A parallel algorithm to produce long polymer chains in molecular dynamics.

Generating initial configurations of polymer melts above the entanglement molecular weight is a challenge in molecular dynamics and Monte Carlo simulations. In this work, we adapt an algorithm mimicking a chemical polymerization to all-atom force fields. The principle of this algorithm is to start from a bath of monomers between which bonds are created and relaxed sequentially. Our implementation is parallel and efficient. The parallelization is that of a classical molecular dynamics code and enables the user to generate large systems, up to 7 × 106 atoms. The efficiency of the algorithm comes from the linear scaling between the simulation time and the chain length in the limit of very long chains. The implementation is able to produce long polymer chains, up to ∼2000 carbon atoms, with thermodynamic and local structural properties in good agreement with their experimental and numerical counterparts. Moreover, the chain conformations are close to being equilibrated right after the end of the polymerization process, corresponding to only a few hundred of picoseconds of simulation, despite a systematical drift from Gaussian-like behavior when the density of reactively available monomers decreases. Finally, the algorithm proposed in this work is versatile in nature because the bond creation can be easily modified to create copolymers, block copolymers, and mixtures of polymer melts with other material.

Anisotropic surface stresses of a solid/fluid interface: Molecular dynamics calculations for the copper/methane interface.

The full tensorial surface stress of an interface between a face-centered cubic crystal (copper) and an isotropic liquid (methane) is computed for two crystal orientations {100} and {110} using molecular dynamics simulations. The bulk crystal orientation {100} is symmetric, whereas the {110} orientation is not. Finite size effects, which can be important in the case of an interface between an isotropic solid and a liquid, are studied in detail for the two crystal orientations. We first show that the symmetry of the surface stress tensor is that of the bulk crystal orientation. In the case of the asymmetric crystal orientation {110}, the relative difference between the components of the surface stress is substantial (∼50%). Finally, we show that finite size effects persist to much larger sizes in the case of the {100} orientation compared to the case of the {110} interface, for instance, through an artificial breakdown of the symmetry of the surface stress tensor.

Distribution of melting times and critical droplet in kinetic Monte Carlo and molecular dynamics.

A kinetic Monte Carlo model on a lattice, based on a reaction-like mechanism, is used to investigate the microscopic properties of the homogeneous melting of a metastable crystal. The kinetic Monte Carlo model relies on nearest-neighbors interactions and a few relevant dynamical parameters. To examine the reliability of the model, careful comparison with molecular dynamics simulations of a hard sphere crystal is drawn. A criterion on the critical nature of a microscopic configuration is deduced from the bimodal character of the probability density function of melting time. For kinetic Monte Carlo simulations with dynamical parameter values which fit the molecular dynamics results, the number of liquid sites of the critical droplet is found to be smaller than 300 and the ability of the critical droplet to invade the entire system is shown to be independent of the droplet shape as long as this droplet remains compact. In kinetic Monte Carlo simulations, the size of the critical droplet is independent of the system size. Molecular dynamics evidences a more complex dependence of melting time on system size, which reveals non-trivial finite size effects.

Molecular dynamics simulations of a hard sphere crystal and reaction-like mechanism for homogeneous melting.

Molecular dynamics simulations of a hard sphere crystal are performed for volume fractions ranging from solidification point to melting point. A local bond order parameter is chosen to assign a nature, liquid or solid, to a particle. The probability for a liquid or solid particle to change state presents a typical sigmoid shape as the nature of its neighbors changes. Using this property, I propose a reaction-like mechanism and introduce a small number of rate constants. A mean-field approach to melting and a kinetic Monte Carlo algorithm on a lattice are derived from these chemical processes. The results of these models successfully compare with molecular dynamics simulations, proving that the main properties of melting can be captured by a small number of dynamical parameters.