🔢 mathematics

Quasiconvexity in the Riemannian setting

This paper introduces a notion of quasiconvexity for continuous functions on the vector bundle of linear maps between tangent spaces of a Riemannian manifold and $\mathbb{R}^m$ , proving that this condition characterizes the sequential lower semicontinuity of the associated integral functional with respect to the weak $^*$ topology of $W^{1,\infty}$ .

Original authors: Aurora Corbisiero, Chiara Leone, Carlo Mantegazza

Published 2026-04-21

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Aurora Corbisiero, Chiara Leone, Carlo Mantegazza

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

The Big Picture: Smoothing Out the Rough Edges

Imagine you are an architect trying to design the most efficient, stable, and "smooth" building possible. In mathematics, this is called the Calculus of Variations. You have a formula (a functional) that calculates the "cost" or "energy" of a design. Your goal is to find the design that minimizes this cost.

However, there's a catch: sometimes, as you tweak your design to get closer to the perfect one, the cost doesn't settle down nicely; it jumps around or behaves erratically. Mathematicians call this a lack of lower semicontinuity. If a formula isn't "lower semicontinuous," you can't guarantee that your "best" design actually exists.

For decades, mathematicians knew exactly when this formula behaves nicely in flat, Euclidean space (like a standard grid on a piece of paper). The secret ingredient is a property called Quasiconvexity.

This paper asks: What happens if we aren't building on a flat piece of paper, but on a curved, bumpy surface like the Earth (a Riemannian manifold)? Does the same secret ingredient still work?

The authors, Aurora Corbisiero, Chiara Leone, and Carlo Mantegazza, say yes. They have successfully translated the rules of "smoothness" from flat space to curved space.

The Core Concepts (With Analogies)

1. The Flat World vs. The Curved World

The Flat World (Euclidean): Imagine a giant, flat chessboard. If you want to measure how a piece moves, you just use a ruler. The rules are simple and consistent everywhere.
The Curved World (Riemannian): Imagine the surface of the Earth. If you are in New York and want to compare a direction with a direction in London, you can't just lay a ruler across the ocean. The ground curves. To compare them, you have to "transport" one direction to the other, which gets tricky because the ground bends.

The authors are dealing with functions defined on these curved surfaces. They need a way to define "Quasiconvexity" that works even when the ground is curving under your feet.

2. What is Quasiconvexity? (The "Rubber Sheet" Analogy)

In the flat world, a function is convex if it looks like a bowl. If you roll a ball in it, it always settles at the bottom.
Quasiconvexity is a slightly more relaxed version. Imagine a rubber sheet stretched over a frame.

Convexity: The sheet is perfectly smooth and bowl-shaped.
Quasiconvexity: The sheet might have tiny ripples or wrinkles, but if you look at the average height of the sheet over a small area, it's never lower than the height at the very center.

In math terms: If you take a specific design (a "slope" or "gradient") and wiggle it slightly with a small, smooth perturbation (like a tiny ripple), the average energy of the wiggled design should never be lower than the energy of the original, flat design. If it were lower, the system would be unstable, and you couldn't find a true minimum.

3. The Problem: Adding Things on Curved Surfaces

Here is the tricky part the authors solved.
In flat space, if you have a slope at point A and a slope at point B, you can just add them together like numbers: $5 + 3 = 8$ .
On a curved surface, the "slope" at point A lives in a different "universe" (tangent space) than the slope at point B. You can't just add them directly. It's like trying to add "miles North" to "miles East" without a map.

The Authors' Solution:
They invented a "moral" way to add these slopes. They use the Exponential Map (a mathematical tool that projects a flat map onto a curved surface).

Imagine you are at point $x_0$ . You want to test a wiggle at a nearby point $x$ .
You take the wiggle at $x$ , and you "roll" it back to $x_0$ using the geometry of the surface.
Now that both slopes are at the same point, you can add them.
Because the surface is curved, this "rolling" process introduces a tiny error. The authors proved that as you get closer and closer to the center point, this error vanishes.

They defined a new version of Quasiconvexity that includes this "correction term" for the curvature.

The Main Result: The "Golden Rule"

The paper proves a beautiful equivalence (a two-way street):

If your function is Quasiconvex (according to their new curved-surface definition), THEN your energy formula is stable. You can find a minimum, and the math behaves nicely.
If your energy formula is stable (sequentially lower semicontinuous), THEN your function must be Quasiconvex.

