Plateau's problem for singular curves
Paul Creutz

TL;DR
This paper extends the solution of Plateau's problem to singular curves with self-intersections, utilizing a general metric space approach, and presents new results even in Euclidean spaces.
Contribution
It provides a novel solution to Plateau's problem for singular, self-intersecting curves in broad metric spaces, including Euclidean spaces.
Findings
Solution applicable to singular curves with self-intersections
Works in very general metric spaces
Main results are new even in Euclidean spaces
Abstract
We give a solution of Plateau's problem for singular curves possibly having self-intersections. The proof is based on the solution of Plateau's problem for Jordan curves in very general metric spaces by Alexander Lytchak and Stefan Wenger and hence works also in a quite general setting. However the main result of this paper seems to be new even in .
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Numerical Analysis Techniques · Computational Geometry and Mesh Generation · Mathematics and Applications
Plateau’s problem for singular curves
Paul Creutz
Paul Creutz, Mathematisches Institut der Universität zu Köln, Weyertal 86-90, 50931 Köln, Germany
Abstract.
We give a solution of Plateau’s problem for singular curves possibly having self-intersections. The proof is based on the solution of Plateau’s problem for Jordan curves in very general metric spaces by Alexander Lytchak and Stefan Wenger and hence works also in a quite general setting. However the main result of this paper seems to be new even in .
The author was partially supported by the DFG grant SPP 2026.
1. Introduction
1.1. Main Result
Classical formulation of Plateau’s problem reads: Does there exist a disc of least area spanning a given closed curve in ? If yes, what can be said about regularity of such discs?
The first question has been positively answered by Douglas and Radó independently for general Jordan curves in 1930 and the second one and its variants have been extensively studied in the last 90 years, see [DHS10] and [DHT10] and the references therein for a detailled account on the literature. However the classical approaches break down if instead of Jordan curves one considers closed curves having self-intersections. Reason for this being that there need not be a disc of least energy spanning . Joel Hass attacked the question of finding a least area disc for such singular curves in [Has91] and showed by cutting into possibly infinitely many curves, that an area minimizing disc for exists in the sense of continuous maps and Lebesgue notion of area. However the resulting disc a priori does not satisfy any regularity at the possibly large preimage of beyond continuouity and does not quite fit into the classical setting as it might not be a Sobolev map. Using results for Plateau’s problem in singular metric spaces given by Alexander Lytchak and Stefan Wenger in [LW17a] we are able to obtain the following result.
Theorem 1.1**.**
Let be a closed rectifiable curve in possibly having self-intersections. Then there exists and a Sobolev disc spanning of least area among all discs spanning . One may choose locally -Hölder continuous on and globally -Hölder continuous on where and .
Here by a closed curve we mean an equivalence class of parametrized closed curves where we identify two parametrized curves if they the same constant speed parametrization up to isometry of . We say that a Sobolev disc spans if has a continuous representative which is an element of . The area of is given by
[TABLE]
Theorem 1.1 is not only interesting for self-intersecting curves but also seems to be new for rectifiable Jordan curves of low regularity. Let be a rectifiable Jordan curve in and be a disc of least area spanning classically obtained by minimizing the Dirichlet energy. Then satisfies Laplace equation and hence is smooth on which is much stronger than local Hölder continuouity, see [DHS10]. If furthermore satisfies a -chord-arc condition, then is known to be globally -Hölder continuous for some depending on , see [HvdM99]. If and is then one may take any by [Nit65, p.238]. Imposing higher regularity on one may gain also higher boundary regularity of , see for example [Nit69], [War70]. However for a general rectifiable Jordan curve in the only boundary regularity of that seems to be known is .
The idea of proof is to replace by the metric space obtained by attaching a collar to along . Now we use the results in [LW17a] to solve Plateau’s problem for a certain regular Jordan curve in . A simple projection argument completes the proof. This way we trade in the singularity of the original curve for the singularity of .
