• Aucun résultat trouvé

A construction algorithm for minimal surfaces

N/A
N/A
Protected

Academic year: 2021

Partager "A construction algorithm for minimal surfaces"

Copied!
7
0
0

Texte intégral

(1)

HAL Id: jpa-00210592

https://hal.archives-ouvertes.fr/jpa-00210592

Submitted on 1 Jan 1987

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

A construction algorithm for minimal surfaces

S. Lidin, S.T. Hyde

To cite this version:

S. Lidin, S.T. Hyde. A construction algorithm for minimal surfaces. Journal de Physique, 1987, 48

(9), pp.1585-1590. �10.1051/jphys:019870048090158500�. �jpa-00210592�

(2)

A construction algorithm for minimal surfaces

S. Lidin (*) and S. T. Hyde

(*) Inorganic Chemistry 2, Chemical Centre, P.O. Box 124, S-22100 Lund, Sweden

(~) Department of Applied Mathematics, Research School of Physical Sciences, P.O. Box 4, Canberra 2601,

Australia

(Requ le 25 mars 1987, accept6 le 7 mai 1987)

Résumé.2014 Nous construisons des surfaces minimales périodiques infinies à partir de fonctions complexes qui

sont simplement reliées à l’orientation des points plats sur la surface. Nous engendrons deux familles

tétragonales de surface, dont nous montrons qu’elles se réduisent dans des cas particuliers à des surfaces minimales classiques : la surface du diamant cubique (surface D) et la surface de Scherk.

Abstract.

2014

Infinite periodic minimal surfaces are constructed from complex functions, which are simply

related to the orientation of flat points on the surface. Two tetragonal families of surfaces are generated, which

are shown to reduce in special cases to classical minimal surfaces : the cubic diamond surface (D surface) and

the Scherk surface. In all cases the construction algorithm for the complex functions yields the expected results, supporting the validity of the procedure. The algorithm can be used to determine new periodic minimal

surfaces.

Classification

Physics Abstracts

02.40

-

61.30

-

61.50E

1. Introduction.

In recent years interest in translationally periodic

minimal surfaces has risen considerably. Minimal

surfaces have been found to be useful for describing

such diverse phenomena as lyotropic liquid crystal-

line phases [1, 2], crystal structures [3, 4] and ion mobility in the solid state [5].

Progress in all these fields has been seriously

retarded by the conspicuous paucity of known infi- nite periodic minimal surfaces (IPMS). With the exception of the P and D surfaces (Schoen’s nomenc-

lature [6]), found last century by Schwarz [7] and

Schoen’s gyroid [6] or (G surface) whose Cartesian coordinates have been computed explicitly [8], the

fifteen-odd other known IPMS have been described in terms of their approximate boundary curves, or, at best, in some implicit parametrisation.

In order to confirm the natural occurrence of IPMS in liquid and solid phases, more detailed knowledge of the multiplicity of surfaces of various space groups and the intrinsic geometry of these solutions is required. Surface coordinates permit

direct comparison of crystal structures with IPMS,

while computation of the surface area, volume

fraction and curvatures is required to match IPMS

with X-ray scattering data from cubic phases of

surfactant/water/oil mixtures [9].

Evidence is mounting that many more IPMS exist than those so far discovered [10]. It is likely that

many solutions exist within each space group [11].

Considerable progress has been made towards devel-

oping techniques for deriving appropriate boundaries

for IPMS of a required symmetry. A method for generating the exact coordinates of IPMS has been

conjectured [9]. In this paper we present some examples of the application of this method to

tetragonal IPMS which support this conjecture.

2. Theory.

The translational periodicity of IPMS invariably

results in the surface being non-analytic in R3.

Weierstrass established that any minimal surface

(except the plane) can be expressed as a line integral

of elliptic functions [12]. The integral is calculated in the complex plane, which is related to real space

(R3) by a pair of mappings.

The surface is mapped into a unit sphere under the

Gauss transformation, which associates each point

on the surface in real space with a point on the

surface of the sphere which corresponds to the point

of intersection of the normal vector of the surface (at

the point to be mapped), centred at the origin of the sphere, with the unit sphere. (Thus, for example, a

horizontal portion of the surface is mapped into the

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019870048090158500

(3)

1586

north pole of the sphere). The Gauss map results in

sphere coordinates :

where the surface is parametrized in real space by f (x, y, z )

=

Const.

The sphere coordinates are then transformed into the complex plane by standard stereographic projec-

tion from the north pole, giving :

where x" and y" are the real and imaginary axes respectively. These coordinates are usually reduced

to the single complex number,

i denoting the complex constant.

The R3 coordinates of the minimal surface are

related to the coordinates of the surface in the

complex plane by Weierstrass’ equations [12] :

where Re refers to the real part of the complex integrals.

These integrals are generally not analytically sol- vable, but once the IPMS is rendered into this form the Cartesian coordinates can be computed to any

degree of accuracy. Thus the problem of determining

. the geometry of the IPMS lies with a suitable

technique for generating the « Weierstrass func- tion », R (co ).

