Surfaces in three-dimensional Lie groups The work is supported by RFBR (grant no. 03-01-00403) and by the programm of fundamental researches of RAS “Mathematical methods in nonlinear dynamics”. , , ,

In the present paper we extend the methods of the Weierstrass (or spinor) representation of surfaces in

R^{3}

[10, 11] and

S U (2) = S^{3}

[12] for surfaces in the three-dimensional Lie groups

N i l, \tilde{S L} 2

, and

S o l

endowed with so-called Thurston's geometries [9] .

The main feature of this approach is that the geometry of a surface is related to the spectral properties of the corresponding Dirac operator. Therewith this approach reveals some unknown before geometric meanings of the Willmore functional and the Willmore conjecture which states that for tori the Willmore functional attains its minimum on the Clifford torus.

A function

ψ

generates a surface in

R^{3}

via the Weierstrass formulas if and only if it meets some equation of the Dirac type where the Dirac operator has, in general, two potentials

U

and

V

which coincide for the case of surfaces in

R^{3}

Given a surface

M

R^{3}

, the integral

E (M) = \int_{M} U V \frac{i d z \land d \bar{z}}{2}

equals to

\frac{1}{4} \int_{M} H^{2} d μ

where

H

is the mean curvature and

d μ

is the induced area form on

M

. The latter integral is equal up to some multiples and constants to the Willmore functional of

M

which is

W (M) = \int_{M} (H^{2} - K) d μ .

For closed oriented surfaces, by the Gauss–Bonnet theorem, we have

\int K d μ = 2 π χ (M)

and therefore

W (M)

and

E (M)

have the same extremals. All that was explained in [10] .

The same was established for surfaces in

S U (2)

in [12] although the Weierstrass formulas have to be replaced by an analogous construction valid for noncommutative Lie groups.

Looking for a physical confirmation of the Willmore conjecture we notice that for the Clifford torus the spectral curve which was defined for general tori in [11] has zero geometric genus. Moreover the deviation of the spectral curve of a torus from the flat curve which is the spectral curve of the Dirac operator with vanishing potentials

U = V = 0

is measured by

E (M)

. This led us to an approach for proving the Willmore conjecture by proving that in each conformal class the minimum of the Willmore functional is attained on a torus with the minimal geometric genus (this part some time ago was announced to be proved by M.U. Schmidt) as the first step and by checking the conjecture for tori with the minimal geometric genus in their conformal classes as the final step.

Recently M. Haskins proposed to treat the genus of the spectral curve as a measure of geometric complexity for some other variational problems of geometry [6] .

In the present paper we derive the analogous formulas for the functional

E (M)

for surfaces in

N i l, \tilde{S L} 2

, and

S o l

. It appears it has very nonexpectable geometric properties. We call it the spinor energy or just the energy of a surface.

Until now we even do not know is it bounded from below or not. For surfaces in

S o l

we do not prove that it is real-valued. However it measures the deviation of the spectral curve of a torus from the flat curve and we think that that is enough for confirming its geometric importance.

The problems of finding the analogs of the Willmore conjecture for such functionals and describing their extremals are very interesting.

In the present paper we also derive the equations for

ψ

corresponding to minimal surfaces in Lie groups and by our means obtain another proof to Abresch's result that for constant mean curvature surfaces in

N i l

and

\tilde{S L} 2

certain quadratic differentials are holomorphic [1] . It would be very intriguing to relate these conditions to integrable systems as it was done for constant mean curvature surfaces in

R^{3}

and

S^{3}

We note that the approach by Abresch is different and based on a representation of

N i l

and

S o l

as line bundles of constant curvature over surfaces of constant curvature (see also a recent paper [3] where this property is used for studying surfaces

N i l

and

S o l

). It looks that the failure of such a representation for

S o l

explains why our approach does not lead to such satisfactory understanding of surfaces as in the cases of

N i l

and

\tilde{S L} 2

We notice that the study of integrable surfaces in arbitrary Lie groups by methods of integrable systems was first performed in [5] . We also hope that our approach will be helpful in studying global properties of minimal and constant mean curvature surfaces in Lie groups in the spirit of [2, 4] .

The results of this paper were partially exposed on the conference on the surface theory in Benediktbeuern (January 2005).

We thank U. Abresch for many helpful conversations and an explanation of his result on holomorphic quadratic differentials for constant mean curvature surfaces in

N i l

and

S o l

2 Preliminary facts

2.1 Left invariant metrics on Lie groups

Recall that a metric

〈 ξ, η 〉

on Lie group

G

is called left invariant if it is invariant with respect to left translations:

L_{g} : G \overset{\times g}{⟶} G : h \to g h, h \in G,

i.e. for any vectors

ξ

and

η

tangent to

G

h

, the inner product of their translations by

g

, and the inner product of

ξ

and

η

coincide:

〈 ξ, η 〉 = 〈 L_{g}^{*} ξ, L_{g}^{*} η 〉, ξ, η \in T_{h} G, g, h \in G .

Clearly every left invariant metric is determined by the inner product of vectors tangent at the unit

1 \in G

of the group

G

, i.e. by the inner product on the Lie algebra

G

G

For calculating the Levi-Civita connection we apply the general tetrad formalism known from mathematical physics which is as follows.

Given vector fields

e_{1}, . . ., e_{n}

on an

n

-dimensional manifold

M

such that at any point the corresponding vectors form an orthonormal base:

〈 e_{i}, e_{j} 〉 = δ_{i j}, i, j = 1, . . ., n,

the Levi-Civita connection is given by the following formulas:

Γ_{j k}^{i} = \frac{1}{2} (c_{k j}^{i} + c_{i k}^{j} + c_{i j}^{k}),

\nabla_{e_{k}} e_{j} = Γ_{j k}^{i} e_{i},

where

[e_{i}, e_{j}] = c_{i j}^{k} e_{k} .

In our case let us take an orthonormal base

ξ_{1}, . . ., ξ_{n}

for the tangent space in the unit of the group

G

and extend it to vector fields on the whole group by left translations:

e_{i} (g) = L_{g}^{*} ξ_{i}, i = 1, . . ., n, g \in G .

Then

c_{j k}^{i}

are just the structure constants of the Lie algebra

G

of the group

G

and putting for simplicity

α_{i j k} = 〈 [e_{i}, e_{j}], e_{k} 〉 = c_{i j}^{k}

we derive

\begin{matrix} \nabla_{e_{k}} e_{j} = \frac{1}{2} \sum_{i} (α_{k j i} + α_{i k j} + α_{i j k}) e_{i} . \end{matrix}

(1)

In particular,

α_{i j k}

is a skew-symmetric tensor for a compact Lie group

G

with the Killing metric and the formula 1 reduces to

\nabla_{X} Y = \frac{1}{2} [X, Y]

where

X

and

Y

are left invariant vector fields. For details we refer to [8] .

2.2 The derivational equations

Let

Σ

be a surface immersed into

G

and

f : Σ \to G

be the immersion. Choose a conformal parameter

z = x + i y

Σ

(or, more precisely, in a domain of

Σ

) and denote by

I = e^{2 α} d z d \bar{z}

the induced metric.

Let us consider the pullback of

T G

to a

G

-bundle over

Σ

G \to E = f^{- 1} (T G) \overset{π}{\to} Σ

and the differential

d_{A} : Ω^{1} (Σ; E) \to Ω^{2} (Σ; E),

which acts on

E

-valued

1

-forms as follows. Let us write down a form

ω

ω = u d z + u^{*} d \bar{z} .

Then

d_{A} ω = d_{A}^{'} ω + d_{A}^{''} ω

where

d_{A}^{'} ω = - \nabla_{\bar{\partial} f} u d z \land d \bar{z}, d_{A}^{''} ω = \nabla_{\partial f} u^{*} d z \land d \bar{z} .

