HPCNano05 Workshop Program

Morning Program

"Introduction to HPC for nanotechnology"
Jun Ni and Jack Dongarra, the University of Iowa and The University of Tennessee
(8:30AM-8:40AM on Nov. 16, Wed; Room 206)

"Computational Nanotechnology - An Introduction"
John E. Savage, Andre' DeHon, Ben Gojman, Charles Lieber, Eric Rachlin, and Yue Wu
Brown University, USA
(8:40AM-9:20AM on Nov. 16, Wed; Room 206)

Abstract: Computational nanotechnology offers the promise of very high chip densities, a promise that will be realized through the introduction of new devices, methods of assembly, and architectures. The new architectural and performance characteristics of these technologies are a) the predominance of regular, rectangular structures, b) randomness in circuit layouts and connections and c) high fault and defect rates. In this talk we give an overview of these issues. In particular, we describe and evaluate stochastic methods of addressing NWs and give a brief introduction to fault and defect tolerance. Opportunities to model nanotechnology systems will be sketched.

"Large-scale Simulation for Finding Novel Properties and Functions on Nanostructured Carbon Material, using the Earth Simulator" (abstract)
Hisashi Nakamura, Research Organization for Information Science & Technology (RIST), Japan
(9:20AM-10:00AM on Nov. 16, Wed; Room 206)

Abstract: Through progress of capability of high-end computing in these days, simulation models of atom and molecular dynamics enables us to reveal novel properties and functions of nano and mesoscale material interacting with surrounding complex environment. Aiming at realizing a practical simulation methodology for nanoscience and technology, we have been so far developing a large-scale simulation software technology utilizing a quantum based tight-binding molecular dynamic model adopting a recursive-technique. By optimizing the model to the Earth Simulator through parallelization and vectorization, we have succeeded in extending the scale of simulation of a carbon nanotube up to ten thousand of atoms without the spatial symmetry and homogeneous condition, and recording the performance of 7.1 Teraflops on 3,480 processors in simulating thermal conductivity of carbon nanotube with 48,600 atoms. This method has lead us to simulate effectively properties and functions of complex nanocarbon structures with a large number of atoms: thermal conductivity of a single walled carbon nanotube (CNT), mechanical strength of CNTs, atomic welding of super-diamond, stability of peapod with a defective wall, thermal dissolution of fullerene and so on. The large-scale simulation could be one of powerful and effective methods for generating and finding new materials and novel functions of nano-complex systems as well as experimental and theoretical approaches.

"Overview of the NSF TeraGrid Initiative"
Charlie Catlett, TeraGrid Director, University of Chicago and Argonne National Laboratory
(10:00AM-10:30AM on Nov. 16, Wed; Room 206)

Abstract:TeraGrid is a collaboration of partners providing a high-performance, nationally distributed capability infrastructure for computational science. The TeraGrid team has utilized multiple surveys of user requirements to develop five-year roadmaps describing new capabilities and services, organized into several new initiatives: Deep, Wide, and Open. TeraGrid is managed by the University of Chicago and includes resources at eight partner “resource provider”
sites (Argonne National Laboratory, Indiana University, National Center for Supercomputing Applications, Oak Ridge National Laboratory, Pittsburgh Supercomputing Center, Purdue University, San Diego Supercomputer Center, and Texas Advanced Computing Center).

TeraGrid Deep aims to assist scientists with applications that require the combination of multiple leadership class systems- including TeraGrid storage, computing, instruments, visualization, etc. – working in concert. A team of roughly 15 staff is providing hands-on assistance to application teams pursuing TeraGrid Deep projects.

TeraGrid Wide is a set of partnerships with peer Grid projects and prototype "science gateways" that are aimed at making TeraGrid resources available to, and tailored to, entire communities of users. Science gateways are driving policy, process, and technology standards to enable web portals, desktop applications, campus clusters, and other grid infrastructure projects to seamlessly use TeraGrid resources. Initial TeraGrid science gateway projects include community portals and desktop tools supporting life sciences and biomedicine, high-energy physics, neutron science, astronomy, nanotechnology, atmospheric and climate sciences, and environmental and emergency decision-support.

