r}, r > 1, and up to natural identifications, B : A0 → A1 is
holomorphic. By the Riemann removable singularity theorem, B extend holomorphically
to the disc A0 ∪ {Pj }. In particular, B(γ0 ) is homotopically trivial, which is absurd and
proves that S+↑ is a compact surface.
By Theorem 4.2.1, χ(S+↑ ) = 0, i.e., S+↑ is a torus and G↑+ =< R, R0 >∼ = Z × Z. Since
0
R and R commute, they are hyperbolic orthochronous screw motions with the same axis
`. As G↑+ acts freely and properly on ` as a discrete group of translations, then the group
< R|` , R0 |` > is cyclic. Thus, there exist n, m ∈ Z such that Rn ◦ R0m fixes the origin,
and henceforth Rn ◦ R0m = Id, which is absurd and proves that G↑+ consists of spacelike
translations.
By Lemma 4.2.2, G↑+ has either one or two generators. In the first case, S+↑ is a cylinder,
and Huber theorem again implies S+↑ is of finite conformal type. In the second case, S+↑ is
a torus. Since S+↑ covers S, then S is either a cylinder, a Möbius strip, a torus or a Klein
bottle.
For the classification part of the theorem, it is important to keep in mind that an
isometry of G reverses the orientation of the spacelike graph S̃ if and only if it is either
positive and non orthochronous or negative and orthochronous.
Suppose N is orientable, i.e., N+ = N .
By the preceding facts, the orthochronous case leads trivially to (a)(i) and (a)(ii).
Assume that N ↑ 6= N . If G↑+ is generated by a spacelike translation, then G is generated
by a non orthochronous (hyperbolic, by Lemma 4.0.2) screw motion R0 . Since R02 ∈ G↑+ ,
the angle of R0 must be zero, and in a suitable coordinate system R0 is given as in (b)(i).
When G↑+ has rank two, G =< T00 , R0 >, where T00 is a spacelike translation, R0 is a non
orthochronous hyperbolic screw motion and G↑+ =< R02 , T00 > (note that G↑+ is an index
two subgroup of G and G ∼ = Π1 (KB), where KB is a Klein bottle). Reasoning as in the
previous case, in a suitable coordinate system R0 ((x1 , x2 , x3 )) = (x1 , −x2 , −x3 ) + (ν, 0, 0),
ν 6= 0, . Label v = (γ, λ, µ) as the translation vector of T00 . Since T000 := R0−1 ◦ T00 ◦ R0 is
a translation of vector (γ, −λ, −µ) and < T000 ◦ T00 , R02 > acts freely and properly on the
x1 -axis, we get γ = nδ, n ∈ Z. As T00 ◦ R0−n has no fixed points we infer that n is even.
Therefore T0 := T00 ◦ R0−n is a spacelike translation of vector orthogonal to (ν, 0, 0) and
G =< T0 , R0 > . Without loss of generality, T0 (x) = (0, λ, 0), λ 6= 0, which proves (b)(ii).
Suppose now that N is non orientable (N+ 6= N ).
If G is cyclic, Lemma 4.2.1 gives G =< R >, where, in a suitable coordinate system,
either R = R1 (orthochronous case) and S is a Möbius strip or R = R2 (non orthochronous

case) and S is a cylinder. This proves (c)(i) and (d)(i).
Assume that G↑+ = G+ and G has rank two. In this case G is generated by a spacelike
translation T 0 and a negative isometry R without fixed points, where G↑+ =< R2 , T 0 > .
As above and in a suitable coordinate system, either R = R1 and S is a Klein bottle
(orthochronous case) or R = R2 and S is a torus (non orthochronous case). Write T 0 (x) =
x + (v1 , v2 , v3 ). In case R = R1 , T 00 = R1 ◦ T 0 ◦ R1−1 is a translation of vector (v1 , −v2 , v3 )
different from T 0 (G ∼ = Π1 (KB) is not commutative) and so v2 6= 0. Since T 00 ∈< R12 , T 0 >
and this group only contains translations of spacelike type, it is not hard to see that v3 = 0,
and so T 0 ◦ T 00 (x) = x + (2v1 , 0, 0). As < T 0 ◦ T 00 , R12 > acts freely and properly on L3 ,
we get v1 = kδ, k ∈ Z. Moreover, since T 00 ◦ R1−k has no fixed points, k is even. Thus,
T1 := T 00 ◦ R1−k is the translation of vector (0, v2 , 0) and G =< R1 , T1 >, proving (c)(ii)
for λ = v2 . In case R = R2 , G is commutative and so T 0 ◦ R2 = R2 ◦ T 0 . This implies that
T2 = T 0 is a horizontal translation, which leads to (d)(ii).
It remains to study the case G has rank two and G↑+ 6= G+ . Observe that in this case
G∼ = Π1 (KB) and G/G+ ∼
↑
= Z2 ×Z2 . Remind that an isometry of G reverses the orientation of
S̃ if and only if it is either positive and non orthochronous or negative and orthochronous.
Therefore, G =< R0 , R >, where R0 is a positive non orthochronous isometry, R is a
negative orthochronous isometry and R02 ◦ R2 = Id (recall that G ∼ = Π1 (KB)). Since
2 ↑
R0 ∈ G+ , in a suitable coordinate system we have R0 (x1 , x2 , x3 ) = (x1 + ν, −x2 , −x3 ),
ν 6= 0.
As G ~ := {R~ : R ∈ G} ∼ = Z2 × Z2 , then R ~ 2 = Id and R ~ and R ~ 0 commute. Thus
we can deduce that either R((x ~ 1 , x2 , x3 )) = (x1 , −x2 , x3 ) or R((x
~ 1 , x2 , x3 )) = (−x1 , x2 , x3 ).
The last case is impossible because R ◦ R0 has no fixed points. Hence R((x1 , x2 , x3 )) =
(x1 , −x2 , x3 ) + (v1 , v2 , v3 ), and using that R02 ◦ R2 = Id we get v1 = −ν and v3 = 0. Defining
R2 = R0 ◦ R, G =< R0 , R2 > and (d)(iii) holds. 2
Remark 4.2.2 The hypothesis in Theorem 4.2.2 of being maximal is crucial as shows the
CSF embedded surface S̃ = {(x, y, z) ∈ L3 : y 2 − z 2 = −1}, which is invariant under a
hyperbolic orthochronous screw motion with axis ` = span{(1, 0, 0)}.
As a matter of fact, and in a more general setting, any globally hyperbolic complete
spacetime admits a smooth Cauchy hypersurface, (see [2] for details).
4.3 The Global Geometry of singly and doubly periodic CMS surfaces 63
4.3 The Global Geometry of singly and doubly peri-