By straightforward computations we obtain the first derivational equation

\begin{matrix} d_{A} (d f) = 0 . \end{matrix}

(2)

By the definition of the tension vector

τ (f)

, we have

d_{A} (* d f) = f \cdot (e^{2 α} τ (f)) d x \land d y = \frac{i}{2} f \cdot (e^{2 α} τ (f)) d z \land d \bar{z}

where

f \cdot τ (f) = 2 H N

N

is the normal vector and

H

is the mean curvature.

Hence we obtain the second derivational equation:

\begin{matrix} d_{A} (* d f) = i e^{2 α} H N d z \land d \bar{z} . \end{matrix}

(3)

For

S = S U (2)

we expose this scheme in [12] (see also the derivation of the harmonicity equation for surfaces in Lie groups in [7] ).

Since the metric is left invariant we write down these equations in terms of

Ψ = f^{- 1} \partial f, Ψ^{*} = f^{- 1} \bar{\partial} f

\begin{matrix} \partial Ψ^{*} - \bar{\partial} Ψ + \nabla_{Ψ} Ψ^{*} - \nabla_{Ψ^{*}} Ψ = 0, \end{matrix}

(4)

\begin{matrix} \partial Ψ^{*} + \bar{\partial} Ψ + \nabla_{Ψ} Ψ^{*} + \nabla_{Ψ^{*}} Ψ = e^{2 α} H f^{- 1} (N) . \end{matrix}

(5)

The equation 4 is equivalent to 2 and the equation 5 is equivalent to 3 .

In the sequel we assume that the Lie group

G

is three-dimensional and we choose an orthonormal basis

e_{1}, e_{2}, e_{3}

for the inner product on the Lie algebra

G

of the Lie group

G

Let us decompose

Ψ

and

Ψ^{*}

in this basis as

Ψ = \sum_{k = 1}^{3} Z_{k} e_{k}, Ψ^{*} = \sum_{k = 1}^{3} {\bar{Z}}_{k} e_{k}

and rewrite the equations 4 and 5 in terms of

Z_{k}

k = 1, 2, 3

, as follows

\begin{matrix} \sum_{j} (\partial {\bar{Z}}_{j} - \bar{\partial} Z_{j}) e_{j} + \sum_{j, k} (Z_{j} {\bar{Z}}_{k} - {\bar{Z}}_{j} Z_{k}) \nabla_{e_{j}} e_{k} = 0, \end{matrix}

(6)

\begin{matrix} \begin{matrix} \sum_{j} (\partial {\bar{Z}}_{j} + \bar{\partial} Z_{j}) e_{j} + \sum_{j, k} (Z_{j} {\bar{Z}}_{k} + {\bar{Z}}_{j} Z_{k}) \nabla_{e_{j}} e_{k} = & 2 i H [({\bar{Z}}_{2} Z_{3} - Z_{2} {\bar{Z}}_{3}) e_{1} + ({\bar{Z}}_{3} Z_{1} - Z_{3} {\bar{Z}}_{1}) e_{2} + ({\bar{Z}}_{1} Z_{2} - Z_{1} {\bar{Z}}_{2}) e_{3}] . \end{matrix} \end{matrix}

(7)

Here we assumed that the basis

{e_{1}, e_{2}, e_{3}}

is positively oriented and therefore

f^{- 1} (N) = 2 i e^{- 2 α} [({\bar{Z}}_{2} Z_{3} - Z_{2} {\bar{Z}}_{3}) e_{1} + ({\bar{Z}}_{3} Z_{1} - Z_{3} {\bar{Z}}_{1}) e_{2} + ({\bar{Z}}_{1} Z_{2} - Z_{1} {\bar{Z}}_{2}) e_{3}]

(for

G = S U (2)

with the Killing metric this formula takes the form

f^{- 1} (N) = 2 i e^{- 2 α} [Ψ^{*}, Ψ]

. ^$1$ Since the parameter

z

is conformal we have

〈 Ψ, Ψ 〉 = 〈 Ψ^{*}, Ψ^{*} 〉 = 0, 〈 Ψ, Ψ^{*} 〉 = \frac{1}{2} e^{2 α}

which is rewritten as

Z_{1}^{2} + Z_{2}^{2} + Z_{3}^{2} = 0, | Z_{1} |^{2} + | Z_{2} |^{2} + | Z_{3} |^{2} = \frac{1}{2} e^{2 α} .

The first equality implies that the vector

Z

can be parameterized in the form

\begin{matrix} Z_{1} = \frac{i}{2} ({\bar{ψ}}_{2}^{2} + ψ_{1}^{2}), Z_{2} = \frac{1}{2} ({\bar{ψ}}_{2}^{2} - ψ_{1}^{2}), Z_{3} = ψ_{1} {\bar{ψ}}_{2} . \end{matrix}

(8)

^$1$ In [12] the term of $2$ in the right hand side of the same formula for $N$ (the formula (47)) was skipped however it was counted in all computations and that does not affect any result.

2.3 The Dirac operator and the energy of a surface

The Weierstrass representation is derived by substituting

ψ

into 6 and 7 (see [10, 12] for such representations of surfaces in

R^{3}

and

S^{3}

In this event the derivational equations takes the form of the Dirac equation

\begin{matrix} D ψ = [(\begin{matrix} 0 & \partial \\ - \bar{\partial} & 0 \end{matrix}) + (\begin{matrix} U & 0 \\ 0 & V \end{matrix})] ψ = 0 \end{matrix}

(9)

which is satisfied by

ψ = (\begin{matrix} ψ_{1} \\ ψ_{2} \end{matrix}) .

The first fundamental form is equal to

I = e^{2 α} d z d \bar{z} = (| ψ_{1} |^{2} + | ψ_{2} |^{2})^{2} d z d \bar{z}

and the left translation of the normal vector is

\begin{matrix} \begin{matrix} f^{- 1} (N) = e^{- α} [i (ψ_{1} ψ_{2} - {\bar{ψ}}_{1} {\bar{ψ}}_{2}) e_{1} - & (ψ_{1} ψ_{2} + {\bar{ψ}}_{1} {\bar{ψ}}_{2}) e_{2} + (| ψ_{2} |^{2} - | ψ_{1} |^{2}) e_{3})] . \end{matrix} \end{matrix}

(10)

Since

Ψ d z = (\sum Z_{k} e_{k}) d z

is a correctly defined

1

-form , we conclude that the

1

-forms

ψ_{1}^{2} d z, {\bar{ψ}}_{2}^{2} d z, ψ_{1} {\bar{ψ}}_{2} d z

are globally defined on the whole surface. By the Dirac equation, this implies that the potentials

U

and

V

are the squares of

d z d \bar{z}

forms and therefore

U V d z \land d \bar{z}

is a well-defined

2

-form on the surface (see [10] for details).

For a surface

M

immersed into

R^{3}

the potentials of the Dirac operator are real-valued and coincide:

U = V

. For such a surface the following equality holds:

E (M) = \int_{M} U V \frac{i d z \land d \bar{z}}{2} = \frac{1}{4} \int_{M} H^{2} d μ

where

d μ = e^{2 α} d x \land d y

is the induced measure on the surface and

H

is the mean curvature. We recall that the Willmore functional equals

W (M) = \int_{M} (H^{2} - K) d μ .

This observation from [10] was the starting point for treating the spectral quantities of this operator

D

as the geometric quantities and for physical explanation of the Willmore conjecture by using the spectral curves of immersed tori (see [10, 12] ).

