The title of this question may seem an oxymoron, but let me describe it and you'll see that perhaps it is not.


I am analysing the following Plateau problem.
Let $G\subsetneq\Bbb R^n$ be a general open set, its topological boundary and $\Bbb B^n(x, R)$ a ball of radius $R>0$ centred in $x\in\partial G$: moreover let the boundary $n-2$-dimensional contour (the Dirichlet data four our Plateau problem) be given by the intersection $\partial G\cap\partial \Bbb B^n(x, R)$.
It is a standard result, found for example in the monograph [1] by Enrico Giusti (§2, p. 17, due to Ennio De Giorgi), that there exist (at least a) minimal (hyper)surface $M$ spanning this contour: possibly $\partial G$ may act as an obstacle, forcing $M$ not to be exactly the spanning minimal surface (see for example [1] §2, remark 1.22, p. 19) and moreover uniqueness does not hold in general.

The question

Assumed that there isn't uniqueness, I am wondering if a phenomenon ow well posedness similar to the one seen for the first time by Calogero Vinti [2] for (nonlinear) hyperbolic equations may occur (see also here for a relevant Q&A). Precisely, Let's suppose that there exist multiple minimal (hyper)surfaces $M_i$ spanning the same contour $\partial G\cap\partial \Bbb B^n(x, R)$ and suppose that I can approximate this "contour", for example

  • by mollyfinig the characteristic function $\chi_G$ and considering a sequence of intersections $\partial G_k\cap\partial \Bbb B^n(x, R)$ ($\partial G_k$ being an appropriate level set of the mollified characteristic function), or
  • by stretching the ball $\Bbb B^n\big(x, R+{1\over k}\big)$ and considering the sequence of intersections $\partial G\cap\partial \Bbb B^n\big(x, R+{1\over k}\big)$.

Is it reasonable to expect the existence of sequences of approximating minimal surfaces $\{M_{i}^j\}_{j\in\Bbb N}$ converging uniformly to each "branch" $M_i$ respect to a proper norm $\|\cdot\|$, i.e. such that $\|M_i^j - M_i\|\to 0 $ uniformly in $x$?

I think this should be possible thanks to the existence of a maximum principle for the minimal surface equation, but not being an expert, I decided to ask here for an expert advice.


  • After a clarification requested by Daniel Asimov, I added this picture (taken from [1], §2, p. 19, picture) I've done in order to clarify (in the next future...) an earlier Math.SE answer I gave a few months ago. In this picture context,

    • $\Omega\equiv G$ is the the starting open set, also acting as an obstacle,
    • $L\equiv \Bbb B^n(x, R)$ is the Caccioppoli set which, jointly with $\partial\Omega$, defines the contour
    • the dashed set $E$ is the minimising Caccioppoli set, the part of whose boundary shown in red is the minimal surface $M$.

    In the picture we see that $M$ (i.e. $\partial E\setminus \partial L$) is a minimal surface in its rectilinear parts, while it coincides with $\partial\Omega$ otherwise. enter image description here

A loosely related note

Recently (26th of March 2024) Enrico Giusti passed away: I believe that the best thing I can do in order to remember him is recalling Karl Weierstraß's words for Sofya Kovalevskaya (as also Francesco Tricomi did in remembering Vito Volterra):

Die Menschen sterben, die Gedanken bleiben.


[1] Enrico Giusti, Minimal surfaces and functions of bounded variation, (English) Monographs in Mathematics, Vol. 80, Boston-Basel-Stuttgart: Birkhäuser, pp. XII+240, ISBN: 0-8176-3153-4, MR0775682, Zbl 0545.49018.

[2] Calogero Vinti "Su una specie di dipendenza continua delle soluzioni dal dato iniziale, per l’equazione $p=f(q)$, in una classe ove manca l’unicità [On a kind of continuous dependence of solutions from the initial data, for the $p=f(q)$ equation, in a class where uniqueness lacks]" (in Italian), Annali della Scuola Normale Superiore di Pisa - Classe di Scienze, Serie 3, Volume 19 (1965) no. 2, p. 251-263, MR185249, Zbl 0133.04602.

  • $\begingroup$ The phrase "forcing 𝑀 not to be exactly the spanning minimal surface" leaves me wondering exactly what the result of De Giorgi states. What is M, then? $\endgroup$ Apr 11 at 16:48
  • $\begingroup$ @DanielAsimov $M$ is partly a minimal surface and partly (and possibly) the boundary of $G$ which acts as an obstacle. I will ad an explanatory picture taken from Giusti's monograph. $\endgroup$ Apr 12 at 8:25

1 Answer 1


I am far from being an expert, but if I understand your question correctly, the answer is "no", at least for the Plateau problem in the realm of Caccioppoli sets.

Vinti's cited result [2] proves (the Theorem at bottom page 254) that for every approximating sequence of boundary datums, you can find a sequence of solutions (with this data at the boundary) converging uniformly. The desired statement for minimal surfaces would be the following.

Say $\Omega = B_1 \subset \mathbb{R}^3$ and you have a $1$-dimensional boundary datum $\gamma \subset \partial B_1$. Then you claim that for every $\gamma_k \to \gamma$ then, there exists a sequence of minimizers $M_k$ with $M_k\cap \partial B_1=\gamma_k$ with least area (not necessarily unique) that converges uniformly.

Take the following example. In $B_1$ there exists exactly one catenoid $C\subset B_1$ whose boundary is two parallel circles in $\partial B_1$ and such that its area is equal to the area of the two circles. That is, the Plateau minimizer for the boundary datum $C\cap \partial B_1$ is not unique. Moreover, if you move the two circles a bit closer, then the minimizer is unique, and it's the catenoid connecting them, and if you move them a bit further, the minimizer is unique and is the union of the two parallel disks. Now take an approximating sequence $\Gamma_k \subset \partial B_1$ of boundary data that oscillates between these two cases; that is, you always swap between two circles slightly closer and slightly further than the critical configuration. Then, for every $k$, there is a unique Plateau minimizer with boundary $\Gamma_k$, and it oscillates between a catenoid and two disks, so it cannot converge uniformly.

  • $\begingroup$ Thank you very much Michele: I suspected that the non uniqueness could be a problem when, for example, $\gamma_n$ changes its connectivity (i.e. when a loop on $\partial B_1$ collapses to a single point or joins with another loop): now I see that the problem is much worser as in the example you described, when $\gamma_n$ has the same connectivity of $\gamma$ but a branch minimal surface disappears suddenly. $\endgroup$ Apr 12 at 14:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.