Robust Estimation of Position-Dependent Anisotropic Diffusivity Tensors from Molecular Dynamics Trajectories

Area of Science:

• Materials Science
• Computational Chemistry
• Nanoscale Physics

Background:

• Confinement alters material properties, making spatial variations crucial for nanoscale applications.
• Accurate characterization of position-dependent material properties is a key research challenge.
• Determining spatial profiles of transport properties like diffusivity is particularly difficult.

Purpose of the Study:

• To adapt and validate filtered covariance estimators (FCEs) for extracting diffusivity profiles from molecular dynamics (MD) trajectories.
• To address challenges in calculating position-dependent transport properties in confined systems.
• To develop a correction scheme for improving the accuracy of normal diffusivity calculations near boundaries.

Main Methods:

• Adaptation of analytically derived filtered covariance estimators (FCEs) for MD simulations.
• Application of the adapted FCEs to a Lennard-Jones fluid confined within a slit pore.
• Implementation of a correction scheme using simulated annealing and diffusion maps to resolve boundary artifacts.

Main Results:

• The MD-adapted FCE accurately estimates lateral diffusivity across the pore.
• A systematic underestimation of normal diffusivity near hard boundaries was observed.
• The correction scheme successfully resolved boundary artifacts, yielding normal diffusivity profiles consistent with van Hove correlation functions.

Conclusions:

• The developed MD-adapted estimator shows significant potential for accurately characterizing spatial diffusivity variations in confined materials.
• The correction scheme effectively improves the accuracy of normal diffusivity measurements near interfaces.
• This work provides a robust computational tool for understanding nanoscale transport phenomena.