Abstract
This research addresses the fundamental challenge of translating discrete atomistic data from Molecular Dynamics (MD) simulations into the continuous fields of momentum and energy fluxes used in continuum mechanics. We propose a two-level consistency framework—encompassing conservation consistency and continuum consistency—to evaluate flux definitions for complex multi-body potentials like the Embedded Atom Method (EAM) and Tersoff potentials. By employing a control-volume approach and verifying results against the Cauchy-Born rule, we provide a rigorous methodology for bridging the gap between atomistic trajectories and macroscopic engineering physics.
Introduction: The Challenge of Multi-Body Fluxes
In multiscale modeling, accurately defining "fluxes"—the movement of momentum (traction/stress) and energy (heat flux) across a surface—is critical for connecting MD simulations to continuum-scale theories. While straightforward for simple pair potentials, modern simulations often utilize multi-body potentials where the force on an atom depends on the relative positions of multiple neighbors. In such systems, there is no unique way to partition forces and energy among atoms, which can lead to definitions that violate conservation laws or fail to match established macroscopic elasticity.
Methodology: A Two-Level Consistency Framework
To ensure the physical validity of flux definitions, we propose two criteria:
- Conservation Consistency: The rate of change of momentum or energy within a spatial cell must exactly equal the sum of the fluxes across its boundaries.
- Continuum Consistency: In the limit of a perfect crystal under uniform deformation, the atomistic traction must match the stress predicted by the Cauchy-Born rule, which links atomic lattice energy to macroscopic elasticity.
We implement these criteria using a control-volume approach, dividing the atomistic system into discrete cells to track the flow of quantities across their interfaces.
Analytical Derivations and Numerical Verification
The paper provides detailed analytical derivations and computational pseudocode for two complex potentials:
- Embedded Atom Model (EAM): We demonstrate that the multi-body "embedding" energy requires specific mathematical treatment to maintain energy balance and translational invariance.
- Tersoff Potential: We account for the complex bond-order formulation of the Tersoff potential, which involves interactions beyond simple three-body terms.
These definitions were verified using a BCC lattice for EAM and a 24,000-atom silicon diamond structure for the Tersoff potential. Numerical tests on a non-homogeneous silicon system featuring a central void confirmed that the change in a cell's momentum and energy matched the calculated fluxes with high precision (errors ranging from $10^{-12}$ to $10^{-9}$).
Results and Discussion
For systems under uniform stretch and shear (up to 5%), the tractions calculated using our MD definitions matched the Cauchy-Born stress to machine precision. Furthermore, in non-homogeneous systems, the calculated tractions correctly converged to zero upon reaching mechanical equilibrium, proving the robustness of the definitions even in the presence of defects like voids.
A key insight from this work is the necessity of translational invariance in force decomposition. If the force between two atoms changes based on the absolute position of the system, the resulting stress definition will fail to align with continuum mechanics.
Conclusion and Future Work
We have established a rigorous framework for defining momentum and energy fluxes in MD simulations with multi-body potentials. While this approach successfully captures the "convective" part of energy flux, defining the heat conduction component—energy transport via phonons—remains a significant challenge for future research in non-equilibrium statistical mechanics. This work provides a roadmap for deriving consistent fluxes for a wide range of exotic interatomic potentials.
Summary updated