Home Existence and symmetry of periodic nonlocal-CMC surfaces via variational methods
Article
Licensed
Unlicensed Requires Authentication

Existence and symmetry of periodic nonlocal-CMC surfaces via variational methods

  • Xavier Cabré ORCID logo , Gyula Csató ORCID logo EMAIL logo and Albert Mas ORCID logo
Published/Copyright: October 5, 2023

Abstract

This paper provides the first variational proof of the existence of periodic nonlocal-CMC surfaces. These are nonlocal analogues of the classical Delaunay cylinders. More precisely, we show the existence of a set in R n which is periodic in one direction, has a prescribed (but arbitrary) volume within a slab orthogonal to that direction, has constant nonlocal mean curvature, and minimizes an appropriate periodic version of the fractional perimeter functional under the volume constraint. We show, in addition, that the set is cylindrically symmetric and, more significantly, that it is even as well as nonincreasing on half its period. This monotonicity property solves an open problem and an obstruction which arose in an earlier attempt, by other authors, to show the existence of minimizers.

Award Identifier / Grant number: MTM2017-84214-C2-1-P

Award Identifier / Grant number: PID2021-123903NB-I00

Award Identifier / Grant number: RED2018-102650-T

Award Identifier / Grant number: MTM2017-83499-P

Award Identifier / Grant number: PID2021-125021NA-I00

Award Identifier / Grant number: CEX2020-001084-M

Funding statement: The three authors are supported by the Spanish grants MTM2017-84214-C2-1-P, PID2021-123903NB-I00, and RED2018-102650-T funded by MCIN/AEI/10.13039/501100011033 and by “ERDF A way of making Europe”. The second and third authors are also supported by the Spanish grant MTM2017-83499-P. The second author is in addition supported by the Spanish grant PID2021-125021NA-I00. This work is supported by the Spanish State Research Agency, through the Severo Ochoa and María de Maeztu Program for Centers and Units of Excellence in R&D (CEX2020-001084-M).

A Periodic minimizers in R 2

In this appendix, we present, when n = 2 , an expression for the functional P s which is simpler than that of Lemma 2.2 in R n . As mentioned in the introduction, this expression allowed us to prove the existence of a minimizer when n = 2 in a simple way. We later found the existence proof of Theorem 1.1 in all dimensions.

The formula for the periodic version of the fractional perimeter when n = 2 reads as follows. If E = { ( x 1 , x 2 ) R 2 : | x 2 | < u ( x 1 ) } for some nonnegative 2 π -periodic function 𝑢, then P s [ E ] can be written in terms of 𝑢 as

(A.1) P s [ E ] = π π d x 1 + d y 1 2 | x 1 y 1 | s × { G ( u ( x 1 ) u ( y 1 ) | x 1 y 1 | ) + H ( u ( x 1 ) + u ( y 1 ) | x 1 y 1 | ) } ,

where

G ( t ) := 0 t d τ ( t τ ) ( 1 + τ 2 ) 2 + s 2 and H ( t ) := G ( + ) t G ( t ) for  t R .

Note that G ( 0 ) = 0 ,

0 < G ( t ) = 0 t d τ ( 1 + τ 2 ) 2 + s 2 < G ( + ) < + ,

and G ′′ ( t ) = ( 1 + t 2 ) 2 + s 2 > 0 for all t R . Therefore, 𝐺 is a nonnegative even function which is increasing in ( 0 , + ) and strictly convex on ℝ. Furthermore, it behaves as a quadratic function near the origin and approaches a linear one as t ± . On the other hand, H ( 0 ) = 0 , H = G ( + ) G > 0 , and H ′′ = G ′′ < 0 in [ 0 , + ) . Thus, 𝐻 is positive, increasing, and strictly concave in [ 0 , + ) . Finally, by Fubini’s theorem,

H ( + ) = 0 + d t H ( t ) = 0 + d t ( G ( + ) G ( t ) ) = 0 + d t t + d τ ( 1 + τ 2 ) 2 + s 2 = 0 + d τ τ ( 1 + τ 2 ) 2 + s 2 < + .

For the sake of completeness we will give the proof of (A.1), even though this expression was not used to prove the results of this paper. But before, let us explain how to use (A.1) to give a simple proof of the existence of a constrained minimizer of P s when n = 2 . The argument goes as follows. Since G ( t ) + C 1 C 2 | t | for some constants C 1 , C 2 > 0 , the term in (A.1) involving 𝐺 bounds the W s , 1 ( π , π ) norm of 𝑢 up to an additive constant. In this way, since H 0 , a minimizing sequence for P s is a bounded sequence in W s , 1 ( π , π ) . By compactness, this yields the existence of the desired constrained minimizer. We refer to [2] for the details on this argument; see also the end of Part 1 in our proof of Theorem 1.1.