Generically

U

and

V

do not necessarily coincide and we have to consider the energy of a closed surface as

\begin{matrix} E (M) = \int_{M} U V d x \land d y = \int_{M} U V \frac{i d z \land d \bar{z}}{2} . \end{matrix}

(11)

It is even not clear that it always be real-valued. We shall check that for some special cases. For that we shall use the following simple proposition which follows from the Dirac equation. Indeed we have

\bar{\partial} ψ_{1} = (R e V + i I m V) ψ_{2}, \bar{\partial} {\bar{ψ}}_{2} = (- R e U + i I m U) ψ_{1}

which implies

Proposition 1 Given

ψ

meeting the Dirac equation 9 , the following identity holds

\begin{matrix} \bar{\partial} (ψ_{1} {\bar{ψ}}_{2}) = (- R e U | ψ_{1} |^{2} + R e V | ψ_{2}^{2} |) + i (I m U | ψ_{1} |^{2} + I m V | ψ_{2} |^{2}) . \end{matrix}

(12)

2.4 Thurston's geometries on Lie groups $N i l, S o l$ and $\tilde{S L} 2$

By Thurston's theorem all three-dimensional maximal simply connected geometries

(X, I s o m X)

admitting compact quotients belongs to the following list:

1) three geometries with constant sectional curvature:

X = R^{3}, S^{3}

, or

H^{3}

; 2) a pair of product geometries:

X = S^{2} \times R

H^{2} \times R

; 3) three geometries modelled on Lie groups

N i l, S o l

, and

\tilde{S L} 2

with certain left invariant metrics.

We refer for details of this classification theorem and for the related famous Thurston geometrization conjecture for three-manifolds to [13, 9] .

In the paper we study surfaces in these geometries modelled on noncompact Lie groups

N i l, S o l

, and

\tilde{S L} 2

. Before going to the surface theory we briefly recall main facts about these geometries referring for an advanced exposition to [9] .

2.4.1 The group $N i l$

This group is formed by all matrices of the form

(\begin{matrix} 1 & x & z \\ 0 & 1 & y \\ 0 & 0 & 1 \end{matrix}), x, y, z \in R,

with the usual multiplication rule and with the left invariant metric

d s^{2} = d x^{2} + d y^{2} + (d z - x d y)^{2} .

The Lie algebra is generated by the elements

e_{1} = (\begin{matrix} 0 & 1 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{matrix}), e_{2} = (\begin{matrix} 0 & 0 & 0 \\ 0 & 0 & 1 \\ 0 & 0 & 0 \end{matrix}), e_{3} = (\begin{matrix} 0 & 0 & 1 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{matrix}),

which meet the following commutation relations

[e_{1}, e_{2}] = e_{3}, [e_{1}, e_{3}] = [e_{2}, e_{3}] = 0 .

We derive from 1 that the connection is as follows:

\begin{matrix} \begin{matrix} \nabla_{e_{1}} e_{2} = - \nabla_{e_{2}} e_{1} = \frac{1}{2} e_{3}, \nabla_{e_{1}} e_{3} = \nabla_{e_{3}} e_{1} = - \frac{1}{2} e_{2}, & \nabla_{e_{2}} e_{3} = \nabla_{e_{3}} e_{2} = \frac{1}{2} e_{1}, \nabla_{e_{1}} e_{1} = \nabla_{e_{2}} e_{2} = \nabla_{e_{3}} e_{3} = 0 . \end{matrix} \end{matrix}

(13)

2.4.2 The group $\tilde{S L} 2$

The group

G = \tilde{S L} 2

is the universal covering of the group

S L (2)

formed by all real

2 \times 2

-matrices with unit determinant. The group

P S L (2) = S L (2) / \pm 1

is the group of orientation preserving isometries of the hyperbolic plane

H^{2}

acting and it is diffeomorphic to the unit tangent bundle

U H^{2}

H^{2}

. Moreover the left invariant metric on

\tilde{S L} 2

corresponding to one of the Thurston geometries (i.e., with the maximal isometry group) is the pullback under the projections

\tilde{S L} 2 \to S L (2) \to P S L (2) \approx U H^{2}

of the metric on

U H^{2}

which is as follows.

Let us realize the hyperbolic plane as the disc

| z | < 1

on the complex plane with the metric

d s^{2} = \frac{4 d z d \bar{z}}{(1 - | z |^{2})^{2}}

and take on

U H^{2}

the metric

d l^{2} = \frac{4 (d z d \bar{z} + d φ^{2})}{(1 - | z |^{2})^{2}}

where

φ

is the angle coordinate on the unitary circles in the tangent spaces to

H^{2}

. We identify the generators of the Lie algebra of

\tilde{S L} 2

with the generators of the following one-parametric subgroups

g_{x} = (\begin{matrix} \frac{1}{\sqrt{1 - x^{2}}} & \frac{x}{\sqrt{1 - x^{2}}} \\ \frac{x}{\sqrt{1 - x^{2}}} & \frac{1}{\sqrt{1 - x^{2}}} \end{matrix}), g_{y} = (\begin{matrix} \frac{1}{\sqrt{1 - y^{2}}} & \frac{i y}{\sqrt{1 - y^{2}}} \\ - \frac{i y}{\sqrt{1 - y^{2}}} & \frac{1}{\sqrt{1 - y^{2}}} \end{matrix}),

g_{φ} = (\begin{matrix} e^{i φ / 2} & 0 \\ 0 & e^{- i φ / 2} \end{matrix})

which act on

H^{2}

by fractional linear transformations

z \to \frac{a z + b}{c z + d}, (\begin{matrix} a & b \\ c & d \end{matrix}) \in S L (2) .

The generators of these subgroups are

f_{1} = (\begin{matrix} 0 & 1 \\ 1 & 0 \end{matrix}), f_{2} = (\begin{matrix} 0 & i \\ - i & 0 \end{matrix}), f_{3} = (\begin{matrix} \frac{i}{2} & 0 \\ 0 & - \frac{i}{2} \end{matrix}) .

These generators meet the commutation relations

[f_{1}, f_{2}] = - 4 f_{3}, [f_{1}, f_{3}] = - f_{2}, [f_{2}, f_{3}] = f_{1} .

The desired metric on

\tilde{S L} 2

is induced by the inner product

〈 f_{j}, f_{k} 〉 = 4 δ_{j k}, j, k = 1, 2, 3,

on the Lie algebra of

\tilde{S L} 2

and, by 1 , this implies

\begin{matrix} \begin{matrix} \nabla_{e_{1}} e_{2} = - \nabla_{e_{2}} e_{1} = - e_{3}, \nabla_{e_{1}} e_{3} = e_{2}, \nabla_{e_{3}} e_{1} = \frac{3}{2} e_{2}, & \nabla_{e_{2}} e_{3} = - e_{1}, \nabla_{e_{3}} e_{2} = - \frac{3}{2} e_{1}, \nabla_{e_{1}} e_{1} = \nabla_{e_{2}} e_{2} = \nabla_{e_{3}} e_{3} = 0, \end{matrix} \end{matrix}

(14)

where

e_{j} = \frac{1}{2} f_{j}, 〈 e_{j}, e_{k} 〉 = δ_{j k}, j, k = 1, 2, 3 .

2.4.3 The group $S o l$

This group consists of all matrices of the form

(\begin{matrix} e^{- z} & 0 & x \\ 0 & e^{z} & y \\ 0 & 0 & 1 \end{matrix}), x, y, z \in R,

with the usual multiplication rule and with the left invariant metric

d s^{2} = e^{2 z} d x^{2} + e^{- 2 z} d y^{2} + d z^{2} .

The Lie algebra is generated by the elements

e_{1} = (\begin{matrix} 0 & 0 & 1 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{matrix}), e_{2} = (\begin{matrix} 0 & 0 & 0 \\ 0 & 0 & 1 \\ 0 & 0 & 0 \end{matrix}), e_{3} = (\begin{matrix} - 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 0 \end{matrix}),

which satisfy the commutation relations

[e_{1}, e_{2}] = 0, [e_{1}, e_{3}] = e_{1}, [e_{2}, e_{3}] = - e_{2} .