1.2. Sketch of proof
Giving the ’right’ definitions, Plateau’s problem can almost verbatim be asked not only in but for a curve in a general complete metric space , see section 2. If is proper and a Jordan curve, Plateau’s problem has been solved by Alexander Lytchak and Stefan Wenger in [LW17a]. This result has been generalized to Jordan curves in many locally non-compact spaces by Chang-Yu Guo and Stefan Wenger in [GWar]. To obtain regularity of the maps they additionally assume the space to satisfy a quadratic isoperimetric inequality. A metric space is said to satisfy a -quadratic isoperimetric inequality if for every parametrized Lipschitz curve there exists such that and
[TABLE]
Instead of cutting the curve as Hass does in [Has91] the idea of proving theorem 1.1 is to resolve the self-intersections by gluing a collar. Assume for simplicity is a closed curve of length in . Let be a constant speed parametrization of and the metric space obtained by gluing a strip to along . Then satisfies a quadratic isoperimetric inequality and the curve corresponding to is a Jordan curve in satisfying a chord-arc condition. So the results in [LW17a] give a ’nice’ disc spanning of least area within the metric space . Now there is a canonical -Lipschitz retraction . Then gives the desired disc spanning . The details may be found in section 4.1.
This proof is not very specific to . It gives the following more general result.
Theorem 1.2**.**
Let be a proper metric space satisfying a -quadratic isoperimetric inequality and a closed rectifiable curve in . Then there exists and a disc of least area spanning .
Theorem 1.2 may be generalized to the class of spaces which are -complemented in some ultracompletion considered in [GWar]. It includes among others Hadamard spaces, dual Banach spaces, injective spaces and -spaces. The proof is the same upon replacing the [LW17a] results by the [GWar] ones. We will only give the proof of the locally compact result and indicate where one has to make changes to obtain the more general version.
A similar construction of gluing a strip has also turned out to be useful in [Cre19] and [Sta18].
1.3. Parametrized version and applications
Let be a proper metric space and a closed curve in . Let
[TABLE]
Now let be a parametrization of . A subtle but apparent question is whether there exists such that and . In the [Has91] setting the answer is always yes. However the procedure discussed there produces a map which might not be in . A positive answer to the question has been given in [LW16] if is a Jordan curve satisfying a chord-arc condition and a Lipschitz parametrization. We use this result and the same trick as in the proof of theorem 1.2 to obtain the following.
Theorem 1.3**.**
Let be a proper metric space satisfying a -quadratic isoperimetric inequality and a closed rectifiable curve in . If is a Lipschitz parametrization of , then there exists a disc such that and for some depending only on .
Especially if a metric space satisfies a -quadratic isoperimetric for every , then it satisfies a -quadratic isoperimetric inequality. This simplifies various things in the works of Alexander Lytchak and Stefan Wenger. For example one may slightly improve theorem in [LWYar] and theorem 1.4 in [LW17b] as follows.
Corollary 1.4**.**
Let and be a sequence of proper, geodesic metric spaces satisfying a -quadratic isoperimetric inequality. If is an ultralimit of , then satisfies a -quadratic isoperimetric inequality.
Corollary 1.5**.**
Let be a complete, geodesic metric space that is homeomorphic to a -dimensional manifold. Then satisfies a -quadratic isoperimetric inequality iff for every Jordan curve in there exists a Jordan domain bounded by such that
[TABLE]
2. Reminder: Sobolev maps
In this section we give a short reminder on metric space valued Sobolev maps. For more details see for example [LW17a], [KS93] and [Res97].
Let a bounded domain, a separable complete metric space and . A measurable, essentially separably valued map belongs to if for every -Lipschitz map the composition belongs to the classical Sobolev space and its Reshtnyak p-energy
[TABLE]
is finite. We say that belongs to if for every precompact one has .
If is a Lipschitz domain, then for there is a canonical almost everywhere defined trace map . If extends continuously to a map , then one may simply take .