odic CMS surfaces
From Theorem 4.2.2, complete flat 3-dimensional spacetimes admitting a CMF (embedded)
surface are quotients L3 /G, where G is a discrete group of spacelike translations of rank
1 or 2. Therefore, we can restrict attention to CMS surfaces in L3 invariant by groups of
spacelike translations. By definition, a CMS surface S̃ in L3 is said to be singly periodic
if it is invariant by a (spacelike) translation T 6= 0 and S := S̃/ < T > is of finite type.
Likewise, S̃ is said to be doubly periodic if it is invariant by two independent (spacelike)
translations T1 and T2 . In the second case S := S̃/ < T1 , T2 > is a torus, and so it is of
finite type.
Throughout this section we always suppose that G ⊂ Iso(L3 ) is a discrete group of
spacelike translations of rank 1 or 2, S̃ ⊂ L3 is a CSS maximal surface invariant under G,
and S := S̃/G is an CMF surface in N := L3 /G. Up to a Lorentzian isometry, we will also
assume that the translations in G are horizontal, i.e., contained in {x3 = 0}..
By Proposition 5.1.1, the conformal support S̃0 of S̃ is conformally equivalent to C −
∪n∈N D̃n , where {D̃n : n ∈ N} are open discs with pairwise disjoint closures and no
accumulation in C. Therefore, the group G can be conformally interpreted as a discrete free
and proper group of translations in C, and the conformal support S0 of S is biholomorphic
to either C∗ − ∪kj=1 Dj or T − ∪kj=1 Dj , where k ∈ N, T is a conformal torus and {Dj : j =
1, . . . , k} are open discs with pairwise disjoint closures.
Let X̃ : S̃0 → L3 be a conformal maximal immersion which parameterizes S̃, and label
(g̃, φ˜3 ) its Weierstrass representation. Since T ∗ (φ̃3 ) = φ̃3 and g̃ ◦ T = g̃, for any T ∈ G,
these Weierstrass data can be pushed out on S0 . We denote by (g, φ3 ) the meromorphic
data on S0 associated to the induced maximal conformal immersion X : S0 → L3 /G. Up
to a Lorentzian isometry, we always assume that |g| ≤ 1 on S0 .
In the singly periodic case, g extends holomorphically to the ends {0, ∞} of S0 ≡
∗
C − ∪kj=1 Dj (that is to say, the tangent plane is well defined at these points). Since
the induced metric on S0 , namely ds2 = 41 (1 − |g|2 )2 | φg3 |2 , is complete, the same holds
for the metric | φg3 |2 . A classical result by Osserman [26] gives that the 1-form ω := φg3
has poles at the ends. This simply means that ω is meromorphic at the compact surface
S 0 := C − ∪kj=1 Dj . In the doubly periodic case, we establish the convention S 0 = S0 . At
this point we state the following lemma:
Lemma 4.3.1 If S̃ is singly periodic, then ω = φg3 has simple poles at the two ends of S0 .
Moreover, any end of S is asymptotic to a totally geodesic spacelike cylinder in L3 /G.
Proof : Let (D ≡ D(0, 1) ⊂ S 0 , z) be a once punctured conformal disc in S 0 centered at an

end of the surface. Up to a Lorentzian isometry we can suppose that the tangent plane at
the end is horizontal, that is to say, g(0) = 0. Write g(z) = z p , p ≥ 1, and put
+∞
!
X
j
ω= cj z dz, c−q 6= 0,
j=−q
on D. Reasoning by contradiction, suppose q ≥ 2.

