Home Mathematics Gluing Karcher–Scherk saddle towers I: Triply periodic minimal surfaces
Article
Licensed
Unlicensed Requires Authentication

Gluing Karcher–Scherk saddle towers I: Triply periodic minimal surfaces

  • Hao Chen ORCID logo EMAIL logo and Martin Traizet
Published/Copyright: January 31, 2024

Abstract

We construct minimal surfaces by gluing singly periodic Karcher–Scherk saddle towers along their wings. Such constructions were previously implemented assuming a horizontal reflection plane. We break this symmetry by prescribing phase differences between the saddle towers. It turns out that, in addition to the previously known horizontal balancing condition, the saddle towers must also be balanced under a subtle vertical interaction. This interaction vanishes in the presence of a horizontal reflection plane, hence was not perceived in previous works. Our construction will be presented in a series of papers. In this first paper of the series, we will explain the background of the project and establish the graph theoretical setup that will be useful for all papers in the series. The main task of the current paper is to glue saddle towers into triply periodic minimal surfaces (TPMSs). Our construction expands many previously known TPMSs into new 5-parameter families, therefore significantly advances our knowledge on the space of TPMSs.

Funding statement: Hao Chen is partially supported by Individual Research Grant from Deutsche Forschungsgemeinschaft within the project “Defects in Triply Periodic Minimal Surfaces”, Projektnummer 398759432. Martin Traizet is supported by the ANR project Min-Max (ANR-19-CE40-0014).

A Rigidity of saddle towers

This section is dedicated to a proof of Theorem 5.5. Let p ˙ ker D Λ ( p ) . We want to prove that, if p ˙ h = 0 for three h 𝖧 , then p ˙ = 0 .

Define

p ( ϵ ) = p + ϵ p ˙ .

Let ( Φ 1 ( ϵ ) , Φ 2 ( ϵ ) , Φ 3 ( ϵ ) ) be the Weierstrass data given by equation (5.1) and X ( ϵ ) the corresponding immersion given by equation (5.2). Note that X ( ϵ ) is harmonic but not minimal. Let H ( ϵ ) be the mean curvature and let N ( ϵ ) be the normal of X ( ϵ ) . To ease computation, the dependence on ϵ will not be written. Instead, we use an superscript to denote the value at ϵ = 0 and a dot to denote the partial derivative with respect to ϵ at ϵ = 0 .

Lemma A.1.

We have H ˙ = 0 on the Riemann sphere minus the points p h for h H .

Proof.

We have for all ϵ

Q = i = 1 n - 2 λ i Φ 1 d z z - ζ i with  λ i = Res ( Q Φ 1 , ζ i ) .

We have λ i = 0 and λ ˙ i = D Λ i ( p ) p ˙ = 0 so Q ˙ = 0 . Let w = x 1 + i x 2 be a local complex coordinate on the Riemann sphere. Recall the standard formula for the mean curvature

(A.1) H = g 22 b 11 + g 11 b 22 - 2 g 12 b 12 2 ( g 11 g 22 - g 12 2 ) ,

where g and b are respectively the matrices of the first and second fundamental forms in the coordinate system ( x 1 , x 2 ) . We have for all ϵ

Q = ( X x 1 2 - X x 2 2 - 2 i X x 1 , X x 2 ) d w 2 = ( g 11 - g 22 - 2 i g 12 ) d w 2 .

Hence from Q = 0 and Q ˙ = 0 we obtain

(A.2) g 11 = g 22 , g 12 = 0 , g ˙ 11 = g ˙ 22 and g ˙ 12 = 0 .

Since X ( ϵ ) is harmonic for all ϵ, we have b 11 + b 22 = 0 for all ϵ, so

b 11 + b 22 = 0  and  b ˙ 11 + b ˙ 22 = 0 .