By 1 , we conclude that

\begin{matrix} \begin{matrix} \nabla_{e_{1}} e_{2} = \nabla_{e_{2}} e_{1} = 0, \nabla_{e_{1}} e_{3} = e_{1}, \nabla_{e_{3}} e_{1} = 0, & \nabla_{e_{2}} e_{3} = - e_{2}, \nabla_{e_{3}} e_{2} = 0, & \nabla_{e_{1}} e_{1} = - \nabla_{e_{2}} e_{2} = - e_{3}, \nabla_{e_{3}} e_{3} = 0 . \end{matrix} \end{matrix}

(15)

2.5 The curvature tensors of three-dimensional Lie groups

By 13 , 14 and 15 we compute the curvature tensor

〈 R (X, Y) Z, W 〉 = R_{l k j i} W^{l} Z^{k} Y^{j} X^{i} = 〈 (\nabla_{Y} \nabla_{X} - \nabla_{X} \nabla_{Y} + \nabla_{[X, Y]}) Z, W 〉

of the groups

N i l, \tilde{S L} 2

, and

S o l

. We skip all these simple computations presenting only the result which we need.

Proposition 2 For the groups

N i l, \tilde{S L} 2

, and

S o l

if there are three different indices among

i, j, k, l

then

R_{i j k l} = 0

. The other components of the curvature tensor are as follows

R_{1212} = {\begin{matrix} - \frac{3}{4} & for N i l \\ - 4 & for \tilde{S L} 2 \\ 1 & for S o l \end{matrix}, R_{1313} = R_{2323} = {\begin{matrix} \frac{1}{4} & for N i l \\ 1 & for \tilde{S L} 2 \\ - 1 & for S o l \end{matrix} .

Recall that if

| X | = | Y | = 1

and these vectors are linearly independent then

K_{X Y} = 〈 R (X, Y) X, Y 〉

equals the sectional curvature of the plane spanned by

X

and

Y

. This proposition admits us to compute the sectional curvature of any plane.

3 The Weierstrass representation for surfaces in Lie groups

3.1 A construction of a surface from $ψ$

In this section we derive the compatibility conditions for a vector function

ψ

, i.e. a criterion for

ψ

to correspond to an immersion of a surface into

N i l, \tilde{S L} 2

, or

S o l

. Given any function

ψ

meeting these conditions for a Lie group

G

from this list we may construct a surface as follows. ^$2$ Let

ψ

be defined on a surface

M

with a complex parameter

z

. Let us pick up a point

P \in M

First, we substitute

ψ

into the formula 8 for the components

Z_{1}, Z_{2}, Z_{3}

Ψ = \sum_{k = 1}^{3} Z_{k} e_{k} = f^{- 1} \partial f

. Second, we solve the following linear equation in the Lie group

G

f_{z} = f Ψ

with a certain initial data

f (P) = g \in G

. Thus we obtain the desired surface as the mapping

f : M \to G .

It follows from the derivation of the compatibility conditions in § 2.1 that any surface

f : M \to G

is obtained by this construction. The induced metric takes the form

d s^{2} = e^{2 α} d z d \bar{z}, e^{α} = (| ψ_{1} |^{2} + | ψ_{2} |^{2}),

and the induced measure on the surface

M

d μ = e^{2 α} \frac{i d z \land d \bar{z}}{2} .

The Hopf quadratic differential equals

ω = A d z^{2}, A = 〈 \nabla_{f_{z}} f_{z}, N 〉 .

Its explicit formula in terms of

ψ

depends on a Lie group and the Levi-Civita connection on it. Indeed, we have

\begin{matrix} \begin{matrix} A = 〈 Ψ_{z}, N 〉 + 〈 \sum_{j, k} Z_{j} Z_{k} \nabla_{e_{j}} e_{k}, N 〉 = & = ({\bar{ψ}}_{2} \partial ψ_{1} - ψ_{1} \partial {\bar{ψ}}_{2}) + 〈 \sum_{j, k} Z_{j} Z_{k} \nabla_{e_{j}} e_{k}, N 〉 . \end{matrix} \end{matrix}

(16)

The compatibility conditions takes the form of the Dirac equation 9 and describe

\bar{\partial} ψ_{1}

and

\partial ψ_{2}

D ψ = [(\begin{matrix} 0 & \partial \\ - \bar{\partial} & 0 \end{matrix}) + (\begin{matrix} U & 0 \\ 0 & V \end{matrix})] ψ = 0 .

Another derivatives which are

\partial ψ_{1}

and

\bar{\partial} ψ_{2}

are derived as follows. We differentiate

e^{α}

and obtain

α_{z} e^{α} = {\bar{ψ}}_{1} \partial ψ_{1} + ψ_{2} \partial {\bar{ψ}}_{2} + (ψ_{1} \partial {\bar{ψ}}_{1} + {\bar{ψ}}_{2} \partial ψ_{2}),

where, by the Dirac equation, the expression in brackets is written as

ψ_{1} \partial {\bar{ψ}}_{1} + {\bar{ψ}}_{2} \partial ψ_{2} = (\bar{V} - U) ψ_{1} {\bar{ψ}}_{2} .

Together with the formula for the Hopf differential this gives a system for

\partial ψ_{1}

and

\partial {\bar{ψ}}_{2}

(\begin{matrix} {\bar{ψ}}_{1} & ψ_{1} \\ {\bar{ψ}}_{2} & - ψ_{1} \end{matrix}) (\begin{matrix} \partial ψ_{1} \\ \partial {\bar{ψ}}_{2} \end{matrix}) = (\begin{matrix} α_{z} e^{α} + (U - \bar{V}) ψ_{1} {\bar{ψ}}_{2} \\ A - 〈 \sum_{j, k} Z_{j} Z_{k} \nabla_{e_{j}} e_{k}, N 〉 \end{matrix}) .

Resolving this system we obtain the expressions for

\partial ψ_{1}

and

\bar{\partial} ψ_{2}

which together with the Dirac equation give us the complete set of the Weingarten equations in terms on

ψ

\begin{matrix} (\partial - A) ψ = 0, (\bar{\partial} - ℬ) ψ = 0 . \end{matrix}

(17)

Now the Codazzi equations are derived from the zero-curvature conditions

\partial \bar{\partial} ψ_{k} = \bar{\partial} \partial ψ_{k}, k = 1, 2 .

In the sequel we call the function

ψ

the generating spinor of a surface (see [12] for an explanation of such a terminology).

^$2$ This generalizes the Weierstrass representation of surfaces in $R^{3}$ [11] and $S U (2)$ [12] for surfaces in these groups.

3.2 The group $N i l$

By 13 , the derivational equations 6 and 7 take the form

\begin{matrix} \begin{matrix} \partial {\bar{Z}}_{1} - \bar{\partial} Z_{1} = 0, \\ \partial {\bar{Z}}_{2} - \bar{\partial} Z_{2} = 0, \\ \partial {\bar{Z}}_{3} - \bar{\partial} Z_{3} + (Z_{1} {\bar{Z}}_{2} - {\bar{Z}}_{1} Z_{2}) = 0, \\ \partial {\bar{Z}}_{1} + \bar{\partial} Z_{1} + (Z_{2} {\bar{Z}}_{3} + {\bar{Z}}_{2} Z_{3}) = 2 i H ({\bar{Z}}_{2} Z_{3} - Z_{2} {\bar{Z}}_{3}), \\ \partial {\bar{Z}}_{2} + \bar{\partial} Z_{2} - (Z_{1} {\bar{Z}}_{3} + {\bar{Z}}_{1} Z_{3}) = 2 i H ({\bar{Z}}_{3} Z_{1} - Z_{3} {\bar{Z}}_{1}), \\ \partial {\bar{Z}}_{3} + \bar{\partial} Z_{3} = 2 i H ({\bar{Z}}_{1} Z_{2} - Z_{1} {\bar{Z}}_{2}) . \end{matrix} \end{matrix}

(18)

Substituting 8 into these formulas we rewrite the first pair of equations as

\partial ψ_{2}^{2} + \bar{\partial} ψ_{1}^{2} = 0,

and the pair consisting of the forth and fifth equations is equivalent to

\partial ψ_{2}^{2} - \bar{\partial} ψ_{1}^{2} + i ψ_{1} ψ_{2} (| ψ_{2} |^{2} - | ψ_{1} |^{2}) = - 2 H ψ_{1} ψ_{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) .

