The John-Nirenberg space: Equality of the vanishing subspaces $VJN_p$ and $CJN_p$

This paper establishes the equality of the vanishing subspaces $VJN_p$ and $CJN_p$ within the John-Nirenberg spaces by demonstrating the decay of specific Morrey-type integrals, while also proving that $JN_{p,p}(\mathbb{R}^n)$ coincides with $L^p(\mathbb{R}^n)$ modulo constants.

The Gist

Imagine you are a city planner trying to understand the "vibe" of a city. In mathematics, there's a famous tool called BMO (Bounded Mean Oscillation) that measures how much a neighborhood's "noise level" (or value) fluctuates compared to its average. If the noise is always within a reasonable range, the neighborhood is "BMO."

In 1961, mathematicians John and Nirenberg invented a slightly more flexible version of this tool, called JNp. Think of JNp as a "super-neighborhood" where the noise can be a bit wilder, as long as it doesn't get too crazy too often. It sits somewhere between a perfectly quiet suburb (Lp) and a chaotic, barely-tolerable zone (Weak Lp).

The Two "Vanishing" Neighborhoods

Within this wild JNp city, mathematicians wanted to find the "quietest" parts. They defined two special zones, VJNp and CJNp, which are supposed to be the places where the noise eventually dies down completely.

• VJNp (The "Vanishing" Zone): This is the area where, if you zoom in on a tiny speck of the city, the noise disappears. It's like looking at a single grain of sand; it's too small to have any meaningful fluctuation.
• CJNp (The "Compact" Zone): This is the area where, if you zoom out to look at the whole horizon (a huge area), the noise also disappears. It's like looking at the entire city from space; the individual chaotic streets blend into a smooth, quiet average.

The Big Mystery:
For a long time, mathematicians knew that CJNp was inside VJNp (because if the noise is gone everywhere, it's definitely gone in the tiny spots). But they didn't know if VJNp was strictly bigger than CJNp.

• The Question: Is there a weird, chaotic neighborhood that is quiet when you zoom in (satisfies VJNp) but still has some weird, lingering chaos when you zoom out (fails C JNp)?
• The Intuition: It felt like there should be such a place. It felt like the two zones might be different.

The Breakthrough: They Are Actually the Same!

The authors of this paper, Riikka Korte and Timo Takala, proved a surprising result: VJNp and CJNp are actually the exact same place. There is no "in-between" weird neighborhood. If a function is quiet when you zoom in, it must also be quiet when you zoom out.

How did they prove it?
They used a clever trick involving Morrey-type integrals. Imagine you are measuring the "average loudness" of a neighborhood, but you multiply that loudness by the size of the neighborhood.

• If the neighborhood is tiny, this number should be tiny.
• If the neighborhood is huge, this number should also be tiny.

The authors showed that for any function in the JNp space, this "size-adjusted loudness" always goes to zero, whether you are looking at a microscopic speck or a massive continent.

The Analogy:
Think of a party.

• VJNp is the rule: "If you look at just one person, they aren't screaming."
• CJNp is the rule: "If you look at the whole crowd, the average volume isn't screaming."
• The paper proves: If no single person is screaming (VJNp), then the whole crowd cannot be screaming on average (CJNp). You can't have a crowd that is silent up close but deafening from a distance. The math forces them to be the same.

A Side Note: The "Constant" Puzzle

The paper also solved a smaller, related puzzle about a specific type of mathematical space called JNp,p.
Imagine you have a function that is "almost" in the Lp space (a very well-behaved, quiet space), but it's off by a constant amount (like a radio station that is playing the right song but at a volume that is just slightly too high or too low).
The authors proved that the space JNp,p is exactly the space of "well-behaved functions plus a constant." It's like saying, "If you strip away the background hum (the constant), what's left is perfectly quiet." This answered a question that had been sitting open for a few years.

Why Does This Matter?

In the world of math, knowing that two things are equal is huge. It means we don't have to check two different rules to see if a function is "vanishing." We only need to check one. It simplifies the map of the mathematical universe, showing us that the "vanishing" neighborhoods are more uniform and predictable than we thought.

In short: The authors looked at a complex mathematical landscape, found two areas that looked different, and proved they are actually the same territory, using a clever measuring stick that works for both tiny and giant scales.

Technical Summary

Here is a detailed technical summary of the paper "The John-Nirenberg Space: Equality of the Vanishing Subspaces $VJN_p$ and $CJN_p$" by Riikka Korte and Timo Takala.

1. Problem Statement and Context

Background:
The John-Nirenberg space, denoted as $JN_p$ (for $1 < p < \infty$), is a generalization of the space of Bounded Mean Oscillation ($BMO$). While$BMO$corresponds to the limit$JN_\infty$, the spaces$JN_p$occupy a non-trivial intermediate position between$L^p$and weak-$L^p$($L^{p,\infty}$). Specifically, for a bounded cube$Q_0$, the inclusions$L^p(Q_0) \subsetneq JN_p(Q_0) \subsetneq L^{p,\infty}(Q_0)$ hold.

The Core Question:
Similar to $BMO$, which has well-known vanishing subspaces $VMO$ (functions with vanishing mean oscillation) and $CMO$ (closure of smooth compactly supported functions in $BMO$), the $JN_p$ space has analogous subspaces:

• $VJN_p$: Defined as the closure of smooth functions with gradients in $L^p$ (analogous to $VMO$).
• $CJN_p$: Defined as the closure of smooth functions with compact support (analogous to $CMO$).

It was previously established that $L^p/\mathbb{R} \subseteq CJN_p \subseteq VJN_p \subseteq JN_p$. While strict inclusions were known for the outer boundaries ($L^p \neq CJN_p$ and $VJN_p \neq JN_p$), it remained an open question whether the intermediate inclusion was strict, i.e., whether $VJN_p \setminus CJN_p$ was non-empty.

