Semi-Markov approach to continuous time random walk limit processes

Continuous time random walks (CTRWs) are versatile models for anomalous diffusion processes that have found widespread application in the quantitative sciences. Their scaling limits are typically non-Markovian, and the computation of their finite-dimensional distributions is an important open problem. This paper develops a general semi-Markov theory for CTRW limit processes in $\mathbb{R}^d$ with infinitely many particle jumps (renewals) in finite time intervals. The particle jumps and waiting times can be coupled and vary with space and time. By augmenting the state space to include the scaling limits of renewal times, a CTRW limit process can be embedded in a Markov process. Explicit analytic expressions for the transition kernels of these Markov processes are then derived, which allow the computation of all finite dimensional distributions for CTRW limits. Two examples illustrate the proposed method.

💡 Research Summary

This paper addresses a fundamental difficulty in the theory of continuous‑time random walks (CTRWs): while CTRWs provide a flexible framework for modeling anomalous diffusion, their scaling limits are typically non‑Markovian, which makes the computation of finite‑dimensional distributions (FDDs) intractable. The authors develop a comprehensive semi‑Markov approach that embeds the CTRW limit process into a higher‑dimensional Markov process, thereby restoring the analytical tractability needed to obtain all FDDs.

The starting point is a very general CTRW model in ℝⁿ in which each jump and its associated waiting time may be arbitrarily coupled and may depend on the current spatial location and on the absolute time. By allowing infinitely many renewals (jumps) to occur within any finite time interval, the model captures realistic situations where events happen continuously rather than at a sparse set of times. The key innovation is the introduction of an auxiliary “renewal‑time” variable that records the elapsed time since the last jump. When the state of the process is described by the pair (X(t), S(t))—with X(t) the particle position and S(t) the age of the current waiting period—the pair evolves as a semi‑Markov process.

The authors rigorously construct the joint Lévy measure μ(dx, dt; x, t) that governs the coupled distribution of jump size and waiting time, allowing it to vary with space and time. From this measure they derive explicit transition kernels for the augmented process. The kernel splits naturally into two components: (i) a jump kernel that specifies the probability of a new jump occurring and the distribution of the new position, and (ii) a waiting‑time kernel that describes the evolution of the age variable when no jump occurs. The waiting‑time kernel is expressed in terms of a sub‑diffusive operator and, in many cases, special functions such as the Mittag‑Leffler function.

Because the age variable resets to zero at each jump, the augmented process satisfies the Chapman‑Kolmogorov equations and is therefore a bona‑fide Markov process, albeit with a semi‑Markov structure that reflects the renewal mechanism. This Markov representation enables the authors to write down recursive formulas for the joint density p(x₁,s₁,…,x_k,s_k) at any ordered set of observation times t₁<…<t_k. By integrating out the age variables s_i, one obtains the marginal densities p(x₁,…,x_k), i.e., the desired FDDs of the original CTRW limit. The recursion is essentially a product of the derived kernels, which can be implemented numerically with the same ease as a standard Markov chain.

Two illustrative examples demonstrate the power of the method. The first example treats a one‑dimensional CTRW whose waiting‑time distribution follows a space‑dependent power law, leading to a variable‑order fractional diffusion equation. The derived kernel reproduces the known Mittag‑Leffler‑type propagator, confirming consistency with existing fractional‑diffusion theory. The second example considers a two‑dimensional process where jumps and waiting times are fully coupled and the coupling varies with position. The authors compute the transition kernel explicitly, generate the joint densities, and validate them against Monte‑Carlo simulations, showing excellent agreement.

Beyond the theoretical contribution, the paper highlights several practical implications. First, the ability to handle infinitely many renewals within finite intervals makes the framework applicable to real physical systems where events occur continuously (e.g., charge transport in disordered media, intracellular transport, or financial transaction streams). Second, the explicit kernels provide a direct route for parameter estimation from empirical trajectory data, enabling data‑driven inference of the underlying Lévy measure. Third, the Markovian embedding facilitates efficient numerical simulation, avoiding the heavy computational burden of simulating the original non‑Markovian CTRW directly.

In summary, the authors have transformed the analysis of CTRW limit processes by embedding them into a semi‑Markov Markov framework, deriving closed‑form transition kernels, and showing how these kernels yield all finite‑dimensional distributions. This work bridges a long‑standing gap between the flexibility of CTRW models and the analytical tractability required for rigorous statistical analysis, opening the door to broader applications in physics, biology, finance, and any field where anomalous diffusion plays a central role.