Loop Weierstrass Representation: Abstract

LOOP WEIERSTRASS REPRESENTATION
THOMAS RAUJOUAN, NICK SCHMITT, AND JONAS ZIEFLE
Abstract. We introduce the Loop Weierstrass Representation for minimal surfaces in Euclidean space
and constant mean curvature 1 surfaces in hyperbolic space by applying integral system methods to the
Weierstrass and Bryant representations. We unify associated families, dual surfaces and Goursat transfor-
mations under the same holomorphic data, we introduce a simple factor dressing for minimal surfaces, and
arXiv:2411.04626v1 [math.DG] 7 Nov 2024
we compute and classify various examples.
Introduction
Minimal surfaces in Euclidean 3-space and constant mean curvature 1 (CMC 1) surfaces in hyperbolic
3-space can be produced from holomorphic data by the Weierstrass and Bryant representations, respec-
tively [2, 17]. In this article, we introduce the Loop Weierstrass Representation (LWR), a single framework
that allows for the construction of both types of surfaces. It recovers the Weierstrass and Bryant repre-
sentations as special cases of a wider class of holomorphic representations that arise by introducing a loop
parameter (originating from associated families) and a gauge freedom in the holomorphic data. Consequently,
all questions posed for the Weierstrass and the Bryant representations – such as closing, end behavior, sym-
metries, total curvature – can be covered in the single framework of the LWR. This unification leads to some
structural insights, technical advantages and new aspects that we expose in the present paper:
‚ The LWR produces various families of related surfaces from the same holomorphic data. It covers
the classical associated family and extends it to a 4-real-parameter family (figure 2). It also covers
the Goursat transformations [9] as well as the dual CMC 1 surfaces introduced by Umehara and
Yamada [15]. In the LWR, all these transformations arise after solving one single ODE, whether the
target space is Euclidean or hyperbolic.
‚ The LWR provides a suitable framework for simple factor dressing, an integrable system transfor-
mation that produces new surfaces from old ones. It adds planar or horospherical ends to a given
surface while preserving its periods (figures 1, 3, 4 and 5).
‚ In the LWR, catenoids and surfaces with similar ends can be constructed from potentials that are
locally gauge equivalent to potentials with simple poles. This brings the theory of Fuchsian systems
into the construction of minimal and CMC 1 surfaces, a strategy that has been fruitful for the DPW
representation [6].
‚ An LWR potential for a surface with symmetries can be pushed down by a covering map to a
potential on a simpler surface. For example, n-noids with Platonic symmetries can be constructed
with an LWR potential on a three-punctured CP 1 , and the Schwarz P -surface on a four-punctured
CP 1 .
Holomorphic Representation of Surfaces

The representation of surfaces as immersions is based on their first and second fundamental forms, encod-
ing the intrinsic and extrinsic geometry of the surface, respectively. The compatibility of the fundamental
forms is given by the Gauss-Codazzi equations. By Bonnet’s theorem, immersions can locally be constructed,
provided the Gauss-Codazzi equations are satisfied.
When studying isometric constant mean curvature H (CMC H) immersions of surfaces into spaces of
constant sectional curvature k, the Lawson correspondence [12] states that the Gauss-Codazzi equations for
two immersions are the same if the quantity H 2 ` k is the same for both. Thereby minimal (i.e. CMC 0)
surfaces in Euclidean space E3 (i.e. k “ 0) correspond to CMC 1 surfaces in hyperbolic space H3 (i.e.
k “ ´1). In the Weierstrass and Bryant representations, the correspondence becomes locally explicit in
1
(a) Side (b) Front (c) Top (d) Back
Figure 1. Simple factor dressing of the doubly-wrapped catenoid as in example 8.9 with
pp, qq “ p1, 1q and pu, ℓq “ p 12 , 1q.
terms of an sl2 C-valued holomorphic one-form η of rank 1 that encodes the first and second fundamental
forms. Considering two different integration procedures for η yields two different holomorphic null-curves.
One represents a minimal immersion in E3 and is obtained by solving the ODE dψ “ η for ψ in sl2 C, the
other one gives a CMC 1 immersion in H3 and is obtained by solving the ODE dΨ “ Ψη for Ψ in SL2 C.
Motivation for the LWR

In the following we explain how the LWR appears as a natural extension of the Weierstrass and Bryant
representation by introducing an associate family parameter and a gauge freedom.
Firstly, we note that scaling the Hopf differential of a CMC 1 immersion in H3 by a complex parameter
2
λ and the metric by |λ| gives new solutions of the Gauss-Codazzi equation. The corresponding family of
CMC 1 immersions is called an associated family. For the Bryant representation this amounts to scaling the
potential
η ÞÝÑ ζλ :“ λη
and considering for the associated family of null-curves the ODE
dΨλ “ Ψλ ζλ .
Secondly, we observe that gauging the null-curve by an SL2 C-valued holomorphic function z ÞÑ gpzq, that
is
(0.1) Ψλ ÞÝÑ Φλ :“ Ψλ g,
changes the potential by
ζλ ÞÝÑ ξλ :“ g ´1 ζλ g ` g ´1 dg.
We thus arrive at the ODE
dΦλ “ Φλ ξλ
for a potential ξλ :“ λα ` β with α an sl2 C-valued holomorphic one-form of rank one. Provided that the
initial value for Ψ0 has been set to the identity, then Φ0 “ g and, from (0.1), one can read off the null curve
Ψ as
Ψ “ Φ1 Φ´1
0 .
Furthermore, any frame of the form Φλ1 Φ´1 λ0 is a holomorphic null-curve and therefore produces a CMC 1
surface. We call this representation of CMC 1 immersions the “Loop Weierstrass Representation” (LWR)
and Φλ the “LWR frame”. The LWR possesses a gauge freedom since gauging the LWR frame by Φλ ÞÑ Φλ h
leaves the null-curve unchanged.
Null-curves ψ for minimal surfaces in Euclidean space can be obtained from the LWR frame Φλ by taking
a logarithmic derivative
9 ´1 q|λ“λ ,
ψ :“ pΦΦ 0
2
where the partial derivative with respect to λ is denoted by the dot. This is reminicent of the transition
from the Lie group SL2 C to the Lie algebra sl2 C. As an interpretation for this formula, minimal surfaces in
Euclidean space can be seen as a blow-up limit of CMC 1 surfaces in hyperbolic space (Proposition 4.7).
Structure of the Paper

‚ Section 1 provides the background for studying minimal immersions in Euclidean space and CMC 1
immersions in hyperbolic space and introduces LWR.
‚ In section 2, gauge freedom for the LWR is explained and used in order to obtain two standard forms
for LWR potentials.
‚ As a follow up, section 3 focuses on the local geometric data of surfaces in terms of the LWR, that
is the first and second fundamental forms and the Gauss map.
‚ Section 4 studies how a surface changes when changing the loop parameters pλ0 , λ1 q and the initial
condition of the LWR frame. This yields associated families, rigid motions and as a new aspect, we
see that holomorphic dressing induces Goursat transformation.
‚ The global aspects of closing a surface in the LWR and surfaces with symmetry are discussed in
section 5.
‚ In section 6 we describe surfaces whose first fundamental form is invariant under rotations. This
provides a classification of the well-known examples of Enneper surfaces and catenoids.
‚ Section 7 is dedicated to n-noids and the classification of irreducible trinoids is recovered in the
context of the LWR and Fuchsian systems.
‚ Finally, in section 8, we introduce simple factor dressing for minimal surfaces in Euclidean space and
CMC 1 surfaces in hyperbolic space. Note that simple factor dressing is treated without recourse to
the Birkhoff splitting.
Contents
1 The Loop Weierstrass Representation 3
2 Gauge freedom and standard forms of LWR potentials 6
3 Local description in the LWR 9
4 Related surfaces 12
5 Symmetries and closing 15
6 Intrinsic surfaces of revolution 18
7 n-noids 23
8 Simple factor dressing 27
1 The Loop Weierstrass Representation

In this section, we introduce the Loop Weierstrass Representation (LWR) for constructing minimal immer-
sions into Euclidean space and CMC 1 immersions into hyperbolic space. LWR relies on the Lie group SL2 C
and its Lie algebra sl2 C, so we first recall in section 1.1 how to describe Euclidean and hyperbolic spaces in
a matrix model. We recall some differential geometry definitions in Section 1.2 and state the constructions
of Weierstrass [17] and Bryant [2] that produce minimal and CMC 1 surfaces out of holomorphic null curves.
In Section 1.3, we come to the LWR itself and show how it can build families of null curves from a given
holomorphic connection, and can thus induce families of minimal and CMC 1 surfaces.
1.1 Ambient spaces

Minkowski space R1,3 is identified with the real 4-dimensional vector space H2 of 2-by-2 Hermitian matrices
equipped with the Lorentzian inner product
1
(1.1) xx, yy :“ ´ trpxp
y q, x, y P H2
2
3
where the hat denotes the adjugate matrix, i.e.:
ˆ ˙ ˆ ˙
d ´b a b
(1.2) x
p“ if x “ .
ć a c d
The Lorentzian norm is then given by the determinant:
(1.3) ∥x∥2 “ ´ det x.
Under this identification, hyperbolic space is
H3 “ tx P H2 | det x “ 1, tr x ą 0u “ tF F ˚ | F P SL2 Cu ,
and Euclidean space is identified with the tangent space of H3 at the identity matrix I, which is nothing else
but the trace-free 2-by-2 Hermitian matrices:
E3 “ TI H3 “ i ¨ su2 .
An orientation of E3 is chosen so that px, y, x ˆ yq is positively oriented, where
i
(1.4) x ˆ y :“ ´ rx, ys , x, y P E3 .
2
The group SL2 C acts isometrically and transitively on H3 via
(1.5) x ÞÑ V xV ˚
where V P SL2 C and x P H3 . This action extends the orientation from E3 “ TI H3 to the entire H3 . The
group SU2 ⋉ E3 acts isometrically and transitively on E3 via
x ÞÑ U xU ´1 ` T
where U P SU2 , T P E3 and x P E3 .
1.2 Null curves and minimal or CMC 1 immersions

For a given function f defined on a Riemann surface Σ with local coordinate z “ xìy, we write fx , fy , fz ,
and fz for the x, y, z and z derivatives of f , respectively. In this paper, all the maps defined on a Riemann
surface Σ are assumed to be real-analytic, but not necessarily holomorphic. Let X3 denote Euclidean space
E3 or hyperbolic space H3 .
Definition 1.1. An immersion f : Σ Ñ X3 is conformal if xfz , fz y “ 0. In such case, the first fundamental
form (or metric) reads
2
(1.6) ds2 :“ xdf, df y “ 2 xfz , fz y |dz| .
The unit normal vector N of a conformal immersion f at a point p “ f pz0 q is the unit vector such that
pfx , fy , N q is a positively oriented orthogonal basis of Tp X3 .
The Hopf differential Qdz 2 and the mean curvature H of a conformal immersion at a point are
defined via the second fundamental form:
(1.7) ´ xdf, dN y “: Qdz 2 ` Hds2 ` Qdz 2 .
A conformal immersion into Euclidean space is minimal if H ” 0. A conformal immersion into hyperbolic
space is CMC 1 if H ” 1. An umbilic of f is a point z0 P Σ such that Qpz0 q “ 0. If H is constant and
Q is constantly vanishing, the immersion is flat. The Gauss-Codazzi equations for minimal surfaces in
2
Euclidean space and CMC 1 surfaces in hyperbolic space are the same. Writing ds2 “ 4e2ω |dz| , they are
2
(1.8) |Q| “ 4e2ω ωzz , Qz “ 0.
Definition 1.2. A map ψ : Σ Ñ sl2 C (resp. Ψ : Σ Ñ SL2 C) is a holomorphic null curve if it is holomor-
phic with nowhere vanishing differential, and if det dψ (resp. det dΨ) is constantly vanishing.
Theorem 1.3 (Weierstrass [17]). Let ψ be a holomorphic null curve into sl2 C . Then ψ ` ψ ˚ is a conformal,
minimal immersion into Euclidean space. Moreover, any conformal, minimal immersion can locally be
obtained this way.
4
Theorem 1.4 (Bryant [2]). Let Ψ be a holomorphic null curve into SL2 C. Then Ψ˚ Ψ is a conformal, CMC
1 immersion into hyperbolic space. Moreover, any conformal, CMC 1 immersion can locally be obtained this
way.
1.3 The Loop Weierstrass Representation
We first introduce the Loop Weierstrass Representation as an algorithm, and then show that it pro-
duces holomorphic null-curves (Lemma 1.6), and hence CMC 1 surfaces in H3 and minimal surfaces in E3
(theorem 1.7).
Definition 1.5. An LWR potential ξ “ pξλ qλPC is an affine linear family of meromorphic 1-forms on a
Riemann surface Σ with values in sl2 C, whose linear term is nilpotent. Locally, away from its poles, for all
λ P C, it can be written as
(1.9) ξλ “ pAλ ` Bqdz
where z P U Ă Σ is a local coordinate, A : U Ñ sl2 C is holomorphic, nilpotent, and B : U Ñ sl2 C is
holomorphic.
An LWR frame Φ “ pΦλ qλPC is a holomorphic family of maps Φλ : Σ Ñ SL2 C such that Φ´1 dΦ is an
LWR potential.
The Loop Weierstrass Representation (LWR) is the following algorithm.
(1) Take an LWR potential ξ defined on some Riemann surface Σ.
(2) Let Σ
r be the universal cover of Σ and solve for Φλ : Σ
r Ñ SL2 C the following initial value problem:
#
Φ´1 dΦ “ ξ,
(1.10)
Φλ pz0 q “ Cλ
where z0 P Σ
r and Cλ P SL2 C is holomorphic with respect to λ P C.
(3) Let λ0 ‰ λ1 P C and define ψ : Σ
r Ñ sl2 C and Ψ : Σ
r Ñ SL2 C as
(1.11) 9 ´1 qλ , Ψ :“ Φλ Φ´1
ψ :“ pλ1 ´ λ0 qpΦΦ 0 1 λ0
where the dot denotes the partial derivative with respect to λ.
r Ñ E3 and f H : Σ
(4) Define f E : Σ r Ñ H3 as
(1.12) f E :“ ψ ` ψ ˚ , f H :“ Ψ˚ Ψ.
The input data pΣ, ξ, Φ, λ0 , λ1 q is called LWR data. The ordered pair pλ0 , λ1 q are the evaluation points.
We will sometimes write f X “ LWRpΣ, ξ, Φ, λ0 , λ1 q to denote the immersion induced by LWR.
Lemma 1.6. Fix λ0 ‰ λ1 P C. For any LWR frame Φ, the maps ψ and Ψ defined in (1.11) are holomorphic
null curves (possibly branched). Moreover, any holomorphic null curve into sl2 C or SL2 C can locally be
obtained this way.
Proof. Let ξ be the LWR potential of Φ. We first show that ψ is a holomorphic null curve. The LWR frame
Φ is holomorphic, so ψ is holomorphic. Compute
9 ´1 q “ dpΦqΦ
dpΦΦ 9 ´1 ´ ΦΦ9 ´1 dΦΦ´1
B pdΦq ´1 9 ´1
“ Φ ´ ΦξΦ
Bλ
B pΦξq ´1 9 ´1
“ Φ ´ ΦξΦ
Bλ
9 ´1 .
“ ΦξΦ
Therefore, with ξλ “ pAλ ` Bqdz,
(1.13) dψ “ pλ1 ´ λ0 qΦλ0 AΦ´1
λ0 dz.
By definition of ξ, det A “ 0, so det dψ “ 0. Similarly, Ψ is holomorphic and
(1.14) dΨ “ pλ1 ´ λ0 qΦλ1 AΦ´1
λ0 dz
so Ψ is a holomorphic null curve.
5
We now prove that all null curves can be obtained this way. Let ψ : D Ñ sl2 C be a holomorphic null curve
from a simply-connected neighborhood D of z0 P C. Let αλ “ λλ´λ 0
1 ´λ0
and let ξλ “ αλ dψ. The potential
ξ is an LWR potential on D because ψ is a holomorphic null curve and α is linear in λ. Define the LWR
9 ´1 qλ .
frame Φ on D as the unique solution of Φ´1 dΦ “ ξ with Φpz0 q “ exp pαψpz0 qq. Let X “ pλ1 ´ λ0 qpΦΦ 0
´1
A computation gives Xpz0 q “ ψpz0 q because α9 “ pλ1 ´ λ0 q . Moreover, αλ0 “ 0, so Xz “ ψz . By the
Picard-Lindelöf Theorem, ψ “ X. The case of Ψ : D Ñ SL2 C is covered similarly with ξ “ αΨ´1 dΨ and
Φpz0 q “ exp pα log Ψpz0 qq, computing that Φλ0 “ I and Φλ1 “ Ψ. □
As a direct consequence of theorems 1.3, 1.4 and lemma 1.6, we have:
Theorem 1.7. Fix λ0 ‰ λ1 P C. For any LWR frame Φ, the maps f E and f H defined in (1.12) are
minimal and CMC 1 conformal immersions, respectively. Moreover, any conformal, minimal (resp. CMC
1) immersion into E3 (resp. H3 ) can locally be obtained this way.
2 Gauge freedom and standard forms of LWR potentials

