Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently—those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors. The algorithms are tested on a standard Lennard-Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers--the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y-MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventional vector supercomputers even for small problems. For large problems, the spatial algorithm achieves parallel efficiencies of 90% and a 1840-node Intel Paragon performs up to 165 faster than a single Cray C9O processor. Trade-offs between the three algorithms and guidelines for adapting them to more complex molecular dynamics simulations are also discussed.

Fast parallel algorithms for short-range molecular dynamics

LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales

The stability of two-dimensional (2D) layers and membranes is subject of a long standing theoretical debate. According to the so called Mermin-Wagner theorem, long wavelength fluctuations destroy the long-range order for 2D crystals. Similarly, 2D membranes embedded in a 3D space have a tendency to be crumpled. These dangerous fluctuations can, however, be suppressed by anharmonic coupling between bending and stretching modes making that a two-dimensional membrane can exist but should present strong height fluctuations. The discovery of graphene, the first truly 2D crystal and the recent experimental observation of ripples in freely hanging graphene makes these issues especially important. Beside the academic interest, understanding the mechanisms of stability of graphene is crucial for understanding electronic transport in this material that is attracting so much interest for its unusual Dirac spectrum and electronic properties. Here we address the nature of these height fluctuations by means of straightforward atomistic Monte Carlo simulations based on a very accurate many-body interatomic potential for carbon. We find that ripples spontaneously appear due to thermal fluctuations with a size distribution peaked around 70 \AA which is compatible with experimental findings (50-100 \AA) but not with the current understanding of stability of flexible membranes. This unexpected result seems to be due to the multiplicity of chemical bonding in carbon.

Intrinsic ripples in graphene

Recent molecular dynamics simulations of the growth of $[{\mathrm{Ni}}_{0.8}{\mathrm{Fe}}_{0.2}/\mathrm{Au}]$ multilayers have revealed the formation of misfit-strain-reducing dislocation structures very similar to those observed experimentally. Here we report similar simulations showing the formation of edge dislocations near the interfaces of vapor-deposited (111) [NiFe/CoFe/Cu] multilayers. Unlike misfit dislocations that accommodate lattice mismatch, the dislocation structures observed here increase the mismatch strain energy. Stop-action observations of the dynamically evolving atomic structures indicate that during deposition on the (111) surface of a fcc lattice, adatoms may occupy either fcc sites or hcp sites. This results in the random formation of fcc and hcp domains, with dislocations at the domain boundaries. These dislocations enable atoms to undergo a shift from fcc to hcp sites, or vice versa. These shifts lead to missing atoms, and therefore a later deposited layer can have missing planes compared to a previously deposited layer. This dislocation formation mechanism can create tensile stress in fcc films. The probability that such dislocations are formed was found to quickly diminish under energetic deposition conditions.

Misfit-energy-increasing dislocations in vapor-deposited CoFe/NiFe multilayers

Atomic scale structure of sputtered metal multilayers

Mechanisms, models and methods of vapor deposition

Atomistic simulations using interatomic potentials are widely used for analyzing phenomena as diverse as crystal growth and plastic deformation in all classes of materials. The potentials for some material classes, particularly those for metal oxides, are less satisfactory for certain simulations. Many of the potentials currently utilized for metal oxides incorporate a fixed charge ionic component to the interatomic binding. However, these fixed charge potentials incorrectly predict the cohesive energy of ionic materials, and they cannot be used to simulate oxidation at metal surfaces or analyze metal/oxide interfaces where the local ion charge can be significantly different from that in the bulk oxide. A recent charge transfer model proposed by Streitz and Mintmire has in part successfully addressed these issues. However, we find that this charge transfer model becomes unstable at small atomic spacings. As a result, it cannot be used for the studies of energetic processes such as ion bombardment (e.g., plasma-assisted vapor deposition) where some ions closely approach the others. Additionally, the Streitz-Mintmire charge transfer model cannot be applied to systems involving more than one metal element, precluding study of the oxidation of metal alloys and dissimilar metal oxide/metal oxide interfaces. We have analyzed the origin of these limitations and propose a modified charge transfer model to overcome them. We then unify metal alloy embedded atom method potentials and the modified form of the charge transfer potential to create a general potential that can be used to explore the oxidation of the metallic alloy and the energetic vapor deposition of oxides, and to probe the structure of dissimilar metal oxide/metal oxide or metal alloy/oxide multilayers. Numerical procedures have been developed to efficiently incorporate the potential in molecular dynamics simulations. Several case studies are presented to enable the potential fidelity to be assessed, and an example simulation of the vapor deposition of aluminum oxide is shown to illustrate the potential utility.

Modified charge transfer–embedded atom method potential for metal/metal oxide systems

Vapor deposited multilayers consisting of a low electrical resistivity conductor sandwiched between ferromagnetic metals such as cobalt or nickel-iron alloys sometimes exhibit giant magnetoresistance (GMR). The GMR properties of these films are a sensitive function of structure and defects in the films and therefore depend upon the processing conditions used for their synthesis. A three-dimensional molecular dynamics method has been developed to simulate the [111] growth of model Ni/Cu/Ni multilayers and was used to investigate the role of vapor atom impact energy upon the film structure and defects. High incident atom energies were found to lower interfacial roughness but promoted intermixing by an atomic exchange mechanism. Low incident energies reduced intermixing, but resulted in films with rough, defective interfaces. The simulations identified an intermediate incident energy between 1 and 2 eV that resulted in both low roughness and intermixing, and an anticipated large GMR effect. The simulation me...

Xiaowang Zhou

Papers

Misfit-energy-increasing dislocations in vapor-deposited CoFe/NiFe multilayers

Atomic scale structure of sputtered metal multilayers

Mechanisms, models and methods of vapor deposition

Modified charge transfer–embedded atom method potential for metal/metal oxide systems

Atomistic simulations of the vapor deposition of Ni/Cu/Ni multilayers: The effects of adatom incident energy