Infinite-variation L'evy processes are either strongly eroded or abrupt

arXiv:2610.20916 · math.PR · Submitted 2026-10-14 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: We're looking at the paper: "Infinite-variation L'evy processes are either strongly eroded or abrupt".

Emmy: Here's the gist of the paper. Every real Lévy process of infinite variation is either abrupt or strongly eroded The Core Dichotomy The paper proves that every real Lévy process of infinite variation is either abrupt or strongly eroded, resolving the conjectured geometric dichotomy on the smoothness of the convex hull of the graph of a Lévy path Theorem 1.1 states that exactly one of two conditions holds: either JX(I) < ∞ for every bounded interval I, or JX(I) = ∞ for every proper interval I, which is equivalent to X being either abrupt or strongly eroded Characterizing Abruptness and Erosion Abruptness is characterized by the condition JX(I) < ∞ for every bounded interval I, which is equivalent to s1 ∈ L1 loc(R) The paper shows that Vigon’s conjecture, which conjectured X is abrupt if and only if X(r) hits points for some r ∈ R, fails for symmetric pure-jump Lévy processes, with counterexamples showing that the set E = E

r ∈ R: s1(r) = ∞: can be a compact perfect set of Lebesgue measure zero Richness of the Strongly Eroded Class The strongly eroded class is shown to be much richer than previously thought, as Theorem 1.7 demonstrates that strong erosion is compatible with essentially arbitrary admissible jump activity indices β− ∈

0, 1: and β+ ∈

1, 2: This means strong erosion is not confined to atomic or singular jump measures The paper constructs such processes using absolutely continuous Lévy measures with strictly positive C∞ densities, showing that the process has bounded C∞ transition densities with bounded derivatives of every order at all positive times Applications to Geometric Properties The dichotomy has significant applications, including the classification of convex-minorant slopes and connections to Burgers turbulence Corollary 2.1 establishes that for an infinite variation process X, exactly one of two alternatives holds regarding its greatest L-Lipschitz minorant Ml, where X is either abrupt (Zl is locally finite almost surely for each l > m) or strongly eroded (Zl has zero Lebesgue measure but is not locally finite almost surely) Corollary 2.2 further distinguishes the two classes in the context of Burgers turbulence, stating that for a fixed t > 0, either X is abrupt (At is locally finite and A0,t is regenerative) or strongly eroded (there are no rarefaction intervals and at is strictly increasing on R) Sequential Gaussian Limits The paper utilizes sequential Gaussian limits to construct strongly eroded processes, as seen in the construction of Orey’s process Proposition 5.1 provides a criterion that turns a bounded subsequence of the running maximum of the characteristic exponent into divergence of s1 at every drift This leads to the construction of processes with arbitrary activity indices β− ∈

0, 1: and β+ ∈

1, 2: Counterexample Construction The paper provides a detailed proof for the counterexample to Vigon’s conjecture, showing that there exist abrupt symmetric pure-jump Lévy processes for which s1(0) = ∞, while s1(r) < ∞ for r ∈ R — except possibly at r=0 This demonstrates that the failure of Vigon’s conjecture can be much richer than a single exceptional slope, with the exceptional set E potentially being uncountable and perfect The construction of such processes exploits lacunary sequences of jump sizes to control the finiteness properties of s1 Smooth Transition Densities The construction in Theorem 1.3 specifies an infinite-variation Lévy process with finite variance where the Lévy measure is defined by a complex sum involving lacunary sequences and antisymmetric perturbations The analysis shows that for the lower index, at yn = 3zn/4, z(In) → ∞ as n → ∞, leading to s1(r) = ∞ for r ∈ C Furthermore, the construction ensures bounded C∞ transition densities with bounded derivatives of all orders for the strongly eroded processes The analysis of Q(u) shows that R R u j e − tQ(u)du 0 and j ≥ 0, ensuring bounded C∞ transition densities with bounded derivatives of all orders The Full Activity-Index Region The necessity of β− ≤ 1 ≤ β+ for strongly eroded processes is established in

