On the equivalence between moderate growth-type conditions in the weight matrix setting II

This paper establishes a new characterization of the moderate growth condition in the mixed weight matrix setting by proving its equivalence to a property of the associated weight function, thereby addressing the limitations of previous generalizations.

The Gist

Imagine you are a chef trying to bake the perfect cake. In the world of advanced mathematics, specifically in the study of "ultradifferentiable functions" (which are like super-smooth, ultra-precise curves used in physics and engineering), mathematicians use special "recipes" called weight sequences. These recipes dictate how fast the ingredients (numbers in a sequence) can grow as you add more of them.

If the ingredients grow too fast, the cake collapses (the math breaks). If they grow too slow, the cake is flat and boring. There is a "Goldilocks" rule called Moderate Growth. It ensures the ingredients grow at a steady, manageable pace so the math works out.

The Problem: The "Two-Recipe" Dilemma

In the past, mathematicians mostly looked at one recipe at a time. But recently, they started dealing with Weight Matrices. Think of a Weight Matrix not as a single recipe, but as a library of recipes that change depending on a parameter (like temperature or pressure).

The big question this paper tackles is: If we have a library of recipes, can we still guarantee that the "Goldilocks" rule (Moderate Growth) holds true for the whole library?

Previously, mathematicians tried to apply the old "one-recipe" rules to these libraries. They hoped that if the library was "equivalent" to a good single recipe, the rules would automatically transfer. However, they hit a wall. It turned out that simply swapping one recipe for an "equivalent" one in the library didn't always preserve the growth rules. It was like trying to use a single-recipe rulebook to manage a whole restaurant kitchen; it just didn't fit.

The Solution: A New "Growth Index"

The author, Gerhard Schindl, introduces a new way to measure these libraries. Instead of asking "Is this recipe perfect?", he asks, "How many times do we need to stretch this recipe before it fits the rules?"

He defines a "Moderate Growth Index" (let's call it the Stretch Factor).

• If the Stretch Factor is 1, the recipe is perfect (it follows the old rules).
• If the Stretch Factor is 2, you have to stretch the recipe twice to make it fit.
• If the Stretch Factor is infinite, the recipe is broken and can't be saved.

The paper proves that this "Stretch Factor" is a robust property. Even if you swap your recipe for a slightly different one that is mathematically "equivalent" (like using metric cups instead of imperial cups), the Stretch Factor remains the same. This is a huge relief because it means the property is stable and reliable.

The Bridge: From Numbers to Functions

The most exciting part of the paper is building a bridge between two different ways of looking at the problem:

1. The Sequence View: Looking at the list of numbers (the ingredients).
2. The Function View: Looking at a smooth curve that describes the growth (the shape of the rising dough).

For a long time, mathematicians knew how to check the growth rules for the list of numbers, but they struggled to translate that into a rule for the smooth curve. It was like knowing how to count the steps on a staircase but not being able to describe the slope of the ramp.

Schindl's main result is a translation manual. He proves that you can check the "Stretch Factor" of the number list by looking at a specific property of the smooth curve (the weight function). Specifically, he shows that if the curve satisfies a certain "doubling" condition (if you double the input, the output doesn't explode), then the number list has a finite Stretch Factor.

The "Counter-Example" Dead End

The paper also tackles a tempting idea: "Can we just pretend the library follows the old rules by swapping it with a 'fake' library that looks the same?"

The author tries to construct a "counter-example"—a scenario where you swap the library for an equivalent one, and suddenly the rules break. However, after doing some heavy mathematical lifting, he discovers that this counter-example cannot exist. The rules are so fundamental that you can't trick them. If a library is equivalent to a good one, it must obey the growth rules. This closes a door on a potential loophole and solidifies the theory.

The Big Picture

In simple terms, this paper does three things:

1. Fixes a broken tool: It updates the old "Moderate Growth" rule so it works for complex libraries of recipes (Weight Matrices), not just single ones.
2. Builds a bridge: It connects the world of discrete number lists with the world of smooth curves, allowing mathematicians to switch between them easily.
3. Confirms stability: It proves that these growth rules are unshakeable. You can't trick the system by swapping equivalent recipes; the math holds up no matter how you look at it.