The system of these two equations at the points where

ψ_{1} ψ_{2} \neq 0

are written in the form of the Dirac equation which by continuity is satisfied on the whole surface:

\begin{matrix} \begin{matrix} D_{N i l} ψ = [(\begin{matrix} 0 & \partial \\ - \bar{\partial} & 0 \end{matrix}) + (\begin{matrix} U_{N i l} & 0 \\ 0 & V_{N i l} \end{matrix})] ψ = 0, & U_{N i l} = V_{N i l} = \frac{H}{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) + \frac{i}{4} (| ψ_{2} |^{2} - | ψ_{1} |^{2}), \end{matrix} \end{matrix}

(19)

where

H

is the mean curvature of the surface.

Remark 1. The tangent plane at a point where

Z_{3} = ψ_{1} {\bar{ψ}}_{2} = 0

is spanned by the left invariant fields generated by

e_{1}

and

e_{2}

. Since the commutator of this fields is transversal to their span (i.e.

[e_{1}, e_{2}] = e_{3}

) the left invariant distribution of

2

-planes generated by

e_{1}

and

e_{2}

is nowhere integrable. Therefore the equality

Z_{3} = 0

can not hold in open subsets of the surface. We indicate that the potentials of the Dirac operator are correctly defined everywhere and therefore the equation

D ψ = 0

is extended by continuity to the closure of

{Z_{3} \neq 0}

which, as we argue above, coincides with the whole surface. The same is true for surfaces in

S U (2)

and

\tilde{S L} 2

but does not hold in the case when

G = S o l

The Hopf differential takes the form

\begin{matrix} A = ({\bar{ψ}}_{2} \partial ψ_{1} - ψ_{1} \partial {\bar{ψ}}_{2}) + i ψ_{1}^{2} {\bar{ψ}}_{2}^{2} \end{matrix}

(20)

and the Weingarten equations 17 consists of the Dirac equation and the following system

\begin{matrix} \begin{matrix} \partial ψ_{1} = α_{z} ψ_{1} + A e^{- α} ψ_{2} - \frac{i}{2} ψ_{1}^{2} {\bar{ψ}}_{2}, & \bar{\partial} ψ_{2} = - \bar{A} e^{- α} ψ_{1} + α_{\bar{z}} ψ_{2} - \frac{i}{2} {\bar{ψ}}_{1} ψ_{2}^{2} . \end{matrix} \end{matrix}

(21)

It follows from the Weingarten equations that

(\partial - A) (\bar{\partial} - ℬ) ψ - (\bar{\partial} - ℬ) (\partial - A) ψ = (A_{\bar{z}} - ℬ_{z} + [A, ℬ]) ψ = 0 .

In the cases

G = R^{3}

S U (2)

the spinor

ψ^{*} = (- {\bar{ψ}}_{2}, ψ_{1})^{⊥}

meets the equation

R ψ^{*} = (A_{\bar{z}} - ℬ_{z} + [A, ℬ]) ψ^{*} = 0

which together with

R ψ = 0

implies that

R = A_{\bar{z}} - ℬ_{z} + [A, ℬ] = 0

(see [12] ). In the case of

G = N i l

as well as for

G = \tilde{S L} 2

S o l

the equation

R ψ^{*} = 0

does not hold and we have to derive the Codazzi equations more carefully.

We have

A = (\begin{matrix} α_{z} - \frac{i}{2} Z_{3} & A e^{- α} \\ - W & - \frac{i}{2} Z_{3} \end{matrix}), ℬ = (\begin{matrix} - \frac{i}{2} {\bar{Z}}_{3} & \bar{W} \\ - \bar{A} e^{- α} & α_{\bar{z}} - \frac{i}{2} {\bar{Z}}_{3} \end{matrix})

where

Z_{3} = ψ_{1} {\bar{ψ}}_{2}

and

W = (\frac{H}{2} - \frac{i}{4}) e^{α}

. The equation

R ψ = 0

is written as a system of two equations:

κ_{1} = ((α_{z \bar{z}} - | A |^{2} e^{- 2 α} + | W |^{2}) + \frac{i}{2} (\partial {\bar{Z}}_{3} - \bar{\partial} Z_{3})) ψ_{1} + (A_{\bar{z}} e^{- α} - {\bar{W}}_{z} + α_{z} \bar{W}) ψ_{2} = 0,

κ_{2} = ({\bar{A}}_{z} e^{- α} - W_{\bar{z}} + α_{\bar{z}} W) ψ_{1} + (- (α_{z \bar{z}} - | A |^{2} e^{- 2 α} + | W |^{2}) + \frac{i}{2} (\partial {\bar{Z}}_{3} - \bar{\partial} Z_{3})) ψ_{2} = 0 .

We recall that surfaces in

N i l

we have

\begin{matrix} \partial {\bar{Z}}_{3} - \bar{\partial} Z_{3} = - \frac{i}{2} (| ψ_{2} |^{4} - | ψ_{1} |^{4}), \partial {\bar{Z}}_{3} + \bar{\partial} Z_{3} = H (| ψ_{2} |^{4} - | ψ_{1} |^{4}) . \end{matrix}

(22)

Now the Codazzi equations is the system

κ_{1} {\bar{ψ}}_{1} - {\bar{κ}}_{2} ψ_{2} = 0, κ_{1} {\bar{ψ}}_{2} + {\bar{κ}}_{2} ψ_{1} = 0

which is rewritten as

α_{z \bar{z}} - | A |^{2} e^{- 2 α} + \frac{H^{2}}{4} e^{2 α} = \frac{1}{16} (3 | ψ_{1} |^{4} + 3 | ψ_{2} |^{4} - 10 | ψ_{1} |^{2} | ψ_{2} |^{2}),

A_{\bar{z}} - \frac{H_{z}}{2} e^{2 α} + \frac{1}{2} (| ψ_{2} |^{4} - | ψ_{1} |^{4}) ψ_{1} {\bar{ψ}}_{2} = 0 .

By using 22 we rewrite this system as follows

\begin{matrix} \begin{matrix} α_{z \bar{z}} - e^{- 2 α} | A |^{2} + \frac{1}{4} e^{2 α} H^{2} = \frac{3}{16} e^{2 α} - | Z_{3} |^{2}, & \bar{\partial} (A + \frac{{Z_{3}}^{2}}{2 H + i}) = \frac{1}{2} H_{z} e^{2 α} + \bar{\partial} (\frac{1}{2 H + i}) {Z_{3}}^{2} . \end{matrix} \end{matrix}

(23)

We summarize these results in the following

Theorem 1 For a surface in

G = N i l

its generating spinor

ψ

meets the Dirac equation 19 .

Moreover any function

ψ

meeting 19 serves as the generating spinor of a surface in

N i l

The Weingarten equations for the surface are given by 19 and 21 . The Hopf differential takes the form 20 and the Codazzi equations are given by 23 .

Corollary 1 The generating spinor

ψ

of a minimal surface in

N i l

meets the following equations

\bar{\partial} ψ_{1} = \frac{i}{4} (| ψ_{2} |^{2} - | ψ_{1} |^{2}) ψ_{1}, \partial ψ_{2} = - \frac{i}{4} (| ψ_{2} |^{2} - | ψ_{1} |^{2}) ψ_{1} .