7, Prop. [Ada: 1.6], and the paper shows that the entire region can be attained by constructing a process X with density υ+, zero Gaussian coefficient, and zero drift parameter The activity indices are bounded as β+(X) = b+ and β−(X) = 0 for this construction The lower index is shown to be 0 because the cutoff excludes all bumps with k ≤ n, giving y p−2n V (yn) → 0 for any p > 0 The process Z = X + Y, where Y has finite variation, is strongly eroded by

7, Prop. [Emmy: 1.5], and the indices are shown to be max(b+, b−) = b+ for the upper index Smooth Transition Densities The construction of smooth transition densities involves a symmetric C∞ density on R - (0) defined by a sum involving lacunary sequences and an even, non-negative function supported in

-1, 1: The analysis of Q(u) shows that R R u j e − tQ(u)du 0 and j ≥ 0, ensuring bounded C∞ transition densities with bounded derivatives of all orders This is achieved by showing that Q(u) log u → ∞ as u → ∞ The Full Activity-Index Region The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by

7, Prop. [Ada: 1.5], and the lower index is shown to be b− because r p−2W(r) → ∞ for p < b−, while along yn = 3zn/4, we have as n → ∞, y p−2n VZ(yn) = y p−2n V (yn) + y p−2n W(yn) → 0 The final result is that β−(Z) = b− The Full Activity-Index Region The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by [7, Prop.

Title and authors: Ada: So we’re looking at this paper from Jorge González Cáceres, David Kramer-Bang, and Aleksandar Mijatović that calls itself "Infinite-variation L'evy processes are either strongly eroded or abrupt". It tackles a really fundamental question about how these infinite variation processes behave.

Emmy: Exactly. The title immediately suggests a division, a dichotomy. It’s about classifying every real Lévy process of infinite variation as one of two types: either it’s abrupt, or it’s strongly eroded. This resolves this long-standing geometric issue concerning the smoothness of the convex hull of the graph of any Lévy path <ref:2610.20916#pg1>.

Kolya: What does that mean for us in practice? If we can classify them cleanly, it helps us understand where they come from, right?

Ada: It means we have a definitive way to check if a process is abrupt or strongly eroded by looking at how the function related to the process’s graph behaves. The paper sets up this distinction by looking at two different conditions on the integrated variation of the process over any interval.

Emmy: Specifically, Theorem one point one states that either the integrated variation over any bounded interval is finite, or it’s infinite for every proper interval <ref:2610.20916#pg1>. This condition is mathematically equivalent to saying the process is either abrupt or strongly eroded <ref:2610.20916#pg2>.

Kolya: So, how do we actually tell which one it is? Is that distinction simple enough to check?

Ada: Well, the paper defines abruptness by a condition called JX over any bounded interval I being finite. This is linked to something called s1 being locally integrable on the real line R <ref:2610.20916#pg2>.

Emmy: And the contrast is that strong erosion means that JX over every proper interval I is infinite. The paper shows this structure implies a very different kind of behavior for the process itself, related to its running maximums and how it grows over time <ref:2610.20916#pg3>.

Kolya: That sounds like a distinction between having sharp jumps that stay within bounds versus having a more continuous but still infinite variation structure. What about those conjectures they mention?

Ada: They tackle Vigon’s conjecture, which suggested that a process is abrupt if it hits points for some real number r. The paper shows this conjecture fails even for symmetric pure-jump Lévy processes <ref:2610.20916#pg2>.

Emmy: That failure means the set of slopes where this point hitting condition fails can be much more than just one single exceptional slope, it can be an uncountable perfect set with zero Lebesgue measure <ref:2610.20916#pg3>.

Kolya: So, that means the structure isn't as simple as a single flaw in the process definition. What about the strong erosion side of things? Is that class just a collection of complicated processes then?

Ada: Not at all. Theorem one point seven shows that strong erosion is compatible with essentially arbitrary jump activity indices, which are parameters we use to describe how often jumps occur <ref:2610.20916#pg3>.

Emmy: That compatibility means strong erosion isn't just confined to processes with very simple jump measures, like atomic ones or singular ones. They can be constructed using absolutely continuous Lévy measures that have strictly positive smooth densities for every positive time <ref:2610.20916#pg3>.

