We used the machine learning technique of Li et al. (PRL 114, 2015) for molecular dynamics simulations. Atomic configurations were described by feature matrix based on internal vectors, and linear regression was used as a learning technique. We implemented this approach in the LAMMPS code. The method was applied to crystalline and liquid aluminum and uranium at different temperatures and densities, and showed the highest accuracy among different published potentials. Phonon density of states, entropy and melting temperature of aluminum were calculated using this machine learning potential. The results are in excellent agreement with experimental data and results of full ab initio calculations.

We report the computational advances that have enabled the first micron-scale simulation of a Kelvin-Helmholtz (KH) instability using molecular dynamics (MD). The advances are in three key areas for massively parallel computation such as on BlueGene/L (BG/L): fault tolerance, application kernel optimization, and highly efficient parallel I/O. In particular, we have developed novel capabilities for handling hardware parity errors and improving the speed of interatomic force calculations, while achieving near optimal I/O speeds on BG/L, allowing us to achieve excellent scalability and improve overall application performance. As a result we have successfully conducted a 2-billion atom KH simulation amounting to 2.8 CPU-millennia of run time, including a single, continuous simulation run in excess of 1.5 CPU-millennia. We have also conducted 9-billion and 62.5-billion atom KH simulations. The current optimized ddcMD code is benchmarked at 115.1 TFlop/s in our scaling study and 103.9 TFlop/s in a sustained science run, with additional improvements ongoing. These improvements enabled us to run the first MD simulations of micron-scale systems developing the KH instability.

We present two interatomic potentials for hydrogen in –iron based on the embedded atom method potentials for iron developed by Mendelev et al. Philos. Mag. 83, 3977 2003 and Ackland et al. J. Phys.: Condens. Matter 16, S2629 2004 . Since these latter potentials are unique among existing iron potentials in their ability to produce the same core structure for screw dislocations as density functional theory DFT calculations, our interatomic potentials for hydrogen in iron also inherit this important feature. We use an extensive database of energies and atomic configurations from DFT calculations to fit the cross interaction of hydrogen with iron. Detailed tests on the dissolution and diffusion of hydrogen in bulk –iron, as well as the binding of H to vacancies, free surfaces, and dislocations, indicate that our potentials are in excellent overall agreement with DFT calculations.

We formulate an energy-based atomistic-to-continuum coupling method based on blending the quasicontinuum method for the simulation of crystal defects. We utilize theoretical results from Ortner and Van Koten (manuscript) to derive optimal choices of approximation parameters (blending function and finite element grid) for microcrack and di-vacancy test problems and confirm our analytical predictions in numerical tests.

We explored a new type alloy EAM potential (CLI-EAM) that the value of atomic electron density and pair potential between distinct atoms are obtained by Chen’s lattice inversion based on first-principles calculations. The alloy CLI-EAM potential acquired from NiAl alloy can also apply in Ni3Al successfully and the results of basic properties agreed with the experiments. The results of formation energy of point defects of NiAl and Ni3Al alloy indicate that the structural defects are anti-site defects of Al when enrichments of Al atoms.

Using classical molecular dynamics simulations, we have investigated the growth of {111} Cu on Nb {110} surface. Our results reveal that the deposited Cu layer initially grows as body-centered cubic (bcc) and Vernier misfits are observed in the interface of bcc Cu and bcc Nb. As it continues to grow, the bcc Cu {110} transforms into face-centered cubic (fcc) Cu {111}. The phase transition starts after the bcc Cu layer has accumulated about 3 monolayers and is finished depending on deposition parameters. Nuclei of fcc Cu {111} form in the top surface of Cu and grow in plane and toward the interface. Partial dislocations in the fcc Cu layer nucleate during the late stage of the transition, and the stacking faults grow as the Cu layer thickens.

University of Minnesota Ph.D. dissertation. August 2012. Major: Mathematics. Advisor: Mitchell Luskin. 1 computer file (PDF); viii, 98 pages.

University Chemical Laboratory, Lensfield Road, Cambridge CB2 1EW, United Kingdom(Dated: February 2, 2008)Zinc and cadmium clusters interacting with a Gupta potential have previously been identified asprototypical metallic systems that exhibiting disordered cluster structures. Here, putative globalminima of the potential energy have been located for these clusters for all sizes up to N ≤ 125.Although none of the usual structural forms are lowest in energy and many of the clusters have nooverall order, strong structural preferences have been identified. Many of the clusters are based ondistorted oblate Marks decahedra, where the distortion involves the bringing together of atoms oneither side of a re-entrant groove of the Marks decahedron.I. INTRODUCTION

This report presents requirements for advanced simulation of nuclear reactor and chemical processing plants that are of interest to the Global Nuclear Energy Partnership (GNEP) initiative. Justification for advanced simulation and some examples of grand challenges that will benefit from it are provided. An integrated software tool that has its main components, whenever possible based on first principles, is proposed as possible future approach for dealing with the complex problems linked to the simulation of nuclear reactor and chemical processing plants. The main benefits that are associated with a better integrated simulation have been identified as: a reduction of design margins, a decrease of the number of experiments in support of the design process, a shortening of the developmental design cycle, and a better understanding of the physical phenomena and the related underlying fundamental processes. For each component of the proposed integrated software tool, background information, functional requirements, current tools and approach, and proposed future approaches have been provided. Whenever possible, current uncertainties have been quoted and existing limitations have been presented. Desired target accuracies with associated benefits to the different aspects of the nuclear reactor and chemical processing plants were also given. In many cases the possible gains associated with a better simulation have been identified, quantified, and translated into economical benefits.

