Kolmogorov$\unicode{x2013}$Riesz compactness in asymptotic $L_p$ spaces

This paper extends the classical Kolmogorov-Riesz compactness theorem to asymptotic $L_p$ spaces on $\mathbb{R}^n$ by proving that relative compactness in these nonlocally convex F-spaces is characterized by natural tail and translation conditions alongside a necessary additional almost equiboundedness requirement.

The Gist

Imagine you are a city planner trying to organize a massive, chaotic festival. You have a huge list of performers (functions) who are all over the place. Your goal is to figure out which groups of performers are "manageable" or "compact." In math terms, "compact" means you can fit them all into a small, neat box without losing any of them. If a group is compact, you can predict their behavior, and they won't run off into the distance or explode in size.

For a long time, mathematicians had a perfect rulebook (the Kolmogorov–Riesz Theorem) for organizing these performers, but it only worked in a very specific, well-behaved neighborhood called $L^p$ space. In this neighborhood, everyone follows strict rules: no one is too loud, and everyone stays within a certain distance.

However, the real world is messier. There are "asymptotic" neighborhoods (called $\Lambda^p$ spaces) where the rules are looser. Here, performers can be incredibly loud or wild, as long as they calm down eventually or only cause trouble in tiny, isolated spots. The old rulebook didn't work here because the math tools used to measure "size" (called norms) behave differently in this chaotic zone.

This paper by Nuno J. Alves is like writing a new, updated rulebook specifically for this messy, chaotic neighborhood.

The Three Golden Rules for a Manageable Group

To prove that a group of performers (a set of functions) is "compact" (manageable) in this new, messy neighborhood, you need to check three specific conditions. If you miss even one, the group will fall apart.

1. The "Stay Close to Home" Rule (Tail Condition)

The Analogy: Imagine the festival grounds are an infinite field. You want to make sure that no performer is wandering off into the far, dark woods where you can't see them.
The Math: For any group, you must be able to draw a circle around the center of the city. Outside this circle, the "energy" of the performers must be so low that it's practically zero.

• Why it matters: If your group has members who keep running off to infinity, you can't pack them into a small box. They are too spread out.

2. The "Smooth Moves" Rule (Translation Condition)

The Analogy: Imagine the performers are dancing. If you nudge the entire dance floor slightly to the left, the dancers should move smoothly with it. They shouldn't suddenly jump, teleport, or change their routine entirely just because you shifted the stage a tiny bit.
The Math: If you shift the function slightly (move it left or right), the difference between the original and the new version should be tiny.

• Why it matters: If a group is "jumpy" or erratic, a tiny shift makes them look completely different. A compact group must be stable and predictable.

3. The "Don't Go Wild" Rule (Almost Equiboundedness)

The Analogy: This is the new rule that the author discovered is necessary for the messy neighborhood. Imagine some performers are allowed to scream very loudly, but only if they do it in a tiny, soundproof booth.

• The Old Rulebook: In the well-behaved neighborhood, if everyone stays close to home and moves smoothly, they naturally can't scream too loud.
• The New Reality: In the messy neighborhood, you can have a group that stays close to home and moves smoothly, but one performer decides to scream at a volume of 1,000,000 decibels for a split second. In the old math, this would break the system.
• The Fix: You must ensure that for any group, there is a "volume limit" (say, 100 decibels) such that almost everyone stays below that limit. If someone does scream louder, they can only do it in a tiny, negligible area (like a whisper in a whispering gallery).
• Why it matters: Without this rule, a group could be "compact" in the old sense but still contain a "monster" function that breaks the math because it gets infinitely large in a tiny spot.

Why This Matters

The author shows that in these complex, non-standard mathematical spaces, you can't just rely on the old two rules. You must add this third rule about controlling the "wild spikes" (the almost equiboundedness).

