Denseness in total variation and the class of rational-infinitely divisible distributions

Basic Information

• Paper ID: 2510.24219
• Title: Denseness in total variation and the class of rational-infinitely divisible distributions
• Author: A. A. Khartov
• Classification: math.PR (Probability Theory)
• Publication Date: October 29, 2025
• Paper Link: https://arxiv.org/abs/2510.24219

Abstract

This paper investigates a new class of probability distributions on the real line called rational-infinitely divisible (or quasi-infinitely divisible) distributions. The characteristic functions of these distributions are ratios of characteristic functions of classical infinitely divisible distributions and admit Lévy-Khinchine-type representations with "signed spectral measures." This class of distributions is quite extensive and possesses many desirable properties. For instance, this class is dense in the weak convergence sense within the family of all (univariate) probability distributions. The paper addresses the question of denseness of this class in the total variation convergence sense, considering different types of probability distributions and their supports. A series of "positive" and "negative" results are obtained.

Research Background and Motivation

1. Problem to be Addressed: This paper investigates the denseness of the class of rational-infinitely divisible distributions Q in the total variation convergence sense. It is known that class Q is dense in the weak convergence sense within the family of all probability distributions F, but its denseness in the stronger total variation convergence sense has not been studied.
2. Importance of the Problem:
   • Infinitely divisible distributions constitute one of the most important classes in probability theory, with profound theoretical significance and broad applications
   • The class of rational-infinitely divisible distributions Q is an important generalization of the class of infinitely divisible distributions I, retaining many desirable properties
   • Total variation convergence is stronger than weak convergence; studying the denseness of Q in the total variation sense has important theoretical value
3. Limitations of Existing Methods:
   • The classical class of infinitely divisible distributions I is closed under weak convergence and cannot approximate non-infinitely divisible distributions
   • It is known that Q is dense in weak convergence, but the denseness problem under total variation convergence remains unresolved
   • Systematic research on different types of distributions (discrete, absolutely continuous, etc.) is lacking
4. Research Motivation: To explore the denseness properties of class Q under stronger convergence modes and provide new theoretical results for probability distribution theory.

Core Contributions

1. Established a theoretical framework for denseness of rational-infinitely divisible distributions in the total variation sense, distinguishing "positive" and "negative" results
2. Proved denseness for absolutely continuous distributions: $Q_a$ is dense in $F_a$ with respect to total variation
3. Proved denseness for discrete lattice distributions: $Q_d^{lat}$ is dense in $F_d^{lat}$ with respect to total variation
4. Proved denseness for mixed distributions: Mixed classes of lattice-discrete and absolutely continuous distributions possess denseness
5. Revealed non-denseness for non-lattice discrete distributions: $Q_d$ is not dense in $F_d^{nl}$, and consequently Q is not dense in F
6. Provided complete technical tools: Including key lemmas and constructive proof methods

Detailed Methodology

Task Definition

Study the denseness of the class of rational-infinitely divisible distributions Q and its subclasses in different families of probability distributions with respect to total variation convergence. Specifically includes:

• Input: Probability distribution F
• Output: Determine whether there exists a sequence in Q converging to F in the total variation sense
• Constraints: Consider the support and type of distributions (discrete, continuous, etc.)

Core Concepts and Definitions

Rational-Infinitely Divisible Distribution: A distribution function F is called rational-infinitely divisible if there exist infinitely divisible distribution functions $F_1$ and $F_2$ such that $F_1 = F * F_2$. Equivalently, its characteristic function satisfies $f(t) = f_1(t)/f_2(t)$.

Lévy-Khinchine-Type Representation: $f(t) = \exp\left\{it\gamma + \int_{\mathbb{R}} \left(e^{itx} - 1 - \frac{it\sin(x)}{1+x^2}\right)\frac{1+x^2}{x^2} dG(x)\right\}$ where G is a bounded variation function (possibly non-monotone).

Main Technical Tools

Lemma 1 (Modified Ibragimov): Let $F_0$ have bounded support, $\gamma_0 \neq c_{ext}(F_0)$ (when $F_0$ is shifted symmetric). Then for almost all $\delta \in (0,1)$, $\delta e^{it\gamma_0} + (1-\delta)f_0(t) \neq 0, \quad \forall t \in \mathbb{R}$

Lemma 2 (Kutlu's Result): The function $\phi(t_1,t_2) = \frac{1}{3}e^{it_1} + \frac{1}{3}e^{it_2} + \frac{1}{3}e^{i(t_1+t_2)}$ cannot be uniformly approximated on $[-\pi,\pi]^2$ by zero-free continuous functions.

Lemma 3 (Recent Result): If $F = c_d F_d + (1-c_d)F_c$ with $F \in Q$ and $c_d > 0$, then $\inf_{t \in \mathbb{R}}|f_d(t)| > 0$.

