Discrete Isothermic Surfaces

Alexander Bobenko y Ulrich Pinkall z Fachbereich Mathematik, Technische Universitat Berlin, Strasse des 17. Juni 136, 10623 Berlin, Germany November 10, 1994
Abstract

Discrete isothermic surfaces are de ned. It is shown that these discrete surfaces possess properties, which are characteristic for smooth isothermic surfaces (Mobius invariance, dual surface). Quaternionic zero-curvature loop group representations for smooth and discrete isothermic surfaces are presented. A Weierstrass type representation for discrete minimal isothermic surfaces is obtained and the discrete catenoid and the Enneper surface are constructed.

1 Introduction
Surfaces in an Euclidean 3-space studied by the theory of integrable systems are usually described as immersions of R2 or of some domain in R2

F : R2 ! R3:
partially supported by the Sonderforschungsbereich 288 e-mail: bobenko@sfb288.math.tu-berlin.de e-mail: pinkall@sfb288.math.tu-berlin.de

(1)

The list includes surfaces with constant negative Gaussian and mean curvature and some others (see the survey 2]). In this case R2 is treated as
y z

a space-time of the corresponding integrable system. One can describe topological planes, cylinders and tori in this way. To study surfaces of more complicated topology one should deal with immersions

F : R ! R3 ;

(2)

where R is some Riemann surface. The traditional theory of integrable systems needs essential improvements to be applied to this case and up to now there exists no real progress. This paper is part of our general program on the discretization of surfaces described by integrable systems. By such a surface we mean a map

F : G ! R3 ;

(3)

where G is some graph, which is treated as a discrete space-time of the corresponding discrete integrable system. When the size of the edges tends to zero such a surface approximates a corresponding smooth surface (2). The case closest to (1) F : Z2 ! R3 (4) is relatively well understood. In 3], 4] the discrete surfaces (1) with constant Gaussian and mean curvature were de ned and their geometrical properties were investigated (for the loop group factorization interpretation of these surfaces see 9]). We hope that the integrable discretization (3) may become a method to understand the structure of its corresponding smooth limit (2). Besides this, the elementary geometry of the case (4) itself is very nice and worth studying. The present paper was initiated by the recent preprint of J. Cieslinski, P. Goldshtein and A. Sym 7]. These authors, based on classical results of P. Calapso 6] and W. Blaschke 1], give a characterization of isothermic surfaces as "soliton surfaces" by introducing a spectral parameter. The geometrical meaning of this spectral parameter was clari ed in 5], where it was shown how pairs of isothermic surfaces are given by curved ats in a pseudo-Riemannian symmetric space and vice versa. Here, using the spinor representation of SO(4; 1), in Section 3 we rewrite the zero-curvature representation for the associated family of isothermic surfaces in terms of 2 2 matrices with quaternionic coe cients. A natural discretization of this representation yields the following de nition of discrete isothermic surfaces. 2

De nition 1 A discrete isothermic surface is a map F : Z2 ! R3 such that

the cross-ratios qn;m = q (Fn;m; Fn+1;m ; Fn+1;m+1 ; Fn;m+1 ) of all elementary quadrilaterals are real qn;m 2 R and satisfy the factorization condition qn;mqn+1;m+1 = qn+1;m qn;m+1: (5) This de nition coincides with the de nition (35) in the Remark after De nition 6 in Section 4. Geometrical properties of discrete isothermic surfaces as well as the analytical description of the corresponding associated family are studied in Sections 4, 5. In Section 7 we de ne discrete isothermic minimal surfaces, derive a Weierstrass type representation for them and construct some simple examples.

Let F be a smooth surface in a 3-dimensional Euclidean space and F (x; y) F = (F1; F2; F3) : R ! R3; a conformal parametrization of F < Fx; Fx >=< Fy ; Fy >= eu; < Fx; Fy >= 0: (6) Here the brackets denote the scalar product < a; b >= a1b1 + a2b2 + a3b3; and Fx and Fy are the partial derivatives @F=@x and @F=@y. The vectors Fx; Fy as well as the normal N , < Fx; N >=< Fy ; N >= 0; < N; N >= 1; (7) de ne a moving frame on the surface. We consider a local theory: R is a domain in R2. If the second fundamental form is diagonal < dF; dN >= eu(k1dx2 + k2dy2) (8) with respect to the induced metric < dF; dF >= eu(dx2 + dy2); (9) 3

2 Isothermic surfaces and their properties

then the parametrization is called isothermic. In this case k1 and k2 are the principal curvatures and the preimages of the curvature lines are the lines x = const and y = const in the parameter domain.