If , then is approximately metrically differentiable almost everywhere. That is for almost every there exists a seminorm on such that
[TABLE]
Define the (Busemann) area of by
[TABLE]
where for a seminorm on one sets its (Busemann) Jacobian to be
[TABLE]
This equals the usual definition in case . We say that a map satisfies Lusin’s property (N) if for every with one has . If satisfies Lusin’s property (N), then the following variant of the area formula holds
[TABLE]
A map is -quasi-conformal if for -almost every one has
[TABLE]
for all . If is -quasi-conformal, then one has
[TABLE]
Also the following variant of the Sobolev embedding theorem holds.
Theorem 2.1**.**
Let , and . Then has a representative that satisfies Lusin’s property (N). If furthermore is a Lipschitz domain, then
[TABLE]
So especially .
3. Plateau’s problem for Jordan curves
3.1. Unparametrized Plateau problem
Consider endowed with angular distance. For let be a shortest path in connecting and . We say that a parametrized curve is of constant speed if there is a constant such that for every one has . For every rectifiable parametrized curve there exists a constant speed curve and a monotone map such that . Such a curve is unique up to precomposing an isometry of and will be called the constant speed parametrization of . Having the same constant speed parametrization defines an equivalence relation on the set of parametrized rectifiable curves . A closed rectifiable curve is an equivalence class with respect to . We say that is a parametrization of if .
Let be a closed rectifiable curve. We say that spans if has a continuous representative that is a parametrization of . We define the filling area of by
[TABLE]
In this formulation Plateau’s problem for reads: Is there a disc spanning such that ?
The classical approach to solve Plateau problem for Jordan curves in is to minimize the energy over all discs spanning . Then it turns out that a minimizer of energy exists and is also a minimizer of area. This breaks down for more general metric spaces as energy minimizers might not be area minimizers and worse for self-intersecting curves as there might not even exist energy minimizers, see [LW17a] resp. [Has91]. Still Alexander Lytchak and Stefan Wenger were able to solve Plateau’s problem for Jordan curves in proper metric spaces.
Theorem 3.1** ([LW17a]).**
Let be a proper metric space and a rectifiable Jordan curve in such that . Then there exists a disc of least area spanning . One may furthermore choose to minimize over all discs of least area spanning . In this case is -quasi-conformal. If satisfies property (ET), is even -quasi-conformal.
Here a metric space is called proper if closed and bounded subsets of are compact. A metric space is said to satisfy property (ET) if for every the metric differential is induced by a (possibly degenerate) inner product -almost everywhere on . Spaces satisfying property (ET) include complete Riemannian manifolds and spaces satisfying curvature bounds in the sense of Alexandrov, see [LW17a].
Theorem 3.1 has been generalized to many locally non-compact spaces including Hadamard spaces, dual Banach spaces and -spaces in [GWar]. It holds for all complete metric spaces that are -complemented in some ultracompletion. For convenience of the reader we stick in this article to proper spaces. However all results stated here for proper spaces also hold true within the class of spaces -complemented in some ultra completion.
To obtain regularity of solutions of Plateau’s problem one has to impose additional conditions on . Let and . We say that satisfies a -quadratic isoperimetric inequality if for every parametrized Lipschitz curve such that there exists a disc such that and
[TABLE]
If we also say that satisfies a -quadratic isoperimetric inequality. For example and more generally spaces satisfy a -quadratic isoperimetric inequality. For proper, geodesic spaces this is an equivalent characterization of the -condition, see [LW18b]. Banach spaces satisfy a -quadratic isoperimetric inequality, see [Crear].
Theorem 3.2** ([LW17a], [LW16]).**
Let be a complete metric space satisfying a -quadratic isoperimetric inequality and a rectifiable Jordan curve in . Let be a disc of least area spanning that is -quasiconformal. Then
- (1)
Then for and . 2. (2)
If satisfies a -chord-arc condition, then where and .