Taking the derivative of (A.1), we obtain H ˙ = 0 . ∎

By Lemma A.1, u = X ˙ , N is a Jacobi field. In a neighborhood of p h , we have

(A.3) X ˙ = Re [ ( cos ( θ h ) , sin ( θ h ) , i σ h ) p ˙ h z - p h ] + bounded terms , N = σ h ( - sin ( θ h ) , cos ( θ h ) , 0 ) + O ( z - p h ) ,

so u is a bounded Jacobi field. Since all saddle towers are rigid, there exists c 3 such that u = N , c . Consider the translated immersion Y ( ϵ ) = X ( ϵ ) - ϵ c . Then Y ˙ , N = 0 so Y ˙ is a tangent vector. Using the local complex coordinate w = x 1 + i x 2 , we decompose Y ˙ in the tangent space as

Y ˙ = ξ 1 X x 1 + ξ 2 X x 2 = ξ X w + ξ ¯ X w ¯ with  ξ = ξ 1 + i ξ 2 .

Lemma A.2.

The map ξ d d w defines a holomorphic vector field on the Riemann sphere. Moreover, ξ ( p h ) = - p ˙ h for h H .

Proof.

Since X is conformal and harmonic, we have

X w , X w = 0 , X w , 2 X w 2 = 0 , and 2 X w w ¯ = 0

where , denotes the -bilinear dot product. For all ϵ

X w , X w = 1 4 ( g 11 - g 22 - 2 i g 12 ) .

Taking the derivative with respect to ϵ and using equation (A.2), we obtain

2 X w , Y ˙ w = 2 X w , w ( ξ X w + ξ ¯ X w ¯ ) = 2 ξ ¯ w X w , X w ¯ = 0 .

Hence ξ ¯ w = 0 and ξ is holomorphic. If w is another local complex coordinate, we can write

Y ˙ = ξ X w + ξ ¯ X w ¯ = ξ X w d w d w + ξ ¯ X w ¯ ( d w d w ) ¯

This gives

ξ = ξ d w d w

so ξ transforms as a holomorphic vector field under change of coordinate. Using the complex coordinate w = z - p h in a neighborhood of p h , we have

X w = 1 2 w ( - cos ( θ h ) , - sin ( θ h ) , - i σ h ) + O ( 1 ) .

Hence by equation (A.3),

ξ ( p h ) = - p ˙ h .

Recall that we have fixed the position of three points p h . Hence ξ has at least three zeros. Now a non-zero holomorphic vector field on the Riemann sphere has two zeros, so ξ = 0 . This implies that p ˙ = 0 and concludes the proof of Theorem 5.5.

B Horizontal rigidity of “triangulated” graphs

This section is dedicated to a proof of Theorem 4.7. Consider, for x , the total length

( x ) = 1 2 h 𝖧 x h .

If P hor ( χ ) = 0 , so that χ = grad f , we have

D ( x ) χ = 1 2 h 𝖧 1 x h x h , f v ( - h ) - f v ( h )
= - h 𝖧 1 x h x h , f v ( h ) (using  h - h  for  f v ( - h ) )
= - v V h b ( v ) 1 x h x h , f v
= - v V F b ( v ) hor , f v .

We have proved:

Proposition B.1.

The graph is balanced if and only if x is a critical point of L restricted to those x A such that P hor ( x ) = P hor ( x ) .

Assume that F hor ( x ) = 0 and let χ = grad ( f ) be in the kernel of ( D F hor ( x ) , P hor ) . Differentiating the above equation

D 2 ( x ) ( χ , χ ) = - v V D F b ( v ) hor ( x ) χ , f v = 0 .

On the other hand, a direct computation gives

D 2 ( x ) ( χ , χ ) = 1 2 h 𝖧 1 x h 3 ( x h 2 χ h 2 - x h , χ h 2 ) .

The summands are all non-negative, hence must be all zero, which means that χ h is parallel to x h for all h 𝖧 .

If all faces have two or three edges, this implies that χ h = λ x h for some λ . Then P c 1 hor ( χ ) = λ T 1 = 0 implies λ = 0 and χ = 0 , which proves Theorem 4.7.

C Classification of balanced configurations of genus 3

As promised in Section 7.1, we prove the following Classification Theorem.

Theorem C.1.