Definition 2.1. An LWR gauge is a meromorphic map g : Σ Ñ SL2 C, independent of λ.
Lemma 2.2. LWR gauges act on LWR frames and potentials without changing the induced immersions.
Proof. Let Φ be an LWR frame on Σ with LWR potential ξ and let g : Σ Ñ SL2 C be an LWR gauge. Define
ξ ¨ g :“ g ´1 ξg ` g ´1 dg.
It is easy to check that Φ ¨ g :“ Φg is a frame for ξ ¨ g. The fact that g does not depend on λ implies that
ξ ¨ g is linear in λ and that the linear factor has vanishing determinant. Therefore, ξ ¨ g is an LWR potential,
and Φg is an LWR frame. The λ-independence of g also implies that the holomorphic null curves ψ and Ψ
are unchanged after gauging. Therefore, the immersions themselves are preserved. □
In this section, we give two useful representatives in the gauge orbit of a given LWR potential. The first
one (section 2.2) is linear in λ with no constant term and is directly related to the Weierstrass and Bryant
representations. The second one (section 2.3) is off-diagonal and is uniquely determined by the Schwarzian
derivative of a spinorial quantity that we define in section 2.1.
Acknowledgement: The authors wish to thank Franz Pedit for suggesting the Schwarz potential.
2.1 Spinors
Definition 2.3. Let A P sl2 C be a nilpotent matrix. A vector x “ pu, vq P C2 is a spinor for A if
ˆ ˙ ˆ ˙
úv u2 u
(2.1) A“ “ p´v uq “: xxK .
´v 2 uv v
A spinor for A is uniquely determined up to a sign. We extend the definition and say that x : D Ñ C2 is a
spinor for an LWR potential ξ if xxK “ A where ξ “ pAλ ` Bqdz on some simply connected domain D Ă Σ.
Remark 2.4. The notation xK has been chosen because for all x “ pu, vq P C2 ,
ˆ ˙
K u
x x “ p´v uq “ 0.
v
We give some properties of xK that can be directly computed:
2 2
Lemma 2.5. For all x :“ pu, vq P C2 , writing ∥x∥2 :“ |u| ` |v| ,
@ K D
(2.2) xx , pxxK q˚ “ ∥x∥2 ,
(2.3) rxxK , pxxK q˚ s “ ´∥x∥2 p∥x∥2 I ´ 2xx˚ q,
(2.4) ∥rxxK , pxxK q˚ s∥ “ ∥y∥4 ,

@ ˚ D 1
(2.5) yy , pyy K q1 “ ´ ∥y∥2 y K y 1 ,
2
(2.6) @Φ P SL2 C, @x P C2 , pΦxqK “ xK Φ´1 .
6
Proposition 2.6. Let ξ be an LWR potential on Σ. Let Φ be a local LWR frame for ξ and let λ P C. Let
y :“ Φλ x where x is a spinor for ξ. Then the function
(2.7) q :“ detpy, yz q
is well-defined on Σ and only depends on the orbit of ξ under LWR gauging (not on Φ, not on λ).
Proof. We first show that q satisfies
(2.8) q “ detpx, Bx ` xz q
where ξ “ pAλ ` Bqdz. To do so, differentiate y “ Φλ x using dΦ “ Φξ and recall that xK x “ 0 to get
dy “ Φξx ` Φdx
“ ΦpλAx ` Bx ` xz qdz
“ ΦpλxxK x ` Bx ` xz qdz
“ ΦpBx ` xz qdz.
Therefore,
q “ detpΦx, ΦpBx ` xz qq
and (2.8) follows from det Φ “ 1. This shows that q is well-defined on Σ and does not depend on Φ nor does
it depend on λ.
To show that q is invariant under LWR gauging, let ξ˜ :“ ξ ¨ g where g is an LWR gauge. Let Φ̃ :“ Φg, let
x̃ be a spinor for ξ˜ and let ỹ :“ Φ̃x̃ for some λ P C. Then by (2.6), x̃ “ g ´1 x, so ỹ “ y and q “ detpỹ, ỹz q. □
Definition 2.7. An LWR potential is degenerate if q is constantly vanishing, and non-degenerate
otherwise.
2.2 Weierstrass potential

Proposition 2.8. Let ξ be an LWR potential on Σ and let Φ be an LWR frame for ξ on a simply-connected
domain D Ă Σ. With Φ0 :“ Φ|λ“0 ,
ˆ ˙
g ´g 2 qdz
(2.9) ξ ¨ Φ´1
0 “ λ
1 ´g gz
where
(2.10) g :“ u{v, q :“ uvz ´ vuz
and
pu, vq :“ Φ0 x
where x is a spinor for ξ.
Proof. Write ξ “ pAλ ` Bqdz so that xxK “ A and let y :“ Φ0 x. By (2.6),
Φ0 AΦ´1 K ´1 K
0 “ Φ0 xx Φ0 “ yy .
Moreover, dΦ “ Φξ, so
Φ0 dpΦ´1 ´1 ´1
0 q “ ´dΦ0 Φ0 “ ´Φ0 BΦ0 dz.
Therefore,
ξ ¨ Φ´1 K
0 “ λyy dz
and (2.9) follows from y “ pu, vq and the definitions of g and q. □
Remark 2.9. Let ξ1 be the potential given by (2.9).
1. The functions g and q in ξ1 do not depend on the choice of sign for the spinor x, and the definition
of q in proposition 2.8 agrees with its definition in proposition 2.6.
2. Since Φ0 is only defined locally on Σ, so is the potential ξ1 .
3. Since Φ0 is uniquely determined by ξ up to conjugation by SL2 C, so is the function g. However,
proposition 2.6 shows that the function q is uniquely determined by ξ.
7
4. As we will see in section 3, the potential ξ1 is directly related to the Weierstrass representation: if
λ0 “ 0, then g is the Gauss map of f E and the hyperbolic Gauss map of f H . Moreover, if λ1 “ 1,
then qdz 2 is the Hopf differential of f E and f H .
2.3 Schwarz potential
Definition 2.10. Let g : C Ñ C be a meromorphic function. The map
ˆ 2 ˙2 ˆ 2 ˙1
g g
Srgs “ 1
´
2g 2g 1
is the Schwarzian derivative of g. It is invariant under Möbius transformations of g. It acts as a second
order operator under pre-composition: for any meromorphic ϕ : C Ñ C,
(2.11) Srg ˝ ϕs “ pSrgs ˝ ϕqpϕ1 q2 ` Srϕs.
The following lemma justifies our normalization for the Schwarzian derivative and is merely a computation.
Lemma 2.11. Let u, v be two holomorphic solutions of the equation
y 2 “ Sy
where S is a meromorphic function. If u{v is not constant, then S “ Sru{vs.
Theorem 2.12. Let ξ “ pAλ ` Bqdz be a non-degenerate LWR potential in the coordinate z. Then there
exists an LWR gauge k, unique up to a sign, such that
ˆ ˙
0 1
ξ¨k “ dz
λq ` s 0
for some functions q and s.
Moreover, if pu, vq :“ Φx where x is a spinor for A and Φ is an LWR frame for ξ at λ “ 0, then
q “ uvz ´ vuz and s “ Sru{vs|λ“0 .
Proof. We first show existence. Gauge ξ with k1 :“ Φ´1
0 so that ξ1 :“ ξ ¨ k1 is a Weierstrass potential as in
proposition 2.8. With g :“ pu{vq|λ“0 , compute the successive LWR gauge transformations:
ˆ ˙ ˆ ˙
1 g 0 g1
k2 :“ , ξ2 :“ ξ1 ¨ k2 “ dz,
0 1 λq{g 1 0
ˆ? 1 ˙ ˆ ˙
g 0 h 1 g2
k3 :“ 1 , ξ3 :“ ξ 2 ¨ k3 “ dz, h :“ 1 ,
0 ? 1 λq ´h 2g
ˆ ˙g ˆ ˙
1 0 0 1
k4 :“ , ξ4 :“ ξ3 ¨ k4 “ dz, s :“ Srgs.
´h 1 λq ` s 0
Therefore, with k :“ k1 k2 k3 k4 , the gauged potential ξ ¨ k has the expected form.
We now show uniqueness. Let r1 and r2 be two LWR gauges such that
ˆ ˙ ˆ ˙
0 1 0 1
η1 :“ ξ ¨ r1 “ dz and η2 :“ ξ ¨ r2 “ dz.
λq1 ` s1 0 λq2 ` s2 0
Then η2 “ η1 ¨ r with r :“ r1´1 r2 . This implies that
(2.12) dr “ rη2 ´ η1 r.
With ˆ ˙
a b
r :“ ,
c d
(2.12) is equivalent to
ˆ 1 ˙ ˆ ˙ ˆ ˙
a b1 bq2 0 ć a´d
“λ ` .
c1 d1 dq2 ´ aq1 bq1 ds2 ´ as1 c
But η1 and η2 are gauge-equivalent, so by proposition 2.6, q1 “ q2 “ q. Equating the λ1 -coefficients and
recalling that q is not constantly vanishing because ξ is non-degenerate, b “ 0 and a “ d. We deduce that
c “ 0 and that r “ ˘I because det r “ 1. Therefore, r1 “ ˘r2 and s1 “ s2 . □
8
Corollary 2.13. Let ξ be an LWR potential. Let Φ be a frame for ξ, let x be a spinor for ξ and let y :“ Φx.
With y “ pu, vq, the function Sru{vs only depends on the orbit of ξ under LWR gauging and satisfies for all
λ P C,
(2.13) Sru{vs “ λq ` s
where q and s are as in theorem 2.12.
Proof. We first show that Sru{vs does not depend on the initial condition for Φ. Let Φ̃ :“ CΦ where
C P SL2 C is holomorphic in λ P C and let ỹ :“ Φ̃x. Then ỹ “ Cy. With ỹ “ pũ, ṽq, ũ{ṽ and u{v differ by a
λ-dependent Möbius transformation. Therefore, their respective Schwarzian derivative are the same.
We now show that Sru{vs is invariant under LWR gauging. Let g be an LWR gauge and let Φ̃ :“ Φg and
ξ˜ :“ ξ ¨ g. Then, by (2.6), x ˜ Let ỹ :“ Φ̃x̃. Then ỹ “ Φx “ y, so Sru{vs is invariant
r :“ g ´1 x is a spinor for ξ.
under LWR gauging.
We now prove (2.13) by assuming, without loss of generality, that ξ is given in its Schwarz form, as in
theorem 2.12. Let y “ Φx where x is a spinor for ξ and Φ is an LWR frame associated with ξ. Considering
?
the form of ξ, we have x “ i qp0, 1q. Writing
ˆ ˙
a c
Φ“ ,
b d
we get
?
y “ i qpc, dq
and
u c
“ .
v d
Now differentiate the equation dΦ “ Φξ componentwise in z to get
#
c2 “ a1 “ pλq ` sqc
d2 “ b1 “ pλq ` sqd.
By hypothesis, ξ is non-degenerate, so by proposition 2.6, detpy, yz q ‰ 0, therefore c{d is not constant. By

lemma 2.11, Sru{vs “ λq ` s. □
Remark 2.14. Theorem 3.1 and corollary 3.6 will give a geometric interpretation of q and s in terms of the
the induced immersion’s Hopf differential and its Gauss map’s Schwarzian derivative.
3 Local description in the LWR