Here a rectifiable Jordan curve is said to satisfy a -chord-arc condition where is a constant if its constant speed parametrization satisfies
[TABLE]
for all .
3.2. Parametrized Plateau’s problem
Let be a rectifiable curve in and a parametrization of . We define the filling area of to be
[TABLE]
Then by definition one has . A first apparent question is whether equality holds. This cannot be true because even in as a rectifiable curve will usually have bad parametrizations that do not admit discs spanning them. However this is essentially the only restriction by the following result of Lytchak and Wenger.
Theorem 3.3** ([LW18a], lemma ).**
Let be a complete metric space satisfying a -quadratic isoperimetric inequality, a closed rectifiable curve in and a parametrization of . If , then .
However the proof of theorem 3.3 in [LW18a] is quite tricky and does not give the existence of a disc such that and . The question whether such exists could be considered a parametrized version of Plateau’s problem. The following partial result has been obtained in [LW16].
Theorem 3.4** ([LW16]).**
Let be a proper metric space satisfying a -quadratic isoperimetric inequality, a rectifiable -chord-arc curve in and a Lipschitz parametrization of . If , then there exists and a disc such that and . If furthermore , then one may also achieve
[TABLE]
where .
The following simple gluing lemma is crucial in the proofs of theorem 3.3 and theorem 3.4 and will also play an important role in the proof of theorem 1.2.
Lemma 3.5**.**
Let be a complete metric space, be parametrized curves and . Let be a homotopy between and and let such that . Then there exists a disc such that ,
[TABLE]
and
[TABLE]
where .
See for example section 2.2 in [LW16].
4. Plateau’s problem for singular curves
4.1. Unparametrized variant
Let be a complete metric space and a rectifiable curve in . Let , a geodesic circle of circumference and be a constant speed parametrization of . Consider a -Lipschitz map . Let be the metric quotient of along the relation generated by for . Let be given by the identity on and on .
Lemma 4.1**.**
Let as described.
- (1)
* is -Lipschitz, the inclusion map is an isometric embedding and the inclusion map is -Lipschitz and locally isometric on .* 2. (2)
Let be the curve in corresponding to in . Then is a -chord-arc curve of length . 3. (3)
If is proper, then is proper. 4. (4)
If satisfies -quadratic isoperimetric inequality, then satisfies a -quadratic isoperimetric inequality and a -quadratic isoperimetric inequality where . 5. (5)
If satisfies property (ET), then satisfies property (ET).
Remark 4.2**.**
For the locally non-compact version of the theorems one has to note: if is -completed in a ultra completion , then is -complemented in . This is an easy consequence of equation (12) and the fact that is compact.
Proof.
Let and and be the distance functions of and respectively. The distance on is given as follows, see [Cre19].
[TABLE]
1. and 2. are direct consequences of (12). 3. is a consequence of 1. For 4. see [Cre19]. To see 5. note that has property (ET) by assumption and has property (ET) as it is locally isometric to . ∎
We prove the following stronger variant of theorem 1.2.
Theorem 4.3**.**
Let be a complete metric space satisfying a -quadratic isoperimetric inequality and a closed rectifiable curve in such that . Then there is a disc of least area spanning . One may choose
[TABLE]
where , and . If satisfies property (ET), then may be improved to .
Proof.
Let be the constant speed parametrization of . Then gives a homotopy between and of area . So by theorem 3.3 and lemma 3.5
[TABLE]
Note here that as is a -Lipschitz retract of the quantities and are the same when considered within and within .
By theorem 3.1 there exists a disc of least area spanning that is -quasiconformal, where if satisfies property (ET) and otherwise. By lemma 4.1, theorem 3.2 and theorem 2.1 we may furthermore assume that
[TABLE]
and satisfies Lusin’s property (N) where ,
[TABLE]
and . By continuity of on one has . Then is a disc spanning having the desired regularity properties. As and satisfy Lusin’s property (N) by (5) one has
[TABLE]
[TABLE]
So which proves theorem 4.3. ∎
Theorem 1.1 follows from theorem 4.3 by noting that has property (ET) and satisfies a -quadratic isoperimetric inequality.
