Internal Trajectories and Observation Effects in Langevin Splitting Schemes

TL;DR

The paper addresses how Langevin splitting schemes behave beyond generator-based analysis by focusing on internal trajectories and when observations are recorded. It demonstrates that merging/splitting updates and cyclic permutations can yield identical internal trajectories, while the placement of observation points induces biases in measured distributions, especially at large friction $\xi$ and time steps $\Delta t$. By analyzing ABO- and AP-based schemes and their symmetric/non-symmetric variants, the work provides a unified view of which schemes share trajectories and how observation timing affects configurational sampling, free-energy estimates, and transition rates. The findings show that many modern integrators are remarkably stable under typical MD conditions, with biases only appearing under deliberately large $\xi$ and $\Delta t$, offering a practical framework to understand and control observation effects in Langevin dynamics.

Abstract

Langevin integrators based on operator splitting are widely used in molecular dynamics. This work examines Langevin splitting schemes from the perspective of their internal trajectories and observation points, complementing existing generator-based analyses. By exploiting merging, splitting, and cyclic permutation of elementary update operators, formally distinct schemes can be grouped according to identical or closely related trajectories. Accuracy differences arising from momentum updates and observation points are quantified for configurational sampling, free-energy estimates, and transition rates. While modern Langevin integrators are remarkably stable under standard simulation conditions, subtle but systematic biases emerge at large friction coefficients and time steps. These results clarify when accuracy differences between splitting schemes matter in practice and provide an intuitive framework for understanding observation effects.

Figures (6)

• Figure 1: Visualization and classification of Langevin splitting operators. Elementary update operators are applied right to left.
• Figure 2: Momentum distributions of a mass $m=12.0\, \mathrm{g}/\mathrm{mol}$ at $T=300\, \mathrm{K}$ and $\xi = 1\, \mathrm{ps}^{-1}$ obtained by propagating with full O-steps and with half O-steps compared to the analytical momentum distribution Vertical dashed line indicates the initial condition $p_0$.
• Figure 3: a) Prefactor $c$ (eq. \ref{['eq:generalMomentumUpdate']}) that scales the force kick in eqs. \ref{['eq:BOB']}-\ref{['eq:BO']} compared to the exact force-kick prefactor in $\mathcal{P}$ (eq. \ref{['eq:P']}). b) Relative error of $c$ compared to $\mathcal{P}$. The vertical dotted line indicates $\xi \Delta t = 0.002$, corresponding to typical MD settings of $\Delta t = 2\, \mathrm{fs}$ and $\xi = 1\, \mathrm{ps}^{-1}$. c) Time evolution of the momentum distribution with simulation parameters $m = 12\, \mathrm{g}/\mathrm{mol}$, $T=300\, \mathrm{K}$, $\xi = 10\, \mathrm{ps}^{-1}$, $\Delta t = 10\, \mathrm{fs}$, $\kappa = 1000\, \mathrm{kJ}/ (\mathrm{mol} \cdot \mathrm{nm})$. The vertical dotted line indicates the initial condition $p_0$. d) Time evolution of the position distribution.
• Figure 4: Relative error of $\langle q^2 \rangle$ in a harmonic potential $V(q) = \frac{1}{2}\kappa x^2$ with $\kappa = 1.2\cdot 10^{5}\mathrm{kJ}/ (\mathrm{mol} \cdot \mathrm{nm}^2)$ with simulation parameters: $m = 12\, \mathrm{g}/\mathrm{mol}$, $T=300\, \mathrm{K}$, and (a): $\xi = 1\, \mathrm{ps}^{-1}$, (b): $\xi = 10\, \mathrm{ps}^{-1}$, (c): $\xi = 100\, \mathrm{ps}^{-1}$.
• Figure 5: (a) Trajectories produced by ABOBA and BOBA. The system is the particle in the untilted double-well potential presented in Fig. \ref{['fig:doubleWell']}. (b) Sketch of the path through phase space for $\mathcal{A}$-half-step and $\mathcal{A}$-half-step schemes.