Steven J. Plimpton

Granular flow down an inclined plane: Bagnold scaling and rheology

We have performed a systematic, large-scale simulation study of granular media in two and three dimensions, investigating the rheology of cohesionless granular particles in inclined plane geometries, i.e., chute flows. We find that over a wide range of parameter space of interaction coefficients and inclination angles, a steady-state flow regime exists in which the energy input from gravity balances that dissipated from friction and inelastic collisions. In this regime, the bulk packing fraction (away from the top free surface and the bottom plate boundary) remains constant as a function of depth z, of the pile. The velocity profile in the direction of flow vx(z) scales with height of the pile H, according to vx(z) proportional to H(alpha), with alpha=1.52+/-0.05. However, the behavior of the normal stresses indicates that existing simple theories of granular flow do not capture all of the features evidenced in the simulations.

Dynamical Heterogeneities in a Supercooled Lennard-Jones Liquid

We present the results of a molecular dynamics computer simulation study in which we investigate whether a supercooled Lennard-Jones liquid exhibits dynamical heterogeneities. We evaluate the non-Gaussian parameter for the self part of the van Hove correlation function and use it to identify {open_quotes}mobile{close_quotes} particles. We find that these particles form clusters whose sizes grow with decreasing temperature. We also find that the relaxation time of the mobile particles is significantly shorter than that of the average particle, and that this difference increases with decreasing temperature. {copyright} {ital 1997} {ital The American Physical Society}

Equilibration of long chain polymer melts in computer simulations

Several methods for preparing well equilibrated melts of long chains polymers are studied. We show that the standard method in which one starts with an ensemble of chains with the correct end-to-end distance arranged randomly in the simulation cell and introduces the excluded volume rapidly, leads to deformation on short length scales. This deformation is strongest for long chains and relaxes only after the chains have moved their own size. Two methods are shown to overcome this local deformation of the chains. One method is to first pre-pack the Gaussian chains, which reduces the density fluctuations in the system, followed by a gradual introduction of the excluded volume. The second method is a double-bridging algorithm in which new bonds are formed across a pair of chains, creating two new chains each substantially different from the original. We demonstrate the effectiveness of these methods for a linear bead spring polymer model with both zero and nonzero bending stiffness, however the methods are applicable to more complex architectures such as branched and star polymer.

Implementing Molecular Dynamics on Hybrid High Performance Computers - Short Range Forces

Abstract The use of accelerators such as graphics processing units (GPUs) has become popular in scientific computing applications due to their low cost, impressive floating-point capabilities, high memory bandwidth, and low electrical power requirements. Hybrid high-performance computers, machines with more than one type of floating-point processor, are now becoming more prevalent due to these advantages. In this work, we discuss several important issues in porting a large molecular dynamics code for use on parallel hybrid machines – (1) choosing a hybrid parallel decomposition that works on central processing units (CPUs) with distributed memory and accelerator cores with shared memory, (2) minimizing the amount of code that must be ported for efficient acceleration, (3) utilizing the available processing power from both multi-core CPUs and accelerators, and (4) choosing a programming model for acceleration. We present our solution to each of these issues for short-range force calculation in the molecular dynamics package LAMMPS, however, the methods can be applied in many molecular dynamics codes. Specifically, we describe algorithms for efficient short range force calculation on hybrid high-performance machines. We describe an approach for dynamic load balancing of work between CPU and accelerator cores. We describe the Geryon library that allows a single code to compile with both CUDA and OpenCL for use on a variety of accelerators. Finally, we present results on a parallel test cluster containing 32 Fermi GPUs and 180 CPU cores.

General formulation of pressure and stress tensor for arbitrary many-body interaction potentials under periodic boundary conditions

Three distinct forms are derived for the force virial contribution to the pressure and stress tensor of a collection of atoms interacting under periodic boundary conditions. All three forms are written in terms of forces acting on atoms, and so are valid for arbitrary many-body interatomic potentials. All three forms are mathematically equivalent. In the special case of atoms interacting with pair potentials, they reduce to previously published forms. (i) The atom-cell form is similar to the standard expression for the virial for a finite nonperiodic system, but with an explicit correction for interactions with periodic images. (ii) The atom form is particularly suited to implementation in modern molecular dynamics simulation codes using spatial decomposition parallel algorithms. (iii) The group form of the virial allows the contributions to the virial to be assigned to individual atoms.

A constant-time kinetic Monte Carlo algorithm for simulation of large biochemical reaction networks.

