8:30AM-8:40AM
Opening ( Welcome Message) from Workshop Chair
Dr. Pinaki Mazumder, Professor of Electrical Engineering, University of Michigan; Program Director, NSF

8:40AM-9:00AM
Update of TeraGrid Facilities and Services
Dane Skow, Director TeraGrid Grid Infrastructure Group

Abstract: TeraGrid consists of NSF-funded HPC resources provided by 9 (soon 11) institutions in a common operational framework. In the course of the past year several significant upgrades to computational resources and operations were made, with more planned for the coming year. This talk will give a brief overview of the facility and those upgrades and plans.

9:00AM-9:30AM
Metascalable Molecular Dynamics Simulation of Mechanochemical Processes
Aiichiro Nakano, Rajiv K. Kalia, Ken-ichi Nomura, Fuyuki Shimojo, and Priya Vashishta, Collaboratory for Advanced Computing and Simulations (CACS), Department of Computer Science, Department of Physics & Astronomy, Department of Chemical Engineering & Materials Science, University of Southern California, USA

Abstract: We have designed scalable parallel algorithms for first-principles based simulations of mechanochemical processes on emerging petaflops architectures based on unified divide-and-conquer (DC) algorithmic and tunable hierarchical cellular decomposition (CD) parallelization frameworks for developing reusable "design once, scale on new architectures" (i.e., metascalable) applications. The DCCD frameworks expose and express maximal concurrency and data locality, thereby achieving parallel efficiency as high as 0.998 for 1.06 billion-atom reactive force-field molecular dynamics (MD) and 11.8 million-atom (1.04 trillion electronic degrees-of-freedom) quantum-mechanical (QM) MD in the framework of the DC density functional theory (DFT) on adaptive multigrids, in addition to 134 billion-atom nonreactive MD, on 131,072 IBM BlueGene/L processors. We have also used the DCCD frameworks for automated execution of adaptive hybrid DFT/MD simulation on a Grid of 6 supercomputers in the US and Japan, in which the number of processors changed dynamically on demand and tasks were migrated according to unexpected faults. I will present the application of the frameworks to study: (1) shock and reactive nanojets in energetic materials; and (2) damage of glasses in corrosive environment. I will also discuss the extension of time scales for these simulations based on a space-time-ensemble parallel (STEP) approach.

9:30AM-10:00AM
Scalable O(N) Density Functional Theory Calculations by Finite Differences on Large Parallel Computers
Jean-Luc Fattebert, Center for Applied Scientific Computing Lawrence Livermore National Laboratory, USA

Abstract: Density Functional Theory (DFT) by the Plane-Waves approach typically scale cubically with the number of electrons simulated.
Thus even the most efficient parallelization of such codes does not allow to grow the scale of practical simulations at the same rate as the growth of supercomputers power. However, representing the electronic structure in DFT by a set of localized wave functions discretized on a real-space mesh essentially leads to a linear scaling of the computational cost with the size of the physical system. This can be achieved by formulating the DFT energy functional in terms of general non-orthogonal orbitals which are then optimized under localization constraints (spatial confinement). Multigrid preconditioning is used to accelerate convergence towards the ground state. For localization regions large enough, truncation error can be reduced to a value smaller than discretization error and achieve the level of accuracy of a Plane Waves calculations. Using finite differences and spatial domain decomposition, most communications between processors happen between nearest neighbor sub-domains. Also, non-overlapping localized orbitals can be optimized in parallel. This leads to a very scalable algorithm with O(N) complexity, enabling simulations of size directly proportional to the number of CPUs.Our implementation scales over a thousand processors and becomes competitive with Plane Waves codes around 500 atoms for dense systems.

References: [1] J.-L. Fattebert and F. Gygi, Phys. Rev. B 73, 115124 (2006) [2] J.-L. Fattebert and F. Gygi, Comput. Phys. Comm. 162, 24 (2004); Prepared by LLNL under Contract DE-AC52-07NA27344

10:00AM-10:10AM
Introduction to Consortium of Petascale Computing for Nanotechnology (cPCN)
Dr. Jun Ni, University of Iowa, and Dr. Thom Dunning, Professor of UIUC and Director of NCSA, USA

10:30AM-11:00AM
Opto-Electronic and Thermoelectric Properties of Si and SiGe Nanowires
Trinh Vo, U.S. Department of Energy by Lawrence Livermore National Laboratory, USA

Abstract: We report the results of a first principles study of the structural, electronic, and optical properties of hydrogen-passivated silicon nanowires with [001], [011], and [111] growth directions and diameters ranging from 1 to 3 nm.  We show that the growth direction, diameter, and surface structure all have a significant effect on the structural stability, electronic band gap, band structure, and band edge effective masses of the nanowires. In addition, we have used a combination of density functional theory, classical molecular dynamics and cluster expansion techniques to optimize the thermoelectric figure of merit (ZT) of Si and SixGe1-x nanowires. The electrical conductivity (s) and Seebeck coefficient (S) were obtained using the Boltzmann transport equation with the constant relaxation time approximation, and with first principles electronic structure calculations. A range of SiGe nanowires with different Ge concentrations and distributions were investigated.   We found that the transport coefficients , S, and thus ZT strongly depend on the wire growth direction, surface structure, as well as the concentration and distribution of Ge. This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.