• Without Rule 3: You might think a group is safe, but it contains a "monster" that ruins the compactness.
• With All Three: You have a guarantee that the group is truly manageable, predictable, and fits neatly into your mathematical "box."

The Takeaway

Think of this paper as upgrading a security system. The old system checked if people were in the building (Rule 1) and if they were walking calmly (Rule 2). But in a new, chaotic building, you realized you also needed to check if anyone was holding a firecracker (Rule 3). Even if they are in the building and walking calmly, a firecracker can still blow the whole place up.

Nuno J. Alves has given us the complete checklist to ensure that even in the most chaotic mathematical neighborhoods, we can still find order and predictability.

Technical Summary

Here is a detailed technical summary of the paper "Kolmogorov–Riesz Compactness in Asymptotic $L^p$ Spaces" by Nuno J. Alves.

1. Problem Statement

The paper addresses the problem of characterizing relative compactness (or total boundedness) for subsets of function spaces on unbounded domains, specifically extending the classical Kolmogorov–Riesz compactness theorem to the setting of asymptotic $L^p$ spaces, denoted as $\Lambda^p(\mathbb{R}^n)$.

• Context: The classical theorem provides necessary and sufficient conditions for a family of functions in $L^p(\mathbb{R}^n)$ ($1 \le p < \infty$) to be totally bounded. These conditions typically involve:
  1. Tail decay: Uniform smallness of the integral outside large balls.
  2. Translation continuity: Uniform smallness of the $L^p$ norm of the difference between a function and its translation for small shifts.
• The Gap: The space $\Lambda^p(\mathbb{R}^n)$ is a nonlocally convex F-space (a complete metric vector space with a translation-invariant metric) that contains $L^p(\mathbb{R}^n)$ as a dense subspace. Its topology is generated by a non-homogeneous F-norm defined as $\|f\|_{\alpha p} = \| \min(|f|, 1) \|_p$.
• Challenge: The lack of homogeneity in the F-norm prevents a direct adaptation of the classical proof. The author investigates whether the classical two conditions (tail and translation) are sufficient or if additional constraints are required to characterize compactness in this non-locally convex setting.

2. Methodology

The author employs a combination of functional analysis techniques tailored to F-spaces and metric space compactness arguments:

• Space Definition: The paper utilizes the definition of $\Lambda^p(\mathbb{R}^n)$ as the space of measurable functions that are "almost" in $L^p$ (i.e., they belong to $L^p$ outside sets of arbitrarily small measure). The topology is induced by the F-norm $\|f\|_{\alpha p}$.
• Approximation by Dense Subspaces: Since $L^p(\mathbb{R}^n)$ is dense in $\Lambda^p(\mathbb{R}^n)$, the proofs rely heavily on approximating elements in a totally bounded family $F \subset \Lambda^p(\mathbb{R}^n)$ by a finite set of functions in $L^p(\mathbb{R}^n)$ (and subsequently by smooth functions $C_c^\infty$).
• Truncation Techniques: In the sufficiency proof, the author introduces a truncation operator $T_M$ to map functions in $\Lambda^p$ to bounded functions in $L^p \cap L^\infty$. This allows the application of the classical Kolmogorov–Riesz theorem (Theorem 1.1) to the truncated family.
• Contradiction Arguments: The necessity of the new condition (almost equiboundedness) is proven via contradiction, analyzing the measure of sets where functions exceed a certain threshold.

3. Key Contributions

The primary contribution is the formulation and proof of Theorem 3.1, a new compactness criterion for $\Lambda^p(\mathbb{R}^n)$.