Goal:
The primary objective of this paper is to resolve this open question by proving that $VJN_p = CJN_p$ on $\mathbb{R}^n$. Additionally, the authors aim to clarify the structure of the generalized space $JN_{p,q}$ in the borderline case where $p=q$.

---

2. Methodology

The authors employ a combination of functional analysis, geometric measure theory, and iterative construction techniques.

A. Characterization via Morrey-Type Integrals:
The proof relies on characterizing the subspaces $VJN_p$ and $CJN_p$ through the behavior of specific integrals.

• $f \in VJN_p$ if and only if the sum of oscillations over disjoint cubes vanishes as the cube sizes approach zero.
• $f \in CJN_p$ if and only if $f \in VJN_p$ and a specific "large cube" condition holds:
  $$\lim_{a \to \infty} \sup_{Q: l(Q) \ge a} |Q|^{1/p} \int_Q |f - f_Q| = 0$$
  where $f_Q$ is the average of $f$ over $Q$.

B. The Key Technical Tool (Proposition 3.1):
The central technical innovation is Proposition 3.1, which establishes a lower bound for the oscillation of a function on specific sub-cubes.

• Hypothesis: If a non-negative function $f$ has a "large" Morrey-type integral on a cube $Q$ (specifically $|Q|^{1/p} \int_Q f \ge A(1-\epsilon)$) but "small" integrals on slightly larger/smaller concentric cubes, then there exist two disjoint sub-cubes $Q_1, Q_2$ within a neighborhood of $Q$ where the oscillation $|Q_i|^{1/p} \int_{Q_i} |f - f_{Q_i}|$ is bounded below by a constant times $A$.
• Geometric Construction: The proof involves a sophisticated geometric decomposition of cubes in $\mathbb{R}^n$. For $n \ge 2$, the authors use a dyadic decomposition and projection arguments (Lemma 3.9) to locate disjoint cubes in "opposite directions" relative to a central cube, ensuring that the oscillation cannot be arbitrarily small without contradicting the $JN_p$ norm definition.

C. Contradiction via Disjoint Sequences:
To prove the vanishing of the integrals, the authors assume the limit is non-zero ($A > 0$). Using Proposition 3.1, they iteratively construct an infinite sequence of pairwise disjoint cubes $(Q'_j)$ such that the sum of their oscillation terms diverges:
$$\sum_{j=1}^\infty |Q'_j| \left( \int_{Q'_j} |f - f_{Q'_j}| \right)^p = \infty$$
This contradicts the definition of $f \in JN_p$, which requires this sum to be finite for any collection of disjoint cubes.

---

3. Key Contributions and Results

Result 1: Equality of Vanishing Subspaces ($VJN_p = CJN_p$)

• Theorem 3.12: For any $f \in JN_p(\mathbb{R}^n)$, there exists a constant $b$ such that the Morrey-type integral over large cubes vanishes:
  $$\lim_{a \to \infty} \sup_{l(Q) \ge a} |Q|^{1/p} \int_Q |f - b| = 0$$
• Theorem 3.16: Similarly, for any $f \in JN_p(X)$ (where $X$ is a bounded cube or $\mathbb{R}^n$), the integral over small cubes vanishes:
  $$\lim_{a \to 0} \sup_{l(Q) \le a} |Q|^{1/p} \int_Q |f| = 0$$
• Corollary 3.15: Combining these results with the definitions of $VJN_p$ and $CJN_p$, the authors prove that $VJN_p(\mathbb{R}^n) = CJN_p(\mathbb{R}^n)$. This resolves the open question posed in previous literature (e.g., [15], [17]).

Result 2: Characterization of $JN_{p,p}(\mathbb{R}^n)$

The paper investigates the generalized John-Nirenberg space $JN_{p,q}$, where the $L^1$ oscillation is replaced by an $L^q$ oscillation.

• Previous work established that for $X=\mathbb{R}^n$:
  • If $q < p$, $JN_{p,q} = JN_p$.
  • If $q > p$, $JN_{p,q}$ contains only constant functions.
• Proposition 2.7: The authors settle the borderline case $q=p$. They prove that:
  $$JN_{p,p}(\mathbb{R}^n) = L^p(\mathbb{R}^n) / \mathbb{R}$$
  This means the space consists of functions $f$ such that $f - b \in L^p(\mathbb{R}^n)$ for some constant $b$. This answers a specific question raised in [19, Remark 2.9].

---

4. Significance and Impact

1. Structural Clarity: The equality $VJN_p = CJN_p$ simplifies the understanding of the John-Nirenberg space hierarchy. It confirms that the "vanishing" property in $JN_p$ is robust and does not depend on whether one uses the closure of compactly supported functions or the closure of functions with $L^p$ gradients.
2. Resolution of Open Problems: The paper definitively answers questions raised in multiple recent studies ([15], [17], [18], [19]), closing gaps in the theory of John-Nirenberg spaces.
3. Methodological Advancement: The development of Proposition 3.1 and the geometric lemmas (3.7, 3.9) provides new tools for analyzing oscillation integrals in higher dimensions. The technique of constructing disjoint sequences to force a contradiction with the $JN_p$ norm is a powerful method that may be applicable to other function spaces involving oscillation conditions.
4. Connection to Morrey Spaces: The results imply that $JN_p$ functions satisfy vanishing Morrey-type conditions, linking the John-Nirenberg space more closely to the theory of Morrey spaces ($V L^{1, n-n/p}$).

In summary, Korte and Takala provide a rigorous proof that the two natural definitions of vanishing subspaces for the John-Nirenberg space coincide, while simultaneously completing the classification of the generalized $JN_{p,q}$ spaces.