A new theoretical study published in the open-access journal Results in Physics offers a unified mathematical framework for predicting not only when stochastic systems fail, but also which of several competing failure mechanisms ultimately destroys them. The paper, authored by Santosh Kudtarkar and appearing in Volume 87 of the journal, addresses a long-standing gap in nonequilibrium statistical physics: the simultaneous treatment of catastrophic, nonlocal jumps, imperfect absorption, and mode-resolved failure selection within a single exactly solvable formalism. The work arrives at a moment of growing interest in first-passage phenomena, the branch of probability theory and physics concerned with the question of how long a stochastic process survives before it first reaches a target, a boundary, or a state of terminal failure.
First-passage problems are among the most fundamental constructs in statistical physics. They describe escape over energy barriers, the firing of neurons, the nucleation of droplets, the spread of populations, the extinction of species, and the failure of load-bearing materials. In the textbook setting, the dynamics is local, typically a random walk with small incremental steps, and the boundary acts as a perfect trap: touch it once, and the process ends. The new study argues that this classical picture is inadequate for a broad class of modern problems, from fracture and crackling dynamics in materials to ecological populations battered by rare environmental shocks. In such systems, single catastrophic events can remove a finite fraction of the system’s state in one leap, jumping across many intermediate configurations without ever visiting them, while the nominal failure region often behaves not as a perfect absorber but as a partially reactive boundary from which trajectories can be rescued.
The framework developed in the paper is built on the concept of a killed generator, a matrix object that governs the evolution of a continuous-time Markov chain from which probability mass is continually removed by absorption. The state variable can be interpreted as a resource level, a population size, or the residual load-bearing capacity of an engineered component. The transient region lies above a critical threshold, while the absorbing set below that threshold is partitioned into distinct failure modes, each representing a different terminal channel, such as shallow functional failure versus catastrophic rupture. Catastrophic downward jumps follow a heavy-tailed Zipf–Mandelbrot distribution, a discrete, regularized analogue of Pareto and Lévy statistics whose tail exponent is allowed to vary with the state of the system, thickening dangerously as the process approaches the failure threshold. Recovery is modeled through upward jumps, either one-step or geometric bursts, that compete against the catastrophic losses.
The crucial technical innovation is the treatment of absorption as genuinely imperfect and mode-resolved. When a catastrophe lands the system inside a failure region, absorption occurs with a probability of the form one minus the exponential of the negative product of an absorption strength and the jump size, so that deeper penetrations below the threshold are more likely to be terminal than shallow excursions. If absorption does not occur, the process is rescued back into the transient region by a stochastic kernel, allowing temporary sub-threshold excursions without ending the dynamics. This mechanism captures near-miss events in reliability engineering, temporary population crashes that do not cause extinction, and partially reactive boundaries in diffusion-controlled chemistry. The construction satisfies an exact bookkeeping identity, equating the negative row sums of the killed generator with the sum of mode-specific killing vectors, which casts the entire problem as a competing-risks decomposition embedded directly in the generator formalism.
From this matrix representation, the paper derives exact operator formulas for every observable of interest. Survival probabilities are expressed as matrix exponentials of the killed generator; cause-specific first-passage-time densities appear as the exponential acting on individual killing vectors; mean first-passage times and splitting probabilities reduce to linear solves involving the inverse of the generator. Because the matrices involved are finite, these expressions are exact for the truncated model rather than approximate. They also connect first-passage theory to mature numerical linear algebra, and the author validates the entire machinery against statistically exact Gillespie simulations, the event-driven Monte Carlo algorithms familiar from chemical kinetics. Agreement between the analytic matrix predictions and the stochastic simulations is demonstrated for survival curves, first-passage-time densities, mode-specific densities, splitting probabilities, and the full mode-steering sweeps, confirming that subtle implementation details such as overshoot clipping and rescue mapping are handled consistently.
Perhaps the most conceptually significant results concern the long-time behavior. Conditioning the process on non-absorption leads to a quasi-stationary distribution, the limiting occupancy profile of a system that has survived long enough to reach a metastable conditional equilibrium. Under standard Perron–Frobenius theory for Metzler matrices, the principal eigenvalue of the killed generator governs an exponential survival tail, and the hazard function, the instantaneous failure rate, converges to a plateau equal to this principal decay rate. The paper proves an elegant identity: the principal decay rate is exactly the total killing rate averaged over the quasi-stationary distribution. Consequently, the asymptotic fractions of failures occurring through each competing mode become time-independent and are given simply by the normalized quasi-stationary-averaged killing rates. Failure-mode selection thus emerges not as a short-time transient accident but as a robust property of the conditioned, metastable dynamics.
