Analysis and Simulation of Generalized Langevin Equations with Non-Gaussian Orthogonal Forces

Henrik Kiefer, Benjamin J. A. Héry, Lucas Tepper, Benjamin A. Dalton, Cihan Ayaz, Roland R. Netz · 2025

Read Analysis and Simulation of Generalized Langevin Equations with Non-Gaussian Orthogonal Forces by Henrik Kiefer, Benjamin J. A. Héry, Lucas Tepper, Benjamin A. Dalton, Cihan Ayaz, Roland R. Netz (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-2505-15665.txt.

Introduction

Henrik Kiefer and co-authors analyze a framework that cuts complex molecular motion down to a few coordinates in a 2025 paper. The generalized Langevin equation models many-body dynamics through chosen reaction coordinates plus a memory and noise term. The form changes with the method used to project.

Using simulations of butane's twisting angle, the team set parameters side by side across forms. The random force turned out to be non-Gaussian in all versions, most strongly in the Mori form. Knowing where the simple noise picture fails helps those who build such models make honest reduced models of such motion.

Catalog record

CREATORHenrik Kiefer, Benjamin J. A. Héry, Lucas Tepper, Benjamin A. Dalton, Cihan Ayaz, Roland R. Netz
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE106 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-15665

Browse the full library · About the Tower of Babel

Full text

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

Analysis and Simulation of Generalized Langevin Equations with Non-Gaussian Orthogonal Forces Henrik Kiefer, Benjamin J. A. Héry, Lucas Tepper, Benjamin A. Dalton, Cihan Ayaz, Roland R. Netz Abstract The generalized Langevin equation (GLE) is a useful framework for analyzing and modeling the dynamics of many-body systems in terms of low-dimensional reaction coordinates, with its specific form determined by the choice of projection formalism. We compare parameters derived from different GLE formulations using molecular dynamics simulations of butane’s dihedral angle dynamics. Our analysis reveals non-Gaussian contributions of the orthogonal force in different GLEs, being most enhanced for the Mori-GLE, where all non-linearities are relegated to the orthogonal force. We establish a simulation technique that correctly accounts for non-Gaussian orthogonal forces, which is critical for accurately predicting dihedral-angle mean first-passage times. We find that the accuracy of GLE simulations depends significantly on the chosen GLE formalism; the Mori-GLE offers the most numerically robust framework for capturing the statistical observables of the dihedral angle dynamics, provided the correct non-Gaussian orthogonal force distribution is used. Preprint: AIP/123-QED I Introduction Coarse-graining of complex systems in terms of low-dimensional observables provides useful models for the description of many-body kinetics; examples are protein folding [ 1 , 2 , 3 , 4 ] or chemical reactions between individual molecules [ 5 , 6 , 7 , 8 ] . The resulting reduced equations of motion, typically stochastic, not only give the advantage of numerical efficiency and speed over all-atomistic simulations of a complex system, but also provide the possibility of analysis and interpretation of atomistic simulations and experiments. The generalized Langevin equation (GLE), comprising a deterministic force, non-Markovian friction, and an orthogonal force, serves as a powerful framework for constructing coarse-grained models of many-body systems [ 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ] . By reducing the system’s full degrees of freedom to a few relevant observables, the GLE enables accurate modeling and kinetic predictions of conformational transitions [ 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 ] , reaction dynamics [ 30 , 31 ] , and the vibrational spectra of molecules in the presence of non-linear bond-length and bond-angle potentials [ 8 , 32 ] . First introduced by R. Zwanzig [ 33 ] and H. Mori [ 34 ] , different forms of the GLE arise from different projection operator formalisms. It has been argued that the Mori-projection technique is less suitable for modeling non-linear observables because the resulting GLE includes a deterministic force linear in the coordinate and a frictional force linear in the velocity. However, since the Mori-GLE is an exact decomposition of the total force acting on an observable, non-linearities in the observable are incorporated into the orthogonal force, which becomes non-Gaussian and thereby challenging to parameterize [ 35 , 36 ] . By contrast, GLEs derived using the Zwanzig-projection method are considered more suitable, as they explicitly include the term associated with the potential of mean force (PMF) [ 37 , 38 , 39 ] . To explicitly account for non-linearities in the deterministic forces and to simplify the friction kernel compared to the Zwanzig formulation, a hybrid projection scheme was recently introduced [ 40 ] . This approach combines the linear Mori and conditional Zwanzig projection operators to derive a hybrid GLE, which comprises a memory friction term that splits into two components: one that is linear in the past velocity and another that is generally non-linear in the coordinate but independent of velocities. Numerical methods for extracting all parameters of this GLE from trajectories have been proposed [ 40 ] , however methods to simulate the hybrid GLE are currently not known. Until recently, it was widely assumed that a GLE incorporating the PMF and a friction force only linear in velocity [ 15 ] , which is suitable for various applications [ 41 , 18 , 26 , 32 , 27 , 28 , 42 ] , could only be derived approximately by neglecting position-dependent friction terms. However, it was demonstrated [ 43 ] that such a GLE can be rigorously derived by sequentially applying the Mori and hybrid projection techniques during different steps of the derivation. With the various forms of GLEs discussed, a key question arises: which GLE formulation is most robust for simulating non-linear systems? Here, we evaluate robustness based on the ability to accurately reproduce commonly used statistical observables of the modeled coordinates, such as two-point correlation functions and mean first-passage times. Previous work on coarse-grained simulations relied on replacing the GLE by coupled Markovian Langevin equations with Gaussian noise, a method commonly referred to as Markovian embedding [ 44 , 37 , 20 , 31 ] . These studies concluded that faithfully reproducing statistical observables requires involving the correct non-linear PMF [ 26 , 32 , 39 ] . Furthermore, although the orthogonal force resulting from projection is, in principle, a deterministic function with well-defined initial conditions [ 34 ] , practical simulations assume that it follows a stationary Gaussian distribution, leveraging a relationship between the orthogonal force autocorrelation and the memory kernel. However, experimental studies [ 45 , 46 , 47 , 48 ] and simulations across various fields [ 49 , 16 , 50 , 39 ] frequently reveal deviations from Gaussian behavior of the orthogonal force. The modeling of the system’s kinetics can be strongly influenced by the probability distribution and higher-order correlations of the orthogonal force [ 51 , 52 , 53 , 54 , 55 , 56 , 57 ] . Accurate modeling of the orthogonal force in GLE simulations is, therefore, key for capturing the underlying physics of soft-matter systems, particularly when rare events dominate the long-time dynamics. This is the topic of the present paper. After reviewing the derivations and relationships between the various GLEs discussed, using dihedral angle trajectories from molecular dynamics (MD) simulations, we compare the parameters extracted for the different GLEs. In particular, we explore the behavior of the orthogonal force in the different GLE formulations. Our extraction methods reveal significant non-Gaussian features in the orthogonal force for all GLEs. Ultimately, we present a GLE simulation algorithm for arbitrary orthogonal force distributions. This non-Gaussian force (NGF) technique samples the orthogonal force from MD simulations and operates without additional approximations when simulating the GLE, in contrast to methods used in previous work [ 44 , 26 , 58 , 39 ] . Contrary to common Markovian embedding techniques, we can precisely reproduce the mean first-passage times of the butane dihedral angle when using an orthogonal force with the correct distribution. By comparing simulations of different GLEs, we show that the Mori-GLE, which relegates all non-linearities in the deterministic and friction forces into the orthogonal force, is the numerically most accurate approach for simulating dihedral angle dynamics. Consequently, accurately modeling the statistics of the orthogonal force is more critical than including the appropriate PMF, contrary to previous suggestions [ 40 , 59 , 39 ] . II Theoretical Background II.1 Different generalized Langevin equations We consider a dynamical system of N interacting particles in three-dimensional space. The time

Read the complete text