De nition 2 Surfaces which admit isothermic coordinates are called isothermic.

Surfaces of revolution, quadrics and constant mean curvature surfaces without umbilics are examples of isothermic surfaces. We consider isothermic immersions, i.e. suppose that u(x; y) is nite. Due to (6, 7) the moving frame = (Fx; Fy ; N )T satis es the following Gauss-Weingarten equations: 1 1 0 0 uy =2 ux=2 0 C ux=2 uy =2 k1eu C B 0 A ; y = B ux=2 uy =2 k2eu A : (10) @ x = @ uy =2 ux =2 0 k2 0 k1 0 0 The compatibility conditions of this system are the Gauss-Codazzi equations uxx + uyy + 2k1 k2eu = 0; 2k2 x (k1 k2)ux = 0; (11) 2k1 y + (k1 k2)uy = 0: Our goal is to nd a proper discrete version of isothermic surfaces. Let us mention two important geometrical properties of isothermic surfaces which we want to retain in the discrete case. Property 1 (Mobius invariance). Let F : R ! R3 be an isothermic im~ mersion and M a Mobius transformation of Euclidean 3-space. Then F 3 is also isothermic. M F :R!R Proof. Isothermic coordinates (8),(9) can be characterized in terms of F only jjFxjj = jjFy jj; < Fx; Fy >= 0; Fxy 2 span fFx; Fy g: (12) Clearly, if is enough to prove Property 1 for the case of an inversion M F ~ F = < F; F > : 4

~ ~ ~ Calculating Fx; Fy we see the conformality of F . A direct calculation proves ~ the third property of (12) of F > ~ > ~ ~ Fxy = ( 2 < F; Fy > )Fx + ( 2 < F; Fx > )Fy ; < F; F < F; F where and are de ned by

Fxy = Fx + Fy:

Property 2 (Dual surface). Let F : R ! R3 be an isothermic immersion. Then the immersion F : R ! R3 de ned by the formulas
Fx = e u Fx ; Fy = e u Fy
is isothermic. The Gauss maps of F and F are antipodal

(13)

N= N :
The fundamental forms of F are as follows

< dF ; dF > = e u (dx2 + dy2); < dF ; dN > = k1dx2 + k2dy2: F.

(14)

De nition 3 The immersion F : R ! R3 de ned above is called dual to

Proof of Property 2. The de nition (13) of F makes sense since the equality 1 Fxy = Fyx is equivalent to (e u Fx)y = 2 e u(uxFy uy Fx) = ( e uFy )x. Here we use (13) and the Gauss-Weingarten equation for Fxy . The conformality of F is evident. The expressions (14) are obtained by straightforward calculation. Remark F = F .

To discretize isothermic surfaces (see Sect.5) and for investigation by analytical methods it is more convenient to use 2 2 matrices instead of 3 3 matrices (10), therefore rst we rewrite equations (10) for the moving frame in terms of quaternions. 5

Let us denote the algebra of quaternions by H, the multiplicative quaternion group by H = H n f0g and the standard basis of H by f1; i; j; kg ij = k; jk = i; ki = j: (15) Using the standard matrix representation of H the Pauli matrices are related to this basis as follows: ! ! 0 1 = i i; = 0 i = i j; 1= 1 0 2 i 0 ! ! 1 0 = i k; 1 = 1 0 : (16) 3= 0 1 0 1 We identify 3-dimensional Euclidean space with the space of imaginary quaternions Im H 3 X (17) X = i X 2 Im H ! X = (X1; X2; X3) 2 R3: The scalar product of vectors in terms of matrices is then 1 < X; Y >= 2 trXY: (18) We also denote by F and N the matrices obtained in this way from the vectors F and N . Let us take a unitary quaternion 2 H ; det = 1; (19) which transforms the basis i; j; k into the basis Fx; Fy ; N : Fx = eu=2 1i ; Fy = eu=2 1 j ; N = 1k : (20) A simple calculation proves the following form of the Gauss-Weingarten system. Theorem 1 The moving frame Fx; Fy ; N of an isothermic surface is described by the formulas (20), where the unitary quaternion satis es the equations (21) x=A ; y=B with A; B of the form
=1

A = uy k + k1 eu=2j; 4 2

B = ux k k2 eu=2i: 4 2
6

(22)

The compatibility condition of (21)

Ay Bx + A; B ] = 0
is the Gauss-Codazzi system (11).

