Willmore Surfaces

U. Pinkall and I. Sterling*

C o m p a c t surfaces e m b e d d e d (or immersed) in ~3 have

b e e n studied in differential g e o m e t r y since its v e r y beginnings. Various special surfaces have been investigated, such as ellipsoids, tori of revolution, etc. Surprisingly, with one exception [2], we do not k n o w any
compact surface in ~3 of g e n u s g > 1 that appears in
the literature because of its special differential geometric properties. The same applies to compact n o n o r ientable surfaces in ~3.
Recently m a n y such surfaces have been obtained as
solutions to a variational problem. The problem is to
find c o m p a c t surfaces, of prescribed topological type,
that (on the average) have the least possible curvature
- - i . e . , that are "as s m o o t h as possible." As a m e a s u r e
for the average curvature, one chooses for an immersion f:M 2 --~ ~3 (M 2 an abstract compact surface) the
functional

become c u s t o m a r y to study the equivalent functional

W(T) = 88

+ ~x(M2),

w h e r e x(M 2) d e n o t e s the Euler characteristic of M 2.

Because of the Gauss-Bonnet theorem, w e have

W(f) = ~ H2dA,
JM2
w h e r e H = 89 1 + k2) is the m e a n curvature of the
surface.
For a g i v e n abstract surface M 2, w e will discuss
three problems:
(1) Determine W(M2): = inf W(f) over all immersions
f:M 2 ~ ~3.
(2) Classify all f for which W(f) equals the minimal
value W(M2).

l,V(f) = [ (k12 + k22)dA"

JM2
Here, kl,k 2 are the principal curvatures and dA is the
v o l u m e e l e m e n t i n d u c e d from ~3. Instead of 1,9 it has

* The authors would like to thank H. Karcher for helpful discussions

and the Max-Planck Institut fiir Mathematik for its support and hospitality.

Ivan Sterling received his B.A. from U.C. Santa Cruz in

1980 and his Ph.D. from U.C. Berkeley in 1985 under
W. Y. Hsiang. He spent last year at the Max-Planck Institut f~ir Mathematik in Bonn.
Ulrich Pinkall, born in 1955, is professor of mathematics at
the Technische Universit~it Berlin. He got his Ph.D. in
Freiburg (Black Forest) and spent two years at the MPI
f~ir Mathematik in Bonn. He has worked on various
aspects of conformal and projective differential geometry.
He is married and has three children.
38

Ivan Sterling

T H E M A T H E M A T I C A L INTELLIGENCER VOL. 9, N O . 2 9 1987 S p r i n g e r - V e r l a g N e w York

Ulrich Pinkall

Figure 1.

Figure 2.

(3) Determine all critical points f of W and the corres p o n d i n g values W(f).
Critical points of W are called Willmore surfaces and
are characterized by the Euler equation
AH + 2H(H 2 - K) = 0,
where K = klk 2 is the Gaussian curvature and A is the
Laplace-Beltrami operator. Willmore surfaces were
studied by W. Blaschke and G. Thomsen in 1923 [3],
[16]. 1 They also established the most important property of W: The functional W is invariant u n d e r conformal m a p p i n g s g:~3 U {oo}--+ ~3 U {or For example,
if f:M 2 --+ ~3 is an immersion such that 0 ~ f(M2), then

Figure 3.

For M 2 ~ S2 and M 2 ~ ~ p2, the above problems (1),

(2), and (3) are completely solved. We have W(S2) =
4"rr, a n d the m i n i m u m is a t t a i n e d only for r o u n d
spheres (Willmore 1965 [18]). Recently R. Bryant [5]
classified all Willmore immersions [:S 2 --~ Na. The possible v a l u e s of f s 2 H 2 d A are 4rrn, w i t h n a n a t u r a l
n u m b e r , where either n = I or (n >~ 4 and n even) or (n
/> 9 a n d n odd). Figure 1 shows a Willmore sphere
with n = 4.
A n y immersed projective plane in ~3 must have a
triple point, so, by a result of P. Li and S. T. Yau [12],
W(~P 2)/> 12-rr. Recently Bryant [6] and i n d e p e n d e n t l y
1 Thomsen attributes the modern form of the Euler equation to W.
Schadow; R. Kusner informed us that as early as 1810, Sophie Germain p r o p o s e d J'/-/=dA as the "virtual work" in her study of vibrating curved plates, although whether she carried this program
through is unclear. It is interesting to note that Poisson was aware
(around 1815) of the equivalence between V~/and W for the variational problem (long before the Gauss-Bonnet theorem).