In this section, we compute the first and second fundamental forms as well as the Gauss map of immersions
in terms of their LWR data. This allows us in corollary 3.6 to complete the geometric interpretation of the
Schwarz potential introduced in section 2.3.
3.1 First and second fundamental forms

Theorem 3.1. Let Φ be an LWR frame with LWR potential ξ. Let x be a spinor for ξ and let y :“ Φx. The
metric of f X “ LWRpΣ, ξ, Φ, λ0 , λ1 q is
#
2 2
|λ1 ´ λ0 | ∥yλ0 ∥4 |dz| if X3 “ E3 ,
(3.1) ds2 “ 2 2
|λ1 ´ λ0 | ∥yλ1 ∥4 |dz| if X3 “ H3 .
In both cases, the Hopf differential of f X is

(3.2) Qdz 2 “ pλ1 ´ λ0 qqdz 2
where q :“ detpy, yz q.
9
Proof. We start with Euclidean space and compute the metric given by (1.6). Consider ψ defined in (1.11)
and compute
(3.3) fz “ ψz , fz “ pψz q˚ .
By (1.13), with y0 :“ Φλ0 x (to ease the notation),
(3.4) ψz “ pλ1 ´ λ0 qy0 y0K .
The metric in E3 follows from (2.2).
To compute the Hopf differential, we first compute the normal. Write fx “ fz ` fz and fy “ ipfz ´ fz q
and recall (1.4) to get
(3.5) fx ˆ fy “ ´2ifz ˆ fz “ ´ rfz , fz s .
Use equations (3.3), (3.4), (2.3) and (2.4) to get
(3.6) N “ I ´ 2∥y0 ∥´2 y0 y0˚ .
To get the Hopf differential, compute
fzz “ pλ1 ´ λ0 qpy0 y0K qz .
Therefore, with (2.5),
xfzz , N y “ pλ1 ´ λ0 q detpy0 , py0 qz q.
By proposition 2.6, this gives the expected formula for the Hopf differential in E3 .
In hyperbolic space, we have
` ˘ ` ˘˚
(3.7) fz “ Ψ˚ Ψz Ψ´1 Ψ, fz “ Ψ˚ Ψz Ψ´1 Ψ.
By (1.14) and with y1 :“ Φλ1 x,
(3.8) Ψz Ψ´1 “ pλ1 ´ λ0 qy1 y1K .
Because Ψ acts as an isometry of H3 ,
@ D 2@ D
xfz , fz y “ Ψz Ψ´1 , pΨz Ψ´1 q˚ “ |λ1 ´ λ0 | y1 y1K , py1 y1K q˚ .
The metric in H3 follows from (2.2).
To compute the normal in H3 , we define n with
(3.9) pfz , fz , N q “ Ψ˚ pΨz Ψ´1 , pΨz Ψ´1 q˚ , nqΨ.
By equations (3.8), (2.3) and (2.4),
(3.10) n “ I ´ 2∥y1 ∥´2 y1 y1˚ .
To get the Hopf differential, compute
fzz “ pλ1 ´ λ0 qΨ˚ py1 y1K qz Ψ.
Then, by (2.5),
xfzz , N y “ pλ1 ´ λ0 q detpy1 , py1 qz q.
By proposition 2.6, this gives the expected formula for the Hopf differential in H3 . □
Remark 3.2. By proposition 2.6 and (2.8), one can get the Hopf differential of f given in (3.2) without
computing Φ:
Qdz 2 “ pλ1 ´ λ0 q detpx, Bx ` xz qdz 2 .
Remark 3.3. For fixed pλ0 , λ1 q the immersions f E and f H induced by Φ are not necessarily isometric, but
they have the same Hopf differential.
10
3.2 Gauss map
Let f H : Σ Ñ H3 be a CMC 1 conformal immersion and let N be the normal of f . For each z P Σ, the
Riemannian exponential map at f pzq in the direction of N pzq defines a geodesic γ : R Ñ H3 . In Minkowski
space R1,4 , this geodesic is the intersection between the hyperboloid H3 and the linear plane spanned by
f pzq and N pzq. As t tends to infinity, the asymptotic direction of γptq can be identified with the null line
through f pzq ` N pzq. This null-line has a unit-length representative in the tangent space at I:
2pf ` N q
(3.11) vN :“ ´ I P TI H3 .
trpf ` N q
We identify S2 Ă TI H3 with the Riemann sphere via the stereographic projection St : S2 Ñ C Y t8u defined
by
(3.12) St´1 pu{vq :“ I ´ 2∥ν∥´2 νν ˚ , ν :“ pu, vq.
Definition 3.4. The hyperbolic Gauss map of a CMC 1 immersion is the map G “ St ˝ vN . The Gauss
map of a minimal immersion f E is simply G “ StpN q where N is the unit normal of f E .
Theorem 3.5. Let Φ be an LWR frame with potential ξ, let x be a spinor for ξ and let y :“ Φx. Let f E
and f H be the immersions induced by Φ at the evaluation points pλ0 , λ1 q. The Gauss map of f E and the
hyperbolic Gauss map of f H are both given by
u
G “ , pu, vq :“ yλ0 .
v
Proof. In E3 , the formula is directly given by (3.6).
In H3 , use the definition of f in (1.12) and recall (3.9) to get
f ` N “ Ψ˚ pI ` nqΨ.
Note that for all x P C2 ,
y˚ “ I ´ ∥x∥´2 xx˚ .
xx
Therefore, with (3.10):
I ` n “ 2∥y1 ∥´2 yz ˚
1 y1 .
Noting that
Ψ˚ yz ˚ z˚
1 y1 Ψ “ y0 y0 , y0 :“ yλ0 ,
we get
f ` N “ 2∥y1 ∥´2 yz ˚
0 y0 .
y˚ q “ 1 for all x P C2 ,
Noting that trpxx
2pf ` N q
vN “ Í
trpf ` N q
˚
“ 2yz0y ´ I
0
“ 2I ´ 2∥y0 ∥´2 y0 y0˚ ´ I

“ St´1 pu{vq. □
Corollary 3.6. The Schwarzian derivative of G satisfies

SrGs “ λ0 q ` s
where q and s are as in theorem 2.12.
Proof. This is a direct consequence of corollary 2.13. □
11
4 Related surfaces
Starting with an LWR potential ξ, in order to make an immersion via the LWR, one has to choose a pair
of evaluation points pλ0 , λ1 q and an initial condition Φpz0 q for the LWR frame Φ. This section investigates
the consequences of these choices. We first focus on the evaluation points and show that moving them
describes associated families. We then show how changing the initial condition for the LWR frame amounts
to Goursat transformations of the induced surfaces. Noting that rigid motions induce a special class of
Goursat transformations, we study them in this section.
4.1 Moving the evaluation points

We fix an LWR frame Φ and consider the evaluation points pλ0 , λ1 q as free parameters. We want to study
E H
how the immersions fpλ 0 ,λ1 q
and fpλ0 ,λ1 q
induced by LWR at pλ0 , λ1 q change as λ0 or λ1 moves in C.
4.1.1 Associated family

The classical associated family is parametrized by λ P S1 Ă C. We extend this family to any λ P C˚ .
Definition 4.1. Let f be a minimal or CMC 1 conformal immersion with metric ds2 and Hopf differential
Qdz 2 . The complex associated family pfλ qλPC˚ of f is the family of minimal or CMC 1 immersions whose
metric and Hopf differential read
2
ds2λ “ |λ| ds2 , Qλ dz 2 “ λQdz 2
for all λ P C˚ . By Bonnet’s theorem, the complex associated family is defined up to ambient isometries.
Theorem 4.2. The complex associated family of f E is obtained by fixing λ0 and moving λ1 . The complex
associated family of f H is obtained by fixing λ1 and moving λ0 .
Proof. It is a direct consequence of theorem 3.1. □
4.1.2 Dual associated family

Definition 4.3 ([15]). Let f “ Ψ˚ Ψ be a conformal, CMC 1 immersion into H3 . The immersion f 7 “
pΨ´1 q˚ Ψ´1 is the dual of f . It is also a conformal, CMC 1 immersion into H3 .
Theorem 4.4. The dual of f H is obtained by permuting λ0 and λ1 .
Proof. By (1.11), permuting the evaluation points amounts to taking the inverse null curve, which is what
duality does. □
Definition 4.5. The dual associated family of a conformal, CMC 1 immersion f into H3 is the family
pfˆλ qλPC˚ defined by
˘7
fˆλ :“ pf 7 qλ
`
where p¨q7 denotes the dual and p¨qλ denotes the complex associated family.
Corollary 4.6. The dual associated family of f H is obtained by fixing λ0 and moving λ1 .
Proof. Apply successively theorem 4.4, theorem 4.2 and theorem 4.4. □
The following proposition shows that a minimal immersion can be obtained as the blow-up limit in the
dual associated family of a CMC 1 immersion.
Proposition 4.7 (Blow-up limit of CMC 1 surfaces). Let γ : p´ϵ, ϵq Ñ C such that γp0q “ λ0 and γ 1 p0q “
λ1 ´ λ0 ‰ 0. Let Φ be an LWR frame and let Ψt :“ Φγptq Φ´1 H ˚
γp0q . Define ft :“ Ψt Ψt . Then
d ` H˘
f “ f E,
dt t |t“0
where f E is the minimal immersion induced by the LWR frame at the evaluation points pλ0 , λ1 q.
12
iπ 3iπ
(a) λ1 “ 1 (b) λ1 “ e 4 (c) λ1 “ i (d) λ1 “ e 4 (e) λ1 “ ´1
Figure 2. Five members in the dual associated family of a catenoid (definition 4.5).
The evaluation point λ0 “ 0 is fixed while λ1 moves around the origin. All these immersions
have the same hyperbolic Gauss map.
Proof. Compute
d ` ˘
Φ “ γ 1 p0qΦ9 γp0q “ pλ1 ´ λ0 qΦ9 λ0 .
dt γptq |t“0
Therefore,
d
pΨt q|t“0 “ pλ1 ´ λ0 qΦ9 λ0 Φ´1
λ0 “ ψ
dt
where ψ is the holomorphic null curve induced by the LWR frame Φ at the evaluation points pλ0 , λ1 q.
Furthermore,
d H d d
pf q “ pΨt q˚|t“0 ` pΨt q|t“0 “ ψ ˚ ` ψ “ f E
dt t |t“0 dt dt
because Ψ0 “ I. □
4.2 Holomorphic dressing

Definition 4.8. Let Φ be an LWR frame and let gλ P SL2 C be constant in z (but depending on λ).
‚ Dressing Φ by g is finding a map h (depending on both z and λ) such that Φ̂ :“ gΦh´1 is an LWR
frame.
‚ If g is holomorphic in λ P C, a dressing is given by choosing h “ I. In this case, the map Φ ÞÑ Φ̂
is a holomorphic dressing. Note that holomorphic dressing is a group action of SL2 C-valued
λ-holomorphic maps on LWR frames.
In this section, we study holomorphic dressing and show that they induce Goursat transformations (in-
cluding rigid motions). We will consider simple factor dressing in section 8.
Definition 4.9 ([9], Lemma 5.3.1). Two minimal or CMC 1 immersions are related by a Goursat trans-
formation if they have the same Hopf differential and if the Schwarzian derivative of their Gauss maps are
equal.
Theorem 4.10. Holomorphic dressing induces a Goursat transformation of the induced immersion f . More-
over, any Goursat transformation of f can be obtained by holomorphic dressing, provided that f is not flat.
Proof. Holomorphic dressing amounts to changing the initial condition Φpz0 q in the initial value prob-
lem (1.10). Therefore, it does not change the potential ξ. By remark 3.2, the Hopf differential is unchanged.
Moreover, by theorem 3.1, the Gauss map G p of the dressed immersion fp reads
ˆ ˙
p “ aG ` b , Rλ “ a b P SL2 C
G
cG ` d 0
c d
where G is the Gauss map of f . Therefore, G p and G differ by a Möbius transformation, so they have the
same Schwarzian derivative.
Conversely, let fp be a Goursat transformation of f . The immersions are not flat, so their respective
potential can be gauged into a Schwarz potential (by theorem 2.12). By corollary 2.13, these Schwarz
potentials agree because the respective Hopf differential and Schwarzian derivative of the Gauss map agree
13
for a Goursat pair. Therefore, the gauged LWR frames agree, up to a choice of initial condition, and one
can be obtained from the other via a holomorphic dressing. □
4.3 Rigid motions
Definition 4.11. Let f : Σ Ñ X3 be an immersion. A rigid motion of f is an immersion fp: Σ Ñ X3 such

that there exists an orientation-preserving isometry J of X3 satisfying fp “ J ˝ f .
Theorem 4.12. Let f be the immersion induced by the LWR frame Φ at pλ0 , λ1 q and let Φ̂ “ RΦ be a
holomorphic dressing of Φ.
‚ In E3 , if Rλ0 P SU2 , then the immersion induced by Φ̂ is a rigid motion of f .
‚ In H3 , if Rλ1 P SU2 , then the immersion induced by Φ̂ is a rigid motion of f .
‚ Any rigid motion of f can be obtained this way, provided that f is not flat.
Proof. The null curves ψ̂ and Ψ̂ induced by Φ̂ read
ψ̂ “ Rλ0 ψRλ´1
0
9 ´1 qλ
` pλ1 ´ λ0 qpRR 0
and
Ψ̂ “ Rλ1 ΨRλ´1
0
.
Therefore, if Rλ0 P SU2 , then the induced immersion in E3 reads
(4.1) fÊ “ Rλ0 f E Rλ´1
0
`T
where
(4.2) T “ c ` c˚ , 9 ´1 qλ .
c “ pλ1 ´ λ0 qpRR 0
If Rλ1 P SU2 , then