Corollary 2 (Abresch) For a constant mean curvature surface in

N i l

the quadratic differential

\tilde{A} d z^{2} = (A + \frac{{Z_{3}}^{2}}{2 H + i}) d z^{2}

is holomorphic.

Proposition 3 If the differential

\tilde{A} d z^{2}

is holomorphic then the surface in

N i l

has constant mean curvature.

Proof. We prove that by contradiction. Let us assume that

H_{z} \neq 0

in some domain. Then we have

\frac{1}{2} e^{2 α} = \frac{2 H_{\bar{z}}}{(2 H + i)^{2}} Z_{3}^{2}

and which implies the equality for the moduli of both sides:

\frac{1}{2} e^{2 α} = \frac{2 | ψ_{1} |^{2} | ψ_{2} |^{2}}{4 H^{2} + 1} .

e^{α} = | ψ_{1} |^{2} + | ψ_{2} |^{2}

, this is rewritten as

(4 H^{2} + 1) (| ψ_{1} |^{+} | ψ_{2} |^{2})^{2} = 4 | ψ_{1} |^{2} | ψ_{2} |^{2}

which implies

4 H^{2} e^{2 α} + (| ψ_{1} |^{2} - | ψ_{2} |^{2})^{2} = 0 .

It is clear that this equality is valid if and only if

| ψ_{1} | = | ψ_{2} |

and

H = 0

, i.e.

the surface is minimal. This contradiction proves the proposition.

Finally we compute the energy of a compact surface in

N i l

Proposition 4 Given a closed oriented surface

M

in the group

N i l

, its energy is real-valued and equals

\begin{matrix} E (M) = \int_{M} (\frac{H^{2}}{4} (| ψ_{1} |^{2} + | ψ_{2} |^{2})^{2} - \frac{1}{16} (| ψ_{2} |^{2} - | ψ_{1} |^{2})^{2}) \frac{i d z \land d \bar{z}}{2} \end{matrix}

(24)

Proof. Let us substitute the formulas for

U_{N i l}

and

V_{N i l}

into the identity 12 and obtain

\bar{\partial} (ψ_{1} {\bar{ψ}}_{2}) = \frac{H}{2} (| ψ_{2} |^{2} - | ψ_{1} |^{2}) + i \frac{1}{4} (| ψ_{2} |^{4} - | ψ_{1} |^{4}) .

By the Stokes formula we have

\begin{matrix} \int_{M} \frac{H}{2} (| ψ_{2} |^{2} - | ψ_{1} |^{2}) \frac{i d z \land d \bar{z}}{2} + i \int_{M} \frac{1}{4} (| ψ_{2} |^{4} - | ψ_{1} |^{4}) \frac{i d z \land d \bar{z}}{2} = 0 . \end{matrix}

(25)

The real part of the left hand side of this formula is a multiple of

I m E (M)

which implies that the energy is real-valued. Now we substitute

U_{N i l}

and

V_{N i l}

into

R e E (M)

and derive 24 . This proves the proposition.

Notice that the vanishing of the the imaginary part of the left hand side of 25 together with 10 implies the equality

\int_{M} 〈 f^{- 1} (N), e_{3} 〉 d μ = \int_{M} (| ψ_{2} |^{4} - | ψ_{1} |^{4}) \frac{i d z \land d \bar{z}}{2} = 0 .

Proposition 5 The energy of a surface

M

in the group

N i l

equals

\begin{matrix} E (M) = \frac{1}{4} \int_{M} (H^{2} + \frac{\hat{K}}{4} - \frac{1}{16}) d μ, \end{matrix}

(26)

where

\hat{K}

is the sectional curvature of the tangent plane at a point.

Proof. It is computed by using Proposition 2 that the sectional curvature of a tangent plane at the unit of group equals

\hat{K} = \frac{1}{4} - {cos}^{2} φ

where

φ

is the angle between the normal to the plane and

e_{3}

. By 10 , the integrand in 24 is equal to

\frac{1}{4} e^{2 α} H^{2} - \frac{1}{16} (| ψ_{2} |^{2} - | ψ_{1} |^{2})^{2} = \frac{1}{4} e^{2 α} H^{2} - \frac{1}{16} e^{2 α} 〈 f^{- 1} (N), e_{3} 〉^{2} =

\frac{1}{4} e^{2 α} (H^{2} - \frac{1}{4} {cos}^{2} φ) = \frac{1}{4} e^{2 α} (H^{2} - \frac{1}{4} (\frac{1}{4} - \hat{K})) =

= \frac{1}{4} e^{2 α} (H^{2} + \frac{1}{4} \hat{K} - \frac{1}{16}) .

The proposition is proved.

3.3 The group $\tilde{S L} 2$

Substituting 14 into the derivational equations 6 and 7 we obtain

\begin{matrix} \begin{matrix} \partial {\bar{Z}}_{1} - \bar{\partial} Z_{1} + \frac{1}{2} (Z_{2} {\bar{Z}}_{3} - {\bar{Z}}_{2} Z_{3}) = 0, \\ \partial {\bar{Z}}_{2} - \bar{\partial} Z_{2} + \frac{1}{2} (Z_{3} {\bar{Z}}_{1} - {\bar{Z}}_{3} Z_{1}) = 0, \\ \partial {\bar{Z}}_{3} - \bar{\partial} Z_{3} - 2 (Z_{1} {\bar{Z}}_{2} - {\bar{Z}}_{1} Z_{2}) = 0, \\ \partial {\bar{Z}}_{1} + \bar{\partial} Z_{1} - \frac{5}{2} (Z_{2} {\bar{Z}}_{3} + {\bar{Z}}_{2} Z_{3}) = 2 i H ({\bar{Z}}_{2} Z_{3} - Z_{2} {\bar{Z}}_{3}), \\ \partial {\bar{Z}}_{2} + \bar{\partial} Z_{2} + \frac{5}{2} (Z_{1} {\bar{Z}}_{3} + {\bar{Z}}_{1} Z_{3}) = 2 i H ({\bar{Z}}_{3} Z_{1} - Z_{3} {\bar{Z}}_{1}), \\ \partial {\bar{Z}}_{3} + \bar{\partial} Z_{3} = 2 i H ({\bar{Z}}_{1} Z_{2} - Z_{1} {\bar{Z}}_{2}) . \end{matrix} \end{matrix}

(27)

By substituting 8 into these equations we reduce the system of first two of them to

\partial ψ_{2}^{2} + \bar{\partial} ψ_{1}^{2} - \frac{i}{2} ψ_{1} ψ_{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) = 0,

and the fourth and fifth equations together are reduced to

\partial ψ_{2}^{2} - \bar{\partial} ψ_{1}^{2} + \frac{5}{2} i ψ_{1} ψ_{2} (| ψ_{1} |^{2} - | ψ_{2} |^{2}) = - 2 H ψ_{1} ψ_{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) .

These two equations are written in terms of the Dirac equation (the derivation of that is completely similar to the

N i l

case, see 19 in § 3.2 ]:

\begin{matrix} \begin{matrix} D_{S L} ψ = [(\begin{matrix} 0 & \partial \\ - \bar{\partial} & 0 \end{matrix}) + (\begin{matrix} U_{S L} & 0 \\ 0 & V_{S L} \end{matrix})] ψ = 0, & U_{S L} = \frac{H}{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) + i (\frac{1}{2} | ψ_{1} |^{2} - \frac{3}{4} | ψ_{2} |^{2}), & V_{S L} = \frac{H}{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) + i (\frac{3}{4} | ψ_{1} |^{2} - \frac{1}{2} | ψ_{2} |^{2}), \end{matrix} \end{matrix}

(28)

where

H

is the mean curvature of the surface.