(23)

3 Lax representation for smooth isothermic surfaces

To put a class of surfaces into frames of the theory of integrable equations one should "insert" a spectral parameter into the Gauss-Codazzi system

Ay ( ) Bx( ) + A( ); B ( )] = 0:
Geometrically describes some deformation of surfaces preserving their properties (one can nd many examples in 2]). In some sense, the case of isothermic surfaces is more complicated. A characterization of isothermic surfaces as "soliton surfaces" by introducing a spectral parameter was given in a short note by J. Cieslinski, P. Goldstein and A. Sym 7]. They described a Lax pair with coe cients in so(4; 1). The Lax representation (24, 25) we present below results by using a 4-dimensional spinor representation of so(4; 1) (see, for example, the Appendix in 8]). By a direct calculation one can check the following statement Theorem 2 The system ) ; y =V( ) ; (24) ! ! B jeu=2 ; A ieu=2 ; V ( ) = (25) U ( ) = ie u=2 A je u=2 B where A; B are the matrices (22), is compatible if and only if the GaussCodazzi equations of isothermic surfaces (11) are satis ed. Whereas for = 0 the system (24, 25) is just a doubled Gauss-Weingarten system (21, 22), the geometrical meaning of the Lax representation with 6= 0 is more complicated, and we do not discuss it in this paper. For this interpretation we refer to the recent preprint 5] by F. Burstall, U. HertrichJeromin, F. Pedit and U. Pinkall, where they showed how pairs of isothermic 7
x = U(

surfaces are given by curved ats in a pseudo-Riemannian symmetric space and vice versa. Let us note the following symmetries of the system (24, 25). If (x; y; ) is a solution of the Cauchy problem with (x0; y0; ) = I , then ! ! I 0 I 0 ( )= 0 I ( ) 0 I (26) and ! 0 ; ( = 0) = 0 (27) where 2 H , solves (21, 22). Theorem 3 (Sym formula) Let (x; y; ) be a solution of (24, 25) satisfying (26, 27), then the isothermic immersion F and its dual F are given by the formula ! 1 @ j =0 F = 0 F : ^ (28) F 0 @ The corresponding Gauss maps are ! ! 0 N : 1 0 k ^ =0 N = N 0 k 0 ^ Proof. For Fx formulas (27, 28) imply ! ! 0 eu=2 1 i = 0 Fx ; 1 @U ^ Fx = =0 = e u=2 1 i Fx 0 0 @ ^ ^ where one should use (20, 13). The proof for Fy and N is the same.

4 Discrete isothermic surfaces

To de ne discrete isothermic surfaces we need a notion of a cross-ratio of a quadrilateral (X1; X2; X3; X4) in 3-dimensional Euclidean space. One can easily extend the notion of the cross-ratio of complex numbers to points in R3 identifying a sphere1 S , passing through X1; X2; X3; X4 with the Riemann sphere CP1. It turns out that there is a nice quaternionic description of the cross-ratio if one uses the isomorphism (17).
1

a plane is a special case

De nition 4 Let X1; X2; X3; X4 2 ImH be 4 points in R3 (see Fig. 1). The unordered pair fq; qg of eigenvalues of the quaternion
Q = (X1 X2)(X2 X3) 1(X3 X4)(X4 X1) 1 is called the cross-ratio of the quadrilateral (X1 ; X2 ; X3 ; X4 ).
X4 a X3 b X1 a X2

(29)

Figure 1: Vertices and edges of a quadrilateral

Q is invariant with respect to translations and dilatation of R3; rotations of R3 act on Q as Q ! RQR 1 and therefore preserve fq; qg. We usually will just speak of the "cross-ratio q 2 C". One has to keep in mind that q is only well de ned up to complex conjugation. Denoting the quaternions corresponding to the vector-edges as in Fig.1 one gets Q = aa0 1b0 1b: If the quadrilateral (X1 ; X2; X3; X4) is planar, it can be put to the i; j plane and its edges a1i + a2j = (a11 + a2k)i can be identi ed with complex numbers a = a1 + ia2 2 C: (30) For the cross-ratio this implies 0 q = aa0 2 C (31) bb

Lemma 1 The cross-ratio is invariant with respect to the Mobius transformations of R3.
9

Proof. Only for inversions the proof needs some calculation. Describing the inversion by the inverse matrix

X ~ X ! X = < X; X > = X 1

(32)

we get ~ Q = (X1 1 X2 1 )(X2 1 X3 1 ) 1(X3 1 X4 1 )(X4 1 X1 1 ) 1 = X1 1 QX1: ~ The eigenvalues of Q and Q coincide.
one-parametric family of quadrilaterals F = (F1; F2; F3; F4) with vertices

De nition 5 Let F : R ! R3 be an immersion and F; Fx; :::; Fyy the values of the immersion function and its derivatives at some point (x; y) 2 R. A
F1 = F + ( Fx Fy ) + 2 (Fxx + Fyy + 2Fxy ); 2 F2 = F + (Fx Fy ) + 2 (Fxx + Fyy 2Fxy ); F3 = F + (Fx + Fy ) + 2 (Fxx + Fyy + 2Fxy ); 2 F4 = F + ( Fx + Fy ) + 2 (Fxx + Fyy 2Fxy )
2 2

is called an in nitesimal quadrilateral at (x; y).