(4.3) fˆH “ pRλ´1
0
q˚ f H Rλ´1
0
.
Therefore, in both spaces, fˆ is a rigid motion of f .
Conversely, let fˆ be a rigid motion of f . Rigid motions are a special case of Goursat transformations, and
f is not flat. Therefore, by theorem 4.10, fˆ can be obtained by dressing. Let Φ̂ “ RΦ inducing fˆ. Because
fˆ is a rigid motion of f , the corresponding metrics are the same. Therefore, in E3 , by theorem 3.1, with
y “ Φx,
∥y∥ “ ∥Rλ0 y∥
constantly on Σ. Using that f is not flat, we deduce that Rλ0 P SU2 in E3 . By the same argument, Rλ1 P SU2
in H3 . □
Corollary 4.13 (Identity). Let f and fˆ be two non-flat immersions obtained via the LWR at the same
evaluation points pλ0 , λ1 q. Then f “ fˆ if and only if the corresponding LWR frames satisfy
Φ “ RΦ̂g
for some LWR gauge g and some holomorphic dressing R such that:
‚ in E3 , Rλ0 “ ˘I and pλ1 ´ λ0 qR9 λ0 P su2 ,
‚ in H3 , Rλ0 “ ˘I and Rλ1 P SU2 .
Proof. If fˆ “ f , then the two immersions have the same Hopf differential and the Schwarzian derivative
of their Gauss map are the same. Therefore, the Schwarz form of their potential is the same. Thus, their
respective potential belong to the same gauge class: there exists a gauge g such that
ξˆ “ ξ ¨ g.
Solving the corresponding differential system yields
Φ̂ “ RΦg
for some λ-holomorphic R : C Ñ SL2 C. By the same argument as in the proof of theorem 4.12, in E3 ,
Rλ0 P SU2 and fˆ is given by (4.1). But fˆ “ f and the immersion is not flat, so Rλ0 “ ˘I and c ` c˚ “ 0
14
9 ´1 qλ , implying that pλ1 ´ λ0 qR9 λ P su2 . The same arguments can be applied in
where c :“ pλ1 ´ λ0 qpRR 0 0
3
the case of H .
The converse is a consequence of lemma 2.2, theorem 4.12, equations (4.1)–(4.2) and (4.3). □
We end this section by exhibiting a group homomorphism between holomorphic dressings and rigid mo-
tions. This homomorphism will be useful in section 6.4 in order to distinguish rotations from general screw
motions.
Proposition 4.14. Fix pλ0 , λ1 q. Let Gλ0 be the subgroup of λ-holomorphic maps R : C Ñ SL2 C such that
Rλ0 P SU2 . Consider the map ϕ : Gλ0 Ñ Iso` pE3 q where for all X P E3 ,
ϕpRqpXq :“ Rλ0 XRλ´1
0
` ρpRq ` ρpRq˚
with
9 ´1 qλ .
ρpXq :“ pλ1 ´ λ0 qpRR 0
The map ϕ is a group homomorphism.

Proof. Let R, S P Gλ0 and X P E3 .
ϕpRSqpXq “ pRSqλ0 XpRSq´1 ˚
λ0 ` ρpRSq ` ρpRSq .
But
9 λ pRSq´1 “ ρpRq ` Rλ ρpSqR´1 .
9 ` RSq
ρpRSq “ pλ1 ´ λ0 qpRS 0 λ0 0 λ0
With Rλ0 P SU2 ,
ρpRSq ` ρpRSq˚ “ ρpRq ` ρpRq˚ ` Rλ0 pρpSq ` ρpSq˚ qRλ´1
0
.
Therefore, ϕpRSq “ ϕpRqϕpSq. □
Proposition 4.15. Fix pλ0 , λ1 q. Let Gλ1 be the subgroup of λ-holomorphic maps R : C Ñ SL2 C such that
Rλ1 P SU2 . Consider the map ϕ : Gλ1 Ñ Iso` pH3 q where for all X P H3 ,
ϕpRqpXq :“ pRλ´1
0
q˚ XRλ´1
0
.
The map ϕ is a group homomorphism.
Proof. The proof amounts to checking that
ppRSq´1 q˚ “ pR´1 q˚ pS ´1 q˚
for all R, S P SL2 C. □
5 Symmetries and closing

In this section, we fix the LWR data pΣ, ξ, Φ, λ0 , λ1 q and look for conditions on the frame Φ such that the
induced immersion f admits a symmetry.
Definition 5.1. Let f : Σ Ñ X3 be a conformal immersion. A symmetry of f is a pair pτ, Jq where τ is a
diffeomorphism of Σ and J is an ambient isometry such that
τ ˚ f “ J ˝ f.
In Euclidean space, four cases can occur depending on the orientations of τ and J. They are described in
section 5.2 and summarized in Table 1. In hyperbolic space (section 5.1), only the diagonal entries of Table 1
occur (see remark 5.8). We give closing conditions for f in terms of the monodromy of Φ in Section 5.3.
τ holom. τ anti-holom.
J orient. preserving RΦg pRΦgq´λ
J orient. reversing pRΦgq´λ pRΦgqλ
Table 1. τ ˚ Φ, when τ ˚ f “ J ˝ f .
15
5.1 Symmetries in H3
Theorem 5.2. Assume that f H is not flat. The following statements are equivalent.
(1) There exist a biholomorphism τ of Σ and an orientation-preserving isometry J of H3 such that
τ ˚f H “ J ˝ f H.
(2) There exist an LWR gauge g and a holomorphic dressing R such that Rλ1 P SU2 and
τ ˚ Φ “ RΦg.
Proof. Assume that p1q holds. Then τ ˚ f H is a rigid motion of f H , so by theorem 4.12, there exists a
holomorphic dressing R
r such that R
rλ P SU2 and
1
τ ˚ f H “ LWRpΣ, ξ, RΦ,
r λ0 , λ1 q.
On the other hand,
τ ˚ f H “ LWRpΣ, τ ˚ ξ, τ ˚ Φ, λ0 , λ1 q.
By corollary 4.13, there exists an LWR gauge g and a holomorphic dressing R
p such that R
pλ P SU2 and
1
τ ˚Φ “ R
pRΦg.
r
Conclude with R :“ R
pR.
r
Conversely, assume that p2q holds. Then the LWR potential associated with τ ˚ Φ is ξ ¨ g. Therefore,
τ ˚ f H “ LWRpΣ, τ ˚ ξ, τ ˚ Φ, λ0 , λ1 q
“ LWRpΣ, ξ ¨ g, RΦg, λ0 , λ1 q
“ LWRpΣ, ξ, RΦ, λ0 , λ1 q,
and by theorem 4.12, there exists an orientation-preserving isometry J of H3 such that τ ˚ f H “ J ˝ f H . □
A similar statement holds with an anti-holomorphic τ :
Theorem 5.3. Assume that f H is a not flat and obtained via the LWR at pλ0 , λ1 q P R2 . The following
statements are equivalent.
(1) There exists an anti-holomorphic diffeomorphism τ of Σ and an orientation-reversing isometry J of
H3 such that
τ ˚f H “ J ˝ f H.
τ ˚ Φλ “ Rλ Φλ g.
Proof. Assume that p1q holds. Then there exists an orientation-preserving isometry L such that
τ ˚f H “ L ˝ f H.
By theorem 4.12, there exists a holomorphic dressing R
r such that R
rλ P SU2 and
1
τ ˚ f H “ LWRpΣ, ξ, RΦ,
r λ0 , λ1 q.
On the other hand, defining ξpλ “ τ ˚ ξλ and Φ

p λ “ τ ˚ Φ , we have that ξp and Φ
λ
p are respectively LWR potential
and frame, and going through the steps of LWR gives
Ψ
p “Φ pλ Φ p ´1 “ τ ˚ Ψ
1 λ0
because λ0 , λ1 P R. Moreover,
fp “ Ψ
pΨp ˚ “ τ ˚f H.
Therefore,
τ ˚ f H “ LWRpΣ, ξ,
p Φ,
p λ0 , λ1 q.
We deduce that τ ˚ f H is induced by both RΦr and τ ˚ Φ . By corollary 4.13, there exist an LWR gauge g and
λ
a holomorphic dressing R such that Rλ P SU2 and
p p
1
τ ˚ Φλ “ R
pRΦg.
r
16
Conclude with R :“ RpR.
r
The converse is proved with the same arguments as in the proof of theorem 5.2. □
5.2 Symmetries in E3
Each of theorems 5.4 to 5.7 correspond to an entry of Table 1.
Theorem 5.4. Assume that f E is not flat. The following statements are equivalent.
(1) There exist a biholomorphism τ of Σ and an orientation-preserving isometry J of E3 such that
τ ˚f E “ J ˝ f E.
τ ˚ Φ “ RΦg.
Proof. Follow step by step the proof of theorem 5.2 to prove this theorem. □
E 2
Theorem 5.5. Assume that f is not flat and obtained via the LWR at pλ0 , λ1 q P R . The following
statements are equivalent.
(1) There exist an anti-holomorphic diffeomorphism τ of Σ and an orientation-reversing isometry J of
E3 such that
τ ˚f E “ J ˝ f E.
τ ˚ Φλ “ Rλ Φλ g.
Proof. Follow step by step the proof of theorem 5.3 to prove this theorem. □
We get two more types of symmetries in the Euclidean case.
Theorem 5.6. Assume that f E is not flat and obtained via the LWR with λ0 “ 0. The following statements
are equivalent.
(1) There exist a biholomorphism τ of Σ and an orientation-reversing isometry J of E3 such that
τ ˚f E “ J ˝ f E.
(2) There exist an LWR gauge g and a holomorphic dressing R such that R0 P SU2 and
τ ˚ Φ´λ “ Rλ Φλ g.
Proof. Again, the ideas are the same as in theorem 5.3, noting that the isometry X ÞÑ ´X in E3 is orientation
reversing, and that with Φ
p λ “ Φ´λ and λ0 “ 0, the induced null curve reads
p9 λ Φ
ψp “ pλ1 ´ λ0 qΦ p ´1
0 λ0
“ pλ0 ´ λ1 qΦ9 λ0 Φ´1

λ0
“ ´ψ. □
Theorem 5.7. Assume that f E is not flat and obtained via the LWR at pλ0 , λ1 q with λ0 “ 0 and λ1 P R.
The following statements are equivalent.
(1) There exist an anti-holomorphic diffomorphism τ of Σ and an orientation preserving isometry J of
E3 such that
τ ˚f E “ J ˝ f E.
(2) There exist an LWR gauge g and a holomorphic dressing R such that R0 P SU2 and
τ ˚ Φ´λ “ Rλ Φλ g.
Proof. Combine theorems 5.5 and 5.6. □
Remark 5.8. Theorems 5.6 and 5.7 do not have an equivalent in H3 because if τ is anti-holomorphic and
J is orientation preserving (or if τ is holomorphic and J is orientation reversing), then the mean curvature
of τ ˚ f and the mean curvature of J ˝ f have opposite sign. Therefore, only one of them can be obtained
via the LWR.
17
5.3 Closing conditions
Let π1 pΣq denote the first fundamental group of Σ based at some point z0 P Σ.
Definition 5.9. For any loop γ P π1 pΣq, there exists a holomorphic dressing M pγq such that
τ ˚ Φ “ M pγqΦ
where τ is the deck transformation associated with γ. The λ-holomorphic matrix M pγq is the monodromy
of Φ along γ.
Theorem 5.10. Let M be the monodromy of Φ.
‚ f E descends to a well-defined immersion of Σ if and only if for all γ P π1 pΣq,
Mλ0 pγq P t˘Iu and pλ1 ´ λ0 qM9 λ0 pγq P su2 .
‚ f H descends to a well-defined immersion of Σ if and only if for all γ P π1 pΣq,
Mλ0 pγq P t˘Iu and Mλ1 pγq P SU2 .
Proof. It is a direct consequence of theorems 5.4 and 5.2 applied to every deck transformation τ of Σ,
r and
of corollary 4.13. □
6 Intrinsic surfaces of revolution

Ă˚ is the universal cover of C˚ .
In this section, C
Ă˚ Ñ X3 is an intrinsic surface of revolution if there exists a function
Definition 6.1. An immersion f : C
Ă˚ Ñ Rą0 , depending only on |z|, such that the metric of f reads
m: C
2
ds2 “ m2 |dz| .
We construct all minimal and CMC 1 intrinsic surfaces of revolution with the LWR. We first solve the
Gauss-Codazzi equations for these surfaces. proposition 6.2 together with Bonnet’s theorem implies the
existence of various minimal and CMC 1 surfaces of revolution. We then show how to explicitly construct
them via LWR (section 6.2). Special cases occur when they descend to C˚ and conformally extend to z “ 0
(Enneper surfaces, see section 6.3). Another special case are extrinsic surfaces of revolution (catenoids, see
section 6.4).
6.1 Solution to the Gauss-Codazzi equations

Proposition 6.2. Let f be a minimal or CMC 1 intrinsic surface of revolution.
‚ If f is flat, then there exist a ą 0 and α P R such that the metric of f is
4α 2
(6.1) ds2 “ a4 |z| |dz| .
‚ If f is not flat, then there exist a, b ą 0, α ă β P R and ν P r0, 2πq such that the metric and Hopf
differential of f are
´ ¯2
2α 2β 2
(6.2) ds2 “ a2 |z| ` b2 |z| |dz| ,
(6.3) Qdz 2 “ eiν abpβ ´ αqz α`β´1 dz 2 .

