Linear Analysis of Stochastic Verlet-Type Integrators for Langevin Equations
Niels Grønbech-Jensen · 2025
Read Linear Analysis of Stochastic Verlet-Type Integrators for Langevin Equations by Niels Grønbech-Jensen (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-2505-04100.txt.
Introduction
Niels Grønbech-Jensen grades step tools in a 2025 paper for shake-and-drag sums. Such sums track how grains drift and jig in a bath, a base chore in grain lore. Step tools march the path in set ticks, and each tool bends the truth in its own way.
His frame scores any tool of one wide clan on three plain acts: spread on a flat floor, drift on a slope, and rest in a bowl. Closed sums give each score by tick size. Twelve tools from five decades are ranked by that light. Grain teams gain a straight scale for tools long picked by habit.
Catalog record
| CREATOR | Niels Grønbech-Jensen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-04100
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Linear Analysis of Stochastic Verlet-Type Integrators for Langevin Equations Niels Grønbech-Jensen Abstract We provide an analytical framework for analyzing the quality of stochastic Verlet-type integrators for simulating the Langevin equation. Focusing only on basic objective measures, we consider the ability of an integrator to correctly simulate two characteristic configurational quantities of transport, a) diffusion on a flat surface and b) drift on a tilted planar surface, as well as c) statistical sampling of a harmonic potential. For any stochastic Verlet-type integrator expressed in its configurational form, we develop closed form expressions to directly assess these three most basic quantities as a function of the applied time step. The applicability of the analysis is exemplified through twelve representative integrators developed over the past five decades, and algorithm performance is conveniently visualized through the three characteristic measures for each integrator. The GJ set of integrators stands out as the only option for correctly simulating diffusion, drift, and Boltzmann distribution in linear systems, and we therefore suggest that this general method is the one best suited for high quality thermodynamic simulations of nonlinear and complex systems, including for relatively high time steps compared to simulations with other integrators. Journal of Statistical Physics 193 , 12 (2026) https://doi.org/10.1007/s10955-025-03553-3 I Introduction The prevailing numerical methods for time-dependent computational statistical mechanics and molecular dynamics are based on the simple and efficient Newton-Størmer-Verlet discretization of Newton’s equation of motion. The reason is that this framework strikes a seemingly optimal balance between accuracy, efficiency, and simplicity in addition to exhibiting critical conservation properties that ensure consistency throughout long simulations. While this very simple temporal discretization had been introduced by Newton (see, e.g., Ref. [ 1 ] ), and later further explored by Størmer (see, e.g., Ref. [ 2 ] ), the usefulness for modern simulations grew dramatically after Verlet’s re-introduction of this configurational method in 1967 [ 3 ] , and the method has since then been commonly referred to as the Verlet method. The next few years produced augmented algorithms that possessed the same configurational evolution, but with the addition of central difference approximations to the corresponding velocity variable at either the time step of the configurational coordinate (on-site) [ 4 , 5 ] or at the half-step between two time steps of the coordinate [ 6 , 7 ] , the former typically denoted the velocity-Verlet method and the latter the leap-frog method. A formulation that naturally produces both half-step and on-site velocities is shown in Ref. [ 8 ] . Applications and implementations, especially those relating to molecular modeling, are detailed in, e.g., Refs. [ 9 , 10 , 11 , 12 , 13 ] . Regardless of the formulations that include different velocity approximations, the underlying discrete-time configurational evolution is the same Verlet method [ 3 ] . For Langevin modeling, where Newton’s equation of motion is revised to satisfy the fluctuation-dissipation balance with an imaginary heat-bath, the Verlet method has similarly been revised to include friction and thermal noise contributions in discrete time. While the Verlet method for Newtonian deterministic evolution contains only the selected time step as a free parameter, incorporating friction and noise into the discrete-time Verlet framework can be done in a vast number of seemingly similar ways through different derivation strategies. This leads to a large number of stochastic numerical integrators (thermostats) with different properties, as exemplified by the descriptions in Refs. [ 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 ] , which contain the subset of integrators specifically analyzed in this work. Given the number of available integrators, which all converge to the same (correct) behavior in the small time step limit, but all diverge from each other as the time step is increased, valuable contributions to the literature have been the occasional publications that seek to compare algorithm performance of a few select integrators of interest (see, e.g., Refs. [ 28 , 29 , 30 , 31 ] ), thereby providing some insight to which integrator to choose. Such investigations often select a model problem from a complex system, such as a molecular dynamics ensemble, and use the results to draw conclusions. Comparisons are typically made for both configurational and kinetic coordinates, extracting relevant thermal statistics of potential and kinetic energies, diffusion constants obtained from both configurational and kinetic evolution, etc. Yet, the mutual dependence between configurational and kinetic coordinates that exists in continuous time does not exist in discrete time, as evidenced by, e.g., the Verlet algorithm mentioned above. Indeed, the same configurational evolution can be obtained with or without a velocity variable, and many different velocity variables can be defined to accompany any given configurational trajectory [ 29 , 27 , 32 ] . Additionally, many, if not most, Langevin simulations seek to obtain configurational statistics, such as sampling distribution of phase space or transport in the form of either diffusion or drift. Thus, we submit that the core of an evaluation of an integrator is to first establish its configurational properties, and if those properties are desirable, then decide on the design of a velocity variable that can be consistently associated with the integrator in a manner in which, e.g., simulated kinetic and configurational temperatures coincide (see Refs. [ 29 , 32 ] ). Following the spirit of Ref. [ 28 ] , the presumption of this work is that the quality of a numerical Langevin integrator applied to a nonlinear, complex problem starts with understanding its configurational features for simple, linear problems. The rationale for this approach is that 1) linear analysis can be done objectively for all stochastic Verlet-type integrators, as will be done here, 2) since a thermodynamic system seeks the free energy minimum of an energy surface, the linear component of a problem is often a significant contributor to the sampling, and 3) good algorithm performance for complex problems is expected to be correlated to good performance for simple, linear problems; or expressed in reverse, it is unlikely that an algorithm that is inaccurate for sampling the statistics of linear problems will be generally reliable for nonlinear ones. It is the purpose of this paper to develop a general set of expressions to evaluate stochastic integrators. Within this framework we select a set of commonly discussed methods for configurational evaluation of their basic linear properties relevant for computational statistical mechanics, and thereby provide an objective, problem-independent sense of what one should expect from these integrators. To that end, for any stochastic Verlet thermostat expressed in its pure configurational form, we analytically derive closed expressions for the three most basic properties that are characteristic of the three kinds of linear forces; namely a) diffusion constant for an object on a flat potential, b) drift velocity of an object on a tilted planar potential, and c) the sampling distribution of the location of an object in a harmonic potential. These expressions provide a direct path for basic algorithm analysis, and they provide insight to how an algorithm must be structured in order to yield statistically accurate simulation results as a function of the applied time-step. Twelve representative stochastic Verlet-type integrators are selected