Up to terms of order O( 3) the vertices F1; F2; F3; F4 of the in nitesimal quadrilateral coincide with F (x ; y ); F (x + ; y ); F (x + ; y + ); F (x ; y + ) respectively. The following remark is trivial:

Lemma 2

1) q (F ) = 1 + O( ) () Q(F ) = I + O( );

2) q (F ) = 1 + O( 2) () Q(F ) = I + O( 2 ):

Theorem 4 Conformal and isothermic immersions F are characterized in terms of in nitesimal quadrilateral as follows:
1) Q(F ) = I + O( ) () F is conformal,

10

2) Q(F ) = I + O( 2) () F is isothermic. Proof. To calculate the cross-ratio of the in nitesimal quadrilateral we note, that F up to scaling is a translation of the quadrilateral with vertices at X1 = 0; X2 = Fx Fxy ; X3 = Fx + Fy ; X4 = Fy Fxy : Inverting it by the transformation (32) we send one of the points to in nity F F F +F F F ~ ~ ~ ~ X1 = 1; X2 = kF x F xyk2 ; X3 = kF x+ F yk2 ; X4 = kF y F xyk2 : x xy x y x xy The condition ~ ~ ~ ~ Q(X1; X2; X3; X4) = I + O( 2) is equivalent to the equality of vectors

The zero oder term in (33) yields (6) and the term of oder implies that Fxy lies in the tangential plane, i.e. the third condition in (12). The following geometrical properties of the cross-ratio are standard and can be easily checked: Lemma 3 If the cross-ratio of a quadrilateral (X1; X2; X3; X4) is real q 2 R; Q = qI , then X1; X2; X3; X4 lie on a circle. The cross-ratio of a square is q = 1. Now we are in a position to give a de nition of discrete isothermic surfaces. By a discrete surface we mean a map F : Z2 ! R3 . We use the following notations for the elements of discrete surfaces (n; m are integer labels): Fn;m - for the vertices, Fn+1;m; Fn;m]; Fn;m+1; Fn;m] - for the edges, (Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) - for the elementary quadrilaterals. Motivated by Theorem 4 above we de ne discrete isothermic surfaces as follows: 11

or equivalently 2 Fx + Fy 2 = Fx Fxy 2 + Fy Fxy 2 + O( 2) = (33) kFx + Fy k kFx Fxy k ! kFy Fxy k ! Fx Fxy 1 + 2 < Fx; Fxy > + Fy Fxy 1 + 2 < Fy ; Fxy > + O( 2): = kF k2 kF k2 kF k2 kF k2
x x y y

~ ~ ~ ~ X2 X3=X3X4 +O( 2);

which all elementary quadrilaterals have cross-ratio 1:

De nition 6 A discrete isothermic surface is a map F : Z2 ! R3, for

Q(Fn;m; Fn+1;m; Fn+1;m+1 ; Fn;m+1) = I for all n; m 2 Z:
(34)

Remark. More generally discrete isothermic surfaces can be de ned by the property that Qn;m is a product of two factors

Q(Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) =

m I n

for all

n; m 2 Z; (35)

where n does not depend on m and m not on n. This condition is reformulated in another way in De nition 1. We restrict ourself in this paper to the simplest symmetric case (34). Lemma 1 implies Theorem 5 (Mobius invariance). Let F : Z2 ! R3 be a discrete isothermic ~ surface and M a Mobius transformation of Euclidean 3-space. Then F M F : Z ! R3 is also isothermic. Property 2 of smooth isothermic surfaces also persists in the discrete case: Theorem 6 (Dual surface). Let F : Z2 ! R3 be a discrete isothermic surface. Then the discrete surface F : Z2 ! R3 de ned (up to translation) by the formulas Fn Fn;m Fn+1;m Fn;m = kF+1;m Fn;m2 ; Fn;m+1 Fn;m = kF +1 Fn;m2 (36) n;m Fn+1;m k n;m Fn;m+1 k is isothermic. Proof. It is convenient to use the complex language (30) (elementary quadrilaterals are planar). Thus we assume a; b; a0; b0 2 C. For the elementary quadrilateral (Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) we have

a + b0 = b + a0;
Formulas (36) imply

aa0 = bb0:

(37) (38)

1 1 1 a = a ; a0 = a0 ; b = 1 ; b0 = b0 b 12

a*

b*

b*

a*

Figure 2: An elementary quadrilateral and its dual for the edges of the dual quadrilateral. Combining (37) with (38) one gets

a + b0 = b + a0 ; a a0 = b b0 ; the dual quadrilateral closes up and has a cross-ratio 1.