2
Proof. Write z “ reiθ and ds2 “ 4e2ω |dz| with ω : C
Ă˚ Ñ R. By assumption on f , the function ω only
depends on r. The Gauss-Codazzi equations (1.8) read
2
(6.4) ωrr ` r´1 ωr “ |Q| e´2ω , Qz “ 0.
If f is flat, Q is constantly vanishing. In this case, all the solutions to (6.4) are given by
ωprq “ c1 ` c2 logprq, c1 , c2 P R.
With a :“ 2ec1 ą 0 and α :“ c2 P R, the metric reads as in (6.1).
18
If f is not flat, then Q is not constantly vanishing. The Codazzi equation ((6.4)) implies that Q is
holomorphic. Therefore, there exists a domain D Ă C Ă˚ on which Q never vanishes. On D, consider h “ |Q|2
and compute
´ ¯
(6.5) h´1 hθ “ ih´1 pzhz ´ zhz q “ i zQ´1 Qz ´ zQ´1 Qz .
2
The Gauss equation in (6.4) implies that |Q| only depends on r, so hθ “ 0 and by (6.5),
zQ´1 Qz “ zQ´1 Qz .
The left-hand side is holomorphic whereas the right-hand side is anti-holomorphic, so there exists γ P R such
that
zQ´1 Qz “ γ.
Solving this equation for Q yields for all z P D,
Qpzq “ eiν cz γ , ν P R, c ą 0.
Ă˚ . Going back to the Gauss equation in (6.4), we have
By holomorphicity of Q, this holds on C
(6.6) ωrr ` r´1 ωr “ c2 r2γ e´2ω
and a two-parameter family of solutions is given by
ˆ ˙
1 ` 2 2α ˘
(6.7) ωprq “ log a r ` b2 r2β
2
where a ą 0, b ą 0 and α ă β are defined by
1´ c¯ 1´ c¯
α“ 1`γ´ , β“ 1`γ` .
2 ab 2 ab
With such ω, the metric and Hopf differential of f read as in equations (6.2) and (6.3).
Note that all solutions to (6.6) are given by (6.7). Indeed, let ω be given by (6.7). The function
ψ : R˚` ˆ R˚` Ñ R2 , pa, bq ÞÑ pωp1q, ω 1 p1qq
is surjective, and the ODE is given by
F : Rą0 ˆ R2 Ñ R2 , pr, x, yq ÞÑ py, c2 r2γ e´2x ´ r´1 yq
which is locally Lipschitz with respect to the variable px, yq. By the Picard-Lindelöf theorem, (6.7) gives all
solutions for a, b ą 0. □
6.2 LWR potential for intrinsic surfaces of revolution

Let a, b ą 0, α ă β P R, ν P r0, 2πq and x “ paz α , bz β q. Consider the following LWR data on Σ:
r
#
p0, eiν q in E3 ,
(6.8) ξ “ λxxK dz, Φp1q “ I, pλ0 , λ1 q “
péiν , 0q in H3 .
Theorem 6.3. The LWR data (6.8) induces an intrinsic surface of revolution with metric and Hopf differ-
ential as in (6.2)–(6.3). Moreover, any non-flat, minimal or CMC 1 intrinsic surface of revolution can be
obtained this way, up to an isometry and a coordinate change.
Proof. At λ “ 0, the potential ξ is constantly vanishing, so the frame Φ is constantly the identity. Therefore,
the metric of the induced immersion f can be computed with theorem 3.1 and agrees with (6.2). We deduce
that f is an intrinsic surface of revolution. Theorem 3.1 also gives the Hopf differential as in (6.3) and the
surface is not flat.
Conversely, let f : CĂ˚ Ñ X3 be a minimal or CMC 1, non-flat, intrinsic surface of revolution. By propo-
sition 6.2, there exist a, b ą 0, α ă β and ν P R such that the metric and Hopf differential of f are given by
Equations (6.2)–(6.3). By the first part of the proof together with Bonnet’s Theorem, up to a rigid motion,
f “ LWRpΣ, ξ, Φ, λ0 , λ1 q with λ1 ´ λ0 “ eiν . □
19
6.3 Enneper surfaces
Definition 6.4. A minimal or CMC 1 immersion is an Enneper surface if it is a non-flat, intrinsic surface
of revolution that descends from the universal cover of C˚ to a conformal immersion on C.
Let r ą 0 and n P Ně0 . Consider the following LWR data on C:
#
p0, 1q in E3 ,
ˆ ˙
0 rz n
(6.9) ξ“ dz, Φp0q “ I, pλ0 , λ1 q “
λ 0 p1, 0q in H3 .
Theorem 6.5. The LWR data (6.9) induces an Enneper surface. Moreover, any Enneper surface can be
obtained this way, up to an isometry and coordinate change.
Proof. We first show that the LWR data (6.9) yields an Enneper surface. Let f “ LWRpC, ξ, Φ, λ0 , λ1 q. One
can compute explicitly:
ˆ n`1 ˙
1 rzn`1
(6.10) Φ0 pzq “ .
0 1
By theorem 3.1, the metric of f is
˜ ¸2
2n`2
2 r2 |z| 2
(6.11) ds “ 1` |dz| .
pn ` 1q2
The metric is rotationally invariant, so f is an Enneper surface. Note that the Hopf differential of f is
(6.12) Qdz 2 “ pλ1 ´ λ0 qrz n dz 2 .
We now show that any Enneper surface f : C Ñ X3 can be obtained this way. By definition, the metric
of f is radial and f is not flat. By theorem 6.3, f can be obtained with the LWR data (6.8). The metric of
f is given by
´ ¯2
2α 2β 2
ds2 “ a2 |z| ` b2 |z| |dz| .
But f is conformal at z “ 0, so αβ “ 0 and α ` β ě 0. Because α ă β, we have α “ 0 and β ą 0. The Hopf
differential is
Qdz 2 “ eiν abβz β´1 dz 2 .
The immersion f is well-defined on C, so there exists n P Ně0 such that β ´ 1 “ n. Therefore, the metric
and Hopf differential of f are
´ ¯2
2n`2 2
ds2 “ a2 ` b2 |z| |dz| ,
Qdz 2 “ eiν abpn ` 1qz n dz 2 .

The immersion is not flat, so Q ‰ 0 and a ‰ 0. After the coordinate change
iρ
w “ e n`2 a2 z
where ρ P R is defined by
eiρ :“ pλ1 ´ λ0 qeiν “ ˘eiν ,
and after letting
pn ` 1qb
r“ ą 0,
a2n`3
one can compute that the metric and Hopf differential of f read as equations (6.11) and (6.12) in the
coordinate w. By Bonnet’s theorem, f “ LWRpC, ξ, Φ, λ0 , λ1 q, up to a coordinate change and an isometry.
□
Corollary 6.6. Any Enneper surface admits an extrinsic rotational symmetry of order n ` 2.
20
Proof. Let f be an Enneper surface. By theorem 6.5, f is given by the LWR data (6.9). Let θ :“ 2π{pn ` 2q,
τ pzq :“ eiθ z and compute ˆ íθ{2 ˙
˚ e 0
τ ξ “ ξ ¨ g, g “ ,
0 eiθ{2
so that
(6.13) τ ˚ Φ “ RΦg
where R is holomorphic in λ and independent of z. Evaluating (6.13) at z “ 0 and recalling that Φp0q “ I
gives R “ g ´1 . Therefore, R P SU2 is independent of λ. By theorems 5.2 and 5.4, τ induces a symmetry of
f and a direct computation shows that this symmetry is a rotation of angle θ in both ambient spaces. □
6.4 Catenoids
In this section, we review catenoids in the framework of the LWR. We exhibit a family of Fuchsian
potentials that induce catenoids and show that any catenoid can be obtained this way.
Ă˚ Ñ X3 is a catenoid if it is a non-flat
Definition 6.7. A minimal or CMC 1 conformal immersion f : C
extrinsic surface of revolution: for all t P R,
τt˚ f “ Rptq ˝ f
where τt is a lift of z ÞÑ eit z and R : R Ñ IsopX3 q is a 1-parameter group of rotations.
Remark 6.8. We do not assume that all catenoids are closed on C˚ . However, noting that the group R is
Ă˚ . With t0 :“ mintt ą 0 | τ ˚ f “ f u, we define the
compact, any catenoid closes on some angular sector in C t
wrapping number of f as r :“ 2π{t0 .
Proposition 6.9. Let Φ be an LWR frame inducing a catenoid at the evaluation points pλ0 , λ1 q “ p0, 1q.
Up to an LWR gauge, the LWR potential of Φ reads
ˆ ˙
0 1
(6.14) ξ “ Kz ´1 dz, K “
qλ ` p 0
where q P R˚ and p ą 0. Moreover, the Hopf differential of the catenoid is qz ´2 dz 2 and its wrapping number
?
is 2 p.
Ă˚ Ñ X3 be the catenoid induced by Φ. By assumption, f is not flat, so there exists a Schwarz
Proof. Let f : C
potential inducing f locally (by theorem 2.12), and up to a gauge,
ˆ ˙
0 1
ξ“ dz.
Qλ ` S 0
By corollary 2.13, Qdz 2 is the Hopf differential of f and S is the Schwarzian derivative of its Gauss map.
Let t P R and τ pzq “ eit z. Then ˆ ˙
˚ 0 1 it
τ ξ“ e dz.
τ ˚ Qλ ` τ ˚ S 0
We put this potential into its Schwarz form with a diagonal LWR gauge g P SU2 :
ˆ ˙
0 1
τ ˚ ξ ¨ g “ 2it ˚ dz.
e pτ Qλ ` τ ˚ Sq 0
But f is a catenoid, so τ induces a rotation of the surface. By theorems 5.2 and 5.4, there exists an LWR
gauge g̃ such that τ ˚ ξ “ ξ ¨ g̃. Therefore, ξ and τ ˚ ξ ¨ g are Schwarz potentials that lie in the same gauge
class. By uniqueness of the Schwarz potential (theorem 2.12), this implies that ξ “ τ ˚ ξ ¨ g, i.e.
(6.15) τ ˚ Q “ e´2it Q, τ ˚ S “ e´2it S.
But Q is holomorphic and not identically zero, so by (6.15) Q never vanishes. Therefore there exists q P C˚
such that Q “ qz ´2 . For the same reason, there exists s P C such that S “ sz ´2 . The Schwarz potential
reads ˆ ˙
0 1
ξ“ dz
z ´2 pqλ ` sq 0
21
and ξ extends meromorphically to C˚ . With p :“ s ` 1{4,
ˆ ? ˙ ˆ ˙
z 0 0 1 dz
(6.16) ξ¨ 1 ?1
“ .
´ 2? z z qλ ` p 0 z
Up to a gauge, we can therefore assume that ξ “ Kz ´1 dz where
ˆ ˙
0 1
K“
λq ` p 0
with q P C˚ and p P C.
We now show that q P R˚ and p ą 0. The frame Φ satisfies Φ´1 dΦ “ Kz ´1 dz, so there exists C “ pCλ qλPC
independent of z such that Φ “ Cz K . Therefore, τ ˚ Φ “ RΦ with
R “ C exppitKqC ´1 .
Recall that for all t, τ induces a rotation of f X with a fixed axis.
In H3 , by theorem 5.2, R1 P SU2 for all t. But this implies that exppitK1 q is unitarizable for all
` t, which
˘
in turn implies that det K1 ď 0. Assume by contradiction that det K1 “ 0. Then exppitK1 q “ 10 it1 and
this matrix is not unitarizable
?
by SL2 C-conjugation for all t. Therefore det K1 ă 0, i.e. p ` q P Rą0 . The
? ?
eigenvalues of R0 are e˘it p , so by corollary 4.13, p P R˚ and t0 “ 22π ? , i.e. the wrapping number is 2 p
p
and p P Rą0 . Finally, q P R˚ because p ` q P R and q ‰ 0 because a catenoid is not flat.
In? E3 , p ą 0 for the same reason as in H3 : theorem 5.4 implies that det K0 ă 0. The eigenvalues of R0 are
?
e ˘it p
, so by corollary 4.13, p P R˚ and t0 “ 22π ? . Using det K0 ă 0, we note that K is holomorphically
p
diagonalizable in a neighborhood of λ “ 0 and write R “ U DU ´1 where U is independent of t, U0 P SU2 ,
and ˆ itµ ˙
e 0
D“ , µ2 “ qλ ` p.
0 eítµ
Because U is independent of t, by proposition 4.14, R induces a 1-parameter group of rotations if and only if
D induces a 1-parameter group of rotations. Consider ϕpDq and ρpDq given by proposition 4.14. The linear
part of ϕ is a rotation whose axis consists of the diagonal elements of E3 . Therefore, ϕpDq is a rotation if
and only if the affine part ρpDq ` ρpDq˚ is off-diagonal. By an explicit computation,
ˆ ˙
itq 1 0
ρpDq “ ? ,
2 p 0 ´1
so q is real and q P R˚ . □
Theorem 6.10. Let q P R˚ and let p ą 0. Consider the LWR data pC˚ , ξ, Φ, 0, 1q with ξ as in (6.14) and
#?
1
in E3 ,
ˆ ˙
1 1 µ ,
p
Φp1q :“ ? µ :“ ?
2 ´µ 1 p ` q in H3
and assume that p ` q ą 0 in the case of H3 . Then the induced immersion is a catenoid with Hopf differential
?
qz ´2 dz 2 , wrapping number 2 p and metric
q 2 ´ ´1 2µ ´2µ
¯2
´2 2
(6.17) ds2 “ µ |z| ` µ |z| |z| |dz| .
4µ2
Moreover, any catenoid can be obtained this way, up to a rigid motion and a coordinate change.
Proof. In both cases, the LWR frame reads
Φ “ Cz K , C :“ Φp1q.
it
For all t P R, let τt pzq “ e z. Then τt˚ Φ “ RΦ with
R “ C exppitKqC ´1 .
In E3 , the matrix C is a diagonalizer of K at λ0 “ 0, and thus a diagonalizer of exppitK0 q. Therefore
R0 : R Ñ SU2 is a 1-parameter group. By theorem 5.4, τt induces a 1-parameter group of symmetries Jt
?
on the immersion f . Moreover, this 1-parameter group is closed, with period t0 “ π{ p. Therefore J is a
22
1-parameter group of rotations with a common axis, and the induced immersion is a catenoid of wrapping
?
number 2 p.
Similarly, in H3 , the matrix C is a diagonalizer of K at λ1 “ 1, and thus a diagonalizer of exppitK1 q.
Therefore R1 : R Ñ SU2 is a 1-parameter group. By theorem 5.2, τt induces a 1-parameter group of symme-
?
tries Jt on the immersion f . Moreover, this group is closed with period π{ p, so it is a 1-parameter group
?
of rotations with common axis. Therefore, the induced immersion is a catenoid of wrapping number 2 p.
The metric is given by theorem 3.1, computing explicitly the spinors at λ “ 0 in E and λ “ 1 in H3 :
3
?
? i q
(6.18) x “ p0, i qz ´1{2 q, y “ ? pµ´1{2 z ´1{2`µ , µ1{2 z ´1{2´µ q. □
2µ
We now show that any catenoid can be obtained this way. Let f : C Ă˚ Ñ X3 be a catenoid. By theorem 1.7,
f can be obtained locally via LWR at the evaluation points pλ0 , λ1 q “ p0, 1q. By proposition 6.9, up to a
gauge, ξ is as in (6.14). Note that the proof of proposition 6.9 implies that p ` q ą 0 if X3 “ H3 . By
?
proposition 6.9 again, the Hopf differential of f is qz ´2 dz 2 and its wrapping number is 2 p. Let f˜ be the
catenoid given by the data of theorem 6.10 with the same q and p. Let Φ “ Cz K and Φ̃ “ C̃z K be the
frames for f and f˜ respectively. Up to a rigid motion, assume that f and f˜ share the same axis of symmetry.
Then, at λk (k “ 0 in E3 , k “ 1 in H3 ),
ˆ ˙
µ 0
pCKC ´1 qλk “ pC̃K C̃ ´1 qλk “ .
0 ´µ
We deduce that there exists ρ P C˚ such that
ˆ ˙
ρ 0
Cλk “ C̃λk .
0 ρ´1
One can then compute the spinor of Φ at λk :
ˆ ˙
ρ 0
yλk “ ỹ .
0 ρ´1 λk
2
By (6.18), the metrics ds2 and ds
r agree up to a coordinate change. Moreover, the two catenoids have the
´2 2
same Hopf differential qz dz . By Bonnet’s Theorem, they agree up to a rigid motion.
7 n-noids
In this section, we give a standard form for LWR potentials that induce n-noids. We then show how to
construct generic trinoids by solving a period problem that amounts to the unitarization of a monodromy
representation.
Definition 7.1. A genus zero n-noid is a conformal, minimal or CMC 1 immersion of the n-punctured
sphere obtained from an LWR potential ξ which is
(1) of Fuchsian type: around each puncture pk , there exists an LWR gauge gk such that ξ ¨ gk has a
simple pole at pk , and a
(2) of catenoid type: the residue at pk of ξ ¨ gk has eigenvalues ˘ qk λ ` 1{4. The parameter qk P C˚ is
the weight of the puncture pk .
Our definition of a pole of catenoid type is justified by the following lemma.
Lemma 7.2. Let ξ “ Kz ´1 dz be an LWR potential on C˚ where the residue K is independent of z and
satisfies
det K “ ´pqλ ` pq
for some q P C˚ and p P C. Then there exists an LWR gauge g such that
ˆ ˙
0 1 dz
ξ¨g “ .
qλ ` p 0 z
23
Proof. Write K “ Aλ ` B and let x :“ pu, vq be a spinor for A (x does not depend on z). Up to conjugation
by ˆ ˙
0 i
g0 “ ,
i 0
one can assume that v ‰ 0. Recall that q ‰ 0 and compute:
ˆ ´1 ˙ ˆ? ˙
v u p ´q dz
g1 :“ , ξ1 :“ ξ ¨ g1 “ ? ,
0 v ´λ ´ p z
ˆ? ˙ ˆ? ˙
´q 0 p 1 dz
g2 :“ ?1
, ξ2 :“ ξ1 ¨ g2 “ ? ,
0 ´q qλ ´ p z
ˆ ˙ ˆ ˙
1 0 0 1 dz
g3 :“ ? , ξ3 :“ ξ2 ¨ g3 “ .
´ p 1 qλ ` p 0 z
□
7.1 Potentials for n-noids