R. Kusner [9] f o u n d explicit i m m e r s i o n s of ~p2 for

w h i c h the m i n i m a l value 12rr is a t t a i n e d . I n d e e d ,
Bryant classified all minimizing Rp2's in N3 and found
(modulo conformal transformations) a two-parameter
family of such surfaces. Figure 2 shows a Willmore
Np2 with threefold symmetry, thus providing an "optimal" version of a surface first described qualitatively
by W. Boy in 1903 [4].
For the torus T2, there is the long-standing "Willmore Conjecture": W(T2) = 2"rr2. The value 2"rr2 is actually attained for a certain torus of revolution whose
generating circle has radius 1 and distance V'2 - 1
from the axis of revolution (see Fig. 3).
The Willmore i n e q u a l i t y J'/-/2dA t> 2'rr2 has been
proved for various special classes of i m m e r s e d tori
(such as tori of revolution [10]), but, in general, it is
only k n o w n that for a n y immersion f:T2 --+ ~3 we have
W(f) > 4-rr [18]. Recently L. Simon [15] proved the existence of a minimizing immersion f of T2 with W(f) =
W(T2). This implies that W(T2) > 4~
Li and Yau [12] proved that for a n y immersed surTHE MATHEMATICALINTELL1GENCERVOL.9, NO. 2, 1987

39

Figure 4. Two-holed torus M 2.

Figure 7. Bone

Figure 5. Klein bottle.

Figure 6. Patch

Figure 8. Genus 3, two tetrahedral bones.

In 1970, H. B. Lawson [11] discovered many such

face with self-intersections, one has fH2dA >! 8~r.
Moreover, for any genus g, there are compact orient- Willmore surfaces. He found that every compact surable surfaces of genus g in ~3 with fH2dA < 8~ (see face but the projective plane (which is prohibited) can
below). Therefore, all surfaces of genus g that are ab- be minimally immersed into S3 (see Figs. 4 and 5).
solute minima of W (if they exist) are necessarily em- Every compact, orientable surface can be minimally
e m b e d d e d in S3. These surfaces are obtained by
b e d d e d and, of course, are Willmore surfaces.
A minimal surface in ~3 is one where H = 0 every- solving the Plateau problem for certain geodesic quadwhere. There do not exist compact minimal surfaces in rilaterals in S3, extending this solution surface by re~3. Most of the known examples of embedded Will- flection across its geodesic b o u n d a r y arcs, and remore surfaces come from compact minimal surfaces in peating reflections of this type until the surface closes.
For example, one such family of compact surfaces,
the u n i t s p h e r e S3 C ~4. A l r e a d y Blaschke a n d
Thomsen had proved that stereographic projections {Mg} (g = genus), is obtained by starting with such
(~(M) of compact minimal surfaces M in S3 are always quadrilaterals having edge lengths w/2 and angles ~/2,
Willmore surfaces. The area of M 2 C S 3 equals fH2dA ~/(g + 1), ~/2, "~/(g + 1). For these examples, 4 ~ <
f,~(Mg)H2dA < 8~r and g-~
lim f,7(Mg)H2dA = 8"a"[20].
for the stereographic projection of M 2 in ~3.
40