Lemma A.1

Let u : R R be nonnegative, 2 π -periodic, and such that

u W s , 1 ( π ϵ , π + ϵ ) for some (and hence for all) ϵ > 0 .

Let E = { ( x 1 , x 2 ) R 2 : | x 2 | < u ( x 1 ) } . Then (A.1) holds.

Proof

We prove (A.1) using (2.2). Let us first rewrite 𝜙 in (2.3) for n = 2 in a more convenient way by introducing the function G . For p 0 and q 0 , we have (using the change of variables z = τ + w twice)

ϕ ( p , q ) = p p d w { q d z ( 1 + | z w | 2 ) 2 + s 2 + q + d z ( 1 + | z w | 2 ) 2 + s 2 } = p p d w { G ( q w ) G ( ) + G ( + ) G ( q w ) } .

Note that G is odd, 0 < G ( ) = G ( + ) < + , and G ( q ) G ( q ) is an even function. Therefore,

(A.2) ϕ ( p , q ) = 2 0 p d w { G ( q w ) G ( q w ) + 2 G ( + ) } = 4 G ( + ) p + 2 0 p d w G ( q w ) 2 0 p d w G ( q w ) = 4 G ( + ) p 2 q q p d t G ( t ) + 2 q q p d t G ( t ) = 2 ( 2 G ( + ) p + G ( p q ) G ( p + q ) ) ,

where we also used in the last equality that 𝐺 is even. Now, from (2.2) and (A.2), we deduce that

(A.3) P s [ E ] = 2 π π d x 1 + d y 1 | x 1 y 1 | s × { G ( u ( x 1 ) u ( y 1 ) | x 1 y 1 | ) G ( u ( x 1 ) + u ( y 1 ) | x 1 y 1 | ) + 2 G ( + ) u ( x 1 ) | x 1 y 1 | } .

Our goal is to show that, in the term involving 2 G ( + ) u ( x 1 ) | x 1 y 1 | 1 , we can replace u ( x 1 ) by u ( y 1 ) and identity (A.3) remains unchanged. Once this is proved, we get that, in this last term, one can replace 2 u ( x 1 ) by u ( x 1 ) + u ( y 1 ) . Hence, this gives (A.1), using that H ( t ) = G ( + ) t G ( t ) .

Therefore, it only remains to show that we can replace u ( x 1 ) by u ( y 1 ) in the last term on the right-hand side of (A.3). This follows immediately from Lemma 2.1 (used with m = l = 1 ) applied to the integral

π π d x + d y f ( x , y ) | x y | s ,

where

f ( x , y ) := G ( u ( x ) + u ( y ) | x y | ) + 2 G ( + ) u ( x ) | x y | .

To check the hypothesis of the lemma, we need to verify that

(A.4) π π d x + d y | f ( x , y ) | | x y | s < + .

For this purpose, we write f ( x , y ) = A + B , where

A := G ( u ( x ) + u ( y ) | x y | ) + G ( 2 u ( x ) | x y | ) , B := G ( 2 u ( x ) | x y | ) + 2 G ( + ) u ( x ) | x y | = H ( 2 u ( x ) | x y | ) .

Using that 𝑢 is nonnegative, 𝐻 is nonnegative and increasing in ( 0 , + ) , and H ( + ) < + , we have

(A.5) 0 B H ( + ) < + .

Concerning the quantity 𝐴, observe that

(A.6) | A | = | u ( x ) + u ( y ) | x y | 2 u ( x ) | x y | d t G ( t ) | sup t R | G ( t ) | | u ( x ) u ( y ) | | x y | = G ( + ) | u ( x ) u ( y ) | | x y | .

Recall that we are assuming u W s , 1 ( π ϵ , π + ϵ ) for some ϵ > 0 . According to this, to prove (A.4), we split the integral and we use (A.5) and (A.6) to get

(A.7) π π d x + d y | f ( x , y ) | | x y | s = π π d x | x y | ϵ d y | A | + | B | | x y | s + π π d x | x y | > ϵ d y | f ( x , y ) | | x y | s G ( + ) u W s , 1 ( π ϵ , π + ϵ ) + H ( + ) π π d x π ϵ π + ϵ d y | x y | s + π π d x | x y | > ϵ d y | f ( x , y ) | | x y | s .