Theorem 7.3. Any n-noid can be obtained with the LWR at pλ0 , λ1 q “ p0, 1q from the potential
ˆ ˙
0 1
(7.1) ξ“ dz
Qλ ` S 0
where
‚ Qdz 2 is rational on CP 1 with double poles at z1 , . . . , zn and no other poles. Its quadratic residues
are qk , the weight of ξ at zk .
‚ Sdz 2 is rational on CP 1 with double poles at the zeroes uk of Q and no other poles. Its quadratic
residues are pn2k ´ 1q{4 where nk “ 1 ` orduk Q.
Proof. By theorem 2.12, it suffices to compute the Hopf differential and the Schwarzian derivative of the
Gauss map of a given n-noid f : Σ Ñ X3 where Σ “ CP 1 ztz1 , . . . , zn u.
Away from the punctures zk , the Hopf differential is a holomorphic quadratic differential and the Gauss
map is a meromorphic function, so both Q and S are rational on CP 1 . The Hopf differential is holomorphic,
so has no pole in Σ. Assume that G has a pole or zero of order n ‰ 0 at some u P Σ. Then a computation
shows that its Schwarzian derivative has a double pole at u with quadratic residue pn2 ´ 1q{4. Moreover, for
the surface to be immersed at u, Q must have a zero at u of order n ´ 1 (this is in fact a sufficient condition).
Now we compute the behavior of Q and S at the punctures zk . The potential ξ is of Fuchsian type, so it
is by definition gauge-equivalent to a potential η of catenoid type at zk with residue
ˆ ˙
0 1 1
K“ , µ2 “ λqk ` , qk P C˚ .
µ2 0 4
With remark 3.2 applied to η, one can compute that Q admits a double pole at zk with quadratic residue qk .
By computing the gauge putting η into its Schwarz form (using theorem 2.12), one can show that S has at
most a simple pole at zk (its quadratic residue vanishes because µ20 “ 1{4). But the n-noid f closes around
zk , so by theorem 5.10, the monodromy of Φ0 around zk is M0 “ ˘I, and this happens only if the residue of
S vanishes. Therefore, S is holomorphic at zk . □
7.2 Unitarizability on the three-punctured sphere

Definition 7.4.
‚ A subset X Ă SL2 C (resp. sl2 C) is reducible if there exists ℓ P CP 1 which is an eigenline of X for
all X P X.
‚ A subset X Ă SL2 C (resp. sl2 C) is unitarizable if there exists C P SL2 C such that CXC ´1 P SU2
(resp. su2 ) for all X P X.
‚ ν P C is a logarithmic eigenvalue of M P SL2 C if e2πiν is an eigenvalue of M . The logarithmic
eigenvalues of M are defined up to ν ÞÑ ´ν and ν ÞÑ ν ` 1, and can be normalized uniquely so
that Re ν P r0, 1{2s.
24
Theorem 7.5. Let M0 , M1 , M2 P SL2 C satisfy M0 M1 M2 “ I. Let ν0 , ν1 , ν2 P C be corresponding logarithmic
eigenvalues. Then
‚ tM0 , M1 , M2 u is reducible if and only if ν0 ˘ ν1 ˘ ν2 P Z for some choice of signs.
‚ tM0 , M1 , M2 u is irreducible and unitarizable if and only if ν0 , ν1 , ν2 P R and, when normalized to
r0, 1{2s, satisfy the spherical triangle inequalities:
ν0 ă ν1 ` ν2 , ν1 ă ν0 ` ν2 , ν2 ă ν0 ` ν1 , ν0 ` ν1 ` ν2 ă 1.
Proof. Let
1
tk :“ tr Mk , k “ 0, 1, 2,
2
and
φ :“ 1 ´ t20 ´ t21 ´ t22 ` 2t0 t1 t2 .
By [8], tM0 , M1 , M2 u is reducible if and only if φ “ 0. Moreover, tM0 , M1 , M2 u is irreducible and unitarizable
if and only if t0 , t1 , t2 P p´1, 1q and φ ą 0. Let ek :“ e2πiνk so that
ek ` e´1
k
tk “ , k “ 0, 1, 2.
2
Then
4e20 e21 e22 φ “ pe0 e1 e2 ´ 1qpe0 e1 ´ e2 qpe0 e2 ´ e1 qpe1 e2 ´ e0 q.
So φ “ 0 if and only if ν0 ˘ ν1 ˘ ν2 P Z for some choice of signs. This proves the first point.
To prove the second point, note that tk P p´1, 1q if and only if νk P R. Assume that ν0 , ν1 , ν2 are
normalized. Then φ “ 0 if and only if pν0 , ν1 , ν2 q lies on the boundary of the tetrahedron T Ă r0, 1{2s3
defined by
pν0 “ ν1 ` ν2 q or pν1 “ ν0 ` ν2 q or pν2 “ ν0 ` ν1 q or pν0 ` ν1 ` ν2 “ 1q.
Using the continuity of φ, one can check that φ ą 0 if and only if pν0 , ν1 , ν2 q lies in the interior of T, that
is, ν0 , ν1 , ν2 satisfy the spherical triangle inequalities. □
Theorem 7.6. Let A0 , A1 , A2 P sl2 C satisfy A0 ` A1 ` A2 “ 0. Let ia0 , ia1 , ia2 P C be corresponding
eigenvalues. Then
‚ tA0 , A1 , A2 u is reducible if and only if a0 ˘ a1 ˘ a2 “ 0 for some choice of signs.
‚ tA0 , A1 , A2 u is irreducible and unitarizable if and only if a0 , a1 , a2 P R and, when normalized to R` ,
satisfy the Euclidean triangle inequalities:
a0 ă a1 ` a2 , a1 ă a0 ` a2 , a2 ă a0 ` a1 .
Proof. Let
φ :“ detrA1 , A2 s.
We first show that tA0 , A1 , A2 u is reducible if and only if φ “ 0. If tA0 , A1 , A2 u is reducible, then they
share an eigenline and this eigenline is in the kernel of the commutator rA1 , A2 s, which implies that φ “ 0.
Conversely, if φ “ 0, then the commutator rA1 , A2 s is nilpotent and can therefore be written rA1 , A2 s “ xxK
for some x P C2 . If x “ 0, then A1 and A2 commute, so they share an eigenline and tA0 , A1 , A2 u is reducible.
Assume that x ‰ 0. Considering the bilinear extension of equations (1.1) and (1.4), one has
@ D
0 “ xA1 , rA1 , A2 sy “ A1 , xxK “ detpA1 x, xq.
Therefore, A1 x P xxy and x is an eigenvector of A1 . Similarly, x is an eigenvector of A2 . Therefore
tA0 , A1 , A2 u is reducible. We have proved that tA0 , A1 , A2 u is reducible if and only if φ “ 0.
Now a computation gives
2
φ “ 4pxA1 , A1 y xA2 , A2 y ´ xA1 , A2 y q.
By the polarization identity, and using that A1 ` A2 “ Á0 ,
φ “ 4∥A1 ∥2 ∥A2 ∥2 ´ p∥A0 ∥2 ´ ∥A1 ∥2 ´ ∥A2 ∥2 q2 ,
where ∥Ak ∥2 “ xAk , Ak y “ ´ det Ak . But det Ak “ a2k , so
φ “ pa0 ` a1 ` a2 qpa0 ` a1 ´ a2 qpa0 ´ a1 ` a2 qpá0 ` a1 ` a2 q.
Hence, φ “ 0 if and only if a0 ˘ a1 ˘ a2 “ 0 for some choice of signs, and the first point is proved.
25
To prove the second point, we show that tA0 , A1 , A2 u is irreducible and unitarizable if and only if
a0 , a1 , a2 P R and φ ą 0. Let
¨ ˛
xA1 , A1 y xA1 , A2 y 0
T :“ ˝xA1 , A2 y xA2 , A2 y 0 ‚.
2
0 0 xA1 , A1 y xA2 , A2 y ´ xA1 , A2 y
If tA0 , A1 , A2 u is irreducible and unitarizable, then a0 , a1 , a2 P R and T is real symmetric positive definite,
so φ ą 0. Conversely, assume that a0 , a1 , a2 P R and that φ ą 0. Then T is real symmetric positive definite.
Let T “ U T U be the Cholesky decomposition of T , with U real upper triangular. Let S P SO3 C such
that U “ SV where V “ pA1 , A2 , A1 ˆ A2 q. Let C P SL2 C be a lift of S given by the double covering
SL2 C Ñ SO3 C (see exercise 7.17 in [7]). Then C is a unitarizer of tA0 , A1 , A2 u. Therefore, tA0 , A1 , A2 u is
irreducible and unitarizable if and only if a0 , a1 , a2 P R and φ ą 0, which is equivalent to a0 , a1 , a2 P R and
|a0 | , |a1 | , |a2 | satisfying the Euclidean triangle inequalities. □
7.3 Trinoids
References for the classification of trinoids in E3 are [11], [14] and in H3 , [16],[1]. Here, we outline the
strategy to obtain closed trinoids by finding a unitarizer of a monodromy representation on the 3-punctured
sphere.
Corollary 7.7. Any trinoid can be obtained from the LWR with the data
ˆ ˙
0 1
Σ “ CP 1 zt0, 1, 8u, ξ “ dz, pλ0 , λ1 q “ p0, 1q
Qλ ` S 0
where
3
q0 ` pq1 ´ q0 ´ q2 qz ` q2 z 2 4 pu1´ u0 q2
(7.2) Q“ , S“ ,
z 2 pz ´ 1q2 pz ´ u0 q2 pz ´ u1 q2
q0 , q1 , q2 P C and u1 ‰ u2 P C are the zeroes of Q.
Proof. The function Q is as in (7.2) because of theorem 7.3. The total order of the quadratic differential
Sdz 2 is ´4, and by theorem 7.3, S is rational on CP 1 with double poles at the zeroes of Q and no other poles.
Therefore, Q admits two distinct zeroes u1 , u2 P C, each of order 1. By theorem 7.3 again, the quadratic
residue of Sdz 2 at uk is 43 , hence the form of S in (7.2). □
Proposition 7.8. Let Φ be an LWR frame with LWR potential ξ as in corollary 7.7 and let pM0 , M1 , M2 q
be the monodromy of Φ around some loop enclosing p0, 1, 8q with index 1.
(1) The eigenvalues of Mk at λ1 “ 1 are e˘2πiνk where
1 a
(7.3) νk “ ´ qk ` 1{4.
2
(2) The eigenvalues of M9 k at λ0 “ 0 are ˘ν9 k where
(7.4) ν9 k “ 2πiqk .
Proof. First note that these eigenvalues do not depend upon the choice of an initial condition for Φ, nor do
they depend on the chosen loop. Around each singularity, there exists a gauge g such that the potential ξ ¨ g
is a Fuchsian potential whose residue K has eigenvalues ˘µ where
1
µ2 “ λqk ` .
4
Noting that the gauge g itself has monodromy Í, we deduce that the monodromy of Φ has the same eigen-
values as ´ expp2πiKq. This proves (7.3). To compute the eigenvalues of M9 k , note that K is diagonalizable
in a neighborhood of λ0 “ 0. Therefore, with
ˆ 2πiµ ˙
e k
0 1
Nk :“ ´2πiµk , µ2k “ λqk ` ,
0 e 4
there exists C “ Cλ holomorphic for λ in a neighborhood of λ0 “ 0 such that
Mk “ CNk C ´1 .
26
Using that Mk “ Í at λ0 “ 0, we get
pM9 k qλ0 “ Cλ0 pN9 k qλ0 Cλ´1