Consider the universal covering D̃∗ ≡ {u ∈ C : Re(u) < 0} ⊂ S̃0 of D∗ := D − {0},
and write π : D̃∗ −→ D∗ , π(u) = eu , the covering projection.
Denote by π0 : L3 → {x3 = 0} ≡ C the orthogonal projection, and write X̃ : S̃0 →
L3 a conformal parameterization of the surface S̃. Using complex notation and following
equation (2.1), it is not hard to check that
−i c−q (1−q)u
π0 ◦ X̃(u) = e h(u) u ∈ D̃∗ ,
2(1 − q)
where h : D̃∗ → C is bounded and differentiable in D̃∗ and limu→∞ h(u) = 1 on the strips
{u ∈ D̃∗ : |Im(u)| < C}, C ∈ R+ .
∗
Let A : D̃ → R be a smooth branch of the argument A(u) := arg π0 (X̃(u)) , and
3π
observe that Aθ := limr→−∞ A(r + iθ) = 2
− arg (c−q ) + (q − 1)θ.
2πk 2π(k+1)
In particular, the graph S̃(k) := X̃({u ∈ D̃∗ : Im(u) ∈ [ q−1 , q−1 ]} is projected by
π0 over a planar region in {x3 = 0} containing, up to a compact subset, the complement
of a sector of arbitrarily small angle bisected by the half line {(x1 , x2 , 0) : arg (x1 , x2 ) =
3π
2
− arg (c−q )}, and this for any k ∈ Z. As S̃(k) ⊂ S̃ for any k ∈ Z, we contradict that S̃
is a graph and prove the first part of the lemma.
For the second one, use the same notation as above for q = 1 and obtain
−i c−1
X̃(u) = u, 0 + H(u),
2
where H : D̃∗ → R3 ≡ C×R is a bounded differentiable function in D̃∗ and limu→∞ H(u) =
(1, µ0 ) on the strips {u ∈ D̃∗ : |Im(u)| < C}, C ∈ R+ . Therefore, the surface X̃(D̃∗ )
is asymptotic, as Re(u) → −∞, to a horizontal plane, G is generated by a horizontal
translation T, and the end D∗ ⊂ S is asymptotic to a totally geodesic horizontal cylinder,
which concludes the proof. 2
The following corollary is an elementary consequence of Lemma 4.3.1.

4.3 The Global Geometry of singly and doubly periodic CMS surfaces 65
Corollary 4.3.1 Assume that S̃ is singly periodic. Then the convex hull of S̃ is either a
wedge W (if S has non parallel ends) or a slab (if S has parallel ends). In the first case,
the edge of W is parallel to the vectors in G and ∂(W ) is a graph with spacelike faces.
Moreover, S̃ is asymptotic to the faces of W. In the second one the slab is parallel to the
ends and to the vectors in G.
If S̃ is doubly periodic, its convex hull is a slab parallel to the plane generated by the
vectors in G.
Write ∂(S0 ) = ∪kj=1 γj , γj ≡ S1 , and let S0∗ be the mirror or double surface of S0 with
boundary ∂(S0∗ ) = ∪kj=1 γj∗ , γj∗ ≡ S1 (see Lemma 2.2.2). Let S ≡ S0 ∪ S0∗ the conformal
completion of S0 , that is to say, the Riemann surface without boundary obtained by gluing
analytically along γj and γj∗ , j = 1, . . . , k, the surfaces S0 and S0∗ . The natural mirror
involution J : S → S, J(p) := p∗ , fixes pointwise γj ≡ γj∗ ⊂ S, j = 1, . . . , k. Likewise, we
define S as the conformal completion of S 0 , and call also J the antiholomorphic involution
on S (recall that in the doubly periodic case S = S). Note that S is a compact Riemann
surface of genus k0 , where either k0 = k − 1 (singly periodic case) or k0 = k + 1 (doubly
periodic case). ¿From Lemma 2.2.2 the Weierstrass data of S0 extend to meromorphic data
on S satisfying J ∗ (φj ) = −φj , j = 1, 2, 3. The extension of the Weierstrass data of S0 to
S are still denoted by (g, φ3 ).
Theorem 4.3.1 (Analytical representation of CMF surfaces) Let S be a CMF sur-
face in L3 /G with Π1 (S) = G and k ≥ 1 singularities, where G is a discrete group of
horizontal translations of rank 1 or 2 acting freely and properly in L3 .
Label S0 , S and J the conformal support of S, the conformal completion of S0 and the
mirror involution in S, respectively. Let (g, φ3 ) denote the corresponding Weierstrass data
on S.
Then, the following assertions hold:
(i) g ◦ J = 1/ḡ, |g| < 1 on S0 − ∂(S0 ), g : γj → S1 is injective, for any component
γj ⊂ ∂(S0 ), and deg(g) = k.
(ii) The 1-forms φ1 , φ2 , φ3 satisfy: J ∗ (φj ) = −φj , j = 1, 2, 3, they are holomorphic in S

and they have no common zeroes on S. Furthermore, they are holomorphic in S if
and only if S is doubly periodic, and in the singly periodic case, they have at most
simple poles at the ends of S and no other poles. Finally, φ3 has no real periods on
S.
(iii) In the singly periodic case, the ends of S are asymptotic to flat horizontal cylinders
in L3 /G.
R
(iv) The translations in G are given by the vectors {Re γ
(φ1 , φ2 , φ3 ) : γ ∈ H1 (S, Z)}.
Proof : The proof of items (i), (ii) and (iii) is a direct consequence of the definition of the
Weierstrass representation, Lemma 2.2.2 and Lemma 4.3.1. Item (iv) simply express that
Π1 (S) = G. 2
The converse of the preceding theorem holds:
Theorem 4.3.2 (Construction of CMF surfaces) Let S and J : S → S be a com-