Remark. For a discrete isothermic surface in the sense of de nition (35) the dual surface is de ned by Fn Fn+1;m Fn;m = 12 kF+1;m Fn;m2 ; n n;m Fn+1;m k Fn;m+1 Fn;m+1 = 12 Fn;m+1 Fn;m2 : (39)

The cross-ratios of the corresponding quadrilaterals of F and F coincide

m kFn;m Fn;m+1 k

Q(Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) =

m I n

(40)

5 Lax representation for discrete isothermic surfaces

De nition 6 of discrete isothermic surfaces is geometrically motivated and looks natural. In this section we explain how it was found. Namely, we discretize the Lax representation (24, 25) in a natural way preserving all its symmetries and show how De nition 6 appears in this approach. An integrable discretization of the system (24, 25) looks as follows. To each point (n; m) of the Z2-lattice one associates a matrix n;m . These 13

matrices at two neighbouring vertices are related by

n+1;m

= Un;m

n;m ;

n;m+1

= Vn;m

n;m ;

(41)

where the matrices Un;m and Vn;m are associated to the edges connecting the points (n + 1; m); (n; m) and (n; m + 1); (n; m) respectively. Having in mind the continuum limit ( is a characteristic size of edges)

U = I + U + :::;

V = I + V + :::

with U; V of the form (25), and preserving the group structure and the dependency of in the continuous case, it is natural to set ! An;m p0n;m i ; An;m = an;m1 + bn;m k + cn;m j Un;m = p00 i An;m n;m (42) 0 j! q n;m Vn;m = Bq00 j Bn;m ; Bn;m = dn;m 1 + en;mk + fn;mj; n;m n;m where the elds p; q; a; b; c; d; e; f live on the corresponding edges. These matrices can be simpli ed n n;m Un;m ! gg+1;m Un;m; Vn;m ! gg +1 Vn;m n;m n;m by a -independent gauge transformation
n;m

! gn;m

n;m

with gn;m 2 R at vertices. By such a transformation, which preserves the structure (42) of U and V , one can achieve the normalization p0n;m = 1=p00 n;m 0 00 pn;m ; qn;m = 1=qn;m qn;m for all n; m. Then U and V are of the form ! ! An;m pn;m i ; Vn;m = Bn;m qn;m j : (43) Un;m = p 1 i An;m 1 qn;m j Bn;m n;m Let us take four adjacent vertices (n; m); (n+1; m); (n+1; m+1); (n; m+1) and consider the compatibility condition

V 0U = U 0V ;
14

(44)

n,m+1

n+1,m+1

V
n,m

V U
n+1,m

Figure 3: Compatibility condition changing notations for the matrices Un;m; Un;m+1 ; Vn;m; Vn+1;m and their coe cients as it is shown in Fig.3. Then for the coe cients of U ; U 0; V ; V 0 one gets

pp0 = qq0; B0A = A0B; pB0i + q0jA = qA0j + p0 iB; 1 B0i 1 jA = 1 A0j + 1 0iB p q0 q p
With the notation

(45) (46) (47) (48)

un;m( ) = det Un;m ; vn;m( ) = det Vn;m

(49) implies

vn+1;m( )un;m ( ) = un;m+1( )vn;m ( ):

(49)

Now note that the zeros of both sides of (49) (considered as functions of ) should coincide. This implies ( 2 + det Bn+1;m)2( 2 + det An;m)2 = ( 2 + det An;m+1)2( 2 + det Bn;m)2: We suppose that the zeros of un;m+1( ) and un;m( ) coincide. This is equivalent to det An;m = n ; det Bn;m = m; (50) where we have incorporated into the notation that n does not depend on m and m not on n. 15

Now let us de ne discrete surfaces Fn;m and Fn;m in R3 by the Sym formula (28) ! 0 Fn;m ; 1 @ n;m j =0 Fn;m = ^ (51) n;m @ Fn;m 0 where n;m ( ) is the solution of the initial value problem (41, 43) with 0;0( ) = I . Lemma 4 The discrete surface Fn;m de ned by (51) is isothermic in the sense of the generalized de nition (35):