0
and that the eigenvalues of M9 k and the eigenvalues of N9 k are the same at λ0 “ 0. This implies (7.4). □
Definition 7.9. An irreducible trinoid is a trinoid induced by an LWR frame Φ whose monodromy
matrices M0 , M1 , M2 satisfy the following condition:
‚ for minimal trinoids in E3 : tM9 0 , M9 1 , M9 2 uλ0 is irreducible,
‚ for CMC 1 trinoids in H3 : tM0 , M1 , M2 uλ1 is irreducible.
Theorem 7.10. Irreducible trinoids are parametrized by their weights.
Proof. We start with Euclidean space E3 . Let T be the set of irreducible minimal trinoids (modulo rigid
motions) and let Q be the set of triples pq0 , q1 , q2 q P R3 whose absolute values satisfy the Euclidean triangle
inequalities. Let W : T Ñ Q be the map that associates to a trinoid its weights. We show that W is a
bijection.
The map W is well-defined: by corollary 7.7, the weights of a given trinoid are determined by its Hopf
differential, which is invariant under rigid motions. Moreover, we need to show that WpTq Ă Q: let
f “ LWRpΣ, ξ, Φ, λ0 , λ1 q be a trinoid given by corollary 7.7 and let M0 , M1 , M2 be the monodromies of Φ
around the poles 0, 1 and 8 respectively. By definition 7.9 and by theorem 5.10, tM9 0 , M9 1 , M9 2 uλ0 is irreducible
and unitary. By proposition 7.8, the eigenvalues of pM9 k qλ0 are ˘2πiqk . By theorem 7.6, pq0 , q1 , q2 q P Q. Note
that the Hopf differential Qdz 2 given by corollary 7.7 has two distinct zeroes if and only if δpq1 , q2 , q3 q ‰ 0
where
δpq0 , q1 , q2 q :“ q02 ` q12 ` q22 ´ 2q0 q1 ´ 2q0 q2 ´ 2q1 q2 ,
but Q ∩ δ ´1 t0u “ ∅.
The map W is surjective: let q “ pq0 , q1 , q2 q P Q. Let pΣ, ξ, Φ, λ0 , λ1 q be LWR data as in corollary 7.7.
Since q P Q, proposition 7.8 and theorem 7.6 ensure that tM9 0 , M9 1 , M9 2 uλ0 is irreducible and unitarizable. Let
C be a unitarizer. Then the LWR frame CΦ induces an irreducible trinoid.
The map W is injective: let f1 and f2 be two irreducible trinoids with the same weights. By corollary 7.7,
f1 and f2 can be obtained via the LWR from the same potential and the same evaluation points. Therefore,
the corresponding LWR frames Φ1 and Φ2 only differ by an initial condition: there exists C “ pCλ qλPC
such that Φ2 “ CΦ1 . With M and N the monodromies of Φ1 and Φ2 respectively, this implies that
tM9 0 , M9 1 , M9 2 uλ0 and tN9 0 , N9 1 , N9 2 uλ0 are conjugated by Cλ0 (see theorem 5.10). We deduce that Cλ0 P SU2 .
By theorem 4.12, f1 “ f2 up to a rigid motion.
The case of H3 is treated similarly. Let T be the set of irreducible CMC 1 trinoids (modulo rigid motions).
The relevant eigenvalues are now given by (7.3) in place of (7.4). Let V be the set of triples pν0 , ν1 , ν2 q P R3
whose absolute values satisfy the spherical triangle inequalities, let ∆ :“ δ ´1 pt0uq and let
ρ : p´8, 1{2q Ñ p´1{4, `8q, pν0 , ν1 , ν2 q ÞÑ pνk pνk ´ 1qqk“1,2,3
and let Q :“ ρpVqz∆. One can show, with the same arguments as in Euclidean space, that the map
W : T Ñ Q which associates its weights to a trinoid is a bijection. □
8 Simple factor dressing

Darboux transformations of minimal and CMC 1 surfaces are constructed in [5, 13]. The same type of
transformations for non-minimal and CMC H ą 1 surfaces are described via a dressing action on the DPW
data in [4]. We aim at obtaining the surfaces of [5, 13] using the methods of [4] in the framework of the LWR.
We first define a simple factor dressing action on LWR frames and potentials, and show that simple factor
dressing preserves the Hopf differential. We then give a way to control the monodromy of the dressed surface
and give examples obtained from a catenoid.
27
(a) Full surface (b) Half view
Figure 3. Simple factor dressing of the catenoid in E3 , as in example 8.8 with pp, qq “ p 14 , 1q
and pu, ℓq “ p2, 2q.
8.1 Dressed frames

Definition 8.1. Let α P C and ℓ ‰ m P CP 1 . The simple factor SFpα, ℓ, mq is the λ-family of 2-by-2
matrices
SFpα, ℓ, mq :“ S∆S ´1
where ˆ ˙
λ´α 0
∆ :“
0 1
and S P GL2 C is independent of λ and has its first column vector in ℓ and its second column vector in m.
Remark 8.2.
‚ Simple factors are well-defined: multiplying the columns of S by scalars amounts to multiplying S
on the right by a diagonal matrix, which commutes with ∆.
‚ SF is injective because a 2-by-2 diagonalizable matrix is determined by its ordered pairs of eigenvalues
and eigenlines.
‚ Permuting the diagonal entries of ∆ in SFpα, ℓ, mq defines the simple factor SFpα, m, ℓq. Thus, for
all α, ℓ and m,
(8.1) SFpα, ℓ, mq´1 “ pλ ´ αq´1 SFpα, m, ℓq.
Theorem 8.3. Let Φ be an LWR frame and assume that Φ is not flat. For any z-independent simple factor
g, there exists a unique z-dependent simple factor h such that
(8.2) g#Φ :“ gΦh´1
is an LWR frame.
Proof. Write g “ SFpα, ℓ, mq. We first show that the simple factor h is necessarily given by
(8.3) h “ SFpα, x, yq, x “ Φ´1
α ℓ, y “ span w.
where w is a spinor for A in the potential ξ “ pAλ ` Bqdz of Φ.
28
Let h “ SFpβ, x, yq such that Φ̂ :“ gΦh´1 is an LWR frame. Then β “ α because det Φ̂ “ 1. Moreover,
Φ̂ is holomorphic at λ “ α, so on the one hand,
pλ ´ αqgΦh´1 ÝÑ 0.
λÑα
On the other hand, by (8.1),

pλ ´ αqh´1 “ SFpα, y, xq
so
pλ ´ αqgΦh´1 x ÝÑ gα Φα x.
λÑα
Therefore, in order for Φ̂ to be holomorphic at λ “ α, it is necessary that x “ Ker gα Φα , i.e. x “ Φ´1α ℓ.

To determine y, let ξ “ Φ´1 dΦ and ξˆ “ ξ ¨ h´1 . The frame Φ̂ is LWR, so ξˆ is an LWR potential. Writing
h “ S∆S ´1 , S is λ-independent, so both ξ˜ :“ ξ¨pS∆´1 q and ξ¨S are LWR potentials. But ∆ is z-independent,
so
ξ˜ “ ∆pξ ¨ Sq∆´1 .
ˆ ˙
a b
With ξ ¨ S “ ,
c á
ˆ ˙
˜ a bpλ ´ αq
(8.4) ξ“ .
cpλ ´ αq´1 á
Therefore, cα “ 0. Differentiating at λ “ α gives
ˆ ˙
9̃ a9 α bα
ξ“ .
0 á9 α
This matrix is nilpotent, so a9 α “ 0. By linearity of ξ ¨ S, one can write
ˆ ˙
â b̂
(8.5) ξ¨S “
pλ ´ αqĉ á̂
with â, b̂, ĉ independent of λ. With ξ “ pAλ ` Bqdz, this implies that
ˆ ˙ ˆ ˙
0 0 â b̂
SAS ´1 “ and SBS ´1 “ .
ĉ 0 ´αĉ á̂
Therefore, with A “ wwK , w P y, and this determines h “ SFpα, Φ´1 α ℓ, yq uniquely.
To show existence, let Σ be the Riemann surface on which ξ is defined and let Σ̃ be its universal cover,
so that Φ is defined on Σ̃. Let h as in (8.3). Then h is defined on Σ̃zS where
(8.6) S “ tz P Σ̃ | Φα w P ℓu.
By meromorphicity of Φ and w, S is either a set of isolated points or the entire Σ̃. Suppose that S “ Σ̃.
Then
detpΦw, pΦwqz q ÝÑ 0
λÑα
because ℓ is constant in z. By theorem 3.1, this implies that Q “ 0, and that contradicts the non-flatness of
Φ. Therefore, S is a set of isolated points in Σ̃, and h is well-defined on Σ̃zS. Therefore, the frame Φ̂ defined
by (8.2) is a frame on Σ̃zS, and the uniqueness part of this proof shows that it is an LWR frame. □
Definition 8.4. The map Φ ÞÑ g#Φ of (8.2) is the simple factor dressing of Φ by g. It descends, via
the LWR, to a map on the induced immersions, also called simple factor dressing.
Corollary 8.5. Simple factor dressing preserves the Hopf differential.
Proof. Let Qdz 2 and Q̃dz 2 be the Hopf differentials induced by Φ and Φ̃ :“ g#Φ “ gΦh´1 respectively.
The LWR potential of Φ̃ is ξ˜ :“ ξ ¨ h´1 . By theorem 8.3, h is a simple factor, so h “ S∆S ´1 . Therefore,
ξ˜ “ ξ ¨ pS∆´1 S ´1 q, and we have
ξˆ :“ ξ˜ ¨ S “ pξ ¨ Sq ¨ ∆´1 .
29
Let Φ̂ :“ Φ̃S be an LWR frame for ξ˜ and let Q̂dz 2 be the Hopf differential induced by Φ̂. On the one hand,
S is an LWR gauge, so Q̂ “ Q̃. On the other hand, assuming without loss of generality that ξ ¨ S is in its
Schwarz form (theorem 2.12), with Qdz 2 “ pλ1 ´ λ0 qqdz 2 ,
ˆ ˙ ˆ ˙ ˆ ˙
ˆ 0 1 pλ ´ αq´1 0 0 λ´α
ξ“ ¨ dz “ qλ`s dz.
qλ ` s 0 0 1 λ´α 0
But ξˆ is an LWR potential, so qα ` s “ 0. Therefore,
ˆ ˙
ˆ 0 λ´α
ξ“ dz.
q 0
By remark 3.2, Q̂ “ Q. □
Theorem 8.6. Let Φ be a non-flat LWR frame with monodromy M . Let g “ SFpα, l, mq. If l is an eigenline
of Mα , then the monodromy M̂ of g#Φ is
M̂ “ gM g ´1 .
Proof. Let ξ be the potential of Φ and let w be the associated spinor. Assume that ξ is defined on Σ and
let Σ̃ be its universal cover. Let h “ SFpα, x, yq with x “ Φ´1
α ℓ and y “ span w so that the dressed frame
´1
reads Φ̂ “ gΦh as in theorem 8.3. Note that h has no monodromy around the extras singularities defined
by (8.6). Let τ be a deck transformation of Σ̃ Ñ Σ and let M be the corresponding monodromy matrix for
Φ. The potential ξ is well-defined on Σ, so τ ˚ y “ y. Moreover,
τ ˚ x “ τ ˚ Φ´1 ´1 ´1 ´1
α ℓ “ Φ α Mα ℓ “ Φ α l “ x
because ℓ is an eigenline of Mα . Therefore τ ˚ h “ h, and because g is z-independent,