Why does this matter?
Before this paper, if you were working on a curved manifold (like modeling the shape of a protein, the surface of a planet, or the shape of a black hole's event horizon), you didn't have a rigorous checklist to know if your optimization problem was solvable.

Now, you have a checklist:

Check if your function satisfies the "Riemannian Quasiconvexity" condition (the average energy of a wiggle is never lower than the center).
If yes, you are safe to proceed with finding the best shape.

The "Correction Term" (The o(1) Mystery)

In the paper, you see a lot of math with $o(1)$ . Think of this as the "Curvature Tax."

In flat space, the tax is zero.
In curved space, there is a tiny tax because the geometry is bending.
The authors show that this tax is so small that if you zoom in close enough (make the area $r$ very small), the tax disappears. This allows the flat-space rules to apply locally, even on a curved surface.

Summary for the Everyday Reader

Imagine you are trying to find the lowest point in a vast, hilly landscape (the manifold).

Old Math: Told you how to find the lowest point if the landscape was a flat plain.
This Paper: Tells you how to find the lowest point if the landscape is a rolling hill or a sphere.
The Method: They realized that to check if a spot is truly the "best," you have to look at how the ground curves around it. They created a new rule that accounts for the curve, proving that if your landscape follows this rule, you can trust that a "lowest point" actually exists and your calculations won't break.

They didn't just guess; they proved it rigorously, bridging the gap between the simple, flat world of standard calculus and the complex, curved world of modern geometry.

1. Problem Statement and Motivation

The paper addresses a fundamental problem in the Calculus of Variations: characterizing the sequential lower semicontinuity (l.s.c.) of integral functionals defined on Sobolev spaces.

Classical Context: In Euclidean space ( $\mathbb{R}^n$ ), the lower semicontinuity of functionals $F(u) = \int_\Omega f(x, u, Du) \, dx$ with respect to weak (or weak*) convergence in $W^{1,p}$ is classically characterized by the quasiconvexity of the integrand $f$ (introduced by Morrey). This result was significantly refined by Acerbi and Fusco for Carathéodory integrands in $W^{1,p}$ and $W^{1,\infty}$ .
The Gap: While the Euclidean theory is well-established, there was no rigorous, general definition of quasiconvexity for functionals defined on Riemannian manifolds $(M, g)$ , nor a proof of the equivalence between this geometric quasiconvexity and the lower semicontinuity of the associated functional.
Specific Challenge: In a Riemannian manifold, tangent spaces $T_x M$ at different points are distinct vector spaces. Consequently, one cannot simply add differentials $Du(x)$ and $D\phi(x)$ from different points as is done in Euclidean space. The authors must define a notion of "averaging" perturbations that respects the manifold's geometry.

2. Methodology and Definitions

The Geometric Setting

The authors consider a smooth, connected, and complete Riemannian manifold $(M, g)$ of dimension $n$ . They study functionals defined on maps $u: M \to \mathbb{R}^m$ .

The Bundle: They define the vector bundle $L(TM, \mathbb{R}^m)$ consisting of linear maps $\alpha: T_x M \to \mathbb{R}^m$ .
The Functional: The functional of interest is:
$F(u, \Omega) = \int_\Omega f(du) \, d\mu$
where $\Omega \subseteq M$ is a bounded open set, $\mu$ is the Riemannian volume measure, and $f: L(TM, \mathbb{R}^m) \to \mathbb{R}$ is a continuous function depending only on the differential $du$.

Definition of Riemannian Quasiconvexity

The core innovation is Definition 2.1, which generalizes Morrey's quasiconvexity to the manifold setting.
A continuous function $f$ is quasiconvex if for every point $x_0 \in M$ , every linear map $\alpha_{x_0} \in (T_{x_0}^* M)^m$ , and every smooth compactly supported perturbation $\phi \in C_c^\infty(Q_r^{x_0}, \mathbb{R}^m)$ (where $Q_r^{x_0}$ is a geodesic cube), the following inequality holds:

$f(\alpha_{x_0}) \leq \frac{1}{\mu(Q_r^{x_0})} \int_{Q_r^{x_0}} f\left( \alpha_{x_0} + d\phi[x] \circ d(\exp_{x_0})[\exp_{x_0}^{-1}(x)] \right) d\mu(x) + o(1)$