Kolya: Having bounded C-infinity transition densities sounds good for analysis. Does this smoothness help us understand the long-term behavior of these processes?

Ada: It helps a lot, Kolya. The construction in Theorem one point three shows that for these strongly eroded processes, the process has bounded C-infinity transition densities with bounded derivatives of every order at all positive times <ref:2610.20916#pg3>.

Emmy: That smoothness is a key structural feature that separates the two classes when we look at things like convex minorant slopes and their contact sets <ref:2610.20916#pg3>.

Kolya: And what about the applications we talked about? Like those connections to Burgers turbulence or those classifications of minorant slopes? Are those results robust?

Ada: They are, Kolya. Corollary two point one establishes that for an infinite variation process X, exactly one of two alternatives holds regarding its greatest L-Lipschitz minorant M sub l, where X is either abrupt or strongly eroded <ref:2610.20916#pg3>.

Emmy: And Corollary two point two further separates the classes in the context of Burgers turbulence, distinguishing them based on whether there are rarefaction intervals or if the process has strictly increasing paths over all real numbers R <ref:2610.20916#pg3>.

Kolya: That distinction regarding rarefaction intervals sounds like a concrete way to separate the two classes in a physical setting. What about the techniques they use to build these complex strongly eroded processes?

Ada: They use sequential Gaussian limits, which is something we’ve seen used before, like in Orey’s process construction <ref:2610.20916#pg3>.

Emmy: Proposition five point one provides a criterion that turns a bounded subsequence of the running maximum of the characteristic exponent into divergence of s1 at every drift parameter, which is what leads to those processes with arbitrary activity indices beta- in zero one and beta+ in one two <ref:2610.20916#pg3>.

Kolya: So they are constructing processes with a very wide range of jump activities just to show the class is rich enough. What about the specific constraints on those activity indices?

Ada: They establish that for strong erosion, we need beta- to be less than or equal to one and beta+ to be greater than or equal to one <ref:2610.20916#pg3>.

Emmy: The paper shows they can attain the entire region by using a construction with zero drift parameter, zero Gaussian coefficient, and bounded activity indices beta+(X) and beta-(X) <ref:2610.20916#pg3>.

Kolya: That's a lot of machinery to get those bounds on the activity indices right. Does this construction work for processes with lower index b- strictly between zero and one?

Ada: Yes, they define a modified density function upsilon- that preserves the super-logarithmic Fourier growth, which keeps those smooth transition densities intact <ref:2610.20916#pg3>.

Emmy: The resulting process Z plus some finite variation term Y is strongly eroded by using a result from seven Prop one point five, and they show that the lower index beta- can be any value b- between zero and one <ref:2610.20916#pg3>.

Kolya: So they’re showing that you can tune the process to have a specific level of jump activity control, which is useful for modeling things with different jump intensities. What about the smooth transition densities again?

Ada: They use a symmetric C-infinity density on R minus zero defined by a sum involving lacunary sequences and an even function supported in negative one to one <ref:2610.20916#pg3>.

Emmy: The analysis of the function Q of u shows that the term related to u times e minus t times Q of u, scaled by j greater than or equal to zero, tends to infinity as u approaches infinity <ref:2610.20916#pg3>.

Kolya: That divergence estimate is what guarantees those bounded C-infinity transition densities with bounded derivatives of all orders across the board. Does the paper address any limitations in this construction?

Ada: They do acknowledge that the analysis relies on specific constructions, like a complex sum involving lacunary sequences and antisymmetric perturbations to define their Lévy measure <ref:2610.20916#pg3>.

Emmy: And they show that this structure ensures bounded C-infinity transition densities with bounded derivatives of all orders for the strongly eroded processes <ref:2610.20916#pg3>.

Kolya: So, to summarize the core finding of this paper on "Infinite-variation L'evy processes are either strongly eroded or abrupt," we have a complete classification based on local variation properties.

Ada: Precisely. We’ve shown that every real Lévy process of infinite variation falls into one of two categories: it is either abrupt, meaning its integrated variation over any bounded interval is finite, or it is strongly eroded, which has infinite integrated variation for every proper interval <ref:2610.20916#pg1>.