τ ˚ Φ̂ “ gpτ ˚ Φqh´1 “ gM Φh´1 “ gM g ´1 Φ̂. □
8.2 Example: dressed catenoids

Theorem 8.7. For any catenoid, there exists a two-real-parameter family of simple factor dressings, each one
admitting a discrete rotational symmetry and an extra end on every fundamental piece. The first parameter
u ą 0 determines the angle of the rotation, while the second parameter ℓ P RP1 determines the location of
the end. The dressed surface closes over a cover of the catenoid’s domain if and only if u P Q. In this case,
u “ n{r where r is the wrapping number and n is the number of extra ends.
Proof. We first study the Euclidean case. Let f be a catenoid in E3 . By theorem 6.10, f is obtained via the
LWR with the data of theorem 6.10. Thus, Φ “ Cz K with det K “ ´pqλ ` pq with q P R˚ and p ą 0. Let
u ą 0, u ‰ 1, ℓ P RP1 .
?
Write det Kλ “ ´µ2λ with µ0 ą 0. Let α :“ ppu2 ´ 1q{q P R˚ so that µα “ u p ą 0 and µα {µ0 “ u ą 0. Let
1
g :“ SFpα, ℓ, mq for any m P CP . Define
a
Φ̂ :“ ĝΦh´1 , ĝ :“ det g0 g0´1 g.
where h is defined via g#Φ “ gΦh´1 as in theorem 8.3. Let fˆ be the immersion induced by Φ̂ at the
evaluation points pλ0 , λ1 q “ p0, 1q.
To check that fˆ admits a discrete rotational symmetry, let τ : z ÞÑ eiπ{µα z and recall that τ ˚ Φ “ Rpπ{µα qΦ
with Rptq “ C exppitKqC ´1 . Compute Rα pπ{µα q “ Í and deduce with (8.3) that τ ˚ x “ x. Moreover,
τ ˚ y “ y, so τ ˚ h “ h. We then have
τ ˚ Φ̂ “ R̂Φ̂
with
(8.7) R̂ :“ ĝRpπ{µα qĝ ´1 .
Note that R̂ “ g̃Rpπ{µα qg̃ ´1 where g̃ :“ g0´1 g P SL2 C is holomorphic for λ in a neighborhood of λ0 “ 0 and
g̃0 “ I. Moreover, the wrapping number of the catenoid is 2µ0 , so Rpπ{µα q induces a rotation of angle 2πµ
µ .
0
? α
By proposition 4.14, R̂ induces a rotation of the same angle. Therefore, a rotation of angle 2π{p2u pq in
the domain induces a rotation of angle 2π{u in space.
30
One can compute explicitly the location of the singularities of h by solving for z with the notations of
(8.3):
1 pµ0 ` µα qz 2µα ´ pµ0 ´ µα q ℓ1
Φ´1
α ℓ“y ðñ Φα w P ℓ ðñ ´ ¨ 2µ
“ .
µ0 pµ0 ´ µα qz α ´ pµ0 ` µα q ℓ2
Considering z 2µα as the unknown of the equation, it has exactly one solution. Therefore, the dressed surface
admits one extra end on each fundamental domain of its rotational symmetry. Note that the change
ˆ ˙ ˆ ˙ ˆ 1{2 ˙
ℓ1 ωℓ1 ω 0
ÞÑ , Φ ÞÑ Φ pω P S1 q,
ℓ2 ℓ2 0 ω ´1{2
induces a rotation of the surface, and justifies our normalization ℓ P RP1 . This parameter will determine the
location z1 of the extra singularity in each fundamental domain. By (3.1), this singularity is an end because
det h ÝÑ 0 ùñ ∥h´1 ∥ ÝÑ 8 ùñ ds2 ÝÑ 8
zÑz1 zÑz1 zÑz1
and the end is planar because the Hopf differential holomorphically extends to z1 by corollary 8.5.
The proof in H3 follows the same arguments with
?
p`q
u ą 0, u ‰ 1, u ‰ ? , ℓ P RP1
p
?
so that α R t0, 1u. Let ĝ :“ det g1 g1´1 g. Use proposition 4.15 to show that the surface admits a discrete
rotational symmetry. □
Figure 4. Simple factor dressing of the catenoid in H3 viewed in the Poincaré ball model,
as in example 8.8 with pp, qq “ p 14 , ´0.1q and pu, ℓq “ p2, 2q.
Example 8.8. Dressing of a singly-wrapped catenoid (figures 3–4). The original catenoid is given by
1
p“ and q P p´1{4, 8qzt0u,
4
and the parameters for dressing are
u P Zě2 and ℓ P RP1 .
31
Then the dressed immersion fˆ has a discrete rotational symmetry of order u both in the domain and in
space. In particular, it closes on C˚ ztzk uk“1,¨¨¨ ,u where the zk are the solutions of
up2ℓ2 ´ ℓ1 q ´ p2ℓ2 ` ℓ1 q
zku “ , ℓ :“ spanpℓ1 , ℓ2 q.
upℓ1 ´ 2ℓ2 q ´ p2ℓ2 ` ℓ1 q
Example 8.9. Dressing of a doubly-wrapped catenoid (figures 1–5). The original catenoid is given by
p“1 and q P p´1, 8qzt0u,
and the parameters for dressing are
1
u“ and ℓ P RP1 .
2
The fundamental piece is C˚ ztz1 u where
3ℓ1 ´ ℓ2
z1 “ , ℓ :“ spanpℓ1 , ℓ2 q
3ℓ2 ` ℓ1
and the dressed immersion fˆ closes on C˚ ztz1 u. This construction is inspired by [3].
Figure 5. Simple factor dressing of the doubly-wrapped catenoid in E3 , as in example 8.9

with pp, qq “ p1, 1q and pu, ℓq “ p 12 , 1q (see also figure 1).
Remark 8.10. By computing explicitly the Gauss maps of the examples above, and with Corollary 1 of
[10], one can show that the dressed catenoids are Darboux transformations of the standard catenoid.
References
[1] Alexander I. Bobenko, Tatyana V. Pavlyukevich, and Boris A. Springborn, Hyperbolic constant mean curvature one sur-
faces: spinor representation and trinoids in hypergeometric functions, Math. Z. 245 (2003), no. 1, 63–91. MR2023953
[2] Robert L. Bryant, Surfaces of mean curvature one in hyperbolic space, 1987, pp. 12, 321–347, 353. Théorie des variétés
minimales et applications (Palaiseau, 1983–1984). MR955072
32
[3] Joseph Cho, Katrin Leschke, and Yuta Ogata, New explicit cmc cylinders and same-lobed cmc multibubbletons, arXiv
preprint arXiv:2205.14675 (2022).
[4] Joseph Cho and Yuta Ogata, Simple factor dressings and bianchi–bäcklund transformations, Illinois Journal of Mathematics
63 (2019), no. 4, 619–631.
[5] A. V. Corro, W. Ferreira, and K. Tenenblat, Minimal surfaces obtained by Ribaucour transformations, Geom. Dedicata
96 (2003), 117–150. MR1956836
[6] Josef Dorfmeister, Franz Pedit, and Hongyong Wu, Weierstrass type representation of harmonic maps into symmetric
spaces, Communications in analysis and geometry 6 (1998), no. 4, 633–668.
[7] William Fulton and Joe Harris, Representation theory: a first course, Vol. 129, Springer Science & Business Media, 2013.
[8] William M. Goldman, Topological components of spaces of representations, Invent. Math. 93 (1988), no. 3, 557–607.
[9] Udo Hertrich-Jeromin, Introduction to möbius differential geometry, London Mathematical Society Lecture Note Series,
Cambridge University Press, 2003.
[10] Udo Hertrich-Jeromin and Atsufumi Honda, Minimal darboux transformations, Beiträge zur Algebra und Geome-
trie/Contributions to Algebra and Geometry 58 (2017), 81–91.
[11] Shin Kato, Masaaki Umehara, and Kotaro Yamada, General existence of minimal surfaces of genus zero with catenoidal
ends and prescribed flux , Comm. Anal. Geom. 8 (2000), no. 1, 83–114. MR1730895
[12] H. Blaine Lawson Jr., Complete minimal surfaces in S 3 , Ann. of Math. (2) 92 (1970), 335–374. MR270280
[13] M. Lemes, P. Roitman, K. Tenenblat, and R. Tribuzy, Lawson correspondence and Ribaucour transformations, Trans.
Amer. Math. Soc. 364 (2012), no. 12, 6229–6258. MR2958934
[14] Francisco J López, The classification of complete minimal surfaces with total curvature greater than-12π, Transactions of
the American Mathematical Society (1992), 49–74.
[15] Masaaki Umehara and Kotaro Yamada, A duality on CMC-1 surfaces in hyperbolic space, and a hyperbolic analogue of
the Osserman inequality, Tsukuba J. Math. 21 (1997), no. 1, 229–237. MR1467234
[16] , Metrics of constant curvature 1 with three conical singularities on the 2-sphere, Illinois Journal of Mathematics
44 (2000), no. 1, 72–94.
[17] Karl Weierstrass, über die flächen, deren mittlere krümmung überall gleich null ist, Monatsberichte der Königlichen
Preussische Akademie des Wissenschaften zu Berlin (1866), 612–625.
33

Loop Weierstrass Representation: Abstract

Uploaded by

Copyright:

Available Formats

Loop Weierstrass Representation: Abstract

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

Loop Weierstrass Representation: Abstract

Uploaded by

Copyright:

Available Formats

LOOP WEIERSTRASS REPRESENTATION

THOMAS RAUJOUAN, NICK SCHMITT, AND JONAS ZIEFLE

we compute and classify various examples.

Holomorphic Representation of Surfaces

Motivation for the LWR

Structure of the Paper

1 The Loop Weierstrass Representation

1.1 Ambient spaces

1.2 Null curves and minimal or CMC 1 immersions

2 Gauge freedom and standard forms of LWR potentials

(2.3) rxxK , pxxK q˚ s “ ´∥x∥2 p∥x∥2 I ´ 2xx˚ q,

(2.4) ∥rxxK , pxxK q˚ s∥ “ ∥y∥4 ,

2.2 Weierstrass potential

By hypothesis, ξ is non-degenerate, so by proposition 2.6, detpy, yz q ‰ 0, therefore c{d is not constant. By

3 Local description in the LWR

3.1 First and second fundamental forms

In both cases, the Hopf differential of f X is

“ 2I ´ 2∥y0 ∥´2 y0 y0˚ ´ I

Corollary 3.6. The Schwarzian derivative of G satisfies

4.1 Moving the evaluation points

4.1.1 Associated family

4.1.2 Dual associated family

4.2 Holomorphic dressing

4.3 Rigid motions

Definition 4.11. Let f : Σ Ñ X3 be an immersion. A rigid motion of f is an immersion fp: Σ Ñ X3 such

If Rλ1 P SU2 , then

The map ϕ is a group homomorphism.

5 Symmetries and closing

On the other hand, defining ξpλ “ τ ˚ ξλ and Φ

“ pλ0 ´ λ1 qΦ9 λ0 Φ´1

6 Intrinsic surfaces of revolution

6.1 Solution to the Gauss-Codazzi equations

(6.3) Qdz 2 “ eiν abpβ ´ αqz α`β´1 dz 2 .

6.2 LWR potential for intrinsic surfaces of revolution

Qdz 2 “ eiν abpn ` 1qz n dz 2 .

7.1 Potentials for n-noids

7.2 Unitarizability on the three-punctured sphere

pM9 k qλ0 “ Cλ0 pN9 k qλ0 Cλ´1

Theorem 7.10. Irreducible trinoids are parametrized by their weights.

ρ : p´8, 1{2q Ñ p´1{4, `8q, pν0 , ν1 , ν2 q ÞÑ pνk pνk ´ 1qqk“1,2,3

8 Simple factor dressing

8.1 Dressed frames

On the other hand, by (8.1),

Therefore, in order for Φ̂ to be holomorphic at λ “ α, it is necessary that x “ Ker gα Φα , i.e. x “ Φ´1α ℓ.

because ℓ is an eigenline of Mα . Therefore τ ˚ h “ h, and because g is z-independent,

8.2 Example: dressed catenoids

(a) Full surface (b) Half view

(a) Full surface (b) Half view

Figure 5. Simple factor dressing of the doubly-wrapped catenoid in E3 , as in example 8.9

You might also like