The prediction of structure is a key problem in computational materials science that forms the platform on which rational materials design can be performed. Without detailed structure information the prediction of properties rapidly becomes irrelevant. We will present an ab initio approach that rapidly finds the stable crystal structure of materials with > 95% of success. The premise of the approach is that for many materials chemistries standard computational quantum mechanics is highly accurate in selecting the true ground state of a system from a small set of candidate structures, though notable exceptions exist. Finding ground states by traditional optimization methods on quantum mechanical energy models is difficult due to the complexity and high dimensionality of the coordinate space. An unusual, but efficient solution to this problem can be obtained by merging ideas from heuristic approaches and ab initio methods: In the same way that scientist build empirical rules by observation of experimental trends, we have developed machine learning approaches that extract knowledge from a large set of experimental information and a database of over 20,000 first principles computations, and used these to rapidly direct accurate quantum mechanical techniques to the lowest energy crystal structure of a material. Knowledge is captured in a Bayesian probability network that relates the probability to find particular crystal structure at a given composition to structure and energy information at other compositions. We show that this approach is highly efficient in finding the ground states of binary metallic alloys and can be easily generalized to more complex systems. We have already used this approach to identify several hundred new compounds

The fundamental challenge of rational compound design, i.e. the reverse engineering of chemical compounds with predefined specific properties, originates in the high-dimensional combinatorial nature of chemical space. Chemical space is the hyper-space of a given set of molecular observables that is spanned by the molecular grand-canonical variables (elementary particle densities of electrons and nuclei) which define chemical composition. A working definition of chemical space has been given within the notion of a molecular grand-canonical ensemble multi-component density functional theory framework [1]. I will discuss this approach, as well as numerical results for controlling molecular properties, described within various levels of theory on the multiscale hierarchy, through variation of chemical composition. Properties include electronic molecular eigenvalues, intermolecular energies, or drug-enzyme affinity [2-5]. Eventually, the effects of erroneous many-body interatomic potentials on cohesive energies, or of details in the pseudopotential approximation on band-gap estimates, shall exemplify the important issue of ensuring sufficient accuracy across the multiple scales in the context of multiscale materials design [6,7].[1] von Lilienfeld and Tuckerman, J Chem Phys, 125, 154104 (2006)[2] von Lilienfeld et al, Phys Rev Lett, 95, 153002 (2005)[3] von Lilienfeld and Tuckerman, J Chem Theory Comput, 3, 1083 (2007)[4] Marcon et al, J Chem Phys, 127, 064305 (2007)[5] von Lilienfeld et al, (in preparation)[6] von Lilienfeld and Schultz, Phys Rev B, 77, 115202 (2008)[7] von Lilienfeld and Tkatchenko, Phys Rev B, accepted (2008)Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000.

In modeling non-equilibrium thermal transport in solids, classical molecular dynamics (MD) has the primary strength of explicitly representing phonon modes and the defects that scatter phonons. On the other hand, electrons and their role in energy transport are missing. These effects are vital in applications such as laser processing, thermoelectrics, and current induced thermal failure. In nanoscale and nanostructured systems, the behavior of the system is complicated further by phonon-confinement, ballistic transport, and discrete defect scattering effects. These effects are absent in phenomenological models of heat transport, but naturally captured by MD. Our goal is to couple a MD treatment of the ionic subsystem with a partial differential equation (PDE)-based model of the electronic subsystem in order to accurately capture the aggregate behavior of the coupled electron-ion system. Along these lines, we have enhanced the LAMMPS MD package by coupling the ionic motions to a finite element (FE) based representation of electronic charge and heat transport. The coupling between the subsystems occurs via a local version of the two-temperature model that allows the ionic and electronic subsystems to exchange energy and eventually come into equilibrium. The rate of equilibration between the ionic and electronic temperatures is calculated for a representative system from first principles using a Time Dependent Density Functional Theory (TDDFT) simulation. The ions are initially in thermal motion, while the electrons are initially in their ground state. During the TDDFT simulation, the electrons leave the Born-Oppenheimer surface and gain energy to equilibrate with the ions. Our approach is intrinsically multiscale and multiphysics. Furthermore, the tight coupling between the MD and FE paradigms utilizes the inherent strengths of each. Initial demonstrations of our approach and capabilities have focused on heat transport in representative carbon nanotubes, and these results will be discussed.Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE- AC04-94AL85000.

A phase-field approach to the dynamics of liquid-solid interfaces that evolve due to solute precipitation and/or dissolution is presented. In contrast to solidification processes controlled by a temperature field that is continuous across the solid/liquid interface (with a discontinuous temperature gradient), precipitation/dissolution is controlled by a solute concentration field that is discontinuous at the solid/liquid interface. The Gibbs-Thomson effect on the interface dynamics was included in the phase-field model, and a sharp-interface asymptotic analysis of the phase-field equations was performed for precipitation/dissolution processes to demonstrate that the phase-field equations converge to the proper sharp-interface limit. The mathematical model was validated by numerically solving the coupled phase-field equations for a one- and two- dimensional precipitation/ dissolution problems and by comparison with the analytical solutions.

A coarse-graining method has been proposed [1] for a crystalline system of atoms to describe the propagation of relatively long-wavelength waves, in which the inter-particle interaction is obtained through coarse-graining of the partition function of the atomic Hamiltonian in the harmonic approximation.Though the method has attractive features such as its natural incorporation of atomistic phonons and its potential suitableness to connection to both atomistic and continuum simulation methods bridging the wide scale-gap, the original formulation limits its application to periodic systems without surfaces.In this paper, we advance the method to be applicable to realistic systems of meso-scales with various shapes under stressed conditions at relatively low computation costs. The points of advancement include: (i) recursive coarse-graining procedure [2] for both homogeneous and inhomogeneous systems, (ii) application to pseudo-2D systems such as the graphene sheet and the carbon nanotube. Accuracy of the present method is analyzed through the phonon spectra in homogenous systems, the propagation and scattering of elastic waves in both homogenous and inhomogenous systems, the elastic properties, and so on. Also we will demonstrate its concurrent hybridization with both the molecular dynamics for atoms and the lattice Boltzmann method for fluids, for multiscale simulation of interesting systems.[1] R.E. Rudd and J.Q. Broughton, Phys. Rev. B 58, R5893 (1998).[2] R. Kobayashi and S. Ogata, Mat. Trans. (2008), in press.

The concept of configurational forces is a powerful computational tool for the quantitative description of the behavior of defects in materials and structural components. It enables us to(1)evaluate the crack driving force in arbitrary micro- or macroscopically inhomogeneous materials and components, (2)take into account the influences of eigenstrains and residual stresses,(3)estimate the crack growth direction using the criterion of maximum dissipation,(4)assess the shielding and anti-shielding effects of near-tip and remote plasticity. In this presentation, first a short overview shall be given about theory and computational aspects. Then specific applications are shown, e.g. coated steels, ceramic multilayered materials, or biological materials. These examples demonstrate that the concept of configurational forces - applied in combination with sophisticated experimental methods to determine the spatial variations of local material properties and residual strains - opens new prospects for the design of future damage resistant materials and components.