Emmy: This classification ties the geometric smoothness of the path's graph to these intrinsic properties of the process itself. It’s a way to understand why certain structures appear in other areas, like convex hull smoothness and Burgers shock geometry <ref:2610.20916#pg3>.

Kolya: And on the Vigon conjecture failure, they show that even when we relax symmetry conditions, the set of slopes where point hitting fails can be uncountable but still have zero Lebesgue measure <ref:2610.20916#pg3>.

Ada: They’ve also given us concrete counterexamples for symmetric pure-jump processes where the failure set E is much richer than just a single point, potentially being an uncountable perfect set <ref:2610.20916#pg2>.

Emmy: The robustness of the strongly eroded class is shown by its ability to support arbitrary jump activity indices, which means we can model processes with a huge variety of jump behaviors <ref:2610.20916#pg3>.

Kolya: And from an application standpoint, they provide clear distinctions in areas like Burgers turbulence by looking at rarefaction intervals and path monotonicity <ref:2610.20916#pg3>.

Ada: So, to wrap up this discussion on "Infinite-variation L'evy processes are either strongly eroded or abrupt," we’ve established a complete dichotomy based on local variation behavior.

Emmy: This paper gives us the tools to classify these processes by checking the finiteness of JX over bounded intervals versus every proper interval <ref:2610.20916#pg3>.

Kolya: Kolya's final thought is that understanding these structural differences in path smoothness and jump activity, as detailed in "Infinite-variation L'evy processes are either strongly eroded or abrupt," provides a solid framework for analyzing more complex stochastic systems <ref:2610.20916#pg3>.

Ada: That’s it for this paper. We’ve seen how the structure of the process dictates its behavior, whether it's abrupt or strongly eroded <ref:2610.20916#pg1>.

The paper's summary: Ada: So, we just saw that every real Lévy process of infinite variation fits neatly into one of two boxes: either it’s abrupt or strongly eroded <ref:2610.20916#pg1>.

Emmy: That means we've got a complete classification system now. We look at how the integrated variation behaves over different intervals, and that tells us whether the process is abrupt or strongly eroded <ref:2610.20916#pg3>.

Ada: It’s really about geometric smoothness, Emmy. If you're dealing with an infinite variation process, you can't just have any path shape; it has to adhere to one of these two structural rules <ref:2610.20916#pg2>.

Emmy: Exactly. The abrupt ones are the simpler cases, where the variation over any bounded interval is finite <ref:2610.20916#pg3>. They have a certain kind of local finiteness everywhere <ref:2610.20916#pg3>.

Kolya: And then you have the strongly eroded ones, which are all about that variation being infinite for every proper interval <ref:2610.20916#pg3>. That’s where the complexity really kicks in <ref:2610.20916#pg3>.

Ada: Kolya, when you say complexity, are we talking about jump activity indices or something else? Because the paper shows that strong erosion isn't just one single type of process anymore <ref:2610.20916#pg3>.

Kolya: It’s definitely jump activity. They show that strong erosion is compatible with almost any jump activity indices, which are parameters describing how often jumps happen <ref:2610.20916#pg3>. This means the strongly eroded class is much richer than we thought before <ref:2610.20916#pg3>.

Emmy: That richness comes from the construction methods they use, involving absolutely continuous Lévy measures with strictly positive smooth densities for every positive time <ref:2610.20916#pg3>. So, you can build processes that look very smooth but still have infinite variation <ref:2610.20916#pg3>.

Ada: And that smoothness is key when we talk about transition densities, Emmy. The paper proves that these strongly eroded processes have bounded C-infinity transition densities with bounded derivatives of every order at all positive times <ref:2610.20916#pg3>. That's a big deal for analysis <ref:2610.20916#pg3>.

Kolya: So, if you’re trying to model something where you need smooth behavior but infinite variation, this paper gives you a class of processes that actually has that regularity <ref:2610.20916#pg3>. What about the stuff we discussed earlier, like those applications in Burgers turbulence?