The first two terms on the right-hand side of (A.7) are finite. To estimate the third one, we use that 0 G ( t ) G ( + ) t for all t 0 , and hence

(A.8) π π d x | x y | > ϵ d y | f ( x , y ) | | x y | s G ( + ) π π d x | x y | > ϵ d y 3 u ( x ) | x y | 1 + s + G ( + ) π π d x | x y | > ϵ d y u ( y ) | x y | 1 + s .

The first double integral in the right-hand side of (A.8) is finite since s > 0 and

u W s , 1 ( π ϵ , π + ϵ ) L 1 ( π , π ) .

For the second double integral, since 𝑢 is 2 π -periodic and nonnegative, using Lemma 2.1, we deduce

π π d x | x y | > ϵ d y u ( y ) | x y | 1 + s = π π d x R d y χ { | x y | > ϵ } | x y | 1 + s u ( y ) = π π d y ̄ u ( y ̄ ) | x ̄ y ̄ | > ϵ d x ̄ | x ̄ y ̄ | 1 + s .

As before, this is finite since s > 0 and u L 1 ( π , π ) . Hence (A.4) follows from (A.7) and the finiteness of the integrals in (A.8). ∎

B Strict monotonicity of the periodic heat kernel

In this section, we address the proof of an important ingredient in the proof of Theorem 1.2, namely, that the function defined in (3.13) is decreasing in ( 0 , π ) . As shown there, this property follows from the fact that the periodic heat kernel is decreasing with respect to the spatial variable in ( 0 , π ) . This is a well-known fact [1, 3, 14, 15, 24], whose proofs are commented next. Here, for the sake of completeness, we provide a simple proof based on maximum principles for the heat equation.

Prior to this, let us briefly comment on the proofs given in the above-mentioned references. The one in [1] follows exactly the same lines as the one presented here, but it is carried out within the context of hypersurfaces of revolution. The one in [15] is also based on the strong maximum principle for the heat equation, but it is more technically involved than [1] since a more general class of manifolds is considered there. The proof in [3] is based on an explicit computation. The one in [14] follows by expressing the derivative of the kernel in terms of the heat kernel in smaller dimensions and using an induction argument; see [14, §6.3]. Finally, the one in [24] is a straightforward application of the sharp estimates for the heat kernel on the sphere S n found there.

The fundamental solution of the heat equation in ( π , π ) with periodic boundary conditions is given by

Γ ( z , t ) := 1 4 π t k Z e ( z + 2 k π ) 2 4 t for  z R  and  t > 0 .

Given a 2 π -periodic initial data 𝑔, the function

u ( x , t ) = π π d y Γ ( x y , t ) g ( y )

satisfies t u x x u = 0 in R × ( 0 , + ) , that u ( , t ) is 2 π -periodic for all t > 0 , and that u ( , 0 ) = g in ℝ. This follows immediately from the properties of the standard (nonperiodic) heat kernel and the fact that

k Z π π d y e ( x y + 2 k π ) 2 4 t g ( y ) = k Z π π d y e ( x y + 2 k π ) 2 4 t g ( y 2 k π )
= k Z π 2 k π π 2 k π d y ̄ e ( x y ̄ ) 2 4 t g ( y ̄ )
= R d y ̄ e ( x y ̄ ) 2 4 t g ( y ̄ ) .

We now state and prove the result on the periodic heat kernel that we used in the proof of Theorem 1.2.

Theorem B.1

For every t > 0 , the function z Γ ( z , t ) is decreasing in ( 0 , π ) .

Proof

Let us first show that Γ ( , t ) is nonincreasing in ( 0 , π ) . For this, we take functions g ϵ C c ( ϵ , ϵ ) , where 0 < ϵ < π / 2 , approximating the Dirac delta as ϵ 0 , and we consider their 2 π -periodic extensions from [ π , π ] to ℝ (which we also denote by g ϵ ). In particular, we assume that

ϵ ϵ g ϵ = 1 for all  0 < ϵ < π / 2 .

We additionally assume that the g ϵ are nonnegative, even, and nonincreasing in ( 0 , π ) .

Let u ϵ be the solution to

{ t u ϵ x x u ϵ = 0 in R × ( 0 , + ) , u ϵ ( , 0 ) = g ϵ in R .