pact Riemann surface and an antiholomorphic involution. Suppose that the fixed point set
of J consists of k pairwise disjoint analytical regular Jordan curves γ1 , . . . , γk , and the do-
main S − ∪kj=1 γj consists of two biholomorphic components whose closures, namely S 0 and
∗
J(S 0 ) := S 0 , are topologically equivalent to either a sphere or a torus with k holes. In case
S 0 is a sphere with k holes, we also consider two points P, Q ∈ S 0 − ∪kj=1 γj and denote
S0 := S 0 −{P, Q}, and S := S−{P, Q}. As above, S0 ≡ S 0 and S ≡ S if S 0 is a torus with
k holes. Let S denote the quotient surface of S0 identifying each boundary circle γj with a
single point qj , where qn 6= qm provided that n 6= m and F := {qj : j = 1, . . . , k} ∩ S0 = ∅.
Let (g, φ3 ) be meromorphic data in S satisfying the conditions (i), (ii) and (iii) in
Theorem 4.3.1, and label G as the discrete group of translations generated by the vectors
in Theorem 4.3.1,(iv).
Then, G has either rank 1 (if S 0 is a sphere with holes) or rank 2 (if S 0 is a torus with
holes). Moreover, the maximal immersion
Z p
3
X : S0 → L /G, X(p) := Re (φ1 , φ2 , φ3 ), (4.2)
p0
where p0 ∈ S0 and (φ1 , φ2 , φ3 ) are given accordingly to equation (2.1), is well defined and
gives a complete maximal embedding of S in L3 /G with k singularities.
R
Proof : Since J∗ (γj ) = γj and J ∗ (φi ) = −φi , we infer that Re γj φi = 0, for any j and
R
i. Therefore, the real periods of (π1 , φ2 , φ3 ) are the vectors in {Re γ (φ1 , φ2 , φ3 ) : γ ∈
H1 (S, Z)}, and since H1 (S, Z) has either rank 1(if S 0 is a sphere with holes) or rank 2 (if
S 0 is a torus with holes), G has either rank 1 or 2 as well.
R
Obviously, the immersion X : S0 → L3 /G, X := Re (φ1 , φ2 , φ3 ), is well defined, and
from Lemma 2.2.2 it leads to an immersion of S in L3 with k conelike singularities. The
completeness of the metric ds2 induced by X on S0 is an elementary consequence of the
expression ds2 = 41 (1 − |g|2 )2 | φg3 |2 . Indeed, just take into account that either S0 is compact
or otherwise φg3 has a single pole at the ends of S0 . Finally, Proposition 4.1.1 yields that
the immersion of S in L3 /G induced by X is an embedding, which concludes the proof. 2
4.4 Examples of CMF surfaces in quotients of L3 67
4.4 Examples of CMF surfaces in quotients of L3

In this section we present some families of singly and doubly periodic highly symmetric
maximal surfaces. These surfaces are given in terms of a compact Riemann surface S and
meromorphic data (g, φ3 ) on S satisfying the hypothesis in Theorem 4.3.2. Moreover, they
provide examples of all the cases described in Theorem 4.2.2.
Family I: Singly periodic maximal surfaces with non parallel ends and 1 sin-
gularity in the quotient.
Consider the following data:
S = C, J : C → C, J(z) = 1/z̄ S0 = D(0, 1) − {b, −b} b ∈]0, 1[

zdz
g(z) = z, φ3 =
(z 2 − b2 )(b2 z 2 − 1)
Then the immersion given by Equation (4.2) is a CMF surface in L3 /G, where G is the
group generated by the translation T of vector
−π(b2 − 1)(b4 − 1)
v = Re(2πiRes(Φ, b)) = ( , 0, 0)
2b
Moreover, the antiholomorphic transformation of S0 given by A(z) := −z̄ satisfies
∗
A (φ1 , φ2 , φ3 ) = (φ̄1 , −φ̄2 , φ̄3 ). Thus, the lifted surface in L3 , S̃, is invariant by the isometry
Z A(0)
R(x1 , x2 , x3 ) = (x1 , −x2 , x3 ) + Re Φ = (x1 , −x2 , x3 )
0
and so the quotient of S̃ by the group generated by the isometry R1 := T ◦ R is a CMF

surface of type (c)(i) in Theorem 4.2.2 with one singularity.
A surface in this family has been illustrated in Figure 4.1.
Family II: Singly periodic maximal surfaces with parallel ends with 2 singu-
larities in the quotient.
Consider the data:
2 (z − a)(z − b)
S = {(z, w) ∈ C : w2 = }, J : S → S, J(z, w) = (1/z̄, 1/w̄)
(az − 1)(bz − 1)
Figure 4.1: Example of singly periodic maximal surface of Scherk type.
S0 = {(z, w) ∈ S : |z| ≤ 1, z 6= 0} b < a < 1, a > 0, b 6= 0

dz
g(z, w) = z φ3 =
w(az − 1)(bz − 1)
Then the immersion given by Equation (4.2) is a CMF surface in L3 /G, where G is
generated by the translation T of vector
−π −iπ
v = Re(2πiRes(Φ, 0)) = Re(( √ , √ , 0))
ab ab
Thus, the translation vector is in the direction of the x1 −axis in case b > 0 or in the
direction of the x2 −axis if b < 0.
Moreover, if we choose b = −a, the transformations of S0 given by A0 (z, w) = (−z̄, w̄),
A1 (z, w) = (z̄, w̄) and A2 (z, w) = (−z, w), lift to the following isometries in L3 :
Z A0 (a)
v
R0 (x1 , x2 , x3 ) = (−x1 , x2 , −x3 ) + Re Φ = (−x1 , x2 , −x3 ) + ,
a 2
R1 (x1 , x2 , x3 ) = (−x1 , x2 , x3 ) and