Emmy: That’s where the distinction becomes physical. Corollary two point one shows that for an infinite variation process, you can tell if it's abrupt or strongly eroded by looking at its greatest L-Lipschitz minorant <ref:2610.20916#pg3>.

Ada: Abrupt means the variation over any bounded interval is finite, and strongly eroded means that the path isn't locally finite almost surely, but it still has a specific kind of control on its contact sets <ref:2610.20916#pg3>.

Kolya: And Corollary two point two makes it even clearer in the context of Burgers turbulence. It tells you whether there are rarefaction intervals or if the path is strictly increasing over all real numbers <ref:2610.20916#pg3>. That’s a way to tell them apart in practice <ref:2610.20916#pg3>.

Emmy: It means that when we look at turbulence, we can categorize the solutions based on these path properties <ref:2610.20916#pg3>. This helps us understand the underlying geometry of those solutions <ref:2610.20916#pg3>.

Ada: It’s a strong classification tool, because it connects the abstract math of Lévy processes to concrete geometric features and physical phenomena like turbulence <ref:2610.20916#pg3>.

The paper's improvements: Ada: We just looked at how we classify infinite variation Lévy processes as either abrupt or strongly eroded <ref:2610.20916#pg1>. Now, let’s talk about what the authors suggest they can do better with this framework.

Emmy: They are proposing ways to make that classification even more robust, especially concerning the underlying structure of these processes <ref:2610.20916#pg3>.

Ada: Specifically, they want a clearer way to distinguish between those two classes based on how their convex minorant slopes behave <ref:2610.20916#pg3>. They want to see if the variation of that minorant is purely atomic or if it’s more diffuse <ref:two thousand six hundred ten point two zero nine one six#pg3.

Kolya: That distinction between atomic and diffuse variation sounds like it would give us a better handle on the geometry of these paths <ref:2610.20916#pg3>. If we can classify them by that, does it help with anything else?

Ada: It helps with connections to things like Burgers turbulence <ref:2610.20916#pg3>. The current classification is good, but the authors are looking for a way to use these geometric properties more directly in those physical models <ref:two thousand six hundred ten point two zero nine one six#pg3.

Emmy: They’re also exploring how to pin down the exact parameters of strong erosion, like those activity indices beta- and beta+ we talked about <ref:two thousand six hundred ten point two zero nine one six#pg3. They want to show that we can precisely identify these indices for any strongly eroded process <ref:two thousand six hundred ten point two zero nine one six#pg3.

Kolya: So, it’s not just about the existence of the class anymore, it's about being able to measure exactly where a process falls within that class <ref:two thousand six hundred ten point two zero nine one six#pg3. That would be useful for statistical applications <ref:two thousand six hundred ten point two zero nine one six#pg3.

Ada: Exactly, Kolya. And they’re also trying to solidify the bounds on those activity indices using their specific constructions <ref:two thousand six hundred ten point two zero nine one six#pg3. They are showing that the entire possible range of beta- and beta+ can be reached with a simple process setup <ref:two thousand six hundred ten point two zero nine one six#pg3.

Emmy: It seems like they are moving from just proving something exists to building explicit, controllable examples for every parameter in the class <ref:two thousand six hundred ten point two zero nine one six#pg3. That’s a big step in terms of structural understanding <ref:two thousand six hundred ten point two zero nine one six#pg3.

Kolya: If they can show that the bounds survive when we apply it to something like a specific turbulence model, that makes the whole thing much more useful for real-world problems <ref:two thousand six hundred ten point two zero nine one six#pg3.

Conclusion: Ada: So we’ve just covered the core findings of "Infinite-variation Lévy processes are either strongly eroded or abrupt" <ref:2610.20916#pg1>. It boils down to a complete classification system based on how integrated variation behaves over different intervals <ref:2610.20916#pg3>.

Emmy: That’s right, Ada. We established that every real Lévy process of infinite variation must fit into one of two categories, either abrupt or strongly eroded <ref:2610.20916#pg3>. It ties the geometric smoothness of the path directly to these intrinsic properties <ref:2610.20916#pg3>.

Kolya: From a probabilistic standpoint, this means we have a way to filter out the infinite variation processes and know exactly what kind of structure they possess <ref:2610.20916#pg3>. It gives us concrete tools to analyze their jump activity <ref:2610.20916#pg3>.