Remark 4.4**.**
Let be a space and a rectifiable curve in of finite total geodesic curvature . In this case one may improve the constant in theorem 4.3 to and hence the map will be locally Lipschitz on . However the constant will depend on in this case. To achieve this one performs the same proof using the funnel extension discussed in section 3.1 of [Sta18] instead of .
4.2. Parametrized variant
The proof of the following variant of theorem 1.3 is very similar to the proof discussed in the previous subsection.
Theorem 4.5**.**
Let be a proper metric space satisfying a -quadratic isoperimetric inequality, a closed rectifiable curve in such that and be a Lipschitz parametrization of .
Then there there exists and a disc such that and . If , then one may furthermore achieve
[TABLE]
where .
Proof.
First assume is the constant speed parametrization of . Consider again the space as in the previous subsection. By theorem 3.4 there exists and a disc such that and . If , then one may furthermore achieve
[TABLE]
where . Set . Then . Hence and the same argument as in the proof of theorem 4.3 implies .
Now for a general Lipschitz curve by [LWYar, lemma 3.6] there exists a universal constant and a Lipschitz homotopy between and such that and . So the proof is completed by applying lemma 3.5 to the disc constructed for and the homotopy . ∎
As a consequence of theorem 4.5 if a proper space satisfies a -quadratic isoperimetric for every then it satisfies a -quadratic isoperimetric inequality. Corollary 1.5 is an immediate consequence of this fact and theorem in [LW17b].
Corollary 4.6**.**
Let be a sequence of proper, geodesic metric spaces satisfying a -quadratic isoperimetric inequality. If is an ultralimit of , then satisfies a -quadratic isoperimetric inequality.
Proof.
By [LWYar, theorem 5.1] satisfies a -quadratic isoperimetric inequality for every . So if is proper we are done. If not one may perform the same proof as in [LWYar, theorem 5.1] but applying our theorem 4.5 instead of theorem 5.2 therein. ∎
5. Acknowledgements
I would like to thank my PhD advisor Alexander Lytchak for great support in everything.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[Cre 19] Paul Creutz. Space of minimal discs and its compactification. preprint arxiv, 2019.
- 2[Crear] Paul Creutz. Majorization by Hemispheres & Quadratic Isoperimetric Constants. Trans. Amer. Math. Soc. , to appear. preprint ar Xiv:1810.01340 v 2.
- 3[DHS 10] Ulrich Dierkes, Stefan Hildebrandt, and Friedrich Sauvigny. Minimal surfaces . Grundlehren der mathematischen Wissenschaften 339. Springer-Verlag Berlin Heidelberg, 2 edition, 2010.
- 4[DHT 10] Ulrich Dierkes, Stefan Hildebrandt, and Anthony J. Tromba. Regularity of minimal surfaces . Grundlehren der mathematischen Wissenschaften 340. Springer-Verlag Berlin Heidelberg, 2 edition, 2010.
- 5[G War] Chang-Yu Guo and Stefan Wenger. Area minimizing discs in locally non-compact metric spaces. Comm. Anal. Geom. , to appear. preprint ar Xiv:1701.06736 v 2.
- 6[Has 91] Joel Hass. Singular curves and the Plateau problem. Int. J. Math. , 2:1–16, 1991.
- 7[Hvd M 99] Stefan Hildebrandt and Heiko von der Mosel. On two-dimensional parametric variational problems. Calc. Var. Partial Dierential Equations , 9(3):249–267, 1999.
- 8[KS 93] Nicholas J. Korevaar and Nicholas J. Schoen. Sobolev spaces and harmonic maps for metric space targets. Comm. Anal. Geom. , 1(3-4):561–659, 1993.