The Hopf differential equals

\begin{matrix} A = ({\bar{ψ}}_{2} \partial ψ_{1} - ψ_{1} \partial {\bar{ψ}}_{2}) - \frac{5 i}{2} ψ_{1}^{2} {\bar{ψ}}_{2}^{2} \end{matrix}

(29)

and we have

\begin{matrix} \begin{matrix} \partial ψ_{1} = α_{z} ψ_{1} + A e^{- α} ψ_{2} + \frac{5 i}{4} ψ_{1}^{2} {\bar{ψ}}_{2}, & \bar{\partial} ψ_{2} = - \bar{A} e^{- α} ψ_{1} + α_{\bar{z}} ψ_{2} + \frac{5 i}{4} {\bar{ψ}}_{1} ψ_{2}^{2} . \end{matrix} \end{matrix}

(30)

The matrices

A

and

ℬ

are given by the formulas

A = (\begin{matrix} α_{z} + \frac{5 i}{4} Z_{3} & A e^{- α} \\ - W & \frac{5 i}{4} Z_{3} \end{matrix}), ℬ = (\begin{matrix} \frac{5 i}{4} {\bar{Z}}_{3} & \bar{W} \\ - \bar{A} e^{- α} & α_{\bar{z}} + \frac{5 i}{4} {\bar{Z}}_{3} \end{matrix})

where

W = \frac{1}{2} (H + i) e^{α}

. The equation

R ψ = 0

is equivalent to the equations

κ_{1} = ((α_{z \bar{z}} - | A |^{2} e^{- 2 α} + | W |^{2}) - \frac{5 i}{4} (\partial {\bar{Z}}_{3} - \bar{\partial} Z_{3})) ψ_{1} + (A_{\bar{z}} e^{- α} - {\bar{W}}_{z} + α_{z} \bar{W}) ψ_{2} = 0,

κ_{2} = ({\bar{A}}_{z} e^{- α} - W_{\bar{z}} + α_{\bar{z}} W) ψ_{1} + (- (α_{z \bar{z}} - | A |^{2} e^{- 2 α} + | W |^{2}) - \frac{5 i}{4} (\partial {\bar{Z}}_{3} - \bar{\partial} Z_{3})) ψ_{2} = 0 .

Since the generating spinors of surfaces in

\tilde{S L} 2

meet the equalities

\partial {\bar{Z}}_{3} - \bar{\partial} Z_{3} = i (| ψ_{2} |^{4} - | ψ_{1} |^{4}), \partial {\bar{Z}}_{3} + \bar{\partial} Z_{3} = H (| ψ_{2} |^{4} - | ψ_{1} |^{4}) .

we rewrite the equations

κ_{1} {\bar{ψ}}_{1} - {\bar{κ}}_{2} ψ_{2} = 0

and

κ_{1} {\bar{ψ}}_{2} + {\bar{κ}}_{2} ψ_{1} = 0

as the following system

\begin{matrix} \begin{matrix} α_{z \bar{z}} - e^{- 2 α} | A |^{2} + \frac{1}{4} e^{2 α} H^{2} = e^{2 α} - 5 | Z_{3} |^{2}, & \bar{\partial} (A + \frac{5 Z_{3}^{2}}{2 (H - i)}) = \frac{1}{2} H_{z} e^{2 α} + \bar{\partial} (\frac{5}{2 (H - i)}) {Z_{3}}^{2} . \end{matrix} \end{matrix}

(31)

We derive

Theorem 2 Given a surface in

G = \tilde{S L} 2

, its generating spinor

ψ

satisfies the Dirac equation 28 .

Any function

ψ

meeting 28 is the generating spinor of a surface in

\tilde{S L} 2

The Weingarten equations for the surface is the system consisting of 28 and 30 . The Hopf differential is given by the formula 20 and the Codazzi equations take the form 23 .

Corollary 3 The generating spinor of a minimal surface in

\tilde{S L} 2

satisfies the equations

\bar{\partial} ψ_{1} = i (\frac{3}{4} | ψ_{1} |^{2} - \frac{1}{2} | ψ_{2} |^{2}) ψ_{2}, \partial ψ_{2} = - i (\frac{1}{2} | ψ_{1} |^{2} - \frac{3}{4} | ψ_{2} |^{2}) ψ_{1} .

Corollary 4 (Abresch) For a constant mean curvature surface in

\tilde{S L} 2

the quadratic differential

\tilde{A} d z^{2} = (A + \frac{5}{2 (H - i)} Z_{3}^{2}) d z^{2}

is holomorphic.

Remark 2. The method of proving Proposition 3 does not work in the case of

\tilde{S L} 2

and we do not know if the holomorphicity of

\tilde{A}

implies that the surface has constant mean curvature.

We finish our study of surfaces in

\tilde{S L} 2

by computing the energy functional.

Proposition 6 Given a closed oriented surface

M

in the group

\tilde{S L} 2

, its energy is real-valued and equals

\begin{matrix} \begin{matrix} E (M) = \int_{M} [\frac{H^{2}}{4} (| ψ_{1} |^{2} + | ψ_{2} |^{2})^{2} - & (\frac{1}{2} | ψ_{1} |^{2} - \frac{3}{4} | ψ_{2} |^{2}) (\frac{3}{4} | ψ_{1} |^{2} - \frac{1}{2} | ψ_{2} |^{2})] \frac{i d z \land d \bar{z}}{2} . \end{matrix} \end{matrix}

(32)

The proof of this proposition is straightforward and is analogous to the proof of Proposition 4 . Hence we skip it and only mention that as for surfaces in

N i l

we have

\int_{M} 〈 f^{- 1} (N), e_{3} 〉 d μ = \int_{M} (| ψ_{2} |^{4} - | ψ_{1} |^{4}) \frac{i d z \land d \bar{z}}{2} = 0

As for the group

N i l

we can write down the formula for the energy in common geometric terms.

Proposition 7 The energy of a surface

M \subset \tilde{S L} 2

equals

\begin{matrix} E (M) = \frac{1}{4} \int_{M} (H^{2} + \frac{5}{16} \hat{K} - \frac{1}{4}) d μ, \end{matrix}

(33)

where

\hat{K}

is the sectional curvature of the tangent plane at a point.

This proposition is derived straightforwardly from 24 and Proposition 2 .

3.4 The group $S o l$

By 15 , the derivational equations 6 and 7 for surfaces in the group

S o l

take the form

\begin{matrix} \begin{matrix} \partial {\bar{Z}}_{1} - \bar{\partial} Z_{1} + (Z_{1} {\bar{Z}}_{3} - {\bar{Z}}_{1} Z_{3}) = 0, & \partial {\bar{Z}}_{2} - \bar{\partial} Z_{2} - (Z_{2} {\bar{Z}}_{3} - {\bar{Z}}_{2} Z_{3}) = 0, & \partial {\bar{Z}}_{3} - \bar{\partial} Z_{3} = 0, & \partial {\bar{Z}}_{1} + \bar{\partial} Z_{1} + (Z_{1} {\bar{Z}}_{3} + {\bar{Z}}_{1} Z_{3}) = 2 i H ({\bar{Z}}_{2} Z_{3} - Z_{2} {\bar{Z}}_{3}), & \partial {\bar{Z}}_{2} + \bar{\partial} Z_{2} - (Z_{2} {\bar{Z}}_{3} + {\bar{Z}}_{2} Z_{3}) = 2 i H ({\bar{Z}}_{3} Z_{1} - Z_{3} {\bar{Z}}_{1}), & \partial {\bar{Z}}_{3} + \bar{\partial} Z_{3} - 2 (| Z_{1} |^{2} - | Z_{2} |^{2}) = 2 i H ({\bar{Z}}_{1} Z_{2} - Z_{1} {\bar{Z}}_{2}) . \end{matrix} \end{matrix}

(34)

As in § 3.2 and § 3.3 we substitute 8 in these equations and derive the following pair of equations

\partial ψ_{2}^{2} + \bar{\partial} ψ_{1}^{2} - {\bar{ψ}}_{1} {\bar{ψ}}_{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) = 0,

\partial ψ_{2}^{2} - \bar{\partial} ψ_{1}^{2} + {\bar{ψ}}_{1} {\bar{ψ}}_{2} (| ψ_{1} |^{2} - | ψ_{2} |^{2}) = - 2 H ψ_{1} ψ_{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) .

