Classification for dynamics of Markov chains on non-negative integers with arbitrary transition rates and its application

Minjun Kim, Seokhwan Moon, Jinsu Kim · 2025

Read Classification for dynamics of Markov chains on non-negative integers with arbitrary transition rates and its application by Minjun Kim, Seokhwan Moon, Jinsu Kim (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-20000.txt.

Introduction

Minjun Kim, Seokhwan Moon, and Jinsu Kim classify the long-term behavior of continuous-time Markov chains that count upward from zero. Such chains model biological systems, populations, and queues, and birth-and-death chains are the best-known examples. Earlier criteria covered polynomial transition rates, while this work allows arbitrary rate functions.

The classification is written in terms of how the transition rates grow, and it decides properties such as explosivity, recurrence, and exponential ergodicity. For rates with certain expansion forms, including all rational functions, the authors give exact if-and-only-if conditions. Only three easy quantities are needed: the top degree and the mean and variance of the rates.

Catalog record

CREATORMinjun Kim, Seokhwan Moon, Jinsu Kim
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-20000

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Classification for dynamics of Markov chains on non-negative integers with arbitrary transition rates and its application Minjun Kim, Seokhwan Moon, Jinsu Kim Abstract Continuous-time Markov chains on non-negative integers can be used for modeling biological systems, population dynamics and queueing models. Qualitative behaviors of birth-and-death models, typical examples of such a one-dimensional continuous-time Markov chains, have been substantially studied. For one-dimensional Markov chains with polynomial transition rates, recent studies provided criteria for their long-term behavior. In this paper, we provide a classification of Markov chains on non-negative integers when the transition rates are arbitrary functions. The criteria are written with asymptotics of the transition rates. This classification implies their dynamical properties, including explosivity, recurrence, positive recurrence, and exponential ergodicity. As an application, we derive a complete classification (if and only if conditions) for those dynamical features when the transition rates have certain expansion forms, which include all rational functions, so that our classifications cover mass-action kinetics, Michaelis–Menten kinetics, and Haldane equations. Our classification solely relies on easily computable quantities: the maximal degree of the expansion of the transition rates, the mean and the variance of the transition rates. We demonstrate the utility of this classification framework using the approximation of high-dimensional mass-action systems by a one-dimensional reaction system with general rational kinetics. 1 Introduction One-dimensional continuous-time Markov chains on non-neg- ative integers have been widely used to model population growth in various biological and ecological systems, particularly describing the time evolution of copy numbers of proteins, bacteria, cells, and chemical species [ 1 , 2 , 3 ] . Thanks to their simplicity, these models are particularly useful and commonly applied. Typical examples include birth-and-death processes. They are sophisticated enough to serve as powerful tools for studying the copy numbers of biochemical species, yet simple enough to allow detailed analyses of their qualitative and quantitative behaviors. However, when the Markov chain allows multiple jumps with arbitrary jump sizes, their qualitative properties such as positive recurrence and exponential ergodicity have not been fully characterized for general transition rates. In this paper, we provide simple criteria for determining various dynamical features of continuous-time Markov chains on non-negative integers. When the transition rates of these stochastic processes are rational functions, our criteria yield a complete classification based on their dynamical properties. A critical motivation of our work is that one-dimensional Markov chains with non-polynomial transition rates can often be derived as reductions of higher-dimensional Markov chains governed by mass-action kinetics, which are fundamental polynomial laws for biochemical reaction systems. A typical example is the reduction of an enzyme–substrate model to the Michaelis–Menten kinetics model via a quasi–steady-state approximation [ 4 ] . Such time-scale separation–based model reductions can lead to more general one-dimensional Markov chains with non-polynomial transition rates. For instance, consider two biochemical species that catalyze each other under two distinct timescales described by the following biochemical reaction network: \displaystyle\emptyset\to 2A,\quad B+2A\to B+A,\quad B+A\to B \displaystyle\text{(slow