11:00AM-11:30AM
Productive Electronic Structure Calculations for Nanoscience on the Cray XT4
Paul R. C. Kent, Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, USA

Abstract: Today's nanoscience research routinely demands first principles calculations involving thousands of atoms, treatment of heavy atoms, and dynamical simulations involving long time scales. To meet this challenge we have optimized an implementation of the plane wave projector augmented wave formalism on the Cray XT4 platform. In this talk we review our experience on a wide range of nanoscience calculations including long time Born-Oppenheimer molecular dynamics calculations of thousand atom systems. Routine scalabilities of up to
2-4 processors per atom are obtained. We show that scaling is currently limited by global linear algebra operations and communications related to subspace diagonalization and orthogonalization, and not plane wave operations as is often assumed for this class of simulation. We discuss options for improving scalability on the Cray XT4 and future machines, including use of multicores, asynchronous communications, and mixed precision algorithms. Sponsored by the Division of Scientific User Facilities of the US Department of Energy and the Laboratory Directed Research and Development program at Oak Ridge National Laboratory.

11:30AM-12:00PM
Transport Properties through Nanoscale Materials by First-principles Calculations and Nonequilibrium Green’s Function Formalism
Hiroshi Mizuseki and Yoshiyuki Kawazoe, Institute for Materials Research, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan

Abstract: Nanoscale molecular devices are potential candidates for this next step, and they would make it possible to realize the most advantageous devices. However, source of expenditure is necessary that such a large number of organic molecules can be obtained by synthetic chemistry, so any means of exploring their properties and behavior in order to predict the relevant properties of a molecule in advance of its synthesis would be extremely useful. Our group has covered a wide range of nanoscale materials[1] such as self-assembled nanowires on Si(001) [2, 3], quantum length dependence of conductance in oligomers [4] and single-molecule rotation switch [5] and so on. In this presentation, we will present our recent study on the transport properties of these nanoscale materials using the nonequilibrium Green’s function formalism for quantum transport and the density functional theory (DFT) of electronic structures using local orbital basis sets. References: [1] http://www-lab.imr.edu/~mizuseki/nanowire.html, [2] J.-T. Wang, C. Chen, E. G. Wang, D.-S. Wang, H. Mizuseki, and Y. Kawazoe, Phys. Rev. Lett., 97 (2006) 046103; [3] R. V. Belosludov, A. A. Farajian, H. Mizuseki, K. Miki, and Y. Kawazoe, Phys. Rev. B, 75 (2007) 113411; [4] Y. X. Zhou, F. Jiang, H. Chen, R. Note, H. Mizuseki, and Y. Kawazoe, Phys. Rev. B, 75 (2007) 245407; [5] Y. Y. Liang, F. Jiang, Y. X. Zhou, H. Chen, R. Note, H. Mizuseki, and Y. Kawazoe, J. Chem. Phys. 127 (2007) 084107.

Lunch

1:30PM-2:00PM
New Large Scale Simulation for Evaluating Superconductivity
Satoshi Nakamura, Syogo Tejima, Hisashi Nakamura Research Organization for Information Science and Technology 2-2-54, Nakameguro, Meguro-ku, Tokyo, 153-0061, Japan

Abstract: Discovering new superconductors with a higher Tc (superconducting transition temperature) is one of the most challenging areas in the research of high-Tc superconductivity. However, there has been no further increase of Tc since mercury-based copper oxide superconductor with Tc = -138 ℃ was discovered in 1993. In order to produce a breakthrough under such situations, we have developed a new simulation scheme for evaluating superconductivity in carbon-based materials. As an application, we investigated that at what temperature ideally hole-doped diamond becomes superconducting . For quantitative analyses using our scheme, it is inevitable to perform large scale tight-binding molecular dynamics simulation consisting of at least several thousand atoms; therefore we executed our scheme on the Earth Simulator. In the present simulation, the number of carbon atoms in the diamond was 4096 and the tight-binding molecular dynamics simulation was performed up to 8.2 ps with 0.5 fs time step. The sustained performance of 4.82 Tfops was achieved by utilizing 130 nodes (1040 PEs) of the Earth Simulator. Our simulation suggests that hole-doped diamond will be promising superconductor under the condition where holes are ideally and heavily doped into pure diamond.