Proof Strategies

Proof Strategy for Positive Results:

1. Convolution approximation: Construct continuous Bernoulli distributions $B_q$ for convolution
2. Apply Lemma 1: Ensure the characteristic function of the mixed distribution is zero-free
3. Total variation estimates: Control approximation error through $L^1$ norm

Proof Strategy for Negative Results:

1. Counterexample construction: Use irrational numbers to construct non-lattice discrete distributions
2. Linear independence: Utilize basis representation and periodic function theory
3. Proof by contradiction: Combine Lemmas 2 and 3 to derive contradiction

Experimental Setup

This paper is purely theoretical research with no numerical experiments. All results are obtained through rigorous mathematical proofs.

Theoretical Verification Methods

• Constructive proofs: Provide explicit construction methods for positive results
• Counterexample construction: Construct specific counterexamples for negative results
• Lemma application: Systematically apply existing theoretical results

Main Results

Positive Results

Theorem 1: Let S be a subset of $\mathbb{R}$ such that $F_a(S) \neq \emptyset$. Then for any $\tau > 0$, the classes $Q_a([S]_\tau^+)$ and $Q_a([S]_\tau^-)$ are dense in $F_a(S)$ with respect to total variation.

Corollary 1: $Q_a$ is dense in $F_a$ with respect to total variation.

Theorem 2: Let $L \subset \{a + bk : k \in \mathbb{Z}\}$ be non-empty. Then $Q_d(L)$ is dense in $F_d(L)$ with respect to total variation.

Corollary 2: $Q_d^{lat}$ is dense in $F_d^{lat}$ with respect to total variation.

Theorem 3: The class $Q_{d,a}^{mix}(L,S)$ is dense in $F_d(L) \oplus F_a(S)$ with respect to total variation.

Negative Results

Theorem 4: $Q_d$ is not dense in $F_d^{nl}$ with respect to total variation.

Theorem 5: Q is not dense in $F_d^{nl}$ with respect to total variation.

Theorem 6: Q is not dense in F with respect to total variation.

Key Counterexample

Construct the distribution: $F(x) = \frac{1}{3}1_1(x) + \frac{1}{3}1_\alpha(x) + \frac{1}{3}1_{1+\alpha}(x)$ where $\alpha$ is a positive irrational number. Prove that no sequence in $Q_d$ converges to F in the total variation sense.

Related Work

Infinitely Divisible Distribution Theory

• Lévy-Khinchine formula and its applications
• Self-decomposable and stable distributions
• Applications in stochastic processes

Quasi-Infinitely Divisible Distributions

• Pioneering work by Lindner et al. 16
• Discrimination criteria for discrete distributions 1,8
• Extensions to mixed distributions 3,4,10

Denseness Theory

• Denseness results under weak convergence
• Negative results for multivariate case by Kutlu 13
• Denseness under uniform convergence

Conclusions and Discussion

Main Conclusions

1. Classification Completeness: Completely characterize the total variation denseness of class Q in different distribution families
2. Clear Positive and Negative Results: Absolutely continuous and lattice discrete distributions possess denseness, while non-lattice discrete distributions do not
3. Technical Contributions: Provide systematic proof techniques and counterexample construction methods

Limitations

1. Multivariate Generalization: Results are limited to the univariate case; the multivariate case is more complex
2. Constructiveness: The construction of positive results is quite involved with limited practical applications
3. Boundary Cases: Characterization of certain boundary cases lacks refinement

Future Directions

1. Multivariate Extension: Study denseness of multivariate rational-infinitely divisible distributions
2. Convergence Rates: Investigate convergence rates and error bounds of approximations
3. Application Exploration: Seek practical applications of theoretical results

In-Depth Evaluation

Strengths

1. Theoretical Completeness: Systematically resolves an important theoretical problem with complete and profound results
2. Technical Innovation: The modification and application of Lemma 1 demonstrates sophisticated technical skill
3. Rigorous Proofs: All proofs are rigorous and complete with clear logic
4. Clear Structure: The classification of positive and negative results provides a clear theoretical landscape

Weaknesses

1. Limited Practical Utility: Pure theoretical results lacking direct practical applications
2. High Technical Threshold: Requires deep background in probability theory and real analysis
3. Difficult Generalization: Generalization to the multivariate case faces fundamental difficulties

Impact

1. Theoretical Contribution: Provides important new results for probability distribution theory
2. Methodological Value: Proof techniques serve as reference for related problems
3. Foundation for Future Research: Establishes foundation for further investigation

Applicable Scenarios

• Theoretical probability research
• Infinitely divisible distribution theory
• Distribution approximation theory
• Stochastic process theory

References

The paper cites 23 important references covering classical and recent results in infinitely divisible distributions, quasi-infinitely divisible distributions, real analysis, and other related fields, reflecting the author's deep understanding and solid foundation in relevant areas.