Key Technical Features of the Definition:

Parallel Transport via Exponential Map: Since $d\phi[x]$ lives in $T_x^* M \otimes \mathbb{R}^m$ and $\alpha_{x_0}$ lives in $T_{x_0}^* M \otimes \mathbb{R}^m$ , they cannot be added directly. The term $d(\exp_{x_0})[\exp_{x_0}^{-1}(x)]$ acts as a "transport" mechanism, pulling the differential of the perturbation back to the tangent space at $x_0$ .
Correction Term ( $o(1)$ ): Because the manifold is curved, the exponential map is not the identity (unlike in $\mathbb{R}^n$ ). The definition includes a remainder term $o(1)$ that vanishes as the radius $r \to 0$ . This term depends monotonically on the $L^\infty$ norm of $d\phi$ .
Consistency: The authors prove that if $(M, g) = \mathbb{R}^n$ , this definition reduces exactly to the classical Euclidean quasiconvexity (Theorem 1.1).

3. Key Results

The paper establishes a complete equivalence between the geometric definition of quasiconvexity and the variational property of lower semicontinuity.

Theorem 3.1 (Sufficiency)

If $f$ is continuous and quasiconvex in the sense of Definition 2.1, then the functional $F(u, \Omega)$ is sequentially lower semicontinuous with respect to the weak* topology of $W^{1,\infty}(\Omega, \mathbb{R}^m)$ .

Proof Strategy: The proof adapts the Fonseca-Müller argument. It involves:
- Trivializing the bundle locally using an orthonormal frame.
- Constructing a specific perturbation sequence $\phi_j = \psi(u_j - u)$ to test the functional.
- Utilizing the modulus of continuity of $f$ and the properties of the exponential map to bound the error terms.
- Using the Radon-Nikodym theorem and the properties of weak* convergence of measures to pass to the limit.

Theorem 3.2 (Necessity)

If the functional $F(u, \Omega)$ is sequentially lower semicontinuous in $W^{1,\infty}(\Omega, \mathbb{R}^m)$ for every bounded open $\Omega$ , then $f$ must be quasiconvex in the sense of Definition 2.1.

Proof Strategy: The authors construct a specific sequence of oscillating perturbations $\phi_h$ (using periodic extensions and scaling via the exponential map) that converge weak* to zero. By testing the lower semicontinuity condition on the sequence $u + \phi_h$ , they recover the integral inequality required for quasiconvexity.

Theorem 3.3 (Main Characterization)

Combining the above, the paper proves:

$f$ is quasiconvex (Def 2.1) if and only if $F(u, \Omega)$ is sequentially lower semicontinuous in $W^{1,\infty}(\Omega, \mathbb{R}^m)$ .

This result is the Riemannian analogue of the Acerbi-Fusco theorem for the case $p = +\infty$ .

4. Significance and Contributions

Generalization of Variational Calculus: The paper successfully extends the foundational theory of quasiconvexity from flat Euclidean spaces to curved Riemannian manifolds. This is crucial for applications in geometric analysis, elasticity on curved surfaces, and shape optimization on manifolds.
Resolution of the "Addition" Obstacle: The paper provides a rigorous mathematical framework for handling the non-linear geometry of manifolds in variational problems. By introducing the exponential map correction and the $o(1)$ term, it resolves the issue of comparing differentials at different points.
Foundation for Future Research:
- The authors outline a path to extend these results to $W^{1,p}$ spaces ( $1 \leq p < \infty$ ).
- They propose combining their results with the work of Focardi and Spadaro (who studied maps $u: \Omega \to M$ ) to define quasiconvexity for maps between two arbitrary Riemannian manifolds.
- They suggest investigating the relationships between quasiconvexity, polyconvexity, and rank-one convexity in the Riemannian context.

5. Conclusion

Corbisiero, Leone, and Mantegazza have established that the classical characterization of lower semicontinuity via quasiconvexity holds true in the Riemannian setting, provided the definition of quasiconvexity is appropriately modified to account for the manifold's curvature via the exponential map. This work bridges the gap between geometric analysis and the calculus of variations, providing the necessary tools to study minimization problems on curved spaces.