🔢 mathematics

An introduction to the vectorial Calculus of Variations in $\textbf{L}^\infty$ AND Aronsson PDE systems

This expository article introduces the vectorial Calculus of Variations in $L^\infty$ , focusing on supremal functionals and the associated Aronsson PDE systems that arise as extremality conditions, while highlighting the distinct challenges and tools required compared to classical integral variational problems.

Original authors: Hussien Abugirda, Nikos Katzourakis

Published 2026-08-18

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hussien Abugirda, Nikos Katzourakis

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Mathematicians have long sought to understand how things change, how shapes settle, and how systems find their most efficient state. This quest, known as the calculus of variations, usually involves adding up tiny bits of energy across a whole region to find a total average. Imagine trying to smooth out a crumpled sheet of paper; the traditional method looks at the total amount of crumpling everywhere and tries to minimize that sum. This approach has worked beautifully for centuries, guiding everything from the design of bridges to the paths of planets. However, there is a different kind of problem where the average doesn't matter as much as the worst single spot. In these scenarios, the goal is not to lower the total energy, but to ensure that no single point ever spikes too high. This is the realm of the supremum, a mathematical way of looking at the absolute maximum value a function can reach. While the old methods of adding things up fail completely here, a new field has emerged to tackle these "worst-case" scenarios, offering a way to solve problems where a single large error is unacceptable.

In this new field, researchers are exploring a specific type of problem involving maps that stretch from one space into another, often with multiple dimensions. The authors of this paper, H. Abugirda and N. Katzourakis, are guiding us through the latest developments in this area, focusing on what happens when we try to minimize the maximum energy of these multi-dimensional maps. They explain that while the theory for simple, single-line problems has been understood for decades, the more complex, multi-dimensional versions have only recently begun to be unraveled. The central challenge is that the standard tools used to find solutions in the old world of averages simply do not work when looking for the absolute maximum. The equations that describe these optimal shapes are strange and difficult, often breaking down or behaving unpredictably in ways that traditional mathematics cannot easily handle.

The paper introduces a set of complex equations, known as Aronsson systems, which act as the rules for finding these optimal shapes. These equations are the result of pushing a familiar mathematical process to its extreme limit, effectively asking what happens when the "average" calculation is replaced by a "maximum" calculation. The researchers show that for maps with multiple dimensions, the solution is not a single, smooth curve but often a structure that splits into different phases. In some parts of the map, the shape behaves like a simple, one-dimensional line, while in other parts, it behaves like a complex, multi-dimensional surface. These different regions are separated by sharp boundaries where the behavior of the map changes abruptly. This phenomenon, called phase separation, means that the mathematical rules governing the shape are not the same everywhere; they switch depending on which phase the map is currently in. This creates a system with coefficients that can jump or become discontinuous, making the equations much harder to solve than their simpler counterparts.

One of the most significant findings in the paper is that the traditional idea of a "best" solution, where a shape is the absolute minimum everywhere, does not work for these multi-dimensional problems. In the simpler, single-dimensional cases, finding the best solution was straightforward, but in the complex, multi-dimensional world, the researchers discovered that the standard definition of a minimum fails to capture the true nature of the solution. Instead, they propose a new way of thinking about these solutions, breaking them down into two distinct parts: one that moves along the surface of the shape and another that moves perpendicular to it. Each part has its own specific rules for what makes it optimal. This distinction is crucial because it reveals that a shape can be optimal in one direction while not being optimal in another, a nuance that was previously missed.

The paper also addresses the difficult question of whether these complex equations actually have solutions at all. Because the equations are so irregular and the coefficients can jump, standard methods for proving that a solution exists do not apply. To overcome this, the authors describe a new, sophisticated approach that treats the derivatives of the shape not as single, fixed values, but as a spread of possibilities, similar to how a cloud of particles might behave. This method, which uses a concept called Young measures, allows mathematicians to find solutions even when the shape is not perfectly smooth or differentiable. They prove that for a wide class of these problems, there is not just one solution, but an infinite number of valid solutions that satisfy the conditions. This lack of uniqueness is not a failure of the method but a fundamental feature of the problem itself, reflecting the complex and flexible nature of these multi-dimensional shapes.

