Equi-integrable approximation of Sobolev mappings between manifolds

This paper establishes that limits of sequences of smooth maps between compact Riemannian manifolds with equi-integrable $W^{1, p}$-Sobolev energy can always be strongly approximated by smooth maps, thereby extending Hang's density result to integer $p \ge 2$ and providing proofs for higher-order and fractional Sobolev spaces as well as cases governed by the Bethuel-Demengel-Colon-Hélein cohomological criterion.

The Gist

Imagine you are trying to paint a picture of a complex, curved surface (like the surface of a sphere or a twisted torus) using only smooth, perfect brushstrokes. In the world of mathematics, this is what we call approximating a "rough" map with "smooth" maps.

The paper you provided, written by Jean Van Schaftingen, tackles a specific problem in this artistic endeavor: Can we always fix a "rough" painting by smoothing it out, provided the "energy" used to create the roughness isn't too wild?

Here is a breakdown of the paper's core ideas using everyday analogies.

1. The Setting: The Canvas and the Target

• The Domain ($M$): Think of this as your canvas. It's a compact shape (like a sphere or a donut).
• The Target ($N$): This is the shape you are trying to paint onto. It's also a curved surface.
• The Map ($u$): This is the painting itself. It tells you, for every point on your canvas, where on the target surface you should paint.
• The Problem: Sometimes, the painting is "rough" (mathematically, it's in a Sobolev space). It has jagged edges or sudden jumps, but it's still a valid function. The question is: Can we replace this rough painting with a sequence of perfectly smooth paintings that look and feel exactly the same?

2. The Three Ways to "Look the Same"

The paper compares three different ways to say a sequence of smooth paintings ($u_n$) is getting closer to the rough target ($u$):

1. Strong Approximation (The "Perfect Match"):

   • Analogy: You are replacing the rough painting with smooth ones. Not only does the final image look identical, but the brushstrokes (the derivatives/slopes) also match perfectly everywhere.
   • Math: The total difference in both the image and the slopes goes to zero.
   • The Catch: Sometimes, due to the shape of the target (like a donut with a hole), you simply cannot do this. The topology (the holes) creates a "knot" that smooth paint can't untie.

2. Bounded Approximation (The "Safe" Match):

   • Analogy: You replace the rough painting with smooth ones. The final image looks identical, and the total amount of paint used (the total energy) stays within a safe budget. However, the brushstrokes might get incredibly wild in tiny spots, as long as the average energy doesn't explode.
   • Math: The image converges, and the total energy is bounded, but the energy isn't necessarily "well-behaved" everywhere.

3. Equi-integrable Approximation (The "Well-Behaved" Match):

   • Analogy: This is the paper's hero. You replace the rough painting with smooth ones. The image looks identical, and the energy is not just bounded, but well-behaved.
   • The Metaphor: Imagine the energy of the painting is like water in a bucket.
     • Bounded: The bucket doesn't overflow (total water is finite).
     • Equi-integrable: The water doesn't suddenly shoot up in a tiny, dangerous spike in one corner. The "spikes" of energy are controlled and spread out reasonably. If you look at the highest energy spots, they don't get infinitely high as you add more paintings.

3. The Big Discovery: "Well-Behaved" = "Perfect"

For a long time, mathematicians knew that if you have a "Well-Behaved" (equi-integrable) sequence of smooth maps, you might be able to get a "Perfect Match" (strong approximation), but it wasn't always guaranteed, especially when the math gets tricky (like when the dimension of the canvas matches the power of the energy).

The Paper's Main Result (Theorem 1.1):
Van Schaftingen proves a beautiful equivalence:

If you can approximate a rough map with smooth maps that have "well-behaved" energy (equi-integrable), then you can always approximate it with smooth maps that are a "perfect match" (strongly).

In plain English:
If the "roughness" of your painting isn't hiding any dangerous, exploding spikes of energy, then that roughness is just an illusion. You can always smooth it out completely without losing the shape or the topology. The "Well-Behaved" condition is the golden ticket to the "Perfect Match."

4. Why Was This Hard? (The Topological Knots)

The paper mentions that sometimes you can't smooth things out.