The Meeks, aG, aH, and aI configurations in Figure 6 are the only balanced configurations whose graphs are orientable with two faces. Hence they are the only possible configurations that give rise to TPMSs of genus 3.

Proof.

If a configuration gives rise to a TPMS of genus three, its graph must have two faces. By Euler’s formula, the average degree of the graph is

2 | 𝖤 | / | 𝖵 | = 2 ( | 𝖵 | + | 𝖥 | ) / | 𝖵 | = 2 + 4 / | 𝖵 | .

By Assumption 3.6, the average degree is at least 4, hence the graph has at most 2 vertices. We discuss two cases

Case | V | = 1 , | E | = 3 . All edges are loops. Since edges are represented by straight segments, no loop is null-homologous. If two loops are homologous, they must be parallel, and represented by the same segment that cuts the torus into an annulus. To form a 2-cell embedding, the remaining edge must be in a different homology class. But then, it is impossible to orient the half-edges alternately incoming and outgoing around the vertex, contradicting the orientability.

So we have three pairwise non-homologous simple loops. They only intersect at the vertex, hence any two of them form a homology basis, and the remaining loop must be homologous to their concatenation. So the graph must be homeomorphic to that of the aH.

Any configuration with this graph is trivially balanced because all edges are loops. It is also trivially rigid as there is only one vertex, so the cut space is trivial.

Case | V | = 2 , | E | = 4 . We first prove that such a graph has no loop. The graph is connected, hence the edges can not be all loops. If exactly one or three edges are loops, the degree of a vertex will be smaller than 4, contradicting Assumption 3.6. If exactly two edges are loops, they must be adjacent to different vertices. So they divide the torus into two annuli. Since the graph is represented as a limit of 2-cell embeddings, the remaining edges must lie in different annuli. But then, it is impossible to orient the half-edges alternately incoming and outgoing around each vertex, contradicting the orientability. This proves that none of the edges is a loop.

So we have four edges between two vertices. Any two of the edges form a cycle. If some of these cycles are null-homologous, we will have parallel edges. If at most one edge is simple, it is not possible to form a 2-cell embedding. If exactly two edges are simple, the only 2-cell embedding does not have a proper orientation.

So we have four simple edges between two vertices. Then the graph must be as shown in Figure 11. This graph is balanced if and only if the half-edges form two collinear pairs around each vertex. So if one vertex is at 0, the other vertex must lie at a 2-division point.

By the period condition (6.1), the phase function must be of the form

c - Ψ 2 2 , c + Ψ 2 2 , c + Ψ 1 + Ψ 2 2 , c + Ψ 1 - Ψ 2 2

on the half-edges around a vertex, where Ψ 1 and Ψ 2 are the fundamental shifts. See Figure 11.

Figure 11

If arg ( T 2 T 1 ) π 2 , then the shortest edges of the graph form a collinear pair. By assuming arg ( T 2 T 1 ) < π 2 as in Figure 11, the phase function is balanced if and only if

sin ( c + Ψ 1 2 ) = 0 or cos Ψ 2 - Ψ 1 2 = 0 .

The solution Ψ 1 = - 2 c gives Meeks’ configuration which is generically rigid. The solution Ψ 2 - Ψ 1 = π gives the aI configurations, which is not rigid because c remains a free variable.

If arg ( T 2 T 1 ) = π 2 , the phase function is balanced if and only if

sin c + sin ( c + Ψ 1 ) = 0 or cos ( Ψ 2 2 ) = 0 .

The solution Ψ 1 = - 2 c gives again Meeks’ configurations. The solutions Ψ 1 = π and Ψ 2 = π give the aG configurations, which is not rigid because c remains a free variable. ∎

D Integral of a holomorphic 1-form through a neck

To prove Lemma 8.3, we first prove a general result, Theorem D.1 below. This was in fact proved in [26, Lemma 1] and has been used in several papers, but was not clearly stated as an independent result and the value at t = 0 was not explicitly given.