Q(Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) =

The surface Fn;m in (51) is its dual (39). Proof. Formula (51) implies for the edges

m I: n

Fn+1;m

! Fn+1;m Fn;m = 1 U 1 @ Un;m n;m n;m @ Fn;m 0 1 1 0 pn;m n;mAn;mi = p 1 1A 1i 0 n;m n;m n;m n;m

n;m n;m

= !

and a similar formula for the edges Fn;m; Fn;m+1]; Fn;m; Fn;m+1]: Here as in the smooth case n;m denotes the components of ! 0 : n;m n;m ( = 0) = 0 n;m Considering the local geometry of the quadrilaterals (Fn;m; Fn+1;m; Fn+1;m+1, Fn;m+1) and (Fn;m ; Fn+1;m; Fn+1;m+1; Fn;m+1) we can neglect a general rotation 1 n;m ::: n;m. Then the edges of these quadrilaterals in notations of Fig.3 are given by the vectors in Fig.4. The cross-ratio of the quadrilateral (Fn;m; Fn+1;m, Fn+1;m+1; Fn;m+1) is equal to

Qn;m = ( pA 1)( q0A 1B0 1jA) 1(p0B 1A0 1iB)(qB 1j) 1 = = A 1iA 1kBjB = m I;
n

16

pB 1 A 1 i B Fn,m+1 Fn+1,m+1 qA 1 B 1 jA pA 1 i F*n,m q1B1 j F*n,m+1 p1A 1 i Fn+1,m (p)1B 1 A 1 i B F*n+1,m (q)1A 1B1 jA F*n+1,m+1

qB 1 j Fn,m

Figure 4: Quadrilaterals (Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) (Fn;m; Fn+1;m; Fn+1;m+1; Fn;m+1) obtained from the Sym formula

and

where we have used equations (45, 46), the equalities 1 jB = m B 1 j n iA = Ai; and the notations (50). One can easily check that the edges of the surface F : Z2 ! R3 are related to the edges of F : Z2 ! R3 by (39). The special case det An;m = det Bn;m = 1; leads to discrete isothermic surfaces as de ned in De nition 6. Now let us show that any discrete isothermic surface generates the Lax representation (43). Consider an elementary quadrilateral (Fn;m; Fn+1;m, Fn+1;m+1; Fn;m+1). In a xed frame its edges (see Fig 5) are represented by the matrices X; Y; X 0 ; Y 0, which satisfy the following identities:

X + Y 0 = X0 + Y XY 0 1X 0Y 0 1 = 1:
17

(52) (53)

n,m+1

n+1,m+1

Y
n,m

Y X
n+1,m

Figure 5: An elementary quadrilateral in R3 The dual quadrilateral is also closed:

X 0 Y 0 1 = X 0 1 Y 1:

(54)

To compare these identities with the system (45-48) let us introduce a moving frame, de ned at each vertex of the surface. This frame is described ~ ~ by unitary quaternions n;m. We denote by X; Y the coordinate matrices of the vectors X and Y in the frame n;m taken at the basic vertex of the vector 1 ~ 1 ~ X = n;mX n;m ; X = n;m Y n;m : ~ ~ Then for the vectors X 0; Y 0 one has 1 ~ 1 1 ~ X 0 = n;m+1 X 0 n;m+1 = n;m Bn;mX 0Bn;m n;m; (55) 1 ~ 1 1 ~ Y 0 = n+1;m Y 0 n+1;m = n;m An;mY 0An;m n;m ; where still have to be de ned.

An;m =

n+1;m n;m ;

Bn;m =

n;m+1 n;m

Theorem 7 Let F : Z2 ! R3 be a discrete isothermic surface,

Xn;m = Fn+1;m Fn;m;
the vectors of its edges and

Yn;m = Fn;m+1 Fn;m qn;m = kYn;m k

18 (56)

pn;m = kXn;m k;

the lengths of these edges. Let n;m be a unitary quaternion describing the frame at the vertex (n; m) and 1 1 ~ ~ Xn;m = n;m Xn;m n;m ; Yn;m = n;mYn;m n;m (57) be a representation of the vectors Xn;m ; Yn;m in this frame. Suppose that the frames at negbouring points are related by
n+1;m

= An;m

n;m ;

n;m+1

= Bn;m

n;m ;

(58) (59)

where Then

~ An;m = i p 1 Xn;m ; n;m

~ Bn;m = j q 1 Yn;m : n;m

A = An;m;

p = pn;m ;

A0 = An;m+1 ;

p0 = pn;m+1 ;

B = Bn;m;

q = qn;m;