This is crucial for fields like physics and engineering where these "super-smooth" functions are used to model real-world phenomena. By ensuring the rules are consistent and translatable, the paper gives scientists a more reliable toolkit for their calculations.

Technical Summary

Here is a detailed technical summary of the paper "On the Equivalence Between Moderate Growth-Type Conditions in the Weight Matrix Setting II" by Gerhard Schindl.

1. Problem Statement and Context

The paper addresses a fundamental problem in the theory of ultradifferentiable function spaces: the generalization of the moderate growth condition from single weight sequences to mixed settings involving weight matrices.

• Background: In classical theory, a weight sequence $M = (M_p)_p$ satisfies the moderate growth condition (mg) if $\exists C \ge 1, \forall p, q \in \mathbb{N}: M_{p+q} \le C^{p+q+1} M_p M_q$. This condition is crucial for the stability of function spaces under ultradifferential operators.
• The Mixed Setting: Recent developments involve weight matrices $\mathcal{M} = \{M(x) : x \in I\}$, where $I = (0, +\infty)$. These matrices are used to define function classes associated with weight functions $\omega$ (in the sense of Braun-Meise-Taylor).
• The Core Issue: It was previously known that the classical equivalent reformulation of (mg)—specifically the comparison between the sequence of quotients $\mu_p = M_p/M_{p-1}$ and the roots $(M_p)^{1/p}$—fails to generalize directly to the mixed setting.
  • The "Quotient-Root Comparison" properties (Roumieu-type and Beurling-type) do not automatically hold for the weight matrix $\mathcal{M}_\omega$ associated with a weight function $\omega$.
  • A weaker condition, denoted as (1.4) or (2.12), was introduced in a previous work [23]:
    $$\exists d \in \mathbb{N}_{>0}, \exists A \ge 1, \forall p \in \mathbb{N}_{>0}: \mu_p \le A (M_{dp})^{\frac{1}{dp}}$$
    This defines a moderate growth index $g(M)$.
• The Gap: While condition (1.4) characterizes the validity of the quotient-root comparison for the matrix, a direct characterization of this condition in terms of the associated weight function $\omega_M$ was missing. The paper aims to bridge this gap.

2. Methodology

The author employs a multi-step analytical approach combining sequence theory, convex analysis, and the theory of associated weight functions:

1. Auxiliary Sequences: The paper introduces and utilizes auxiliary sequences defined by $\widetilde{M}^a_p := (M_{ap})^{1/a}$. This notation allows for the manipulation of indices and powers to relate different growth conditions.
2. Legendre-Fenchel Conjugates: The analysis relies heavily on the properties of the associated weight function $\omega_M(t) = \sup_p \log(t^p/M_p)$ and its relationship with the convex conjugate of $\phi_\omega(t) = \omega(e^t)$.
3. Generalized Lower Legendre Conjugate: A key technical tool is the "generalized lower Legendre conjugate" (or infimal convolution) defined as:
   $$(\sigma \check{\star} \tau)(t) := \inf_{s>0} \{ \sigma(s) + \tau(t/s) \}$$
   This operation relates the weight function of a product of sequences to the sum of their individual weight functions.
4. Equivalence Relations: The study distinguishes between R-equivalence (Roumieu-type) and B-equivalence (Beurling-type) of weight matrices and investigates how growth properties are preserved under these equivalences.

3. Key Contributions and Results

The paper establishes several new characterizations and equivalences:

A. Characterization of the Moderate Growth Index via Weight Functions

The main result (Theorem 5.3 and Corollary 5.5) provides a necessary and sufficient condition for a weight sequence $M$ to satisfy the growth condition (2.12) in terms of its associated weight function $\omega_M$.