Thus, u ϵ ( x , t ) = π π d y Γ ( x y , t ) g ϵ ( y ) and u ϵ ( , t ) is 2 π -periodic for all t > 0 . Notice that u ϵ ( , t ) is even with respect to x = 0 and x = π (by uniqueness, since so is g ϵ ). We deduce that the derivative v ϵ := x u ϵ solves

{ t v ϵ x x v ϵ = 0 in ( 0 , π ) × ( 0 , + ) , v ϵ ( 0 , t ) = v ϵ ( π , t ) = 0 for all t > 0 .

Moreover, by the assumptions on g ϵ , we have that v ϵ ( x , 0 ) = x u ϵ ( x , 0 ) = g ϵ ( x ) 0 for all x [ 0 , π ] . Hence, the maximum principle yields

(B.1) v ϵ 0 in [ 0 , π ] × [ 0 , + ) .

We now claim that, for every given t > 0 and m = 0 , 1 , 2 , ,

(B.2) u ϵ ( , t ) converges to Γ ( , t ) in C m ( [ 0 , π ] ) as ϵ 0 .

This follows by combining the uniform continuity in [ 0 , π ] of Γ ( , t ) and of all its derivatives with the fact that, for every δ > 0 , we have

| x m u ϵ ( x , t ) x m Γ ( x , t ) | ϵ ϵ d y | x m Γ ( x y , t ) x m Γ ( x , t ) | | g ϵ ( y ) | δ

if 𝜖 is chosen small enough.

Now, from (B.1) and (B.2), we conclude that

x Γ ( x , t ) = lim ϵ 0 x u ϵ ( x , t ) = lim ϵ 0 v ϵ ( x , t ) 0

for all x [ 0 , π ] and t > 0 .

Finally, we show that Γ ( , t ) is decreasing in ( 0 , π ) . For this, notice that we already proved that, for all t > 0 , Γ ( , t ) is even with respect to x = 0 and x = π . Therefore, given any t 0 > 0 , we see that v := x Γ solves

{ t v x x v = 0 in ( 0 , π ) × ( t 0 , + ) , v ( 0 , t ) = v ( π , t ) = 0 for all t > t 0 , v ( , t 0 ) = x Γ ( , t 0 ) 0 in [ 0 , π ] .

Now, the strong maximum principle yields that either v < 0 in ( 0 , π ) × ( t 0 , + ) or v 0 in [ 0 , π ] × [ t 0 , + ) . The proof is now complete since v 0 is absurd – it would give that Γ ( , t ) is constant in 𝑥, clearly contradicting the fact that Γ is the fundamental solution of the heat equation with periodic boundary conditions, as shown in the beginning of this appendix. ∎

Acknowledgements

We thank M. Ritoré, E. Valdinoci, and T. Weth for sharing some references and information on the topic of the paper.

References

[1] D. Alonso-Orán, F. Chamizo, A. D. Martínez and A. Mas, Pointwise monotonicity of heat kernels, Rev. Mat. Complut. 36 (2023), no. 1, 207–220. 10.1007/s13163-021-00417-8Search in Google Scholar PubMed PubMed Central

[2] M. Alvinyà, Delaunay cylinders with constant non-local mean curvature, Master’s Thesis, UPC, 2017, https://upcommons.upc.edu/handle/2117/104594. Search in Google Scholar

[3] D. Andersson, Estimates of the spherical and ultraspherical heat kernel, Master’s Thesis, Chalmers University of Technology Gothenburg, 2013. Search in Google Scholar

[4] M. Athanassenas, A variational problem for constant mean curvature surfaces with free boundary, J. reine angew. Math. 377 (1987), 97–107. 10.1515/crll.1987.377.97Search in Google Scholar

[5] A. Baernstein, II, Convolution and rearrangement on the circle, Complex Variables Theory Appl. 12 (1989), no. 1–4, 33–37. 10.1080/17476938908814351Search in Google Scholar

[6] A. Baernstein, II, Symmetrization in analysis, New Math. Monogr. 36, Cambridge University, Cambridge 2019. 10.1017/9781139020244Search in Google Scholar

[7] A. Baernstein, II and B. A. Taylor, Spherical rearrangements, subharmonic functions, and * -functions in 𝑛-space, Duke Math. J. 43 (1976), no. 2, 245–268. 10.1215/S0012-7094-76-04322-2Search in Google Scholar

[8] A. Burchard and H. Hajaiej, Rearrangement inequalities for functionals with monotone integrands, J. Funct. Anal. 233 (2006), no. 2, 561–582. 10.1016/j.jfa.2005.08.010Search in Google Scholar

[9] X. Cabré, G. Csató and A. Mas, On the stability of cylinders as periodic nonlocal-CMC surfaces, to appear. Search in Google Scholar