Fix some numbers 0 < ϵ 1 < ϵ . For t such that 0 < | t | < ϵ 2 , let 𝒜 t be the annulus | t | ϵ < | z | < ϵ and let ψ t : 𝒜 t 𝒜 t be the involution defined by ψ t ( z ) = t z . Let β t be the curve from ϵ 1 to t ϵ 1 parameterized for s [ 0 , 1 ] by

β t ( s ) = ϵ 1 1 - 2 s t s .

Note that β t depends on the choice of the argument of t. Let γ be the circle parameterized by γ ( s ) = ϵ 1 e 2 π i s .

Theorem D.1.

Let ω t be a family of holomorphic 1-forms on A t , depending holomorphically on t D * ( 0 , ϵ 2 ) , and let ω ~ t = ψ t * ω t . Define

α t = 1 2 π i γ ω t = - 1 2 π i γ ω ~ t .

Assume that

lim t 0 ω t = ω 0 and  lim t 0 ω ~ t = ω ~ 0

where ω 0 and ω ~ 0 are holomorphic in D * ( 0 , ϵ ) with at most simple poles at z = 0 , and the limit is uniform on compact subsets of D * ( 0 , ϵ ) . Then β t ω t - α t log t is a well-defined holomorphic function of t 0 which extends holomorphically at t = 0 . Moreover, its value at t = 0 is

lim z 0 [ ( ϵ 1 z ω 0 ) - α 0 log z ] - lim z 0 [ ( ϵ 1 z ω ~ 0 ) + α 0 log z ] .

Proof.

If arg ( t ) is increased by 2 π , the homotopy class of β t is multiplied on the left by γ, so β t ω t is increased by 2 π i α t . On the other hand, log t is increased by 2 π i , so the difference β t ω t - α t log t is a well-defined holomorphic function of t in D * ( 0 , ϵ 2 ) . Using the change of variable rule, we write

(D.1) ( β t ω t ) - α t log t = [ ( ϵ 1 t ω t ) - α t log t ] - [ ( ϵ 1 t ω ~ t ) + α t log t ] .

To estimate the first term, we fix ϵ 1 < ϵ 2 < ϵ and expand ω t in Laurent series in the annulus 𝒜 t as

ω t = n a n ( t ) z n - 1 d z

with a 0 ( t ) = α t and

a n ( t ) = 1 2 π i | z | = ϵ 2 z - n ω t = - 1 2 π i | z | = ϵ 2 ψ t * ( z - n ω t ) = - 1 2 π i | z | = ϵ 2 t - n z n ω ~ t .

Since ω t and ω ~ t are uniformly bounded on the circle | z | = ϵ 2 , this gives the estimates, for n > 0 and a uniform constant C,

(D.2) | a n ( t ) | C ( ϵ 2 ) n and | a - n ( t ) | C | t | n ( ϵ 2 ) n .

Then we have

( ϵ 1 t ω t ) - α t log t = - α t log ϵ 1 + n = 1 [ a n n ( t n 2 - ( ϵ 1 ) n ) - a - n n ( t - n 2 - ( ϵ 1 ) - n ) ] .

Using estimates (D.2) and ϵ 1 < ϵ 2 , it is straightforward to check that the sum is uniformly bounded with respect to t and that we have

lim t 0 [ ( ϵ 1 t ω t ) - α t log t ] = - α 0 log ϵ 1 - n = 1 a n ( 0 ) n ϵ 1 n .

On the other hand, we have

ϵ 1 z ω 0 = α 0 log ( z ϵ 1 ) + n = 1 a n ( 0 ) n ( z n - ( ϵ 1 ) n ) ,

so

(D.3) lim t 0 [ ( ϵ 1 t ω t ) - α t log t ] = lim z 0 [ ( ϵ 1 z ω 0 ) - α 0 log z ] .

The second term in (D.1) is estimated in the exact same way, leading to

(D.4) lim t 0 [ ( ϵ 1 t ω ~ t ) + α t log t ] = lim z 0 [ ( ϵ 1 z ω ~ 0 ) + α 0 log z ] .

The function β t ω t - α t log t is bounded so extends holomorphically at t = 0 by the Riemann Extension Theorem.