There are two atoms per primitive unit cell in a monoatomic HCP lattice. Internal relaxations within the unit cell when a homogeneous strain is applied to the overall HCP crystal structure have been investigated with a short-range embedded-atom method (EAM) model both by computer energy minimization and by analytic expansions. Relaxation displacements proportional to the strain occur for shear strains in the basal plane, i.e., for (C11-C12)/2- and C66-type strains, and these two constants remain equal after relaxation. The internal displacement is comparable in magnitude to the relative displacement of neighbouring atoms arising from the homogeneous applied strain and the decrease in this elastic constant due to relaxation is about 10%, with both results independent of model parameters. Another relaxation effect was found to C44-type shears, but the displacements were proportional to the square of the strain. While this does not affect the value of this elastic constant in equilibrium, i.e., in the limit of small displacements, it does yield a softening with large strains. There are no relaxation effects for strains independent of these two.

The paper describes theoretical and computational studies associated with the interface elastic properties of noncoherent metallic bicrystals. Analytical forms of interface energy, interface stresses, and interface elastic constants are derived in terms of interatomic potential functions. Embedded-atom method potentials are then incorporated into the model to compute these excess thermodynamics variables, using energy minimization in a parallel computing environment. The proposed model is validated by calculating surface thermodynamic variables and comparing them with preexisting data. Next, the interface elastic properties of several fcc-fcc bicrystals are computed. The excess energies and stresses of interfaces are smaller than those on free surfaces of the same crystal orientations. In addition, no negative values of interface stresses are observed. Current results can be applied to various heterogeneous materials where interfaces assume a prominent role in the systems' mechanical behavior.

The modified analytic embedded atom method (EAM) theory considering farther neighbor atoms is improved. We not only adopt an end processing function and an enhanced smooth continuous condition of the pair potential, but also adjust model parameters of multi-body potential by fitting cohesion energies, mono-vacancy formation energies, Rose equation curves, energy differences between several structures, elastic parameters and the equilibrium conditions of crystals. The calculation results of structure energy differences misfit the experiment data for several metallic materials in the unimproved EAM, because its model parameters have not considered these differences. After the modification, the coincidence degree of the structure energy—the reduced distance r/r1 from the present model with that from the Rose equation is greatly improved for Cr. The calculated results of structure stabilities are in good agreement with experiment data and other calculated results for all considered metals, significantly better than the unimproved model with farther neighbor atoms.

The generation of crystal defects in a Cr target irradiated by a short, 200 fs, laser pulse is investigated in computer simulations performed with a computational model that combines the classical molecular dynamics method with a continuum description of the laser excitation of conduction band electrons, electron-phonon coupling, and electron heat conduction. Interatomic interactions are described by the embedded atom method EAM potential with a parametrization designed for Cr. The potential is tested by comparing the properties of the EAM Cr material with experimental data and predictions of density functional theory calculations. The simulations are performed at laser fluences close to the threshold for surface melting. Fast temperature variation and strong thermoelastic stresses produced by the laser pulse are causing surface melting and epitaxial resolidification, transient appearance of a high density of stacking faults along the 110 planes, and generation of a large number of point defects vacancies and self-interstitials. The stacking faults appear as a result of internal shifts in the crystal undergoing a rapid uniaxial expansion in the direction normal to the irradiated surface. The stacking faults are unstable and disappear shortly after the laser-induced tensile stress wave leaves the surface region of the target. Thermally activated generation of vacancy-interstitial pairs during the initial temperature spike and quick escape of highly mobile self-interstitials to the melting front or the free surface of the target, along with the formation of vacancies at the solid-liquid interface during the fast resolidification process, result in a high density of vacancies, on the order of 10 3 per lattice site, created in the surface region of the target. The strong supersaturation of vacancies can be related to the incubation effect in multipulse laser ablation/damage and should play an important role in mixing/alloying of multicomponent or composite targets.

The formation of iron particles from the supersaturated gas phase is investigated by molecular dynamics simulation. The atomic interaction is modelled with a recent parameterization of the embedded atom method which is able to describe the bcc phase of bulk iron. The influence of the state conditions such as temperature and density on the growth mechanisms of the iron particles is analysed. With the common neighbour analysis method the structural changes of the particles over the course of the growth process is traced. Furthermore, the surface fraction is investigated as an order parameter for the morphology of the particles. It appears that in the early state of the growth process the atomic structure is dominated by icosahedral structures which are, over the course of the growth process, transformed into other structures such as fcc or hcp. Iron atoms in the bcc structure were not found for small particles. The structural reorganization depends mainly on the temperature of the system and on the magnitude of the heat removal from the system. The results are in agreement with experimental and other theoretical investigations on the structure of iron nanoparticles.