2:00PM-2:30PM
Theory and Simulation at the Molecular Foundry of the Electronic and Transport Properties of Nanostructures.
Jeffrey B. Neaton, Lead Scientist, Molecular Foundry, LBNL

Abstract. In this talk, I will describe current activities in theoretical and computational nanoscience at the Molecular Foundry, a new national nanoscience center at Lawrence Berkeley National Laboratory funded by the Department of Energy. The Foundry’s mission is to provide scientists in academia, the national labs, and industry with resources —materials, instrumentation, simulation, and access to scientific staff—for synthesis, characterization, and understanding of nanostructures and their emergent behavior. After summarizing the Foundry program, I will present recent results from several on-going computational projects, in collaboration with institutions across the country and worldwide, highlighting both the methods used and the complex nanoscale phenomena explored. These projects include investigations of the electronic properties of nanotube heterojunctions and tapered silicon nanowires; electronic level alignment at the metal-organic interfaces; and measurement and calculation of electrical conductance of single-molecule junctions.

2:30PM-3:00PM
State-of-the-art Eigensolvers for Large Electronic Structure Calculations
Osni Marques, Lawrence Berkeley National Laboratory

Abstract. The solution of the single particle Schrödinger equation that arises in electronic structure calculations often requires solving for interior eigenstates of a large Hamiltonian. The states at the top of the valence band and at the bottom of the conduction band determine the band gap that relates to important physical characteristics such as optical or transport properties. In order to avoid the explicit computation of all eigenstates, a folded spectrum method has been usually employed to compute only the eigenstates near the band gap. In this presentation, we compare the conjugate gradient minimization, the optimal block preconditioned conjugate gradient, the implicit restarted Lanzos, and variants of the (Jacobi-) Davidson algorithms applied to the folded spectrum matrix for the computation of eigenstates of interest. We also show results when some of these algorithms are applied directly to the unfolded spectrum. Acknowledgments. A. Canning, S. Tomov, C. Voemel, L.-W. Wang.

3:00PM-3:30PM
New Methods for Large Scale Computations of the Electronic Structure of Nanosystems
L-W Wang, Computational Research Division, LBNL

Abstract. Experimental nanostructures often contain from a few thousand atoms to tens of thousands of atoms. Large scale supercomputers and linear scaling algorithms are essential to calculate the electronic structures of such systems.
In this talk, I will present some of the recent developments in atomistic ab initio accuracy electronic structure calculations using a variety of methods, focusing on the charge patching method and the linear scaling three dimensional fragment method. I will show how modern supercomputers and these new methods enable us to investigate the electronic and optical properties of large nanosystems, and their applications in energy science.

3:30PM-4:00PM
Prototyping of Meso and Nano Scale Circuits by Using Parallel Processors
Pinaki Mazumder, Department of Electrical Engineering and Computer Science, University of Michigan

Abstract: Meso and nano scale circuits require computation-intensive full-chip modeling in order to account for chip materials, device geometries, and circuit topologies along with layout structures. This talk will refer to the development of a new 3-D device and interconnect solver by the researchers at the University of Michigan in order to optimize the design parameters of me so-scale quantum tunneling based circuits. Pre-fabrication prototyping of medium size circuits require a tremendous amount of computation time that can be reduced significantly by developing parallel version of the simulator running on a petascale machine.

4:00PM-4:30PM
Femtosecond dynamics for surface reconstruction induced by irradiation with highly-charged ion
Yoshiyuki Miyamoto, Nano Electronic Research Laboratories, NEC Corp.

Abstract: In this talk, I mention computational exploration on structural change induced in graphite by irradiating with Ar8+ ion. Injection of hole and collision of ions are simultaneously treated by combining the time-dependent density functional theory for electron dynamics and the classical molecular dynamics for ion motions. The calculation shows that reconstructed structures on the surface are varied as a function of incident energy of Ar8+ ion. An experiment using 400 eV Ar8+ flux reported appearance of diamond-like structure on the graphite surface, while the present calculation with the same incident energy shows appearance of sp3 configuration as a transient geometry. So calculation nicely explains the recent experiment. Furthermore, the present calculation of lower incident energies shows variety of structures; vacancies, inter-layer bridges, which should be tested in future experiments.
New possible simulation expressing partly charged ion, e.g. Ar7+, remaining a valence electron in the ion will also be presented as well as intra-ion electronic excitation which converts dark ions into bright ones.

4:30:5:00PM
Cyber infrastructure-enabled Computations for the Discovery of Innovative Technology in Medical Imaging through Nanotechnology
Jun Ni, Ph.D., Department of Radiology, University of Iowa, USA

Workshop Close (see you at HPCNano08)