TeraGrid Open involves the rapid evolution of the TeraGrid software and services toward interoperation with peer Grids and campus resources. Currently TeraGrid is partnering with the Open Science Grid in the US as well as partners in Europe (e.g. UK eScience,DEISA) and Asia-Pacific (e.g. APAC, Naregi, K*Grid).

Abstract: The Network for Computational Nanotechnology (NCN) is a multi-university, NSF-funded initiative with a mission to lead in nanotechnology research and education as well as outreach to students and professionals by offering a set of cyber services. These services are at the core of a unique web-based infrastructure tailored to serve the nation’s National Nanotechnology Initiative. The primary NCN outreach vehicle is the nanoHUB (http://www.nanoHUB.org), which currently provides interactive online simulation and educational resources such as tutorials, seminars, and online courses packaged using e-learning standards. In the past 12 months, the educational and outreach services were accessed by over 6,800 users. More than 1,100 users performed over 68,000 online simulations thanks to the grid middleware supporting the nanoHUB. Around 30 applications are available online ranging from toy models to sophisticated simulation engines. The NCN provides the resources for modeling, simulating and computing without any software installation to users with access to a web browser. All the NCN services are freely open to the public. The NCN represents a community–based, service–oriented, science architecture that leads the way for new cyberinfrastructure-enabled science. At the core, the NCN research mission is based on four research themes: nanoelectronics, nanoelectromechanics, nano-bio, and high performance computation (HPC). One facet of the NCN goals is the development of new “community codes” that provide the nanoscience research community with new capabilities and that lay a foundation for a new generation of CAD tools that will pave the way to ground breaking nanotechnology devices. The HPC theme is charged with “grand challenges” to enable the three nanotechnology research themes to study realistically complex systems that have previously been computationally impossible or too expensive to explore. The development of numerical techniques and their packaging into portable software, capable of a variety of high accuracy simulations of nano-electronic, nano-biological, and nano-mechanical systems is the overarching goal of this research theme. The NCN organization in such themes as resulted in significant results including 21 million atoms quantum dots simulation on TeraGrid resources. Such simulations were possible through a new TeraGrid initiative called “Science Gateway”. The nanoHUB and the TeraGrid are now partnering to provide state-of-the-art computing resources to the nanoHUB users. With the TeraGrid entering a new phase after its construction, the nanoHUB is getting connected to the TeraGrid to make TeraGrid resources available to nanoHUB users as transparent compute backends. Computationally intensive nanoHUB applications that solve problems across multiple computational scales are being deployed on TeraGrid resources, and NCN partners will be able to access these resources for their own research. Science Gateways represent a shift from the traditional high performance computing use by enabling entire communities of users associated with a common scientific goal to use the national resources through a common interface. Science Gateways are enabled by a community allocation whose goal is to delegate account management, accounting, certificates management and user support to the gateway developers. It represents the core computing infrastructure for the service oriented architecture of the nanoHUB. Finally, the NCN recognizes that scientific codes are really only accessible to a community if they are truly usable. We are developing a framework that enables rapid deployment of legacy applications with appealing graphical user interfaces (GUIs). This framework, called Rappture, lets developers forego the creation of their own I/O routines, and it generates user friendly GUIs automatically. The nanoHUB middleware serves these GUI based applications through the user’s browsers and makes the connection to the compute backends through virtualization technology. In this talk, we will present the NCN, the nanoHUB middleware and its capabilities as well as the educational technology supporting our teaching strategy. Finally, we will show how the nanoHUB infrastructure leads to state-of-the-art research in nanotechnology by showing concrete examples of multimillion atom electronic structure calculations.

Abstract: Electronic Structure calculations based on the density functional theory approach have become one of the biggest consumers of cycles on high performance computers around the world. This is due to the accurate representation of the underlying physics as well as the computational efficiency of this approach. In this talk I will discuss this approach, as used in nanoscience applications on high performance computers as well as new methods that go beyond density functional theory and allow us to simulate larger systems with first principles accuracy. Performance of these methods on high performance computers such as the NEC Earth Simulator, Cray X1E, IBM SPs, and PC clusters will also be discussed.