Z A2 (a)
v
R2 (x1 , x2 , x3 ) = (x1 , x2 , −x3 ) + Re Φ = (x1 , x2 , −x3 ) +
a 2
The quotient of the lifted surface S̃ in L3 by the groups hR0 i, hT ◦ R1 i and hR2 i give
examples of CMF surfaces corresponding to cases (b)(i), (c)(i) and (d)(i) in Theorem 4.2.2,
respectively .
A surface in this family has been illustrated in Figure 4.2 (right).
Family III: Doubly periodic maximal surfaces with 2 singularities in the

quotient.
Figure 4.2: Doubly and singly maximal surfaces
Consider the data:
2 (z 2 − a21 )(z 2 − a22 )

S = {(z, w) ∈ C : w2 = }, a1 , a2 ∈ R∗ , a1 6= a2
(a21 z 2 − 1)(a22 z 2 − 1)
J : S → S, J(z, w) = (1/z̄, 1/w̄), S0 = {(z, w) ∈ S : |z| ≤ 1}
zdz
g(z, w) = z, φ3 =
w(a21 z 2 − 1)(a22 z 2 − 1)
Then, Equation (4.2) leads to a CMF surface in L3 /G, where G is generated by two
R
horizontal translations T1 and T2 of vectors v1 and v2 . Namely, vi = 2Re γi Φ, i = 1, 2,
where γ1 (resp. γ2 ) is a simple arc in S0 joining a1 and −a1 , (resp. a1 and a2 ). Elementary
computations show that in fact v1 = (0, λ, 0), and v2 = (µ, 0, 0), for suitable λ, µ ∈ R∗ .
The surface on the left in Figure 4.2 is an example of this family of surfaces.
As in the previous cases, the transformations of S 0 , A0 (z, w) := (z̄, −w̄), A1 (z, w) :=
(z̄, w̄) and A2 (z, w) := (−z, −w) induce isometries of L3 , R0 , R1 and R2 resp., leaving the
lifted surface S̃ invariant. The expression for these isometries is the following:
R0 (x1 , x2 , x3 ) = (x1 , −x2 , −x3 ), R1 (x1 , x2 , x3 ) = (−x1 , x2 , x3 ) and
R2 (x1 , x2 , x3 ) = (x1 , x2 − λ/2, −x3 )

Thus, the corresponding quotients of S̃ by the groups hT2 ◦ R0 , T1 i, hT1 ◦ R1 , T2 i, hR2 , T2 i
and hT2 ◦ R0 , R2 i provide examples of CMF surfaces of type (b)(ii), (c)(ii), (d)(ii) and
(d)(iii) respectively. A surface in this family has been illustrated in Figure 4.2 (left).
Chapter 5
Parabolicity criteria for CMS

surfaces
In this chapter we are going to study the underlying conformal structure of CMS orientable
surfaces in flat complete 3-dimensional Lorentzian spaces. Our main result asserts that
any such surface is parabolic in the sense that bounded harmonic functions are uniquely
determined by their values on the singular set. We extend this analysis to proper maximal
(resp., minimal) graphs in L3 (resp., R3 ) over starlike regions.
In the following definition we fix the notion of conformal support of a maximal surface
with conelike singularities.
Definition 5.1.1 Let X : M → N be an orientable maximal immersion with (isolated)

conelike singularities at F = {qn : n ∈ Λ}, Λ ⊂ N. Lemma 2.2.2 gives that M − F is
conformally equivalent to a Riemann surface R \ ∪j∈Λ Dj where Dj are pairwise disjoint
analytical closed discs bounded by the curves γj corresponding to the singularities.
By definition, the Riemann surface (with boundary) M0 := R \ ∪j∈Λ Int(Dj ) is said to
be the conformal support of the maximal immersion. Observe that ∂(M0 ) = ∪j∈Λ γj .
If S ⊂ N is an embedded orientable maximal surface, S0 will denote its conformal
support.
For instance, if S ⊂ L3 is CMS embedded surface (∂(S) = ∅) with singular set F =

{qn : n ∈ Λ}, Λ ⊂ N, Lemma 2.2.3 implies that S is simply connected. Moreover Lemma
2.2.2 and [19] gives that S − F is conformally equivalent to a planar circular domain
Ω − ∪n∈Λ Dn , where Ω is either C or the unit disc D and {Dn : n ∈ Λ} are pairwise
disjoint closed discs with no accumulation in Ω. Following Definition 5.1.1, the conformal
support of S is S0 = Ω − ∪n∈Λ Int(Dn ).
71
72 Parabolicity criteria for CMS surfaces
At this point we recall the notion of parabolicity for a Riemann surface R with boundary
(see [18], [1]):
R is a parabolic surface with boundary if the only bounded harmonic function