We see that at points where

Z_{3} = ψ_{1} {\bar{ψ}}_{2} \neq 0

these equations are rewritten in terms of the Dirac equation

\begin{matrix} \begin{matrix} D_{S o l} ψ = [(\begin{matrix} 0 & \partial \\ - \bar{\partial} & 0 \end{matrix}) + (\begin{matrix} U_{S o l} & 0 \\ 0 & V_{S o l} \end{matrix})] ψ = 0, & U_{S o l} = \frac{H}{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) + \frac{1}{2} {\bar{ψ}}_{2}^{2} \frac{{\bar{ψ}}_{1}}{ψ_{1}}, & V_{S o l} = \frac{H}{2} (| ψ_{1} |^{2} + | ψ_{2} |^{2}) + \frac{1}{2} {\bar{ψ}}_{1}^{2} \frac{{\bar{ψ}}_{2}}{ψ_{2}} . \end{matrix} \end{matrix}

(35)

Remark 3. Since the left invariant vector fields generated by

e_{1}

and

e_{2}

commute the equation

Z_{3} = ψ_{1} {\bar{ψ}}_{2} = 0

can be valid in an open subset

B

of a surface. Therefore the Dirac equation can not be extended by continuity onto the whole surface and does not describe

ψ

B

. Since

H = 0

B

it is reasonable to assume that

U_{S o l} = V_{S o l} = 0

inside such a domain. However on the boundaries of the set

{Z_{3} \neq 0}

the potentials

U_{S o l}

and

V_{S o l}

are not always correctly defined due to the indeterminacy of

\frac{{\bar{ψ}}_{1}}{ψ_{1}}

and

\frac{{\bar{ψ}}_{2}}{ψ_{2}}

. The measure of the indeterminacy set is zero and we may correctly define the energy of a surface in

S o l

E (M) = \int_{{Z_{3} \neq 0}} U_{S o l} V_{S o l} d μ .

We did not manage to rewrite this quantity in common geometric terms as we did for surfaces in

N i l

and

\tilde{S L} 2

and its geometric meaning stays unclear for us until now. We even do not know is the energy real-valued (as for surfaces in

S U (2), N i l

, and

\tilde{S L} 2

) or not.

The Hopf differential of a surface in

S o l

equals

\begin{matrix} A = ({\bar{ψ}}_{2} \partial ψ_{1} - ψ_{1} \partial {\bar{ψ}}_{2}) + \frac{1}{2} ({\bar{ψ}}_{2}^{4} - ψ_{1}^{4}) \end{matrix}

(36)

and we complete the Weingarten equations by the following system

\begin{matrix} \begin{matrix} \partial ψ_{1} = α_{z} ψ_{1} + A e^{- α} ψ_{2} - \frac{1}{2} {\bar{ψ}}_{2}^{3}, & \bar{\partial} ψ_{2} = - \bar{A} e^{- α} ψ_{1} + α_{\bar{z}} ψ_{2} - \frac{1}{2} {\bar{ψ}}_{1}^{3} . \end{matrix} \end{matrix}

(37)

The Codazzi equations take the form

\begin{matrix} \begin{matrix} α_{z \bar{z}} - e^{- 2 α} | A |^{2} + \frac{1}{4} e^{2 α} H^{2} = \frac{1}{4} (6 | ψ_{1} |^{2} | ψ_{2} |^{2} - (| ψ_{1} |^{4} + | ψ_{2} |^{4})), & A_{\bar{z}} - \frac{1}{2} H_{z} e^{2 α} = (| ψ_{2} |^{4} - | ψ_{1} |^{4}) ψ_{1} {\bar{ψ}}_{2} . \end{matrix} \end{matrix}

(38)

We summarize these results in

Theorem 3 Let

f : M \to S o l

be a surface. Let us denote by

B

the subset of

M

where

Z_{3} = 〈 f^{- 1} f_{z}, e_{3} 〉 = 0

, denote by

B_{0}

the interior of

B

, and denote by

C

the subset of

M

where

Z_{3} \neq 0

. We have

M = B_{0} \cup \bar{C}

, the set

B \ B_{0}

lies in the closure

\bar{C}

C

and is a zero measure set.

Then the generating spinor

ψ

M

meets the Dirac equation 35 in

C

and meets the Dirac equation with zero potentials:

\bar{\partial} ψ_{1} = \partial ψ_{2} = 0

, in

B_{0}

Any function

ψ

meeting 35 is some set

D \subset M

is the generating function of a surface

f : D \to S o l

The Hopf differential is given by 36 . In

B_{0} \cup C

the Weingarten equations are formed by 37 and by the Dirac equation for

ψ

. The Codazzi equations take the form 38 .

Corollary 5 The generating spinor

ψ

of a minimal surface in

S o l

meets the equations

\bar{\partial} ψ_{1} = \frac{1}{2} {\bar{ψ}}_{1}^{2} {\bar{ψ}}_{2}, \partial ψ_{2} = - \frac{1}{2} {\bar{ψ}}_{1} {\bar{ψ}}_{2}^{2} .

References

Abresch, U.: Generalized Hopf differentials, to appear in Proceedings of the 13th School of Differential geometry, Mat. Contemp., 2004.
Abresch, U., and Rosenberg, H.: The Hopf differential for constant mean curvature surfaces in $S^{2} \times R$ and $H^{2} \times R$ , Acta Mathematica (to appear).
Daniel, B.: Isometric immersions into $3$ -dimensional homogeneous manifolds, arxiv: math.DG/0503500.
Figueroa, C., Mercuri, F., and Pedrosa, R.: Invariant surfaces of the Heisenberg groups, Ann. Math. Pura Appl. 177 (1999), 173–194.
Fokas, A.S., and Gelfand, I.M.: Surfaces on Lie groups, on Lie algebras, and their integrability. Comm. Math. Phys. 177 (1996), 203–220.
Haskins, M.: The geometric complexity of special Lagrangian $T^{2}$ -cones, arxiv: math.DG/0307129.
Hitchin, N.: Harmonic maps from a $2$ -torus to the $3$ -sphere. J. Differ. Geom. 31 (1990), 627–710.
Milnor, J.W.: Curvatures of left invariant metrics on Lie groups. Adv. Math. 21 (1976), 293–329.
Scott, P.: The geometries of $3$ -manifolds. Bull. London Math. Soc. 56 (1983), 401–487.
Taimanov, I.A.: Modified Novikov–Veselov equation and differential geometry of surfaces. Amer. Math. Soc. Transl., Ser. 2, V. 179, 1997, pp. 133–151.
Taimanov, I. A.: The Weierstrass representation of closed surfaces in $R^{3}$ . Functional Anal. Appl. 32:4 (1998), 49–62.
Taimanov, I.A.: Dirac operators and conformal invariants of tori in $3$ -space. Proc. Steklov Inst. Math. 244 (2004), 233–263.
Thurston, W.P.: Three dimensional manifolds, Kleinian groups and hyperbolic geometry. Bull. Amer. Math. Soc. 6 (1982), 357–381.

Surfaces in three-dimensional Lie groups The work is supported by RFBR (grant no. 03-01-00403) and by the programm of fundamental researches of RAS “Mathematical methods in nonlinear dynamics”.

Dmitry A. Berdinsky Department of Mechanics and Mathematics, Novosibirsk State University, 630090 Novosibirsk, Russia; e-mail: mitrich@gorodok.net.

Iskander A. Taimanov Institute of Mathematics, 630090 Novosibirsk, Russia; e-mail: taimanov@math.nsc.ru.