This function can be written as a composition of

two maps, one of which is the inverse of the Gauss

map of the surface [13]. We suggest here that the Weierstrass function is simply related to the Gauss

map of the surface on the sphere, considered as a

Riemann surface. The Weierstrass equations imply

that the function R (w ) diverges at a flat point, for

which the equations are invalid [12]. We thus con-

struct a complex function whose Riemann surface is

equivalent to the Gauss map of the IPMS, with branch points corresponding to flat points on the

surface.

The prescription is :

Fig. 1. - The T surface, AP

=

50.

Fig. 2. - The F surface, T = 14.

where K is a constant, Wa are the images of the flat

points on the IPMS (in the complex plane) and ba are the order of the branch points, the product

taken over all flat point normal vectors in the IPMS.

The branch point order is simply determined from the Gauss map in the vicinity of the flat point. If the

normal vector traces out (n + 1) loops around a flat point on the sphere while traversing a single loop on

the real surface in R3, the flat point gives rise to an n-

th order branch point.

In some cases it is difficult to establish the

presence of flat points. By combining various

topological techniques this can be overcome [9]. A

(4)

Fig. 3. - The T surfaces

=

3.

Fig. 4. - The T surface,

=

2.01.

simple rule of thumb is that both asymptotic lines

and lines of curvature on a minimal surface are

everywhere orthogonal, except at flat points [14].

Whence :

(i) If a straight line (an asymptote) intersects a

plane line of curvature (curve of intersection of the surface with a mirror plane of the boundary) at right angles, the point of intersection is a flat point, and

(ii) If two straight lines or two plane lines of

curvature meet obliquely, their common point is a

flat point.

These criteria suffice to determine flat points for

the IPMS described in this paper.

3. The T surface.

This surface is the earliest discovered IPMS. The

boundary problem was proposed by Gergonne in

1816. The straight line boundary shown in figures 1-

4 results in flat points as shown in figure 5 (by

criteria (i)). As the vertical lines are stretched, the

normal vectors to these flat points change smoothly, resulting in a distinct Weierstrass function for each

tetragonal axis ratio. The Gauss map for a generic

member of this family of surfaces is shown in

figure 5.

Fig. 5.

-

The normal map of the flatpoints to the general

T surface.

The positions of the flatpoint images are of the

form :

giving rise to a Weierstrass function :

-1 1

Substituting 11 for the quadric term, we write this

as :

For symmetry reasons the ba will all be equal. Two special cases are of interest. The first emerges when A is set to B/2/ J3, giving a Weierstrass function of :

If b = 1, this function reduces to the Weierstrass

parametrisation for the P, D and G surfaces. Since the D surface is a special case of the T surface (with

all three cell axes equal), all other cases of the T

(5)

1588

surface are homeomorphic to the D surface. Thus b =1 for general T surfaces, of arbitrary tetragonal

axis ratios.

When A is set to unity, the Weierstrass, function

reduces to :

.

which, for b set to unity, simplifies to :

This is the Scherk surface, a celebrated analytic

minimal surface, which represents the T surface whose tetragonal axis ratio tends to infinity.

4. The CLP surface.

This surface was discovered by Schwarz [7]. Once again, we can stretch the linear boundary unit along

the tetragonal axis, forming distinct Weierstrass functions for each tetragonal axial ratio. Figures 6-9

show the resulting units of IPMS formed by varying

this ratio, together with the general positions of the

flat point normal vectors.

The positions of the flat points will be of the

form :

These result in Weierstrass functions :

Here we also encounter two interesting special

cases. When the axial ratio approaches infinity the

Fig. 6. - The CLP surface, 41

=

0.

Fig. 7. - The CLP surface, 11 = 1.95, K = 1.

Fig. 8. - The CLP surface,1/’= 1.95, K

=

i.

plane lines of curvature perpendicular to the

stretched axis in figure 6 tend to straight lines and

the surface approaches the Scherk surface. As can

be seen from figures 6-9, the parameter A tends to

zero, resulting in the Scherk surface Weierstrass

function.

_

At the other extreme (A = 1/ B/2), the Gauss

map is identical to the vanishing A case, except for a rotation, which is due to different orientations of the surfaces. Two surfaces sharing the same Weierstrass function can only be distinguished through the multiplicative constant of the Weierstrass function,

K. For varying real K, a multiplicity of surfaces are

(6)

Fig. 9.

-

The normal map of the flatpoints to the general

CLP surface.

produced which are related to each other by a scaling factor. If K is complex, varying the argument

(without changing the modulus) results in distinct,

isometric « associate » surfaces. In particular, if K is purely imaginary, the resultant surface is said to be

« adjoint » to the surface parametrised by the purely

real K of the same modulus. This operation of changing the argument of K is known as the Bonnet transformation [15].

Studies of the CLP surface reveal that the surface formed by setting A equal to 1/ B/2 is distinct from the Scherk minimal surface. Schoen states that the CLP surface is self-adjoint [6], except for a tetragon- al distortion. Numerical calculations have confirmed this result. Indeed, setting A to (B/2 - 1 )/2 B/2

results in a surface which is perfectly self-adjoint.