THE MATHEMATICAL INTELLIGENCER VOL. 9, NO. 2, 1987

New examples of compact embedded minimal surfaces in S3, and hence embedded Willmore surfaces,
were recently discovered (Karcher, Pinkall, and Sterling 1986 [8]). These examples are based on the tessellations of S3 into cells having the symmetry of a Platonic solid. For example, S3 is naturally tessellated by
five tetrahedra (as S2 is by four triangles), or S3 is tessellated by two cubes (as S2 is by two hexagons), etc.
Dividing a cell by its planes of symmetry, one obtains
as a fundamental region for the group of symmetries a
tetrahedron T with six dihedral angles. The tetrahedron is determined by its dihedral angles.
To construct a minimal surface in S3, one first finds a
minimal surface with boundary, called a "patch" (Fig.
6), within T, that intersects orthogonally all the plane
faces of T in planar geodesics. From the patch, one
obtains a certain piece of the whole surface, called a
" b o n e " (Fig. 7), by repeatedly reflecting patches
through those plane faces of T that are not contained
in faces of a cell. Finally, one builds the complete surface using reflections through faces of the cells (see
Figs. 8-13).

Figure 9. Genus 6, five tetrahedral bones.

Remarks
1. The computer plays a crucial role in the investigation of minimal surfaces in S3. For example, computer
estimates indicate the area of Lawson's three-holed
torus M 3 is less than that of the genus three surface in
Figure 8. This lends evidence to Kusner's conjecture
[20] that stereographic projections of Lawson's g-holed
tori Mg are "optimal," in the sense that they are absolute minima of the Willmore integral, fMH2dA, among
all genus g surfaces M.
2. It was proved by Lawson that a compact minimal
embedded surface in S3 separates S3 into two diffeomorphic components. Furthermore, there was the following conjecture (]19], problem 98):

Figure 10. Genus 5, two cubical bones.

Conjecture. Any compact minimal embedded

surface in
S3 separates S3 into two components of equal volume.

The surface with dodecahedral symmetry in Figure

12 is a counterexample.
It suffices to prove M stays within a distance ~/2 D (~ 23.8 ~ of its equator of reflection, E, where 4D 2 sin (2D) = ~, since this tube around E contains half
the volume of S3 (note (r(E) is a round $2). This is obvious from Figure 14 and can be rigorously proved
(the actual maximal distance of a point x E M from E
is approximately 7~
3. In 1985 [13], the first author found the first examples of compact embedded Willmore surfaces that
are not stereographic projections of compact emb e d d e d minimal surfaces in S3. Using results of J.
Langer and D. Singer [10] on elastic curves o n S2, an

Figure 11. Genus 7, two octahedral bones.

THE MATHEMATICAL INTELLIGENCER VOL. 9, NO. 2, 1987

41

F i g u r e 12. G e n u s 11, t w o d o d e c a h e d r a l bones.

Figure 15.

F i g u r e 13. G e n u s 19, t w o icosahedral bones.

Figure 16.

F i g u r e 14.

Figure 17.

42 THE MATHEMATICALINTELLIGENCERVOL.9, NO. 2, 1987

Figure 18.

Figure 19.

infinite series of such surfaces is exhibited (see Fig.

15). All of these surfaces are, h o w e v e r , unstable critical points of W a n d h e n c e are n o t candidates for absolute m i n i m a .