f vanishing on ∂(R) is the constant function f = 0. This is equivalent to say
that R admits a proper positive superharmonic function.
If ∂(R) = ∅, parabolicity simply means that positive superharmonic functions are constant.
Anyhow, closed regions of parabolic Riemann surfaces with (possibly empty) boundary are
parabolic [18],[1].
A Riemann surface R is said to be of finite conformal type if ∂(R) = ∅ and R is

conformally equivalent to a compact Riemann surface minus a finite number of points.
Riemann surfaces of finite type are parabolic, and so, by Theorem 4.2.2, we deduce that
the conformal support of any CMF orientable surface in a 3-dimensional complete flat
spacetime N is parabolic.
In this section we will show that this remains true also for CMS surfaces. First, the
following proposition shows that the conformal support of the universal covering of any
CMF surface in N is parabolic is a circular domain.
Proposition 5.1.1 Let X : M → N be a CMF immersion and denote by M̃ the universal

covering of M.
Then, the conformal support M̃0 of M̃ is conformally equivalent to C − ∪n∈Λ Dn , where
Λ ⊂ N, and {Dn : n ∈ Λ} is a collection of open discs with pairwise disjoint closures and
no accumulation in C.
Proof : Up to a two-sheeted covering we can suppose M orientable.

If M has k singular points, the boundary ∂(M0 ) of its conformal support consists of k
pairwise disjoint analytical circles. Label M as the Riemann surface with empty boundary
obtained by filling up the holes of M0 , that is to say, by attaching k discs to M0 in an
analytical way.
From Theorem 4.2.2, M is conformally equivalent to either C − {0}, T, KB or RP2 −
{[0]} ≡ (C − {0})/ < I >, where T is a conformal torus, KB is a conformal Klein bottle
and I : C − {0} → C − {0} is given by I(z) = −1/z.
Therefore, the conformal universal covering of M is C, and so M̃0 is conformally equiv-
alent to a planar domain bounded by a countable family of analytic Jordan curves with
no accumulation in C. By the results in [19], M̃0 is in fact conformally equivalent to a
circular domain as stated in the proposition. 2
The following theorem and subsequent corollaries are devoted to extend the previous
result to a wider family of surfaces.
Theorem 5.1.1 Let S ⊂ L3 be a proper maximal surface with (isolated) singularities and
assume that S is a graph over a closed starlike region in the plane {x3 = 0} centered at
the origin.
Then its conformal support, S0 , is a parabolic surface with boundary.
Before going on, we emphasize that the number of singular points in S could be infinite.
Proof : Take a conformal parameterization X : S0 → L3 such that X(S0 ) = S. Up to a

vertical translation, suppose that S contains the origin. Then, Lemma 4.0.2 gives that
n : S0 → R, n(p) = kX(p)k, is a positive proper function on S0 . Consider the compact set
K = {p ∈ S0 : 0 ≤ kX(p)k ≤ 2}, and note that S0 is parabolic if and only if S0 − int(K) is
parabolic (see [1], [18] for details). Therefore, it suffices to check that the proper positive
function h : S0 − Int(K) → R given by
h(p) = log(kX(p)k2 )
is superharmonic. Take an isothermal parameter z = u + iv on S0 . The corresponding con-

formal parameterization X(u, v) satisfies < Xu , Xv >= 0 and < Xu , Xu >=< Xv , Xv >=
λ2 . Furthermore, from the maximality X is harmonic, and so:
< X, Xu >2 + < X, Xv >2 λ2

∆h := huu + hvv = −4 −
< X, X >2 < X, X >
Since the plane generated by {Xu , Xv } is spacelike and these two vectors determine an
isothermal basis of it, we get
1
X= (< X, Xu > Xu + < X, Xv > Xv ) − < X, N > N,
λ2
where N is the normal vector.
Hence, < X, X >= λ12 (< X, Xu >2 + < X, Xv >2 ) − < X, N >2 , which proves that
< X, N >2
∆h = −4λ2 ≤0
< X, X >2
and concludes the proof.

2
74 Parabolicity criteria for CMS surfaces
Corollary 5.1.1 . Let X : M → N be a CSS orientable maximal immersion with con-

formal support M0 .
Then, M0 is a parabolic Riemann surface.
Proof : Denote by M̃ the universal covering of M, and call X̃ : M̃ → L3 a lifting of X.

Label M̃0 as the conformal support of X̃, and observe that M0 is parabolic provided that
M̃0 is parabolic. For this reason it suffices to deal with the parabolicity problem of CMS
surfaces in L3 , and the corollary follows from Lemma 2.2.3 and Theorem 5.1.1.
2
Corollary 5.1.2 Any minimal graph S̃ ⊂ R3 over a closed starlike planar region with C 1
boundary is parabolic.
Proof : Let X ≡ (X1 , X2 , X3 ) : S̃0 → R3 be a conformal minimal embedding such that