[10] X. Cabré, M. M. Fall, J. Solà-Morales and T. Weth, Curves and surfaces with constant nonlocal mean curvature: Meeting Alexandrov and Delaunay, J. reine angew. Math. 745 (2018), 253–280. 10.1515/crelle-2015-0117Search in Google Scholar

[11] X. Cabré, M. M. Fall and T. Weth, Delaunay hypersurfaces with constant nonlocal mean curvature, J. Math. Pures Appl. (9) 110 (2018), 32–70. 10.1016/j.matpur.2017.07.005Search in Google Scholar

[12] X. Cabré, M. M. Fall and T. Weth, Near-sphere lattices with constant nonlocal mean curvature, Math. Ann. 370 (2018), no. 3–4, 1513–1569. 10.1007/s00208-017-1559-6Search in Google Scholar

[13] L. Caffarelli, J.-M. Roquejoffre and O. Savin, Nonlocal minimal surfaces, Comm. Pure Appl. Math. 63 (2010), no. 9, 1111–1144. 10.1002/cpa.20331Search in Google Scholar

[14] I. Chavel, Eigenvalues in Riemannian geometry, Pure Appl. Math. 115, Academic Press, Orlando 1984. Search in Google Scholar

[15] J. Cheeger and S. T. Yau, A lower bound for the heat kernel, Comm. Pure Appl. Math. 34 (1981), no. 4, 465–480. 10.1002/cpa.3160340404Search in Google Scholar

[16] J. Dávila, M. del Pino, S. Dipierro and E. Valdinoci, Nonlocal Delaunay surfaces, Nonlinear Anal. 137 (2016), 357–380. 10.1016/j.na.2015.10.009Search in Google Scholar

[17] C. Delaunay, Sur la surface de révolution dont la courbure moyenne est constante, J. Math. Pures Appl. 1 6 (1841), 309–315. Search in Google Scholar

[18] E. Di Nezza, G. Palatucci and E. Valdinoci, Hitchhiker’s guide to the fractional Sobolev spaces, Bull. Sci. Math. 136 (2012), no. 5, 521–573. 10.1016/j.bulsci.2011.12.004Search in Google Scholar

[19] R. Friedberg and J. M. Luttinger, Rearrangement inequality for periodic functions, Arch. Ration. Mech. Anal. 61 (1976), no. 1, 35–44. 10.1007/BF00251861Search in Google Scholar

[20] L. Hauswirth, J. Pérez, P. Romon and A. Ros, The periodic isoperimetric problem, Trans. Amer. Math. Soc. 356 (2004), no. 5, 2025–2047. 10.1090/S0002-9947-03-03362-2Search in Google Scholar

[21] M. Koiso and U. Miyamoto, Stability of hypersurfaces with constant mean curvature trapped between two parallel hyperplanes, preprint (2019), https://arxiv.org/abs/1905.01705. Search in Google Scholar

[22] E. H. Lieb and M. Loss, Analysis, 2nd ed., Grad. Stud. Math. 14, American Mathematical Society, Providence 2001. Search in Google Scholar

[23] A. Malchiodi, M. Novaga and D. Pagliardini, On critical points of the relative fractional perimeter, Ann. Inst. H. Poincaré C Anal. Non Linéaire 38 (2021), no. 5, 1407–1428. 10.1016/j.anihpc.2020.11.005Search in Google Scholar

[24] A. Nowak, P. Sjögren and T. Z. Szarek, Sharp estimates of the spherical heat kernel, J. Math. Pures Appl. (9) 129 (2019), 23–33. 10.1016/j.matpur.2018.10.002Search in Google Scholar

[25] R. H. L. Pedrosa and M. Ritoré, Isoperimetric domains in the Riemannian product of a circle with a simply connected space form and applications to free boundary problems, Indiana Univ. Math. J. 48 (1999), no. 4, 1357–1394. 10.1512/iumj.1999.48.1614Search in Google Scholar

[26] M. Ritoré and E. Vernadakis, Large isoperimetric regions in the product of a compact manifold with Euclidean space, Adv. Math. 306 (2017), 958–972. 10.1016/j.aim.2016.11.001Search in Google Scholar

[27] T. I. Vogel, Stability of a liquid drop trapped between two parallel planes, SIAM J. Appl. Math. 47 (1987), no. 3, 516–525. 10.1137/0147034Search in Google Scholar

Received: 2022-11-04
Revised: 2023-07-30
Published Online: 2023-10-05
Published in Print: 2023-11-01

© 2023 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 10.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/crelle-2023-0057/html
Scroll to top button