Abstract: The need for efficient eigensolvers in the computational nanotechnology field is well recognized. This is because most of the simulations to predict the electronic properties of nanostructures lead to discrete eigenvalue problems. Moreover, the accuracy of these simulations, and hence the chances for discovery, are directly related to the size of the problem that can be solved on the available computational resources. The goal of being able to efficiently handle large scale problems requires an iterative solution approach that would also be suitable for parallel processing. An obstacle to the efficiency of the iterative approach is that the convergence depends on the condition number of the matrix at hand. We have developed a preconditioning technique, called bulk
band (BB) preconditioner, that alleviates this problem and gives a significant computational speed improvement compared to a currently used diagonal preconditioner. The basic idea behind BB is to use the electronic properties of the bulk materials constituent for the nanostructure in designing an efficient preconditioner. In this talk we will describe our work on developing efficient iterative eigensolvers for large scale electronic nanostructure calculations. We
will concentrate on their acceleration through preconditioning. In particular, we will define the BB preconditioner, numerically demonstrate its efficiency, and show how this new preconditioner fits into a conjugate gradient (CG) based framework of eigensolvers that we have developed.

Abstract: Nanotechnology is used to create materials, functional or multifunctional structures, and devices on a nanometer (10-9 m) scale. It has emerged from multidisciplinary research fields, one of which is materials science. In the research of nano- mechanics and materials science domain, molecular dynamics plays an important role, especially in the design of novel nanoscale materials such as nanocomposites. However, it has limitations on length and time scales. The material with a cubic volume of 1 ?m3 contains trillions of atoms, and a typical time-step in molecular dynamics simulations is about femtosecond (~10-15 s). Therefore, there is an urgent need for researchers and computational scientists to develop applicable methodology to efficiently simulate large nano systems. Recently, multiscale methods, especially concurrent multiscale methods have been of interest since they are expected to cover a range of physical domains of different length scales from atomic and microscopic/mesoscopic to macroscopic scales. Unfortunately, most multiscale methods still require intensive computation even with existing high-end computers for large nanoscale simulations although such limitations are much smaller in comparison with those in fully molecular dynamics simulations. The above issues spur us to develop an alternative approach – to conduct nanoscale computations with high-performance computing, including Grid computing. Two examples in computational nano-mechanics and materials science domain are given to demonstrate the need of high performance computing. The examples include reliability analysis of nanotubes and mechanics of nanotube/aluminum composites. We also proposed the framework of a Grid-based multiscale method.

"Multiscale Computational Materials Simulations for Carbon Nanotubes and Composites"
Deepak Srivastava, Center for Nanotechnology (CNT), NASA Ames Research Center, USA
(1:40PM-2:20PM on Nov. 16, Wed; Room 206)

Abstract: Quantum mechanical modeling of charge transport in two and three dimensional nanodevices is a very challenging problem. In this talk, we will present an overview of our current approach to evaluate the current-voltage characteristics of nanodevices. The main computational challenge in nanodevice modeling is the evaluation of quantum mechanical charge density, which is obtained by solving the non equilibrium Green’s function equations. In comparison to solving the drift-diffusion equation once to obtain the steady state classical charge density, the non equilibrium Green’s function equations have to be solved at each energy to find the quantum mechanical charge density.

Abstract: We investigate the structure and stability of silicon “nano-trees” using large scale quantum mechanical simulations. Our calculations show that such structures are stable for underlying tetrahedral as well as cage like cores making up the “stems” and “branches” of these nano-trees. Electronic structure analysis indicates that the formation of “nano-trees” are accompanied by a narrowing of the HOMO-LUMO gap (Highest occupied molecular orbital - Lowest unoccupied molecular orbital). This has important implications in their application in molecular devices due to the predicted enhancement in the conducting properties. We also comment on the computational complexity of our simulations and discuss how the application could benefit from improved parallel algorithms.