Bulletin of the American Physical Society
APS March Meeting 2015
Volume 60, Number 1
Monday–Friday, March 2–6, 2015; San Antonio, Texas
Session M23: Focus Session: Petascale Science and Beyond: Applications and Opportunities in Materials Science and Chemistry II |
Hide Abstracts |
Sponsoring Units: DCOMP Chair: Barry Schneider, National Institute for Standards and Technology Room: 202B |
Wednesday, March 4, 2015 11:15AM - 11:51AM |
M23.00001: DAG Software Architectures for Multi-Scale Multi-Physics Problems at Petascale and Beyond Invited Speaker: Martin Berzins The challenge of computations at Petascale and beyond is to ensure how to make possible efficient calculations on possibly hundreds of thousands for cores or on large numbers of GPUs or Intel Xeon Phis. An important methodology for achieving this is at present thought to be that of asynchronous task-based parallelism. The success of this approach will be demonstrated using the Uintah software framework for the solution of coupled fluid-structure interaction problems with chemical reactions. The layered approach of this software makes it possible for the user to specify the physical problems without parallel code, for that specification to be translated into a parallel set of tasks. These tasks are executed using a runtime system that executes tasks asynchronously and sometimes out-of-order. The scalability and portability of this approach will be demonstrated using examples from large scale combustion problems, industrial detonations and multi-scale, multi-physics models. The challenges of scaling such calculations to the next generations of leadership class computers (with more than a hundred petaflops) will be discussed. [Preview Abstract] |
Wednesday, March 4, 2015 11:51AM - 12:03PM |
M23.00002: High performance electronic structure engineering Marco Govoni, Giulia Galli We discuss the efficiency of a recently proposed method for the calculation of energy levels in condensed and finite systems with density functional theory and many-body perturbation theory at the GW level. We present applications of this technique to the calculation of electronic properties of systems with thousands of electrons, including semiconductor nanoparticles, solid/liquid interfaces and defective materials. In addition we discuss the parallel performance and scalability on high performance architectures of a newly developed code [1], implementing the method.\\[4pt] [1] M. Govoni and G. Galli, Large scale GW calculations, submitted. [Preview Abstract] |
Wednesday, March 4, 2015 12:03PM - 12:15PM |
M23.00003: Excited calculations of large scale multiwalled nanotubes using real-space pseudopotential methods Charles Lena, James Chelikowsky, Jack Deslippe, Yousef Saad, Chao Yang, Steven G. Louie One method for calculating excited states is the GW method. The GW method has many computational requirements. One of the bottlenecks is the calculation of numerous empty states. Within density functional theory, we use a real-space pseudopotential method (PARSEC) to calculate these empty states for multiwalled nanotubes. We illustrate the use of these empty states for calculating excited states using the GW method (BerkeleyGW). We demonstrate why using real-space density functional theory is advantageous for calculating empty states. [Preview Abstract] |
Wednesday, March 4, 2015 12:15PM - 12:27PM |
M23.00004: Scalable real space pseudopotential-density functional codes for materials applications James R. Chelikowsky, Charles Lena, Grady Schofield, Yousef Saad, Jack Deslippe, Chao Yang Real-space pseudopotential density functional theory has proven to be an efficient method for computing the properties of matter in many different states and geometries, including liquids, wires, slabs and clusters with and without spin polarization. Fully self-consistent solutions have been routinely obtained for systems with thousands of atoms. However, there are still systems where quantum mechanical accuracy is desired, but scalability proves to be a hindrance, such as large biological molecules or complex interfaces. We will present an overview of our work on new algorithms, which offer improved scalability by implementing another layer of parallelism, and by optimizing communication and memory management. [Preview Abstract] |
Wednesday, March 4, 2015 12:27PM - 1:03PM |