B0 = Bn+1;m ;

q0 = qn+1;m

satisfy the system (45-48), which is equivalent to the Lax representation (43,44) for the discrete isothermic surface F . Proof. Identity (45) follows immediately from (53, 56). Let us show rst that the frames n;m are well-de ned, i.e. the compatibility condition (46) for the system (58,59) is satis ed. Taking into account (55) we see, that B0A = A0B is equivalent to j 10 A Y 0 = j 10 B X 0: (60)

Formulas (57) imply

1 A = i p X;

1 B = j q Y:

Substituting these identities into (60) and using (45) and the obvious relations Y = q2Y 1; X = p2X 1 ; we see that (46) follows from (53). In the same way, using

iB 1 = Bi;
one shows that (52, 54) imply (47, 48). 19

iA 1 = Ai;

6 Construction of discrete isothermic surfaces

The discrete isothermic surfaces can be constructed in the same way as the discrete K-surfaces we discussed in our paper 3]. Using the de nition of discrete isothermic surfaces and starting with some initial stairway loop

Fn;m; Fn+1;m; Fn+1;m+1; Fn+2;m; :::; Fn+N;m+N = Fn;m;

one can recursively determine the coordinates Fn;m of all the points of the discrete isothermic surface in a unique way (see Fig.6).
n+1,m+1

n,m+1

n,m+1

n,m

a)

b)

Figure 6: Discrete isothermic surfaces as evolutions of initial stairways: a) with Qn;m = 1, b) with Qn;m = m= n Indeed, given the points Fn;m, Fn+1;m, Fn+1;m+1 the cross-ratio (34) uniquely determines the point Fn;m+1. To construct the discrete isothermic surface in the sense of the generalized De nition 1 one should start with an initial stairway strip as shown in Fig.6b. The factorization property (35) of the cross-ratio allows us to calculate all n ; m up to a common factor and, as a corollary, the cross-ratios of all the quadrilaterals of the discrete isothermic surface F : Z2 ! R3. After that one can proceed as before in Fig.6a reconstructing the surface step by step. As in the case of a discrete K-surface, one can construct by this elementary method various cylinders and some other interesting surfaces. We should mention here also, that the evolution of the stairway loops described above 20

is unstable. A small variation of the initial loop yields dramatic changes of the surface. This simple geometrical method does not allow us to control the global behaviour of the surface (for example to control the periodicity of the evolution of the initial stairway loop) and to construct compact discrete isothermic surfaces. To construct them one should apply methods from the theory of integrable equations, which are based on an analytic solution of the problem (and are described for the case of the discrete K-surface in 3]). But this could be the subject of a future paper.

7 Discrete isothermic minimal surfaces.

De nition 7 A discrete isothermic surface F : X 2 ! R3 is called minimal if for any vertex F : n; m of this surface there is a plane Pn;m such that the
< Fn+1;m Fn;m; Nn;m >=< Fn 1;m Fn:m; Nn;m >= = < Fn;m+1 Fn;m; Nn;m >= < Fn;m 1 Fn;m; Nn;m >= where Nn;m is a normal vector to Pn;m (see Fig. 7).

distances of the points Fn 1;m; Fn+1;m; Fn;m 1 ; Fn;m+1 to this plane are equal and
n;m ;

Remark. A mean curvature H for discrete isothermic surfaces has been de ned in 4]. De nition 7 corresponds to the special case H = 0 of the de nition of the discrete constant mean curvature surfaces.

Lemma 5 For all points of a discrete isothermic minimal surface

n;m

is constant on Z2 . Proof. Since the formulas (36) for the n and m edges of a dual surface di er by a sign, we introduce 4 vectors

H = Fn+1;m Fn;m; H 0 = Fn 1;m Fn;m; V = Fn;m Fn;m+1 ; V 0 = Fn;m Fn;m 1 with the basic point at the vertex Fn;m. Then all the endpoints of these
21

Nn,m Fn,m+1 Fn1,m Fn,m1 Fn,m Fn+1,m

Figure 7: De nition of discrete minimal isothermic surfaces vectors lie in a plane Pn;m and n;m is the distance between these two planes (Fig. 8). If we identify Fn;m = Fn;m, formulas (36) show that all the points Fn 1;m ; Fn+1;m, Fn;m 1, Fn;m+1 of the dual surface lie on a sphere Sn;m of radius 2 1 , which is the image of Pn;m under inversion n;m of R3 with respect to the unit sphere with a center at Fn;m. The points Fn 1;m 1; Fn+1;m 1; Fn+1;m+1; Fn 1;m+1 also lie in Sn;m. Indeed, since F is discrete isothermic, Fn+1;m+1 lies in a circle determined by Fn;m; Fn+1;m, Fn;m+1, which lies in Sn;m . Six points Fn;m+1, Fn+1;m+1, Fn;m , Fn+1;m, Fn;m 1, Fn+1;m 1 are common for Sn;m and Sn+1;m, therefore these spheres coincide Sn;m = Sn+1;m; n;m = n+1;m : Actually we proved already that the dual surface lies on a sphere. Without loss of generality one can assume the normalization = 1=2: (61) Then the dual surface lies on the unit sphere. Moreover the Gauss map N of F coincides with F .