Ultimately, this work provides a roadmap for understanding a class of problems that are vital for many real-world applications, from optimizing the design of materials to improving weather forecasting models. By moving beyond the limitations of traditional averaging and embracing the complexity of maximum values, the authors have opened the door to solving problems where the worst-case scenario is the only thing that matters. Their work clarifies that the path to the best solution in these complex systems is not a single, smooth line, but a rich, multi-faceted structure that requires new tools and new ways of thinking to understand. The journey from the simple, smooth curves of the past to the jagged, phase-separated landscapes of the present marks a significant step forward in our ability to model and control the most difficult optimization problems in mathematics.

Technical Summary: An Introduction to the Vectorial Calculus of Variations in $L^\infty$ and Aronsson PDE Systems

1. Problem Statement and Context

The paper addresses the Calculus of Variations in $L^\infty$ , a field concerned with variational problems involving supremal functionals rather than the classical integral functionals. While the scalar theory (where the unknown function maps to $\mathbb{R}$ ) was established by G. Aronsson in the 1960s, the vectorial case (where the unknown map $u: \Omega \subseteq \mathbb{R}^n \to \mathbb{R}^N$ with $N \geq 2$ ) is a much more recent development, originating in the early 2010s.

The central problem is the lack of a robust variational framework for vectorial supremal functionals. In the classical integral setting ( $L^p$ ), minimizers are characterized by Euler-Lagrange equations derived via differentiation. However, supremal functionals of the form:
$E_\infty(u, \mathcal{O}) := \text{ess sup}_{x \in \mathcal{O}} H(x, u(x), Du(x))$
are not differentiable. Consequently, standard variational techniques fail. Furthermore, the concept of a "global minimizer" is insufficient in $L^\infty$ due to the lack of additivity of the supremum; the appropriate notion is that of an absolute minimizer, which minimizes the functional on every subdomain with respect to its own boundary values.

The paper focuses on the derivation and analysis of the Aronsson systems, the PDE systems that arise as extremality conditions for these functionals, specifically highlighting the significant complications introduced by the vectorial nature of the problem ( $N \geq 2$ ).

2. Methodology

The authors employ a multi-faceted approach to introduce the field, contrasting it with the classical integral Calculus of Variations:

Limiting Procedures ( $L^p$ Approximation): A primary method for deriving the Aronsson systems is taking the formal limit as $p \to \infty$ of the Euler-Lagrange equations associated with $L^p$ -Dirichlet integrals ( $\int |Du|^p$ ). This reveals the structure of the limiting PDEs.
Geometric Decomposition: In the vectorial case, the authors utilize orthogonal projections onto the range of the gradient $Du$ and its orthogonal complement. This allows the decomposition of the limiting PDE system into tangential and normal components relative to the image of the map.
Generalized Solutions (D-solutions): Recognizing that classical solutions (even $C^2$ ) may not exist or may not satisfy the PDE in a pointwise sense due to discontinuous coefficients, the paper introduces D-solutions. This theory relies on Young measures and probabilistic representations of derivatives (diffuse derivatives) via limits of difference quotients, avoiding integration by parts or maximum principles.
Variational Characterization: The paper investigates whether specific classes of variations (scalar vs. vectorial, tangential vs. normal) can characterize the solutions to the Aronsson systems, distinguishing between "rank-one absolute minimizers" and "minimal area" conditions.

3. Key Contributions and Results

3.1. The Vectorial Aronsson System

The paper derives the full vectorial Aronsson system for a general supremand $H(x, \eta, P)$ . Unlike the scalar case, which yields a single second-order quasilinear equation, the vectorial case yields a system that generally consists of two independent parts:

Tangential Component: Related to the variation along the image of the map.
Normal Component: Related to variations normal to the image.