The last point of Theorem D.1 follows from equations (D.1), (D.3) and (D.4). ∎

Proof of Lemma 8.3.

Recall the definition of the path B h just before Lemma 8.3. On path number (1), ω t depends holomorphically on t h in a neighborhood of 0 so

O v ( h ) w h = δ ω t

is a holomorphic function of t h in a neighborhood of 0. Same for path number (3). Regarding path number (2) we write

(D.5) ( w h = δ w h = t h δ ω t ) - α h log t h = ( δ t h δ ( w h - 1 ) * ω t ) - α h log t h

and we apply Theorem D.1 to the 1-form ( w h - 1 ) * ω t with ϵ 1 = δ , t = t h , observing that

ψ t h * ( w h - 1 ) * ω t = ( ( ψ t h w h ) - 1 ) * ω t = ( w - h - 1 ) * ω t ,

so the hypotheses of Theorem D.1 are satisfied. Hence (D.5) extends holomorphically at t h = 0 and its value there is

lim z 0 [ ( δ z ( w h - 1 ) * ω 0 ) - α h log z ] - lim z 0 [ ( δ z ( w - h - 1 ) * ω 0 ) - α - h log z ]
= lim z p h [ ( w h = δ z ω 0 ) - α h log w h ( z ) ]
    - lim z p - h [ ( w - h = δ z ω 0 ) - α - h log w - h ( z ) ] .

Adding the three terms gives Lemma 8.3. ∎

Acknowledgements

The authors thank Matthias Weber for an unbelievable video of triply periodic minimal surfaces that motivated this project, and the anonymous referee for careful reading and helpful comments.

References

[1] K. Brakke, Triply periodic minimal surfaces, https://kenbrakke.com/evolver/examples/periodic/periodic.html. Search in Google Scholar

[2] H. Chen, Existence of the tetragonal and rhombohedral deformation families of the gyroid, Indiana Univ. Math. J. 70 (2021), no. 4, 1543–1576. 10.1512/iumj.2021.70.8505Search in Google Scholar

[3] H. Chen and M. Traizet, Stacking disorder in periodic minimal surfaces, SIAM J. Math. Anal. 53 (2021), no. 1, 855–887. 10.1137/20M1312137Search in Google Scholar

[4] H. Chen and M. Weber, An orthorhombic deformation family of Schwarz’ H surfaces, Trans. Amer. Math. Soc. 374 (2021), no. 3, 2057–2078. 10.1090/tran/8275Search in Google Scholar

[5] C. Cosín and A. Ros, A Plateau problem at infinity for properly immersed minimal surfaces with finite total curvature, Indiana Univ. Math. J. 50 (2001), no. 2, 847–879. 10.1512/iumj.2001.50.1829Search in Google Scholar

[6] R. Diestel, Graph theory, 5th ed., Grad. Texts in Math. 173, Springer, Berlin 2017. 10.1007/978-3-662-53622-3_7Search in Google Scholar

[7] W. Fischer and E. Koch, On 3-periodic minimal surfaces, Z. Krist. 179 (1987), no. 1–4, 31–52. 10.1524/zkri.1987.179.1-4.31Search in Google Scholar

[8] C. Godsil and G. Royle, Algebraic graph theory, Grad. Texts in Math. 207, Springer, New York 2001. 10.1007/978-1-4613-0163-9Search in Google Scholar

[9] S. J. Gortler, C. Gotsman and D. Thurston, Discrete one-forms on meshes and applications to 3D mesh parameterization, Comput. Aided Geom. Design 23 (2006), no. 2, 83–112. 10.1016/j.cagd.2005.05.002Search in Google Scholar

[10] A. O. Ivanov and A. A. Tuzhilin, Minimal networks, CRC Press, Boca Raton 1994. Search in Google Scholar

[11] H. Jenkins and J. Serrin, The Dirichlet problem for the minimal surface equation, with infinite data, Bull. Amer. Math. Soc. 72 (1966), 102–106. 10.1090/S0002-9904-1966-11438-6Search in Google Scholar