X(S̃0 ) = S̃. Since S̃ is simply connected, the conjugate harmonic function X3∗ of X3 is well
defined. Then Y := (X1 , X2 , X3∗ ) : S̃0 → L3 is a maximal graph over the same region as
S̃. Since S̃0 itself is the conformal support of Y (S̃0 ), it must be parabolic from Theorem
5.1.1. 2
Bibliography
[1] L. V. Ahlfors and L. Sario: Riemann surfaces. Princeton Univ. Press, Princeton, New
Jersey, 1960.
[2] A. N. Bernal, M. Sánchez:On smooth Cauchy hypersurfaces and Geroch’s splitting

theorem. Comm. Math. Phys. 243 (2003), no. 3, 461–470.
[3] J. E. Marsden and F. J. Tipler.: Maximal hypersurfaces and foliations of constant

mean curvature in general relativity. Phys. Rep., Vol. 66 (1980), no. 3, 109-139.
[4] R. Bartnik and L. Simon: Spacelike hypersurfaces with prescribed boundary values and
mean curvature. Comm. Math. Phys., Vol. 87 (1982/83), 131-152.
[5] V. A. Klyachin and V.M. Miklyukov.: Geometric structures of tubes and bands of
zero mean curvature in Minkowski space. Annales Academia Scientiarum Fennicae
Mathematica, 28 (2003) 239-270
[6] M. Umehara and K. Yamada: Maximal surfaces with singularities in Minkowski space.
Preprint.
[7] K. Ecker.: Area maximizing hypersurfaces in Minkowski space having an isolated sin-
gularity . Manuscripta Math., Vol. 56 (1986), 375-397.
[8] F. J. López, R. López and R. Souam.: Maximal surfaces of Riemann type in Lorentz-
Minkowski space L3 . Michigan J. of Math., Vol. 47 (2000), 469-497.
[9] P. Collin, R. Kusner, W.H. Meeks III, and H. Rosenberg: The topology, geometry and
conformal structure of properly embedded minimal surfaces. Preprint.
[10] W. Goldman and G. A. Margulis: Flat Lorentz 3-manifolds and cocompact Fuchsian
groups. Contemp. Math. 262, Amer. Math. Soc., Providence, 2000, 135-145.
75
76 Bibliography
[11] G. A. Margulis: Complete affine locally flat manifolds with a free fundamental group.
J. Soviet Math. 134, (1987), 129-134.
[12] E. Calabi: Examples of the Bernstein problem for some nonlinear equations. Proc.
Symp. Pure Math., Vol. 15, (1970), 223-230.
[13] S. Y. Cheng and S. T. Yau: Maximal space-like hypersurfaces in the Lorentz-Minkowski

spaces. Ann. of Math. (2), Vol. 104 (1976), 407-419.
[14] A. A. Klyachin and V. M. Miklyukov.: Existence of solutions with singularities for

the maximal surface equation in Minkowski space. Russian Acad. Sci. Sb. Math. 80
(1995), 87-104.
[15] T. Barbot and A. Zeghib: Group actions on Lorentz spaces, Mathematical aspects: a
survey. ”50 years of the Cauchy problem”, edited by P. Chrusciel and H. Friedrich, to
appear.
[16] F. J. López and J. Pérez: Parabolicity and Gauss map of minimal surfaces. Indiana
Univ. Math. J. 52 (2003), no. 4, 1017-1026.
[17] V. Charette, T. Drumm., W. Goldman and M. Morrill: Complete flat affine and
Lorentzian manifolds. Geom. Dedicata 97 (2003), 187-198.
[18] A. Grigor’yan: Analytic and geometric background of recurrence and non-explosion of

the Brownian motion on Riemannian manifolds. Bull. Amer. Math. Soc 36 (1999),
135-249.
[19] Z.-X. He and O. Schramm: Fixed points, Koebe uniformization and circle packings.
Ann. of Math. (2) 137 (1993), no. 2, 369-406.
[20] Hoffman, D., Meeks, W.H., III: The strong half space theorem for minimal surfaces.
Inventiones Math., Vol. 101 (1990), 373-377.
[21] A. Huber: On subharmonic functions and differential geometry in the large. Comment.
Math. Helv., 32, (1957), 13-72.
[22] O. Kobayashi: Maximal surfaces with conelike singularities. J. Math. Soc. Japan 36
(1984), no. 4, 609-617.
[23] F. J. López, I. Fernández and R. Souam: The space of complete embedded maximal
surfaces with isolated singularities in the 3-dimensional Lorentz-Minkowski space L3 .
Preprint. ArXiV e-print archive math.DG/0311330 v1.
Bibliography 77
[24] G. Mess: Lorentz spacetimes of constant curvature. Preprint Institut des Hautes
Études Scientifiques (1990).
[25] B. O’Neill: Semmi-riemannian geometry. Academic Press (1983).
[26] R. Osserman: A survey of minimal surfaces. Dover Publications, New York, second
edition, 1986.
[27] J. Wolf: Spaces of Constant curvature. McGraw-Hill, New York (1967).
[28] R. Schoen.: Uniqueness, symmetry and embeddedness of minimal surfaces. J. Differ-

ential Geometry, 18, (1983), 701-809.
[29] A. A. Klyachin.: Description of the set of singular entire solutions of the maximal
surface equation. Sbornik Mathematics, 194 (2003), no. 7, 1035-1054
[30] L. Jorge, W.H. Meeks III.: The topology of complete minimal surfaces of finite total
Gaussian curvature. Topology, vol 2 (1983), 203-221.
[31] H. M. Farkas, I. Kra.: Riemann surfaces. Graduate Texts in Math., 72, Springer
Verlag, Berlin, 1980.
[32] D. Gilbarg and N. S. Trudinger.: Elliptic partial differential equations of second order.
Springer-Verlag, (1977).