Pn,m

Theorem 8 The following statements are equivalent:

22

Nn,m Fn,m1 Fn,m F*n+1,m H Fn+1,m F*n,m+1 V H Fn1,m

Fn,m+1

Pn,m F*n,m1 V P n,m

F*n1,m

Figure 8: Dual discrete minimal isothermic surface

(i) F : Z2 ! R3 is a discrete isothermic minimal surface normalized by (61). (ii) The dual surface F : Z2 ! R3 lies on a sphere and without loss of generality one can assume

F = N : Z2 ! S 2 :
This theorem allows us to parametrize discrete minimal surfaces by "holomorphic" data N .
cross-ratio of all its elementary quadrilaterals are equal to -1

De nition 8 A map g : Z2 ! R2 = C is called discrete holomorphic if the

q(gn;m; gn+1;m ; gn+1;m+1 ; gn;m+1 ) = (gn+1;m gn;m )(gn;m+1 gn+1;m+1 ) = 1; (gn+1;m+1 gn+1;m )(gn;m gn;m+1 )

gn;m are complex number here.

23

A discrete isothermic surface N : Z2 ! S 2 in S 2 can be obtained from a discrete holomorphic function g : Z2 ! C by stereographic projection C ! S2 2 g2 1 (N1 + iN2; N3) = ( 1 + ggj2 ; jjgjj2 + 1 ): j Combining this formula with (36) one gets an analogue of the Weierstrass representation in the discrete case.

Theorem 9 Let g : Z2 ! C be discrete holomorphic. Then the formulas

Fn+1;m Fn;m = 1 1 (1 gn+1;m gn;m ; i(1 + gn+1;m gn;m); gn+1;m + gn;m ) = 2 Re g n+1;m gn;m Fn;m+1 Fn;m = 1 1 (1 gn;m+1 gn;m ; i(1 + gn;m+1 gn;m ); gn;m+1 + gn;m) = 2 Re g n;m+1 gn;m
describe a discrete minimal isothermic surface. All discrete minimal isothermic surfaces are described in this way.

Figure 9: Discrete Enneper surface 24

gn;m = n + im (which is obviously discrete holomorphic) generates the discrete Enneper surface (Fig. 9). The discrete exponential map gn;m = exp( n + i m);
where = 2 arcsinh(sin 2 ); generates the discrete catenoid. (Fig. 10) = 2 =N;

The identical map

Figure 10: Discrete catenoid

References
1] W. Blaschke: Vorlesungen uber Di erentialgeometrie III, Berlin (1929) 2] A. Bobenko: Surfaces in terms of 2 by 2 matrices. Old and new integrable cases, In: Fordy A., Wood J. (eds) "Harmonic Maps and Integrable Systems", Vieweg (1994) 25

3] A. Bobenko, U. Pinkall: Discrete Surfaces with Constant Negative Gaussian Curvature and the Hirota Equation, Sfb 288 Preprint No.127 (1994) (to be published in Journ. of Di . Geom.) 4] A. Bobenko, U. Pinkall: Discrete Surfaces with Constant Curvature and Integrable Systems, (in preparation) 5] F. Burstall, U. Hertrich-Jeromin, F. Pedit, U. Pinkall: Curved Flats and Isothermic Surfaces, Sfb 288 Preprint No. 132 (1994) 6] P. Calapso: Sulla super cie a linee di curvature isotherme, Rend. Circ. Mat. Palermo 17, 275-286 (1992) 7] J. Cieslinski, P. Goldstein, A. Sym: Isothermic Surfaces in E 3 as Soliton Surfaces, Short Report (1994) 8] D. Ferus, F. Pedit, U. Pinkall, I. Sterling: Minimal Tori in S 4, J. Reine Angew. Math. 429, 1-47 (1992) 9] F. Pedit, H. Wu: Discretizing constant curvature surfaces via loop group factorizations: the discrete sine- and sinh-Gordon equations, Preprint GANG (1994)

26