This suggests that the surface generated from

K = 1, A =1 / J2 is identical to K

=

i, A

=

0, so

that this surface is adjoint to the Scherk surface.

5. Conclusions.

The construction procedure for the Weierstrass function results in IPMS and analytical minimal

surfaces (periodic in two dimensions) which are

consistent with the expected results.

The T and CLP surfaces can be described using

the same Weierstrass function. This function is also suitable for describing the Scherk surface and a

further analytic surface, probably the adjoint to the

Scherk surface. The nature of the D surface as a

special case of the family of T surfaces is clearly

illustrated by the construction algorithm for minimal

surfaces. The common Weierstrass function is :

where 11 is a parameter determining the nature of

the surface :

0 > ’ > - 2, K = 1 or i results in the CLP

surface,

W = - 2, K = 1 or i gives the Scherk or adjoint

Scherk surface,

41 - 2, K = 1 : gives the T family of surfaces

(AP = 14 forms the D surface).

The validity of these calculations suggests that we have a powerful constructive procedure for IPMS.

The detailed geometry of the surface is dependent only on the orientation and order of the flat points

on the surface ; all other points on the surface are

uniquely constrained by these singularities.

The defining property of the IPMS flat points

enables the generation of more complicated IPMS, knowing only the orientation of the flat points

-

a parameter which is often constrained by the sym- metry of the surface.

Consequently, we are at last able to produce

IPMS conforming to space group symmetries exhi-

bited by solid and lyotropic liquid crystals, without

recourse to previous « hit and miss » techniques. We

expect this construction procedure to yield a plethora

of new IPMS, and thus increase our poor vocabulary

of translationally ordered surface structures.

Acknowledgments.

We thank the Swedish Research Council for financial support, which enabled us to carry out this work.

One of us (S.T.H.) is grateful to this body for

assistance as a visiting researcher in Lund. Thanks to Prof. Sten Andersson for useful discussions.

References

[1] HYDE, S. T., ANDERSSON, S., ERICSSON, B. and LARSSON, K., Z. Kristallogr. 168 (1984) 213-219.

[2] HYDE, S. T., ANDERSSON, S. and LARSSON, K., Z.

Kristallogr. 174 (1986) 237-245.

[3] ANDERSSON, S., HYDE, S. T. and VON SCHNERING,

H. G., Z. Kristallogr. 168 (1984) 1-17.

[4] ANDERSSON, S., Angew. Chem. Int. Ed. Engl. 22 (1983) 69-81.

[5] ANDERSSON, S., HYDE, S. T. and BOVIN, J.-O., Z.

Kristallogr. 173 (1985) 97-99.

[6] SCHOEN, A., NASA Technical Report No. D-5541 (1970).

[7] SCHWARZ, H. A., Gesammelte Mathematische Abhandlungen (Springer) 1890.

[8] HYDE, S. T. and ANDERSSON, S., Z. Kristallogr. 170

(1985) 225-239.

(7)

1590

[9] HYDE, S. T., to be published.

[10] RUBINSTEIN, H., private communication.

[11] MEEKS, W., Lectures on Plateau’s Problem, Escole

de Geometria Differencial, Universidade Feder- al do Ceara (1978).

[12] NITSCHE, J. C. C., Vorlesungen über Minimalflächen (Springer Verlag, Berlin) 1975.

[13] SPIVAK, M., A Comprehensive Introduction to Diffe-

rential Geometry (Publish or Perish Inc., Ber- keley) 1979, vol. IV, p. 400.

[14] WILLMORE, T. J., An Introduction to Differential Geometry (Oxford University Press) 5th Ed., Delhi, 1985, p. 107.

[15] LIDIN, S. and ANDERSSON, S., to be published.

Références

Documents relatifs

fluid–fluid) triple contact line, two equations are obtained, which represent: (i) the equilibrium of the forces (surface stresses) for a triple line fixed on the body; (ii)

On considère un écoulement potentiel bidimensionnel d’un ‡uide incompressible et non vis- queux, à surface libre sous un barrage d’ouverture inclinée BC faisant un angle

These are the first spin polarized ab initio self-consis- tent energy band studies of ferromagnetic transi- tion metal filme that are thick enough C9-layers for Ni(001) and 7

Beniamin Bogosel Partitions minimales sur des surfaces 8.. Diff´erences finies - n´egliger les points a l’ext´erieur du domaine... probl`emes pr´es

Si l'on se donne une droite A et des plans tangents passant par A et correspondant homographiquemenlaux points de contact, la transformation d'Ampère leur fera correspondre une

Dans celle Note, nous retrouverons ces mêmes surfaces par une méthode toute différente qui fournit également des surfaces d^Ossian Bonnet..

Nous nous limiterons dans cette étude à l’état géométrique de la surface, car il correspond à l’état de surface technologique ; l’état physico-chimique

Dans cette section, j'étudie Qy sur un bout minimal immergé de courbure totale finie dans R 3 ou R 3 /^ On sait qu'un tel bout est conforme à D — {0}, où D est le disque unité de