Ada: Exactly, Kolya. And the robustness of this classification is what’s interesting for applications, especially in areas like Burgers turbulence where we look at rarefaction intervals <ref:2610.20916#pg3>. It tells us whether the paths have those specific properties or not <ref:2610.20916#pg3>.

Emmy: The ability to control the activity indices, those beta minus and beta plus parameters, is what makes this class so rich <ref:2610.20916#pg3>. It means we can construct processes with almost any jump behavior we want <ref:2610.20916#pg3>.

Kolya: I just wonder if those bounds on the activity indices actually hold up when you try to use them in a statistical context, like finding tail behaviors <ref:2610.20916#pg3>. We need those constants to stay finite for the application to work <ref:2610.20916#pg3>.

Ada: That’s a fair point, Kolya. And the authors did their best by showing that they can reach all those activity indices using simple constructions with zero drift and zero Gaussian coefficient <ref:2610.20916#pg3>. The paper lays out exactly what's possible in this space <ref:2610.20916#pg3>.

Emmy: It’s a strong result because it shows that even in the most complex infinite variation processes, there’s an underlying geometric simplicity that we can measure with these specific indices <ref:2610.20916#pg3>.

Kolya: So, to wrap up on this paper, "Infinite-variation Lévy processes are either strongly eroded or abrupt," we have a classification based on variation over intervals <ref:2610.20916#pg3>. It’s a solid framework for understanding the structure of these complex processes <ref:2610.20916#pg3>.

Ada: It really is, Kolya. And this structural understanding is going to be important when we look at how these processes interact with other stochastic systems <ref:2610.20916#pg3>.

Emmy: We’ll keep an eye on how this new classification informs the next set of research into these processes <ref:2610.20916#pg3>.

Department of Probability and Statistics, IIMAS--UNAM · Department of Mathematics, Aarhus University · Department of Statistics, University of Warwick & The Alan Turing Institute

math.PR

Submitted: 2026-10-14

Updated: 2026-11-12

Comments: 25 pages, short YouTube videos presenting the results: https://youtu.be/xWk_RV_JuSc?si=ITKFgS9ib39DonEK and their proofs: \href{https://youtu.be/SzlJq_Vi4xk?si=9TeAl5AjZw9BzFnJ

License: http://creativecommons.org/licenses/by-sa/4.0/

Importance score: 91/100

The gist: Every real Lévy process of infinite variation is either abrupt or strongly eroded The Core Dichotomy The paper proves that every real Lévy process of infinite variation is either abrupt or strongly…

Key concepts

Abruptness
This condition means the process has finite integrated local variation over any bounded time interval. It is equivalent to the integral of the path's variation being finite, which simplifies its geometric properties significantly.
Strongly Eroded Class
This class is much richer than previously thought. It includes processes with arbitrary admissible jump activity indices $eta^-$ and $eta^+$, meaning it covers a wide range of jump behaviors, not just simple atomic or singular measures.
Activity Indices ($eta^-, eta^+$)
These indices quantify the intensity of jumps in the Lévy process. The paper shows that strong erosion is compatible with essentially arbitrary values for these indices, specifically $eta^- o 0$ and $eta^+ o 2$, allowing for complex jump structures.
Smooth Transition Densities
This property means the process has very well-behaved probability density functions at all times. The paper shows that strongly eroded processes possess transition densities with bounded derivatives of every order, indicating a high degree of regularity.