The quantitative consequences of catastrophe-tail thickness form a central theme of the parameter studies. Sweeping the tail exponent that controls the heaviest catastrophic jumps near threshold, the study finds that thickening the tail sharply shortens mean first-passage times, raises the principal decay rate, and weakens the spectral separation between the dominant leakage mode and the first subleading relaxation mode. In physical terms, rare but enormous jumps increase the overall failure rate while simultaneously eroding the clean two-stage structure, rapid conditional relaxation followed by slow exponential leakage, that makes metastable systems analytically tractable and experimentally recognizable. The hazard crossover broadens, and the long-time balance between absorbing channels shifts, demonstrating that the statistics of extreme events determine not only when failure occurs but also which failure pathway is ultimately selected.
A deliberate stress test probes the limits of the quasi-stationary picture by varying the recovery geometry. Reducing the parameter that weights large recovery bursts suppresses effective repair and can drive the spectral-separation ratio from a strongly metastable regime, where it exceeds one hundred, down to values of order unity. Importantly, the paper distinguishes between mathematical validity and practical observability: the asymptotic formulas remain exact whenever the principal eigenvalue is simple, but in the weak-separation regime the plateau becomes visible only after a substantial fraction of trajectories has already been absorbed. The study also emphasizes that the mean first-passage time can be misleading as a summary statistic. Using the coefficient of variation, the author shows that in strongly metastable regimes the first-passage-time distribution is broad, with the mean substantially exceeding the most probable absorption time, whereas weak recovery geometry produces sharply focused, more predictable failure events.
The mode-steering results carry perhaps the clearest practical message. By varying the absorption strength of one failure mode while holding the other fixed, the study demonstrates that imperfect absorption acts as a genuine control parameter, monotonically redistributing splitting probabilities between channels. The effect is stronger in thick-tail regimes, where frequent large catastrophes engage the size-dependent absorption rule more directly. Notably, the absorption strength steers the composition of failure much more powerfully than it alters the total lifetime, which remains dominated by the transport regime. Robustness checks comparing deterministic rescue with randomized near-threshold rescue show nearly indistinguishable results, indicating that the reported regime map reflects genuine physics rather than artifacts of a particular modeling choice. Throughout, the physical interpretation is grounded in damage accumulation: upward jumps represent healing, stress redistribution, or repair, downward jumps represent avalanches of microcrack growth or acoustic-emission bursts, and the absorption strength measures the fragility or irreversibility of each failure mode.
The broader significance of the work lies in its synthesis of previously separate research strands: classical first-passage theory, quasi-stationary analysis of absorbing Markov chains, partially reactive boundary problems, and nonlocal Lévy-flight transport. By working directly at the level of the continuous-time Markov chain generator on a finite state space, the framework avoids the known failures of continuum approximations near critical regimes and in the tails of extinction-time distributions, where matching only local increment statistics is insufficient. The exact special-function representations derived for the killing rates, expressed through Hurwitz-zeta and Lerch functions, together with explicit truncation-error bounds for the heavy-tailed kernel, provide practical tools for future applications. As extreme events assume growing importance in climate science, materials reliability, and population biology, the ability to compute, exactly and mode-resolved, how catastrophe statistics and boundary fragility jointly shape failure represents a substantial contribution to the statistical physics of survival.
Subject of Research: Mode-resolved first-passage theory in catastrophe processes with nonlocal jumps, imperfect absorption, and quasi-stationary failure-mode selection
Subject of Research: Technology and Engineering
Article Title: Mode-resolved first passage in catastrophe processes with imperfect absorption
Article References: Kudtarkar, S. (2026). Mode-resolved first passage in catastrophe processes with imperfect absorption. Results in Physics, 87, Article 108698. https://doi.org/10.1016/j.rinp.2026.108698
Image Credits: AI Generated
DOI: 10.1016/j.rinp.2026.108698
Keywords: first-passage theory, killed generator, quasi-stationary distribution, catastrophe processes, imperfect absorption, heavy-tailed jumps, competing risks, metastability, splitting probabilities, continuous-time Markov chains
Cite Scienmag News
APA MLA Chicago
Denise Maddox. (September 5, 2026). How imperfect absorption shapes first passage in catastrophe processes. Scienmag. https://scienmag.com/how-imperfect-absorption-shapes-first-passage-in-catastrophe-processes/
Copy citation Download RIS
Tags: boundary conditions in first-passage phenomenaboundary conditions in stochastic systemscatastrophe modeling in statistical physicscatastrophe processes in statistical physicsenergy barrier escape dynamicsexactly solvable models for failureexactly solvable models for failure predictionfailure dynamics in complex systemsfailure mechanisms in load-bearing materialsfirst-passage phenomena in complex systemsimperfect absorption in first-passage processesmathematical framework for catastrophic processesmode-resolved failure analysismode-resolved failure mechanismsnonequilibrium statistical physicsnonlocal jump dynamicsnonlocal jumps in stochastic systemsrole of imperfect absorption in system failurestochastic failure mechanismsstochastic failure predictionsurvival probability in stochastic processes
