A note on Alberti’s Luzin-type theorem for gradients

Siran Li

TL;DR

This paper presents a new, "soft" proof of Alberti's Luzin-type theorem for gradients, which states that any Borel vectorfield can be approximated by a gradient of a smooth function on almost all of the domain. The authors leverage geometric measure theory and topology, offering an alternative to Alberti's original constructive proof and opening avenues for applications in rectifiability problems.

What Problem Does This Paper Address?

The core problem this paper addresses is the approximation of arbitrary vectorfields by gradient fields. Specifically, it focuses on Alberti's Luzin-type theorem, which asserts that any "nice" vectorfield (a Borel vectorfield on an open set with finite Lebesgue measure) can be made arbitrarily close to a gradient of a smooth scalar potential, except on a set of small measure. This is a fundamental result with implications for understanding differential forms and their properties, particularly their "nearness" to being exact. While Alberti provided a constructive proof, this paper offers a new perspective using geometric and topological tools, aiming for a more elegant and potentially more generalizable approach.

Methods

The paper's primary contribution is a novel, "soft" proof of Alberti's theorem. Instead of Alberti's constructive, iterative approach involving dyadic cubes and smoothing, this proof relies on:

  • Geometric Measure Theory: Concepts like N-dimensional Hausdorff measure ($H^N$) are central to quantifying the "smallness" of the exceptional sets. The proof utilizes Federer's structure theorem for rectifiable sets.
  • Topology: The proof frames the problem in terms of the graph of the vectorfield. It then uses topological arguments, particularly related to simplicial approximation and the geometry of graphs, to find approximations.
  • Reduction Steps: The proof is structured through a series of reductions:
  • Continuous, Bounded Vectorfields: Reducing the problem from general Borel vectorfields to continuous, bounded ones.
  • Smooth, Compactly Supported Vectorfields: Further reducing to smooth vectorfields with compact support using mollification.
  • Vectorfields with Rectifiable Graphs: Showing that the graph of the vectorfield can be treated as a rectifiable set, which is crucial for applying geometric measure theory tools.
  • Vectorfields Mapping Between PL-Manifolds: Representing the vectorfield as a map between piecewise linear (PL) manifolds, allowing for simplicial approximation.
  • Simplicial Approximation: Employing the simplicial approximation theorem to find a smooth function whose gradient approximates the vectorfield.

The paper also introduces a general framework (Section 3) for deducing Alberti's theorem from a "rough" approximation result, highlighting the importance of extension properties for Banach spaces of functions.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • "Soft" Proof: The most significant insight is the development of a proof that avoids hard analysis and relies instead on geometric and topological arguments. This suggests that the underlying structure of the problem is more geometric than analytical.
  • Graph-Centric Approach: Shifting focus from the vectorfield itself to its graph provides a powerful new perspective. Approximating the graph topologically leads to the desired approximation of the vectorfield.
  • Connection to Rectifiability: The proof deeply connects Alberti's theorem to the concept of rectifiability. The ability to decompose the graph into rectifiable and unrectifiable parts, and then show the unrectifiable part has zero measure, is key.
  • Generalizability: The framework developed in Section 3, which deduces Alberti's theorem from a more general approximation result, hints at potential extensions to other settings, though the paper notes challenges with higher regularity.

Key Results

The paper successfully provides a new proof of Alberti's Luzin-type theorem for gradients (Theorem 1.1). This theorem states that for any Borel vectorfield $v$ on an open set $\Omega \subset \mathbb{R}^N$ with finite Lebesgue measure, and for any $\epsilon > 0$, there exists an open set $A \subset \Omega$ with $H^N(A) \leq \epsilon H^N(\Omega)$ and a function $\phi \in C^1(\Omega)$ such that $v = \nabla\phi$ on $\Omega \setminus A$. The paper also demonstrates how this theorem can be deduced from a more general approximation theorem (Theorem 3.1) within a functional analysis framework. Furthermore, it touches upon applications to the $C^2$-rectifiability problem, suggesting that Alberti's construction can yield rectifiable currents that are not $C^2$-rectifiable.

Limitations and Future Work

  • "Softness" vs. Quantitative Estimates: While the proof is described as "soft" and avoids hard analysis, the authors mention that recovering the quantitative $L^p$-estimates from Alberti's original work would require additional effort.
  • Higher Regularity: The paper notes that generalizing Alberti's theorem to function spaces with higher regularity (e.g., $C^k$ or $C^{k,\alpha}$) is challenging due to the difficulty in satisfying extension properties analogous to Tietze's theorem for continuous functions. This is identified as a significant open problem.
  • Applications to Rectifiability: The paper hints at further applications in geometric measure theory, particularly concerning $C^2$-rectifiability of currents, and suggests that the techniques could be used to construct "natural" rectifiable currents of arbitrary degree that are non-$C^2$-rectifiable.

Summary updated

Abstract

We give a "soft" proof of Alberti's Luzin-type theorem in [1] (G. Alberti, A Lusintype theorem for gradients, J. Funct. Anal. 100 (1991)), using elementary geometric measure theory and topology. Applications to the $C^2$-rectifiability problem are also discussed.
BibTeX
@inproceedings{li2020note,
  author = {Siran Li},
  title = {{A note on Alberti’s Luzin-type theorem for gradients}},
  booktitle = {Ricerche},
  year = {2020},
  doi = {10.1007/s11587-020-00485-w},
}
Figure