• Result: $M$ satisfies condition (2.12) with index $d$ if and only if the associated weight function satisfies a specific growth inequality involving the infimal convolution:
  $$\exists a \in \mathbb{N}_{>0}, \exists A \ge 1, \exists C \ge 0, \forall t \ge 0: \quad \omega_{M \widetilde{M}^a}(t^2) \le \omega_M(At) + C$$
  where $\omega_{M \widetilde{M}^a}$ is the weight function of the product sequence.
• Significance: This translates a discrete condition on the sequence $M$ into a continuous growth restriction on the function $\omega_M$.

B. Equivalence of Quotient-Root Comparison Properties

The paper proves that the "Quotient-Root Comparison" properties (conditions (1.2) and (1.3) in the introduction) for a weight matrix $\mathcal{M}_\omega$ are equivalent to the weight function $\omega$ satisfying the strong non-quasianalyticity condition (often denoted as (ω6)):
$$\exists C \ge 1, \forall y \ge 0: \int_1^{+\infty} \frac{\omega(yt)}{t^2} dt \le C\omega(y) + C$$
This confirms that the "mixed" growth conditions for the matrix are intrinsically linked to the regularity of the underlying weight function.

C. Preservation Under Equivalence

Theorem 4.7 demonstrates that the condition (2.12) (and thus the finite moderate growth index $g(M)$) is preserved under the equivalence of log-convex weight sequences. If $M \approx N$ and $M$ satisfies (2.12), then $N$ also satisfies it (potentially with a different index $d$). This strengthens previous results by removing the need for strong non-quasianalyticity assumptions in certain contexts.

D. Technical Lemmas on Auxiliary Sequences

The paper provides detailed estimates for the moderate growth index $g(\cdot)$ when scaling the parameter of the weight matrix (Lemma 3.1). It shows that $g(W(cx))$ and $g(W(x))$ are comparable (within a factor of 4), ensuring that the growth index is stable under the re-parameterization of the matrix.

4. Discussion on Counter-Examples and Open Questions

In Section 6, the author investigates whether one can always switch to an equivalent weight matrix that satisfies the strict quotient-root comparison properties (1.2) or (1.3) without loss of generality.

• Necessary Conditions: Proposition 6.1 derives necessary conditions (6.1) and (6.2) that any sequence $S(1)$ must satisfy if it is R/B-equivalent to a matrix satisfying the comparison properties.
• Impossibility of Counter-Examples: Lemma 6.2 proves that for any log-convex sequence $M \in LC$, these necessary conditions are always satisfied (with constants $B=B_1=1$).
• Conclusion: This implies that one cannot construct a counter-example by simply finding a sequence that violates these specific inequalities. Consequently, the question of whether the quotient-root comparison holds "w.l.o.g." (without loss of generality) remains difficult; the obstruction lies deeper than simple inequality violations, suggesting that the failure of these properties in the mixed setting is a structural feature of the weight function setting rather than a specific sequence anomaly.

5. Significance

• Unification of Theories: The paper successfully unifies the discrete theory of weight sequences with the continuous theory of weight functions. It clarifies exactly when the "mixed" conditions (involving two sequences or a sequence and a function) hold.
• Practical Application: The characterization of condition (2.12) via $\omega_M$ allows researchers to verify growth conditions directly using the weight function, which is often more tractable in applications involving ultradifferentiable classes and Gelfand-Shilov spaces.
• Refinement of Previous Work: It corrects and refines the understanding of the weight matrix setting introduced in [15] and [20], showing that the "moderate growth index" $g(M)$ is the correct invariant to study rather than the strict (mg) condition.
• Foundation for Future Research: By establishing the equivalence between the matrix properties and the (ω6) condition, the paper provides the technical foundation needed to transfer results from the weight sequence setting (like those in [7]) to the weight function setting.

In summary, Schindl resolves a critical gap in the theory of ultradifferentiable functions by proving that the failure of direct generalization of moderate growth to mixed settings is characterized precisely by the growth behavior of the associated weight function, specifically through the condition (ω6) and the infimal convolution of weight functions.