The time evolution of species concentrations in biochemical reaction networks is often modeled using the stochastic simulation algorithm (SSA) [Gillespie, J. Phys. Chem. 81, 2340 (1977)]. The computational cost of the original SSA scaled linearly with the number of reactions in the network. Gibson and Bruck developed a logarithmic scaling version of the SSA which uses a priority queue or binary tree for more efficient reaction selection [Gibson and Bruck, J. Phys. Chem. A 104, 1876 (2000)]. More generally, this problem is one of dynamic discrete random variate generation which finds many uses in kinetic Monte Carlo and discrete event simulation. We present here a constant-time algorithm, whose cost is independent of the number of reactions, enabled by a slightly more complex underlying data structure. While applicable to kinetic Monte Carlo simulations in general, we describe the algorithm in the context of biochemical simulations and demonstrate its competitive performance on small- and medium-size networks, as well as its superior constant-time performance on very large networks, which are becoming necessary to represent the increasing complexity of biochemical data for pathways that mediate cell function.

Implementing Molecular Dynamics on Hybrid High Performance Computers – Particle-Particle Particle-Mesh

Abstract The use of accelerators such as graphics processing units (GPUs) has become popular in scientific computing applications due to their low cost, impressive floating-point capabilities, high memory bandwidth, and low electrical power requirements. Hybrid high-performance computers, machines with nodes containing more than one type of floating-point processor (e.g. CPU and GPU), are now becoming more prevalent due to these advantages. In this paper, we present a continuation of previous work implementing algorithms for using accelerators into the LAMMPS molecular dynamics software for distributed memory parallel hybrid machines. In our previous work, we focused on acceleration for short-range models with an approach intended to harness the processing power of both the accelerator and (multi-core) CPUs. To augment the existing implementations, we present an efficient implementation of long-range electrostatic force calculation for molecular dynamics. Specifically, we present an implementation of the particle–particle particle-mesh method based on the work by Harvey and De Fabritiis. We present benchmark results on the Keeneland InfiniBand GPU cluster. We provide a performance comparison of the same kernels compiled with both CUDA and OpenCL. We discuss limitations to parallel efficiency and future directions for improving performance on hybrid or heterogeneous computers.

Implementing peridynamics within a molecular dynamics code

Peridynamics (PD) is a continuum theory that employs a nonlocal model to describe material properties. In this context, nonlocal means that continuum points separated by a finite distance may exert force upon each other. A meshless method results when PD is discretized with material behavior approximated as a collection of interacting particles. This paper describes how PD can be implemented within a molecular dynamics (MD) framework, and provides details of an efficient implementation. This adds a computational mechanics capability to an MD code, enabling simulations at mesoscopic or even macroscopic length and time scales.

Confined granular packings: structure, stress, and forces.

The structure and stresses of static granular packs in cylindrical containers are studied by using large-scale discrete element molecular dynamics simulations in three dimensions. We generate packings by both pouring and sedimentation and examine how the final state depends on the method of construction. The vertical stress becomes depth independent for deep piles and we compare these stress depth profiles to the classical Janssen theory. The majority of the tangential forces for particle-wall contacts are found to be close to the Coulomb failure criterion, in agreement with the theory of Janssen, while particle-particle contacts in the bulk are far from the Coulomb criterion. In addition, we show that a linear hydrostaticlike region at the top of the packings unexplained by the Janssen theory arises because most of the particle-wall tangential forces in this region are far from the Coulomb yield criterion. The distributions of particle-particle and particle-wall contact forces P(f) exhibit exponential-like decay at large forces in agreement with previous studies.

Molecular dynamics simulations of low-energy (25–200 eV) argon ion interactions with silicon surfaces: Sputter yields and product formation pathways

The etch yield and subsurface damage are important issues in low energy (200 < eV) ion interactions with surfaces. In particular, atomic layer etching requires etching of electronic materials with monolayer precision and minimal interlayer atomic mixing. In this study, the molecular dynamics technique is used to simulate the impact of argon ions on chlorine-free and chlorine-passivated silicon surfaces, under conditions relevant to atomic layer etching. Thousands of individual ion impact simulations are performed on a massively parallel supercomputer. The silicon sputter yield is obtained for Ar ion energies ranging from 25 to 200 eV. Where possible, simulation results are compared to available experimental data. Volatile product formation during ion bombardment of ordered surfaces tends to follow distinct local trajectories. For example, the formation of products due to 120 eV Ar ions impacting onto Si(001)(2×1) at normal incidence has been found to occur mainly by a mechanism in which the Ar ion impacts...