For the model $\infty$ -Dirichlet functional ($H = |Du|$), the system is:
$\Delta_\infty u := \left( Du \otimes Du + |Du|^2 [[Du]]^\perp \otimes I \right) : D^2 u = 0$
where $[[Du]]^\perp$ is the projection onto the orthogonal complement of the range of $Du$. This system decouples into:

$Du \, D(\frac{1}{2}|Du|^2) = 0$ (Tangential)
$|Du|^2 [[Du]]^\perp \Delta u = 0$ (Normal)

3.2. Phase Separation and Discontinuous Coefficients

A critical finding is that the vectorial Aronsson system possesses discontinuous coefficients, even for smooth solutions. This occurs because the rank of the gradient $Du$ may not be constant throughout the domain.

Phase Separation: The domain decomposes into phases (e.g., $\Omega_1$ where $\text{rk}(Du)=1$ and $\Omega_2$ where $\text{rk}(Du)=2$ ).
Interfaces: At the interfaces between these phases, the projection operator $[[Du]]^\perp$ becomes discontinuous.
Structure Theorem: For $C^2$ solutions in 2D, the map is either an immersion solving the Eikonal equation (rank 2) or essentially a scalar $\infty$ -harmonic function (rank 1).

3.3. Failure of Absolute Minimality

The paper demonstrates that, unlike the scalar case, vectorial absolute minimizers do not necessarily characterize $\infty$ -harmonic maps.

The $\infty$ -Laplace system is not sufficient for absolute minimality in the vectorial case.
A new variational notion, $\infty$ -Minimal maps, is introduced, requiring the map to be a "rank-one absolute minimal" (tangential) and to have " $\infty$ -minimal area" (normal).
Counterexamples exist where smooth $\infty$ -harmonic maps are not global minimizers for the supremal functional.

3.4. Existence of D-solutions

Addressing the non-existence of classical solutions and the failure of viscosity solution theory (which applies to scalar non-divergence equations but not directly to vectorial systems with discontinuous coefficients), the paper presents an existence result for D-solutions.

Theorem: For the Dirichlet problem of the Aronsson system with specific structural assumptions on $H$ , there exists an infinite set of D-solutions in the class of Lipschitz submersions.
Non-uniqueness: The solution is non-unique, a phenomenon also observed in smooth solutions for $n=N=2$ . The proof utilizes the Baire category method (Dacorogna-Marcellini) applied to a differential inclusion.

4. Significance and Applications

The paper positions the Vectorial Calculus of Variations in $L^\infty$ as a distinct and challenging field requiring new tools beyond classical PDE theory. Its significance lies in:

Theoretical Depth: It establishes that the vectorial case is not merely a generalization of the scalar case but possesses unique structural features (phase separation, discontinuous coefficients, non-uniqueness) that fundamentally alter the variational landscape.
Applications: The authors cite several real-world motivations where minimizing the maximum error (supremal) is more realistic than minimizing the average (integral):
- Optimal Lipschitz Extensions: Finding the smoothest extension of boundary data.
- Optimizing Quasi-conformal Mappings: Minimizing the dilation of maps in geometric analysis.
- Variational Data Assimilation: Specifically in weather forecasting (4DVar), where minimizing the maximum deviation of a model from observations is crucial to avoid "spikes" in error.
- Image Processing and Engineering: Applications in optimal shape design, inverse problems, and fracture mechanics.

The paper concludes that while the scalar theory is well-developed, the vectorial theory remains in its infancy, with the Aronsson systems serving as the central object of study for extremality conditions in $L^\infty$ . The introduction of D-solutions provides a rigorous framework to handle the lack of differentiability and the non-elliptic nature of these systems.

Technical Summary: An Introduction to the Vectorial Calculus of Variations in L∞L^\inftyL∞ and Aronsson PDE Systems