M23.00005: Numerical solutions of the time-dependent Schroedinger equation for atoms and molecules in intense laser fields Invited Speaker: Xiaoxu Guan Recent progress in ab initio computational methods allows us to treat the laser-atom, laser-molecule interaction, and other collision processes with improved accuracy. Full-dimensional quantum calculations for even a few particles are extremely demanding because of the unfavorable scaling of the full quantum wave function, but they are of significant importance for understanding the entangled response of electrons and nuclei in a system strongly influenced by intense lasers and particle beams. In this talk I will concentrate on the applications of grid-based approaches to the time-dependent problems of atoms and molecules driven by intense ultrafast laser pulses. The spatial coordinates are discretized via the finite-element discrete-variable representation. Examples include ionization dynamics in complete breakup processes through few-photon absorption in helium atoms and hydrogen molecules, and also time-delayed attosecond transient absorption spectra in helium. [Preview Abstract] |
Wednesday, March 4, 2015 1:03PM - 1:15PM |
M23.00006: Solving the Time Dependent Schroedinger Equation using the FEDVR/SIL Method Barry Schneider In this talk we will explore how the finite element discrete variational representation coupled to the short iterative Lanczos method has enabled our group to make substantial progress in understanding the single and double ionization of atoms and simple diatomic molecules in short, intense electromagnetic fields. Particular attention will be paid to new time-dependent propagation techniques to shorten the computational times. [Preview Abstract] |
Wednesday, March 4, 2015 1:15PM - 1:27PM |
M23.00007: Hybrid MPI/OpenMP First Principles Materials Science Codes for Intel Xeon Phi (MIC) based HPC: The Petascale and Beyond Andrew Canning, Jack Deslippe, James Chelikowsky, Steven G. Louie Exploiting the full potential of present petascale and future exascale supercomputers based on many core chips requires a high level of threading on the node as well as reduced communications between the nodes to scale to large node counts. We will present results for a variety of first principles materials science codes (Berkeley-GW, PARATEC, PARSEC) on Intel Xeon Phi (MIC) based supercomputers for algorithms using hybrid OpenMP/MPI parallelism to obtain both efficiently threaded single chip performance and parallel scaling to large node counts. [Preview Abstract] |
Wednesday, March 4, 2015 1:27PM - 1:39PM |
M23.00008: Time-Dependent Superfluids on Heterogeneous Computing Platforms Kenneth Roche The superfluid local density approximation (SLDA) and its time-dependent extension (TDSLDA) can be used to study strongly interacting fermion systems. SLDA and TDSLDA have been verified, and validated theoretically and experimentally for both homogeneous and inhomogeneous systems. Our numerical implementation enabled the first time-dependent simulations of fermionic superfluid systems within the DFT approach. Recently, we have developed novel algorithms for the efficient evaluation of the theory using hybrid CPU-GPU computer architectures. In this talk, I will focus on the challenges and implementation details of the hybrid CPU-GPU time-dependent code. [Preview Abstract] |
Wednesday, March 4, 2015 1:39PM - 1:51PM |
M23.00009: Identification of metastable ultrasmall titanium oxide clusters using a hybrid optimization algorithm Eric Inclan, David Geohegan, Mina Yoon Nanostructured TiO$_{2}$ materials have interesting properties that are highly relevant to energy and device applications. However, precise control of their morphologies and characterization are still a grand challenge in the field. Using a hybrid optimization algorithm we theoretically explored configuration spaces of energetically metastable TiO$_{2}$ nanostructures. Our approach is to minimize the total energy of TiO$_{2}$ clusters in order to identify the structural characteristics and energy landscape of plausible (TiO$_{2})_{n}$ (n $=$ 1-100). The hybrid algorithm includes a modified differential evolution algorithm, a permutation operator to perform global optimization on a set of randomly generated structures, and then structure refinement using a BFGS Quasi-Newton algorithm. The results were compared against known physical structures and numerical results in the literature as well as our experimentally synthesized structures. Although the global minimum became more computationally expensive to locate with increasing number of TiO$_{2}$ units, the optimizer successfully identified numerous plausible structures along a range of energies close to the global minimum energy structure for all clusters in the given range. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700