Dea

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dea

Uploaded by

Copyright:

Available Formats

Maximal surfaces with conelike

3 Maximal surfaces of finite type in L3 23

4 Maximal surfaces of finite type in L3 /G 53

5 Parabolicity criteria for CMS surfaces 71

1.1 The Lorentzian catenoid and a Riemann type example. . . . . . . . . . . 8

2.1 An upward pointing conelike singularity. . . . . . . . . . . . . . . . . . . . 19

3.1 A catenoidal and a planar end. . . . . . . . . . . . . . . . . . . . . . . . . 24

4.1 Example of singly periodic maximal surface of Scherk type. . . . . . . . . . 68

A maximal hypersurface in a Lorentzian manifold is a spacelike hypersurface with zero

Theorem[Calabi [12], Cheng and Yau [13]] A complete maximal hyper-

It is therefore meaningless to consider global problems on maximal and everywhere

Figure 1.1: The Lorentzian catenoid and a Riemann type example.

Theorem I: The space Gn of all the complete maximal graphs having n + 1

(A) The classification of spacetimes L3 /G containing complete embedded maximal surfaces

Theorem II.A Let N := L3 /G, G 6= {Id}, be a complete flat spacetime, and

(a) N orientable and orthochronous. In this case, either G is generated by a

torus, or G =< R0 , R2 > and S is a Klein bottle, where R2 ((x1 , x2 , x3 )) =

The paper is lead out as follows:

2.1 Notation and main definitions

• C is the extended complex plane C ∪ {∞}.

• D is the unit disc, D = {z ∈ C : |z| < 1}.

• Given N a Riemann surface and k a positive integer, we call

Divk (N ) = {D : D is an integral multiplicative divisor on N of degree k}.

• M will denote a differentiable surface. For convenience we allow ∂(M) 6= ∅, and in

• L3 is the three dimensional Lorentz-Minkowski space (R3 , ds2 = dx2 + dy 2 − dz 2 ) and

• N will denote a flat 3-dimensional Lorentzian manifold (i.e., a 3-dimensional differ-

• H2 is the hyperbolic sphere in L3 of constant intrinsic curvature −1, H2 = {(x1 , x2 , x3 ) ∈

• The stereographic projection σ for H2 is defined as follows:

• Cp is the light cone at a point p ∈ L3 , ie. Cp = {q ∈ L3 : ||q − p|| = 0}.

• Iso(L3 ) is the group of affine isometries of L3 .

• Iso↑+ (L3 ) = Iso+ (L3 ) ∩ Iso↑ (L3 ).

Remark 2.1.1 A vector v ∈ R3 − {0} is said to be spacelike, timelike or lightlike if

Remark 2.1.3 We know that G↑+ (resp., G+ , G↑ ) is a subgroup of index k+ ↑

Definition 2.1.1 An immersion X : M −→ N is spacelike if the tangent plane at any

If N = L3 /G, where G is a (possibly trivial) group of translations acting freely and

Definition 2.1.2 Let M be a differential surface. A maximal immersion X : M −→ N

Using isothermal parameters, M can be endowed with a conformal structure. In the

2.2 Maximal immersions with singularities

Definition 2.2.1 Let N be a flat 3-dimensional spacetime, and let X : M → N be a

Theorem 2.2.1 (Weierstrass Representation) If X : M −→ L3 is a maximal im-

• there exists a holomorphic 1-form φ3 on M such that

The induced Riemannian metric ds2 on M is given by

Since M is spacelike, then |g| 6= 1 on M, and up to a Lorentzian isometry, we always

Lemma 2.2.1 (Local structure of singularities) Let X : D → L3 be a spacelike im-

Proof : Up to an ambient isometry we can assume that Π = {x3 = 0} and P0 coincides

We are interested in embedded singularities, specially in the maximal case. We say

Figure 2.1: An upward pointing conelike singularity.

Lemma 2.2.3 (Global behaviour of spacelike embedded singularities) Let X : M →

Proof : Let Π = {x3 = 0} and π the canonical projection π : L3 → Π. Then, it is enough

is a local homeomorphism. To finish, it suffices to check that π ◦ X has the path-lifting

When the number of singularities is finite, the answer is yes [23].

An interesting point is to determine the asymptotic behavior of complete embedded

Proof : Reason by contradiction, and suppose that S is non flat.

Figure 2.2: The surface S and the catenoids Ct and Cts .

theorem for assuring that S − {x3 ≥ k0 } = ∅. 2

Maximal surfaces of finite type in

Figure 3.1: A catenoidal and a planar end.

3.1 Existence and Uniqueness of CMF graphs

If G = {(x, u(x)) : x ∈ R2 } is an entire maximal graph with a finite number of

|u(x) − u(y)| < |x − y|, for all x, y ∈ R2 , x 6= y.

Theorem 3.1.2 Let N be a compact genus n Riemann surface, and let J : N → N be

where Ω is topologically equivalent (and so conformally) to an open planar circular domain.

(1) g is a meromorphic function on N of degree n + 1, |g| < 1 on Ω and g ◦ J = 1/g on N,

By definition V() := {(v, D) : D ∈ pk (U ) ⊂ Divk (Ω(v)), v ∈ V ()} is a neighborhood

Υ : U () × W () → W(), Υ = p|U ()×W () (3.4)

Ĥ : V () × ξ(pn (U )) → V (0 ) × W (0 )

dist (p, C0 ) = fθ (t) > fθ (δ) > := dist (∂(D), C0 ) > 0,