Sanjay Kodiyalam

Multimillion atom simulation of materials on parallel computers : nanopixel, interfacial fracture, nanoindentation, and oxidation

We have developed scalable space-time multiresolution algorithms to enable molecular dynamics simulations involving up to a billion atoms on massively parallel computers. Large-scale molecular dynamics simulations have been used to study stress domains and interfacial fracture in semiconductor/dielectric nanopixels, nanoindentation, and oxidation of metallic nanoparticles.

Multimillion Atom Simulations of Nanostructured Materials on Parallel Computers Sintering and Consolidation, Fracture, and Oxidation

Multiresolution molecular-dynamics approach for multimillion atom simulations has been used to investigate structural properties, mechanical failure in ceramic materials, and atomiclevel stresses in nanoscale semiconductor/ceramic mesas (Si/Si3N4). Crack propagation and fracture in silicon nitride, silicon carbide, gallium arsenide, and nanophase ceramics are investigated. We observe a crossover from slow to rapid fracture and a correlation between the speed of crack propagation and morphology of fracture surface. A 100 million atom simulation is carried out to study crack propagation in GaAs. Mechanical failure in the Si/Si3N4 interface is studied by applying tensile strain parallel to the interface. Ten million atom molecular dynamics simulations are performed to determine atomic-level stress distributions in a 54nm nanopixel on a 0.1 µm silicon substrate. Multimillion atom simulations of oxidation of aluminum nanoclusters and nanoindentation in silicon nitride are also discussed.

Pressure Induced Structural Transformations in Nanocluster Assembled Gallium Arsenide

Pressure induced structural phase transformation in nanocluster assembled GaAs is studied using parallel molecular dynamics simulations in the isothermal-isobaric ensemble. In this system the spatial stress distribution is found to be inhomogeneous. As a result structural transformation initiates in the high stress regions at the interface between clusters. Structural and dynamical correlations in the nanophase system are characterized by calculating the spatially resolved bond angle and pair distribution functions and phonon density of states and comparing them with those for a single cluster and bulk crystalline and amorphous systems.

Large-scale molecular dynamics simulations of materials on parallel computers

Scalable space-time multiresolution algorithms implemented on massively parallel computers enable large-scale molecular dynamics simulations involving up to a billion atoms, which are applied to the study of nanosystems of great technological importance. These include sintering, structure, and mechanical properties of nanostructured ceramics and nanocomposites, structural transformation in semiconductor nanocrystals, nanoindentation, and oxidation of metallic nanoparticles.

Large Scale Voxel-Based Finite Element Modeling of Normal and Early Glaucomatous Monkey Lamina Cribrosa

Glaucoma is a leading cause of blindness worldwide. Some of the chief clinical hallmarks of glaucoma are the permanent posterior cupping of the optic nerve head, in the posterior pole of the eye, and the accompanying damage to the lamina cribrosa — the fenestrated structure of connective tissue spanning the scleral canal that provides structural support to the axon bundles passing through it. While elevated intraocular pressure (IOP) is associated with this disease, its role remains unclear. It has been hypothesized that IOP-related stress and strain within the laminar connective tissue (LCT) underlie the onset and progression of glaucoma [1] and that they may be used to predict the location of axonal insult and the pattern of damage within the LCT.© 2008 ASME

Scalable multiresolution algorithms for classical and quantum molecular dynamics simulations of nanosystems

Publisher Summary This chapter discusses the scalable multiresolution algorithmsfor Ccassical and quantum molecular dynamics simulations of nanosystems. The chapter illustrates that modern MD simulations of materials stimulates 864 argon atoms on a CDC 3600 computer. Assuming a simple exponential growth, the number of atoms that can be simulated in classical MD simulations has doubled every 19 months to reach 6.44 billion atoms in 2000. The chapter explores petaflops computers which can maintain the growth rates in these “MD Moores Laws,” and it is able to perform 10 12 -atom classical and 10 7 -atom quantum MD simulations on such computers. Multi resolution approaches described in this chapter, combined with cache-conscious techniques, is essential to achieve scalability on petaflop architectures. In addition, atomistic simulations have reached a scale such that they must be performed in a Meta computing environment of a grid o geographically-distributed multiple supercomputers. Such grid-computing efforts are also underway. Another promising direction toward achieving scalability is the development multiscale simulations schemes that seamlessly combine QM and MD calculations with the continuum mechanics calculation based on the linear elasticity and the finite element method in a single simulation program