• Analogy: Imagine trying to untie a knot in a rope (the map) without cutting it. If the rope is tied around a pole (a hole in the target manifold), you can't just smooth it out; the knot is real.
• The paper confirms that the only reason you can't smooth a map is if there is a genuine topological knot (a hole you can't pass through). If there are no knots, and the energy is "well-behaved," the map can be smoothed perfectly.

5. The "Magic" of the Proof

How did the author prove this?

• The Skeleton Trick: He looked at the "skeleton" of the shapes (like the edges of a cube or the frame of a house). He showed that if the energy is well-behaved, the smooth maps and the rough map are "homotopic" (can be morphed into each other) on these skeletons.
• The Truncation: He used a technique called "truncation," which is like cutting off the top of the energy spikes. He proved that if you cut off the spikes, the remaining shape is still close enough to the original to be smoothed out.
• The Jacobian: In the final section, he connects this to "Jacobians" (which measure how much a map stretches or twists space). He shows that if the energy is well-behaved, these twists behave nicely, confirming the map can be smoothed.

Summary

Think of the paper as a guarantee for a sculptor:

"If you have a rough, jagged sculpture made of clay, and you promise that the 'stress' in the clay isn't concentrated in dangerous, exploding points (equi-integrability), then I promise you can melt that clay down and re-sculpt it into a perfectly smooth version that is indistinguishable from the original. The only time you can't do this is if the sculpture is tied in a knot that physically cannot be untied."

This result unifies different areas of math (integer powers, fractional powers, higher dimensions) under one simple rule: Controlled energy implies perfect smoothability.

Technical Summary

Here is a detailed technical summary of the paper "Equi-integrable Approximation of Sobolev Mappings Between Manifolds" by Jean Van Schaftingen.

1. Problem Statement and Context

The paper addresses the strong approximation problem for Sobolev mappings between compact Riemannian manifolds $M$ and $N$. Specifically, it investigates the relationship between the space of smooth maps $C^\infty(M, N)$ and the Sobolev space $W^{1,p}(M, N)$ (and its higher-order/fractional counterparts).

The Core Question:
Is every map $u \in W^{1,p}(M, N)$ the strong limit of a sequence of smooth maps $u_n \in C^\infty(M, N)$?

• Strong Approximation ($H^{1,p}_{St}$): $u_n \to u$ strongly in $W^{1,p}$ (i.e., $\int_M |u_n - u|^p + |Du_n - Du|^p \to 0$).
• Bounded Approximation ($H^{1,p}_{Bd}$): $u_n \to u$ strongly in $L^p$ with $\sup_n \int_M |Du_n|^p < \infty$.
• Weak Approximation ($H^{1,p}_{Wk}$): $u_n \rightharpoonup u$ weakly in $W^{1,p}$.

The Gap:
It is well-known that strong approximation fails when $1 \le p < \dim M$due to topological obstructions (homotopy groups of$N$) and local analytical obstructions. However, a significant contrast existed in the literature regarding the case$p=1$:

• For $p > 1$, bounded and weak sequential closures coincide ($H^{1,p}_{Bd} = H^{1,p}_{Wk}$).
• For $p = 1$, Hang proved that the weak closure of smooth maps equals the strong closure ($H^{1,1}_{Wk} = H^{1,1}_{St}$), even though the bounded closure is strictly larger ($H^{1,1}_{Bd} \neq H^{1,1}_{St}$). This relied on the fact that weak convergence in $W^{1,1}$ is equivalent to $L^1$ convergence plus equi-integrability of the gradients.

The Goal:
Van Schaftingen aims to unify these results by proving that strong approximation is always equivalent to equi-integrable approximation, regardless of the value of $p$. This resolves the apparent contradiction between the $p=1$ case (Hang's result) and the $p>1$ case.

2. Methodology

The author employs a multi-faceted approach combining geometric analysis, homotopy theory, and functional analysis:

A. Equi-integrable Sequential Closure

The paper defines the equi-integrable closure $H^{1,p}_{Ei}(M, N)$ as the set of maps $u$ that are limits of smooth maps $u_n$ such that:

1. $u_n \to u$ in $L^p$.
2. The sequence of gradients $(|Du_n|)_n$ is equi-integrable (uniformly integrable):
   $$\lim_{t \to \infty} \sup_{n} \int_M (|Du_n| - t)_+^p = 0$$
   The main theorem asserts $H^{1,p}_{St} = H^{1,p}_{Ei}$.

