Abstract
We construct a variety of mappings of the unit interval into $\mathcal{L}^p([0,1])$ to generalize classical examples of $\mathcal{L}^p$-convergence of sequences of functions with simultaneous pointwise divergence. By establishing relations between the regularity of the functions in the image of the mappings and the topology of $[0,1]$, we obtain examples which are $\mathcal{L}^p$-continuous but exhibit discontinuity in a pointwise sense to different degrees. We conclude by proving an Egorov-type theorem, namely that if almost every function in the image is continuous, then we can remove a set of arbitrarily small measure from the index set $[0,1]$ and establish pointwise limits for all functions in the remaining image.
Introduction
In classical real analysis, the "Typewriter Sequence" serves as a standard counterexample to the notion that convergence in the $\mathcal{L}^p$ norm implies pointwise convergence. This sequence of functions converges to zero in the mean—the area under the curve vanishes—yet fails to converge at any single point because a "bump" of constant height continuously traverses the domain.
This paper extends this inquiry from discrete sequences to continuous families. We investigate the behavior of a "curve" of functions $f_t$ indexed by the unit interval $t \in [0, 1]$. Specifically, we explore how a family that moves continuously within an $\mathcal{L}^p$ space can still exhibit pathological pointwise divergence, and we identify the specific thresholds of spatial regularity that prevent such behavior.
Constructing Pathological Continuities
The core of the investigation lies in the construction of mappings that are continuous in the $\mathcal{L}^p$ topology but discontinuous pointwise. Using constructive methods, the authors define functions as infinite sums of carefully scaled and shifted "bumps"—utilizing mollifiers and characteristic functions. These bumps are designed to oscillate with sufficient frequency to cause pointwise divergence across the domain while their magnitudes are constrained to ensure the $\mathcal{L}^p$ norm remains small.
Theorem 2.1 demonstrates that it is possible to construct an $\mathcal{L}^p$-continuous curve where the constituent functions are smooth ($C^\infty$) almost everywhere, yet the family fails to converge pointwise almost everywhere. This highlights that even high levels of individual function smoothness do not necessarily guarantee well-behaved pointwise limits for the family as a whole.
In more extreme cases, Theorem 4.1 shows that by relaxing the requirement for individual function continuity, one can obtain divergence on every subset of the index set with positive measure. This suggests that the "Typewriter" phenomenon is not merely a discrete artifact but a robust feature of continuous mappings in functional spaces.
The Regularity Threshold: $W^{1,1}$ vs. $W^{1,q}$
A significant contribution of this work is the identification of a sharp boundary in spatial regularity, governed by Sobolev embeddings. The authors examine the space $W^{1,q}(\Omega)$, which characterizes functions with derivatives in $\mathcal{L}^q$.
They discover a striking transition:
- The $W^{1,1}$ Case: A curve of functions can consist of functions that are all absolutely continuous (belonging to $W^{1,1}$) and still exhibit pointwise divergence.
- The $W^{1,q}$ Case ($q > 1$): If the functions in the family belong to $W^{1,q}$ for any $q > 1$ and are bounded in that norm, the $\mathcal{L}^p$-continuity of the mapping forces the curve to be pointwise continuous on a dense set.
This result identifies $W^{1,1}$ as the limit of pathological behavior; once the regularity exceeds this threshold, the topology of the index set and the spatial regularity of the functions begin to synchronize, constraining the possible pointwise divergence.
Topological Constraints and Egorov-Type Results
The paper also approaches the problem through the lens of Baire category and measure theory. The authors demonstrate that if every function in the family is continuous, pointwise divergence cannot occur everywhere; it is restricted to a "meager" set (a set of the first category).
Furthermore, the authors prove a "slice-wise" version of Lusin’s Theorem. While the standard Lusin’s Theorem allows a measurable function to be made continuous by removing a small set from its domain, Theorem 5.1 establishes that if the spatial functions $f_t$ are continuous for almost all $t$, one can remove a small set of indices $T_\epsilon$ such that the resulting two-variable function $f(t,x)$ is continuous on the remaining "slice" of the space-time domain. This Egorov-type result provides a bridge between the measure of the index set and the simultaneous continuity of the family.
Conclusion
This research clarifies the relationship between global convergence (averaging) and local convergence (pointwise) in continuous families of functions. While smooth, $\mathcal{L}^p$-continuous curves can exhibit surprising pointwise instability, a specific threshold of Sobolev regularity ($W^{1,q}, q > 1$) acts as a restorative force for pointwise continuity. These findings have implications for the study of evolution equations and semigroups in mathematical physics, where the regularity of operators often determines whether such pathological "Typewriter" behaviors can manifest in physical systems.
Summary updated