reactions)}, \displaystyle\emptyset\to B,\quad B+2A\to 2A,\quad B\to\emptyset \displaystyle\text{(fast reactions)}. Here, the scaled copy number of B rapidly converges, in probability, to values that can be expressed as a function of A . Consequently, the dynamics of B can be approximated by a one-dimensional process involving the reactions A\to\emptyset\to 2A , whose transition rates are rational functions (see Section 6.4 for details). Due to such an utility of one-dimensional Markov models, previous studies focused qualitative behaviors of one-dimensional Markov models with non-polynomial transition rates [ 5 , 6 , 7 ] . Several previous studies have investigated one-dimensional Markov chains obtained through such model reductions. In [ 1 ] , an autoregulated bursty gene expression model was analyzed via reduction to a one-dimensional Markov chain. In [ 2 ] , a reduced one-dimensional stochastic model for an enzyme–substrate system was used to derive detailed dynamical properties, including a closed-form solution of the chemical master equation and transient bimodality. Methods for reducing various biochemical systems to one-dimensional models were proposed in [ 3 ] , with examples including bursty gene expression, receptor–ligand interactions, and the flagellar motor of E. coli . Furthermore, it was shown that a high-dimensional Markov chain can be bounded by a coupled one-dimensional Markov chain, and this coupling approach was applied to verify ergodicity of the original system [ 8 ] . Notably, most of the resulting one-dimensional models in these studies admit non-polynomial transition rates. There also exist systematic approaches to model reduction for biochemical reaction systems. In [ 9 ] , the validity of the stochastic quasi–steady-state approximation for a broad class of reaction systems was established. Model reduction techniques employing intermediate species were further developed in [ 10 , 11 ] . Given the broad applicability of one-dimensional Markov models, understanding their long-term qualitative behavior is essential. Various properties, such as explosivity, recurrence, positive recurrence, and exponential ergodicity, have been analyzed for certain classes of one-dimensional population models, providing valuable insights into their stability and dynamics [ 12 , 6 ] . However, the applicability of these previous studies is limited. The criteria for ergodicity and exponential ergodicity proposed in [ 12 , Theorems 3.16, 4.52, and 4.54] and [ 6 ] are valid only for unit-birth processes. Moreover, the criteria for exponential ergodicity given in [ 6 ] are not necessary conditions unless the underlying process is a unit birth-and-death process. In other words, those results are confined to Markov chains where upward transitions are limited to births of size one. For models with multiple birth sizes, criteria for various dynamical features were provided in [ 7 ] . These criteria can be easily checked using the degrees of the transition rates and the corresponding transition vectors. Notably, in that work, the transition rates were restricted to polynomial functions. Therefore, the applicability of such results remains limited, since, as discussed earlier, many realistic biological models can be represented by one-dimensional Markov chains whose transition rates are rational or more general non-polynomial functions. In this paper, we provide general criteria for determining the dynamical features of one-dimensional Markov chains on non-negative integers. Notably, our conditions are valid for arbitrary types of transition rates and for any finite set of possible jump sizes. The dynamical properties of interest include explosivity, transience, null recurrence, positive recurrence, and exponential ergodicity. We first establish general criteria on the arbitrary transition rates that can be used to verify these dynamical features. The key idea underlying our approach is to generalize the classical analysis of simple birth-and-death models, whose dynamical properties can be determined by elementary calculations involving the ratio of the birth and death rates. To achieve this generalization, we primarily employ Lyapunov function techniques. We then apply these general conditions to derive a complete classification of one-dimensional Markov chains whose transition rates \lambda_{i}(x) ’s admit a Laurent-type expansion : there exists R\in\mathbb{Z} such that \displaystyle\lambda_{i}(x)=a_{i}x^{R}+b_{i}x^{R-1}+O(x^{R-2})\quad\text{for each $i$}, (1) The exponent R may be nonpositive, so this class of transition functions includes both polynomial and rational functions. Hence, our classification extends the previous results in [ 7 ] . These classification criteria for the Laurent-type expansion (Theorem 5.1 ) can be checked using four easily computable indices, \alpha , \gamma , \vartheta , and R (see ( 25 )), in a manner analogous to the criteria given

Read the complete text