[12] H. Karcher, Embedded minimal surfaces derived from Scherk’s examples, Manuscripta Math. 62 (1988), no. 1, 83–114. 10.1007/BF01258269Search in Google Scholar

[13] H. Karcher, The triply periodic minimal surfaces of Alan Schoen and their constant mean curvature companions, Manuscripta Math. 64 (1989), no. 3, 291–357. 10.1007/BF01165824Search in Google Scholar

[14] S. G. Markande, M. Saba, G. Schroeder-Turk and E. A. Matsumoto, A chiral family of triply-periodic minimal surfaces derived from the quartz network, preprint (2018), https://arxiv.org/abs/1805.07034. Search in Google Scholar

[15] F. Martín and V. Ramos Batista, The embedded singly periodic Scherk–Costa surfaces, Math. Ann. 336 (2006), no. 1, 155–189. 10.1007/s00208-006-0778-zSearch in Google Scholar

[16] H. Masur, Extension of the Weil–Petersson metric to the boundary of Teichmuller space, Duke Math. J. 43 (1976), no. 3, 623–635. 10.1215/S0012-7094-76-04350-7Search in Google Scholar

[17] W. H. Meeks, III, The theory of triply periodic minimal surfaces, Indiana Univ. Math. J. 39 (1990), no. 3, 877–936. 10.1512/iumj.1990.39.39043Search in Google Scholar

[18] B. Mohar and C. Thomassen, Graphs on surfaces, Johns Hopkins Stud. Math. Sci., Johns Hopkins University, Baltimore 2001. Search in Google Scholar

[19] F. Morabito and M. Traizet, Non-periodic Riemann examples with handles, Adv. Math. 229 (2012), no. 1, 26–53. 10.1016/j.aim.2011.09.001Search in Google Scholar

[20] J. Pérez and M. Traizet, The classification of singly periodic minimal surfaces with genus zero and Scherk-type ends, Trans. Amer. Math. Soc. 359 (2007), no. 3, 965–990. 10.1090/S0002-9947-06-04094-3Search in Google Scholar

[21] H. F. Scherk, Bemerkungen über die kleinste Fläche innerhalb gegebener Grenzen, J. reine angew. Math. 13 (1835), 185–208. 10.1515/crll.1835.13.185Search in Google Scholar

[22] A. H. Schoen, Infinite periodic minimal surfaces without self-intersections, NASA Technical Note, National Aeronautics and Space Administration, 1970. Search in Google Scholar

[23] M. Traizet, Construction de surfaces minimales en recollant des surfaces de Scherk, Ann. Inst. Fourier (Grenoble) 46 (1996), no. 5, 1385–1442. 10.5802/aif.1554Search in Google Scholar

[24] M. Traizet, Weierstrass representation of some simply-periodic minimal surfaces, Ann. Global Anal. Geom. 20 (2001), no. 1, 77–101. Search in Google Scholar

[25] M. Traizet, Adding handles to Riemann’s minimal surfaces, J. Inst. Math. Jussieu 1 (2002), no. 1, 145–174. 10.1017/S147474800200004XSearch in Google Scholar

[26] M. Traizet, An embedded minimal surface with no symmetries, J. Differential Geom. 60 (2002), no. 1, 103–153. 10.4310/jdg/1090351085Search in Google Scholar

[27] M. Traizet, On the genus of triply periodic minimal surfaces, J. Differential Geom. 79 (2008), no. 2, 243–275. 10.4310/jdg/1211512641Search in Google Scholar

[28] M. Traizet, Opening infinitely many nodes, J. reine angew. Math. 684 (2013), 165–186. 10.1515/crelle-2011-0007Search in Google Scholar

[29] R. Younes, Surfaces minimales dans des variétés homogènes, PhD thesis, Université de Tours, 2009. Search in Google Scholar

Received: 2021-10-28
Revised: 2023-10-28
Published Online: 2024-01-31
Published in Print: 2024-03-01

© 2024 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 28.1.2026 from https://www.degruyterbrill.com/document/doi/10.1515/crelle-2023-0086/html?lang=en
Scroll to top button