Terminology

Summary

Every real Lévy process of infinite variation is either abrupt or strongly eroded

The Core Dichotomy

The paper proves that every real Lévy process of infinite variation is either abrupt or strongly eroded, resolving the conjectured geometric dichotomy on the smoothness of the convex hull of the graph of a Lévy path Theorem 1.1 states that exactly one of two conditions holds: either JX(I) < ∞ for every bounded interval I, or JX(I) = ∞ for every proper interval I, which is equivalent to X being either abrupt or strongly eroded

Characterizing Abruptness and Erosion

Abruptness is characterized by the condition JX(I) < ∞ for every bounded interval I, which is equivalent to s1 ∈ L1 loc(R) The paper shows that Vigon’s conjecture, which conjectured X is abrupt if and only if X(r) hits points for some r ∈ R, fails for symmetric pure-jump Lévy processes, with counterexamples showing that the set E = E[r ∈ R: s1(r) = ∞] can be a compact perfect set of Lebesgue measure zero

Richness of the Strongly Eroded Class

The strongly eroded class is shown to be much richer than previously thought, as Theorem 1.7 demonstrates that strong erosion is compatible with essentially arbitrary admissible jump activity indices β− ∈ [0, 1] and β+ ∈ [1, 2] This means strong erosion is not confined to atomic or singular jump measures The paper constructs such processes using absolutely continuous Lévy measures with strictly positive C∞ densities, showing that the process has bounded C∞ transition densities with bounded derivatives of every order at all positive times

Applications to Geometric Properties

The dichotomy has significant applications, including the classification of convex-minorant slopes and connections to Burgers turbulence Corollary 2.1 establishes that for an infinite variation process X, exactly one of two alternatives holds regarding its greatest L-Lipschitz minorant Ml, where X is either abrupt (Zl is locally finite almost surely for each l > m) or strongly eroded (Zl has zero Lebesgue measure but is not locally finite almost surely) Corollary 2. p. 4

Where the paper says it

Suppose that X has infinite variation, then exactly one of the following holds. (a) The process X is abrupt, Zℓ is locally finite almost surely for each ℓ > |m| and [ ℓ>|m| Zℓ = {t ∈ R : t is a local infimum of X} is countable almost surely.

2 further distinguishes the two classes in the context of Burgers turbulence, stating that for a fixed t > 0, either X is abrupt (At is locally finite and A0,t is regenerative) or strongly eroded (there are no rarefaction intervals and at is strictly increasing on R)

Sequential Gaussian Limits

The paper utilizes sequential Gaussian limits to construct strongly eroded processes, as seen in the construction of Orey’s process Proposition 5.1 provides a criterion that turns a bounded subsequence of the running maximum of the characteristic exponent into divergence of s1 at every drift This leads to the construction of processes with arbitrary activity indices β− ∈ [0, 1] and β+ ∈ [1, 2]

Counterexample Construction

The paper provides a detailed proof for the counterexample to Vigon’s conjecture, showing that there exist abrupt symmetric pure-jump Lévy processes for which s1(0) = ∞, while s1(r) < ∞ for r ∈ R — except possibly at r=0 This demonstrates that the failure of Vigon’s conjecture can be much richer than a single exceptional slope, with the exceptional set E potentially being uncountable and perfect The construction of such processes exploits lacunary sequences of jump sizes to control the finiteness properties of s1

Smooth Transition Densities

The construction in Theorem 1.3 specifies an infinite-variation Lévy process with finite variance where the Lévy measure is defined by a complex sum involving lacunary sequences and antisymmetric perturbations The analysis shows that for the lower index, at yn = 3zn/4, z(In) → ∞ as n → ∞, leading to s1(r) = ∞ for r ∈ C Furthermore, the construction ensures bounded C∞ transition densities with bounded derivatives of all orders for the strongly eroded processes The analysis of Q(u) shows that R R u j e − tQ(u)du 0 and j ≥ 0, ensuring bounded C∞ transition densities with bounded derivatives of all orders

The Full Activity-Index Region

The necessity of β− ≤ 1 ≤ β+ for strongly eroded processes is established in [7, Prop. 1.6], and the paper shows that the entire region can be attained by constructing a process X with density υ+, zero Gaussian coefficient, and zero drift parameter The activity indices are bounded as β+(X) = b+ and β−(X) = 0 for this construction The lower index is shown to be 0 because the cutoff excludes all bumps with k ≤ n, giving y p−2n V (yn) → 0 for any p > 0 The process Z = X + Y, where Y has finite variation, is strongly eroded by [7, Prop. p. 3

Where the paper says it

The symmetric Lévy process X with zero Gaussian coefficient, zero drift, and Lévy measure ν, where (7) bn = 2−η n and ν = c X n⩾1 b −α n (δbn + δ−bn ), is strongly eroded with s1 ≡ ∞ and its indices are β− = 0 and β+ = α. Strategy of the proofs and organisation of the remainder of the paper.

