Correction to “An extension of the Vitali-Hahn-Saks theorem”

Onésimo Hernández–Lerma, Jean B. LasserreProceedings of the American Mathematical Society 1998 · DOI

Abstract

This mathematical note provides a critical correction to a previously published extension of the Vitali-Hahn-Saks theorem. The authors identify a missing condition in the original statement of Theorem 2.1 (1996): for a limit measure to inherit absolute continuity from a sequence of measures, the densities of those measures must remain bounded in the $L^\infty$ norm as the sequence progresses. This correction ensures the logical integrity of the theorem within the framework of measure theory.

Introduction

The Vitali-Hahn-Saks theorem is a fundamental result in measure theory, establishing conditions under which the setwise convergence of a sequence of measures implies properties such as uniform absolute continuity. In 1996, an extension of this theorem was proposed to broaden its applicability. However, subsequent analysis revealed that the main result, Theorem 2.1, was incomplete.

Without a specific constraint on the behavior of the densities, the limit measure $\lambda$ may fail to be absolutely continuous with respect to a reference measure $\mu$, even if every measure in the approximating sequence $\{\lambda_n\}$ is absolutely continuous. This paper formalizes the necessary refinement to bridge this logical gap.

The Problem: Preserving Absolute Continuity

In the context of the original extension, the authors considered a sequence of measures $\{\lambda_n\}$ converging to a limit measure $\lambda$. The goal was to prove that if $\lambda_n \ll \mu$ for all $n$, then $\lambda \ll \mu$.

The issue arises from the nature of the convergence used—specifically, integration against continuous functions vanishing at infinity ($C_0(X)$). In this topology, absolute continuity is not automatically preserved. If the densities $u_n$ of $\lambda_n$ (where $d\lambda_n = u_n d\mu$) are allowed to grow without bound on increasingly small sets, the limit measure $\lambda$ could potentially manifest "singular" behavior, concentrating mass where the reference measure $\mu$ is zero.

The Corrected Theorem

The authors provide a refined version of Theorem 2.1. The corrected statement is as follows:

Let $\{\lambda_n\}$ and $\lambda$ be $\sigma$-finite measures such that $\lambda_n \to \lambda$ in the sense of integration against functions in $C_0(X)$. Suppose each $\lambda_n$ is absolutely continuous with respect to a reference measure $\mu$, with Radon-Nikodym derivatives (densities) $u_n$. If the sequence of densities satisfies the condition: $$\liminf_{n \to \infty} \|u_n\|_\infty < \infty$$ then the limit measure $\lambda$ is also absolutely continuous with respect to $\mu$ ($\lambda \ll \mu$).

Methodology and Proof Sketch

The proof of the corrected theorem relies on the regularity of measures and the control provided by the $L^\infty$ norm. The logical steps are summarized below:

  1. Density Representation: Each measure in the sequence is represented via its density $u_n$, such that for any measurable set $B$, $\lambda_n(B) = \int_B u_n d\mu$.
  2. Set Approximation: To prove $\lambda(B) = 0$ for a $\mu$-null set $B$, the authors use the regularity of measures to enclose $B$ within an open set $G_\epsilon$ such that $\mu(G_\epsilon)$ is arbitrarily small.
  3. Applying the $L^\infty$ Bound: By introducing the condition that the limit inferior of $\|u_n\|_\infty$ is finite, the measure of the open set under $\lambda_n$ can be bounded: $\lambda_n(G_\epsilon) \leq \mu(G_\epsilon) \|u_n\|_\infty$.
  4. Limit Argument: By taking the limit as $n \to \infty$ and then letting the size of the open set $G_\epsilon$ shrink to zero ($\epsilon \downarrow 0$), the authors demonstrate that $\lambda(B)$ must be zero, thereby satisfying the definition of absolute continuity.

Discussion and Insights

The primary insight of this correction is the necessity of "well-behaved" densities in preserving measure properties under weak convergence. The $L^\infty$ bound acts as a safeguard against the concentration of mass that would otherwise lead to singularity in the limit.

While the scope of this note is narrow—focusing specifically on non-negative measures and convergence against $C_0(X)$—it clarifies a vital boundary in measure-theoretic convergence. Future research might investigate whether these conditions can be relaxed, such as by exploring bounds in $L^p$ spaces for $p < \infty$ under different topological constraints.

Conclusion

This correction restores the validity of the 1996 extension of the Vitali-Hahn-Saks theorem. By requiring that the densities of a sequence of measures remain bounded in the $L^\infty$ norm, the authors ensure that absolute continuity is preserved in the limit, providing a robust foundation for further applications of the theorem in functional analysis and probability theory.

Summary updated

BibTeX
@inproceedings{hernandezlerma1998correction,
  author = {Onésimo Hernández–Lerma and Jean B. Lasserre},
  title = {{Correction to “An extension of the Vitali-Hahn-Saks theorem”}},
  booktitle = {Proceedings of the American Mathematical Society},
  year = {1998},
  doi = {10.1090/s0002-9939-98-04497-9},
}
Figure