B. Homotopy on Skeletons (The "Screening" Technique)

The core technical machinery involves analyzing the behavior of maps on $p$-dimensional skeletons of the domain manifold.

• Vanishing Mean Oscillation (VMO): The author utilizes the space $VMO(M, N)$ and results by Brezis, Nirenberg, Schoen, and Uhlenbeck. Maps in $VMO$ are homotopic to continuous maps if their mean oscillation is small.
• Truncation and Fubini Arguments: By using Fubini's theorem and truncation techniques, the author shows that if a sequence is equi-integrable, the "bad" sets where homotopy fails (obstructions) have measure zero on generic $p$-dimensional skeletons.
• Generic Screening: The proof relies on the fact that for equi-integrable sequences, the restriction to a generic $p$-dimensional subcomplex (skeleton) preserves homotopy classes in the limit. This allows the construction of a global smooth approximation from local ones.

C. Higher-Order and Fractional Spaces

• Higher-Order ($W^{k,p}$): The author extends the result to $W^{k,p}$ spaces. For $p > 1$, this is achieved by relating $W^{k,p}$-equi-integrability to $W^{1,kp}$-equi-integrability via refined Gagliardo-Nirenberg interpolation inequalities and maximal function estimates.
• Fractional ($W^{s,p}$): For fractional spaces, the result is shown to be a consequence of the fact that pointwise convergent equi-integrable sequences converge strongly in these spaces.

D. Cohomological Criteria

In Section 5, the paper connects the approximation problem to the distributional Jacobian and de Rham cohomology. It demonstrates that the equi-integrable continuity of pullbacks of differential forms allows one to deduce the main theorem from existing cohomological criteria (Bethuel, Coron, Demengel, Hélein).

3. Key Contributions and Results

Main Theorem (Theorem 1.1)

For compact Riemannian manifolds $M, N$ and $p \in [1, \infty)$:
$$H^{1,p}_{St}(M, N) = H^{1,p}_{Ei}(M, N)$$
Significance: This unifies the theory. It implies that the "weak density" result of Hang for $p=1$ is not an anomaly but a specific instance of a general phenomenon: equi-integrability is the precise condition required to upgrade weak/bounded convergence to strong convergence for manifold-valued maps.

Extension to Higher Orders (Theorem 3.1)

The equivalence holds for higher-order Sobolev spaces $W^{k,p}(M, N)$:
$$H^{k,p}_{St}(M, N) = H^{k,p}_{Ei}(M, N)$$
The proof for $p>1$ uses a novel pointwise interpolation inequality (Proposition 3.3) to control lower-order derivatives via higher-order derivatives and $L^\infty$ bounds.

Cohomological Interpretation (Theorem 5.2)

The paper proves that if $u \in H^{1,p}_{Ei}(M, N)$, then the pullback of any closed differential form $\omega$ satisfies the distributional condition:
$$\int_M u^*\omega \wedge d\phi = 0$$
This generalizes Hang's argument for the circle ($N=S^1$) to general manifolds and higher $p$, showing that equi-integrable limits preserve the topological constraints encoded in cohomology.

4. Significance and Impact

1. Resolution of a Theoretical Discrepancy: The paper clarifies why the $p=1$ case (Hang) and $p>1$ cases behave differently regarding weak vs. bounded closures. It establishes that equi-integrability is the unifying concept.
2. Methodological Advancement: The introduction of "generic screening" techniques combined with VMO homotopy theory provides a robust framework for handling topological obstructions in Sobolev spaces without relying solely on specific topological properties of the target manifold.
3. Generalization: The results are not limited to first-order spaces. The extension to higher-order ($k \ge 2$) and fractional Sobolev spaces broadens the applicability of these approximation theorems to problems in elasticity, fluid dynamics, and geometric analysis where higher derivatives are relevant.
4. Connection to Topology: By linking equi-integrable approximation to the continuity of Jacobians and cohomological criteria, the paper strengthens the bridge between nonlinear functional analysis and algebraic topology.

In summary, Van Schaftingen's work provides a definitive answer to the approximation problem: A Sobolev map between manifolds can be strongly approximated by smooth maps if and only if it can be approximated by smooth maps with equi-integrable gradients. This result subsumes previous partial results and offers a comprehensive framework for future research in geometric analysis.