The Main Result (Theorem 3.1):
A subset $F \subseteq \Lambda^p(\mathbb{R}^n)$ is totally bounded if and only if the following three conditions hold:

1. Tail Condition: For every $\epsilon > 0$, there exists $R > 0$ such that $\int_{|x|>R} \min(|f|, 1)^p dx < \epsilon^p$ for all $f \in F$.
2. Translation Condition: For every $\epsilon > 0$, there exists $r > 0$ such that $\int_{\mathbb{R}^n} \min(|\tau_y f - f|, 1)^p dx < \epsilon^p$ for all $|y| < r$ and $f \in F$.
3. Almost Equiboundedness (New Condition): For every $\epsilon > 0$, there exists $M > 0$ such that $|\{x : |f(x)| > M\}| < \epsilon$ for all $f \in F$.

Significance of the New Condition:

• In classical $L^p$ spaces, boundedness is often redundant because it follows from the tail and translation conditions.
• In $\Lambda^p(\mathbb{R}^n)$, due to the non-homogeneous F-norm, Condition (iii) is independent and necessary. It controls the measure of the "tails" of the functions in the vertical direction (large values), ensuring that functions do not spike to infinity on sets of non-negligible measure.
• The paper proves that Condition (iii) is equivalent to almost equiboundedness: for any $\epsilon$, functions in the family can be bounded by a constant $M$ outside a set of measure less than $\epsilon$.

4. Results and Proofs

The paper is structured to rigorously establish the necessity and sufficiency of the three conditions:

• Necessity (Section 4):
  • Lemma 4.1 & 4.2: Prove that total boundedness implies the tail and translation conditions, respectively, using the finite $\epsilon$-net property and the density of $L^p$ and $C_c^\infty$.
  • Lemma 4.3: Proves that total boundedness implies almost equiboundedness. The proof assumes the contrary (that functions can be arbitrarily large on sets of fixed measure) and derives a contradiction by showing that such a family cannot be covered by a finite number of balls in the $\alpha p$-metric.
• Sufficiency (Section 5):
  • The author constructs a truncated family $F_M = \{f_M : f \in F\}$ where $f_M$ is $f$ truncated at level $M$.
  • Using Condition (iii), the author shows that $F$ is arbitrarily close to $F_M$ in the $\Lambda^p$ metric.
  • Using Conditions (i) and (ii), the author demonstrates that $F_M$ satisfies the classical Kolmogorov–Riesz conditions in $L^p(\mathbb{R}^n)$, making $F_M$ totally bounded in $L^p$.
  • Since $F_M$ is totally bounded in $L^p$ and close to $F$ in $\Lambda^p$, $F$ is totally bounded in $\Lambda^p$.
• Sharpness (Section 6):
  • Five examples are provided to demonstrate that none of the three conditions can be deduced from the others:
    • Ex 6.1: Satisfies (i) and (ii) but fails (iii) (unbounded spikes).
    • Ex 6.2: Satisfies (ii) and (iii) but fails (i) (mass escaping to infinity).
    • Ex 6.3: Satisfies (i) and (iii) but fails (ii) (rapid oscillation/Rademacher functions).
    • Ex 6.4 & 6.5: Illustrate sequences that are totally bounded in $\Lambda^p$ but not in $L^p$, highlighting the distinct nature of the asymptotic space.

5. Significance

• First of its Kind: This is the first Kolmogorov–Riesz type compactness criterion established for an unbounded domain within a nonlocally convex F-space generated by a nonhomogeneous F-norm.
• Theoretical Advancement: It clarifies the structural differences between standard $L^p$ spaces and asymptotic $L^p$ spaces. Specifically, it highlights that in non-locally convex settings, control over the "size" of functions (equiboundedness) is not automatically implied by spatial decay and continuity.
• Applications: The result is relevant for the study of partial differential equations (PDEs) and functional analysis in spaces where standard $L^p$ integrability fails but convergence in measure (or asymptotic $L^p$ convergence) is the natural topology. It provides the necessary tools to establish existence results for PDEs in these broader function spaces.
• Broader Context: The work fits into a growing body of literature extending compactness criteria to variable exponent spaces, grand Lebesgue spaces, and other generalized function spaces, but it uniquely addresses the non-locally convex, non-homogeneous case on unbounded domains.