The determination of formation and migration energies of point and clustered defects in SiC is of critical importance to a proper understanding of atomic phenomena in a wide range of applications. We present here calculations of formation and migration energies of a number of point and clustered defect configurations. A newly developed set of parameters for the modified embedded-atom method (MEAM) is presented. Detailed molecular dynamics calculations of defect energetics using three representative potentials, namely the Pearson potential, the Tersoff potential and the MEAM, have been performed. Results of the calculations are compared to first-principles calculations and to available experimental data. The results are analysed in terms of developing a consistent empirical interatomic potential and are used to discuss various atomic migration processes.

The Embedded-Atom Model (EAM) provides a phenomenological description of atomic arrangements in metallic systems. It consists of a configurational energy depending on atomic positions and featuring the interplay of two-body atomic interactions and nonlocal effects due to the corresponding electronic clouds. The purpose of this paper is to mathematically investigate the minimization of the EAM energy among lattices in two and three dimensions. We present a suite of analytical and numerical results under different reference choices for the underlying interaction potentials. In particular, Gaussian, inverse-power, and Lennard-Jones-type interactions are addressed.

Shock wave propagation and the spallation within materials induced by laser shock have been investigated for roughly two decades. With the latest laser technologies evolution, one can access to shorter regimes in durations, going below the picosecond range. Shots performed with the LULI 100 TW facility evidence the possibility to obtain spallation in a few microns thick metallic target. Such conditions provide an experimental data layout which may be directly comparable with molecular dynamic simulations reachable to these scales. Molecular dynamic simulations on a single crystal of Tantalum have been performed with the CEA TERA 10 computer. First, the Hugoniot calculated by the equilibrium molecular dynamics has been compared with experimental data to check the potential (EAM) relevance to reproduce the shock wave propagation. Then, large scale simulations on a micrometric target have been performed. We have observed the microscopic ductile damage process, the pore apparition and their time and space evo...

Severe adhesive wear on a rough aluminum (Al) substrate is simulated by a hard Lennard-Jones asperity impacting an Al-asperity at high speeds using molecular dynamics (MD). Multiple simulations investigate the effects of variations in the inter-asperity bonding, the geometric overlap between two asperities, the relative impact velocity and the starting temperature. The effect of these experimental variables on degree of adhesive wear and the temperature profiles are discussed, and a design of experiments method is used to help interpret the results. The results indicate that increasing the inter-asperity bonding, the geometric overlap and the starting temperature of two asperities will substantially increase the wear rate, while raising the impact velocity slightly decreases the wear rate. It is observed that the deformation mechanism involves local melting and the formation of a liquid like layer in the contact area between two asperities, and the amorphous deformation of the Al-asperity.

Reduced activation steels are considered as structural materials for future fusion reactors. Besides iron and the main alloying element chromium, these steels contain other minor alloying elements, typically tungsten, vanadium and tantalum. In this work we study the impact of chromium and tungsten, being major alloying elements of ferritic Fe-Cr-W-based steels, on the stability and mobility of vacancy defects, typically formed under irradiation in collision cascades. For this purpose, we perform ab initio calculations, develop a many-body interatomic potential (EAM formalism) for large-scale calculations, validate the potential and apply it using an atomistic kinetic Monte Carlo method to characterize the lifetime and diffusivity of vacancy clusters. To distinguish the role of Cr and W we perform atomistic kinetic Monte Carlo simulations in Fe-Cr, Fe-W and Fe-Cr-W alloys. Within the limitation of transferability of the potentials it is found that both Cr and W enhance the diffusivity of vacancy clusters, while only W strongly reduces their lifetime. The cluster lifetime reduction increases with W concentration and saturates at about 1-2 at.%. The obtained results imply that W acts as an efficient 'breaker' of small migrating vacancy clusters and therefore the short-term annealing process of cascade debris is modified by the presence of W, even in small concentrations.

Recently, the machine learning (ML) force field has emerged as a powerful atomic simulation approach because of its high accuracy and low computational cost. However, there have been relatively fewer applications to multicomponent materials. In this study, we construct and compare ML force fields for both an elemental material (Cu) and binary material (SiO2) with varied inputs and regression models. The atomic environments are described by structural fingerprints that take into account the bond angle, and then, different ML techniques, including linear regression, a neural network and a mixture model method, are used to learn the structure-force relationship. We found that using angular structural fingerprints and a mixture model method significantly improves the accuracy of ML force fields. Additionally, we discuss an effective structural fingerprint auto-selection method based on the least absolute shrinkage and selection operator and the genetic algorithm. The atomic simulations conducted for ML force fields are in excellent agreement with ab initio calculations. As a result of the simulation with our ML force field for the structural and vibrational properties of amorphous SiO2, simulated annealing with a slow cooling rate improved the ring statistics in the amorphous structure and the phonon density of states.