4. W. Boy, Uber die Curvatura integra und die Topologie

geschlossener F1/ichen, Math. Ann., Vol. 57 (1903),
151-184.
5. R. Bryant, A duality theorem for Willmore surfaces, J.
Diff. Geom., Vol. 20 (1984), 23-53.
6. R. Bryant, in preparation.
7. A. M. Garcia, On the conformal types of algebraic surfaces of Euclidean space, Comment. Math. Helv., Vol. 17
(1962), 49-60.
8. H. Karcher, U. Pinkall, and I. Sterling, New minimal
surfaces in S3, preprint MPI-86-27.
9. R. Kusner, Conformal geometry and complete minimal
surfaces, preprint.
10. J. Langer and D. Singer, Curves in the hyperbolic plane
and mean curvature of tori in R3 and S3, Bull. London
Math. Soc., Vol. 16 (1984), 531-534.
11. H. B. Lawson, Jr., Complete minimal surfaces in S3,
Ann. of Math, Vol. 92 (1970), 335-374.
12. P. Li and S. T. Yau, A new conformal invariant and its
applications to the Willmore conjecture and the first eigenvalue of compact surfaces. Invent. Math., Vol. 69
(1982), 269-291.
13. U. Pinkhall, Hopf tori in S3, Invent. Math., Vol. 81 (1985),
379-386.
14. R. Rfiedy, Embeddings of open Riemann Surfaces, Comment. Math. Helv., Vol. 46 (1971), 214-225.
15. L. Simon, Existence of Willmore surfaces, pre~,cint 1986.
16. G. Thomsen, Uber Konforme Geometric ~: Grundlagen
der konformen Fl/ichentheorie, Abh. Math. Sere. Hamburg
(1923), 31-56.
17. H. Wente, Counterexample to a conjecture of H. Hopf,
Vol. 121, No. 1 (1986), 193-243.
18. T. J. Willmore, Note on embedded surfaces, An. Stiint.
Univ. "AI. I. Cusd'" Iasi Sect. I, a Mat., Vol. II (1965),
443-496.
19. S. T. Yau, Problem Section, Seminar of differential geometry, Ann. of Math., Vol. 102 (1982), 669-706.
20. R. Kusner and U. Pinkall, Area estimates for surfaces
with applications to the Willmore problem, in preparation.

Finally, w e w a n t to m e n t i o n that the tori of constant

m e a n c u r v a t u r e in ~3, d i s c o v e r e d r e c e n t l y b y H.
W e n t e [17], 2 are related to the Willmore functional: A.
Garcia [7] a n d R. Rfiedy [14] h a d p r o v e d that a n y compact R i e m a n n surface (an oriented surface with a conformal structure) can be c o n f o r m a l l y i m m e r s e d in ~3.
Again, o n e m i g h t ask for an " o p t i m a l " m o d e l in ~3 (in
the s e n s e of the Willmore functional) for a given compact R i e m a n n surface (M,g) (g a R i e m a n n i a n metric
defining the c o n f o r m i n g structure). More generally,
w e are i n t e r e s t e d in " c o n s t r a i n e d Willmore surfaces"
- - i . e . , critical points of the functional W restricted to
the space of all conformal i m m e r s i o n s f:(M,g) ~ ~3. It
h a s b e e n o b s e r v e d that a n y c o m p a c t surface in ~3 of
c o n s t a n t m e a n c u r v a t u r e is a c o n s t r a i n e d W i l l m o r e
surface. This follows f r o m the fact that the G a u s s m a p
of a c o n s t a n t m e a n c u r v a t u r e surface is harmonic, a n d
the f u n c t i o n 1,9 is essentially just the e n e r g y of the
Gauss map.
Figures 1 6 - 1 9 s h o w an i m m e r s e d torus with constant H t h a t w a s explicitly c o n s t r u c t e d by U. Abresch
[1]. T h e w h o l e surface (Fig. 16) consists of six cong r u e n t pieces, which are i m m e r s e d annuli (Fig. 17).
Figures 18 a n d 19 s h o w one-half a n d one-third of the
t o m s , respectively.

References
1. U. Abresch, Constant mean curvature tori in terms of
elliptic functions, preprint 1985.
2. T. F. Banchoff and H. H. Kuiper, Geometrical class and
degree for surfaces in three-space, J. Diff. Geom., Vol. 16
(1981), 559-576.
3. W. Blaschke, Vorlesungen fiber Differentialgeometrie
III, Berlin: Springer, 1929.
2 These surfaces answered, negatively, the long-standing question
of Hopf: Is a closed surface in ~3 with constant mean curvature necessarily the round sphere?

Mathematics Department (Sterling)

University of California
Santa Cruz, CA 95064 USA
Fachbereich Mathematik (Pinkall)
T.U. Berlin
1000 Berlin 12
Federal Republic of Germany
THE MATHEMATICAL INTELLIGENCER VOL. 9, NO. 2, 1987

43