1.5], and the indices are shown to be max(b+, b−) = b+ for the upper index

The construction of smooth transition densities involves a symmetric C∞ density on R - (0) defined by a sum involving lacunary sequences and an even, non-negative function supported in [-1, 1] The analysis of Q(u) shows that R R u j e − tQ(u)du 0 and j ≥ 0, ensuring bounded C∞ transition densities with bounded derivatives of all orders This is achieved by showing that Q(u) log u → ∞ as u → ∞

The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by [7, Prop. 1.5], and the lower index is shown to be b− because r p−2W(r) → ∞ for p < b−, while along yn = 3zn/4, we have as n → ∞, y p−2n VZ(yn) = y p−2n V (yn) + y p−2n W(yn) → 0 The final result is that β−(Z) = b−

The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by [7, Prop. 1.5], and the lower index is shown to be b− because r p−2W(r) → ∞ for p < b−, while along yn = 3zn/4, we have as n → ∞, y p−2n VZ(yn) = y p-2n V (yn) + y p-2n W(yn) → 0 The final result is that β−(Z) = b−

The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by [7, Prop. 1.5], and the lower index is shown to be b− because r p−2W(r) → ∞ for p < b−, while along yn = 3zn/4, we have as n → ∞, y p-2n VZ(yn) = y p-2n V (yn) + y p-2n W(yn) → 0 The final result is that β−(Z) = b−

The construction of a process with lower index b− ∈ (0, 1] involves defining a modified density υ−(x) that preserves the super-logarithmic Fourier growth and hence the smooth transition densities The resulting process Z = X + Y is strongly eroded by [7, Prop. 1.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Enhanced classification of infinite variation Lévy processes: The system can now classify every real Lévy process of infinite variation as either abrupt or strongly eroded by checking if JX(I) is finite for every bounded interval I, which resolves the conjectured geometric dichotomy on the smoothness of the convex hull of a Lévy path.

  2. Verification against Vigon's conjecture: The system can now detect failures of Vigon’s point-hitting conjecture by checking if there exist abrupt symmetric pure-jump Lévy processes where s1(0) is infinite while s1(r) is finite for all r not equal to 0, or even cases where the exceptional set E = the set of slopes at which s1(r) = infinity is an uncountable perfect set of Lebesgue measure zero.

  3. Geometric and contact set analysis: The system can perform classification based on the nature of convex minorants, distinguishing between abrupt processes where dC'T is purely atomic with isolated atoms and strongly eroded processes where dC'T is diffuse, singular continuous and CT is continuously differentiable a.s., independent of the time horizon T.

  4. Characterization of strong erosion parameters: The system can determine the precise indices beta- and beta+ for strongly eroded processes by analyzing Pruitt and Blumenthal–Getoor indices, allowing it to construct or identify symmetric strongly eroded processes with arbitrary admissible activity indices b- in [0, 1] and b+ in [1, 2].

  5. Shock structure analysis for Burgers turbulence: The system can distinguish between the abrupt case (where the set At is locally finite and A0,t is regenerative) and the strongly eroded case (where there are no rarefaction intervals and at is strictly increasing on R) when analyzing variational solutions of inviscid Burgers equations driven by Lévy noise.

  6. Detection of smooth transition densities: The system can verify that strongly eroded processes possess bounded C-infinity transition densities with bounded derivatives of all orders by checking the divergence estimates for Q(u) log u which tend to infinity as u approaches infinity.

Abstract

We prove that every real Lévy process of infinite variation is either abrupt or strongly eroded, resolving the conjectured geometric dichotomy on the smoothness of the convex hull of the graph of a Lévy path [Ann. Inst. H. Poincaré Probab. Stat. 60(4) (2024), 2352-2381]. We give counterexamples to the stronger point-hitting conjecture of Vigon. As applications, we classify convex-hull smoothness, Lipschitz-minorant contact sets and Burgers shock geometry.

Related papers