CSMD

The Computer Science and Mathematics Division (CSMD) is ORNL's premier source of basic and applied research in high-performance computing, applied mathematics, and intelligent systems. Basic and applied research programs are focused on computational sciences, intelligent systems, and information technologies.

Our mission includes working on important national priorities with advanced computing systems, working cooperatively with U.S. Industry to enable efficient, cost-competitive design, and working with universities to enhance science education and scientific awareness. Our researchers are finding new ways to solve problems beyond the reach of most computers and are putting powerful software tools into the hands of students, teachers, government researchers, and industrial scientists.

News

Research team stays ahead of the computing curve in monumental climate modeling project (Sep 24, 2014)

CSMD: Rick Archibald, Marcia Branstetter, Kate Evans, Abigail Gaddis, Forrest Hoffman, Tianyu Jiang, Salil Mahajan, Ben Mayer, Galen Shipman, and Pat Worley

Oak Ridge National Laboratory (ORNL) is one of eight Department of Energy (DOE) laboratories that will use high-performance computing (HPC) to develop the most sophisticated Earth system model to date for climate change research with scientific and energy applications. The national labs are collaborating with the National Center for Atmospheric Research, four academic institutions, and one private-sector company on this long-term project, known as Accelerated Climate Modeling for Energy, or ACME. Many of the ORNL team members are also Climate Change Science Institute (CCSI) researchers.

Read more at: http://phys.org/news/2014-09-team-monumental-climate.html#jCp

 

CSMD's Nageswara Rao Wins R&D 100 Award (July 2014)
Diagnosis Using the Chaos of Computing Systems, or DUCCS, was developed at ORNL by Nageswara Rao.

DUCCS ultra-lightweight software quickly and nonintrusively detects a variety of hardware faults in processing units, accelerators, memory elements and interconnects of large-scale high-performance computing systems such as supercomputers, clusters and server farms. The software combines chaotic map theory with advanced CPUs and CPU systems to detect component faults in systems that handle large computational problems such as scientific computations, weather predictions and web data processing. DUCCS software provides critical diagnosis information that contributes to the resilience of computing systems in terms of error-free computations and sustained capacity.

The research was funded by the DOE's Mathematics of Complex, Distributed, Interconnected Systems Project, Applied Mathematics Program, Office of Advanced Scientific Computing Research.

 

 

Science Highlights

A Spectral Transform Dynamical Core Option within the Community Atmosphere Model

K. J. Evans, S. Mahajan, M. Branstetter, J. McClean, J. Caron, M. Maltrud, J. Hack, D. Bader, R. Neale, J. Leifeld

Researchers demonstrated a reasonable mean climate as a benchmark for follow-on higher resolution simulation.

An ensemble of simulations covering the present day observational period using forced sea surface temperatures and prescribed sea-ice extent is configured with a spectral transform dynamical core (T85) within the Community Atmosphere Model (CAM), version 4, and is evaluated relative to observed and model derived datasets. The spectral option is well-known and its relative computational efficiency for smaller computing platforms allows it to be extended to perform high-resolution climate length simulations.

The simulation quality is equivalent to the standard one-degree finite volume dynamical core. The spectral core, which is computationally efficient for smaller computing platforms, is shown to be a viable option for CAM and fully coupled Community Earth System Model simulations.

Work was performed at ORNL, and used OLCF as part of the DOE BER Ultra-High Resolution project

Full Document: http://onlinelibrary.wiley.com/doi/10.1002/2014MS000329/abstract

Multilevel Monte Carlo Method with Application to Uncertainty Quantification in Oil Reservoir Simulation

D. Lu, G. Zhang, C. Webster, and C. Barbier

The rational management of oil and gas reservoirs requires understanding of their response to existing and planned schemes of exploitation and operation. Such understanding requires analyzing and quantifying the influence of the subsurface uncertainty on predictions of oil and gas production. As the subsurface properties are typically heterogeneous causing a large number of model parameters, the dimension independent Monte Carlo (MC) method is usually used for uncertainty quantification (UQ).
By using a multilevel Monte Carlo (MLMC) method to improve computational efficiency in uncertainty quantification for high dimensional problems, researchers can significantly reduce the computational cost, which helps decision-makers make economic and management decisions in a reasonable time.

This work is supported by LDRD program of Oak Ridge National Laboratory.

Full Document: http://www.csm.ornl.gov/newsite/documents/MLMC_DanLu.pdf

Interfacial Properties and Design of Functional Energy Materials

Bobby G. Sumpter, Liangbo Liang, Adrien Nicolaï, Vincent Meunier

Researchers advanced the understanding and provided results that map out a possible way to successfully manipulating self-assembly of functional materials that can deliver improved energy transport, conversion and storage properties.

This control of interfacial and nanostructure (orientation and bonding) impacts a broad range of present and future energy materials, including organic photovoltaics, energy storage, fuel cells, membranes for CO2 capture, gas and water purification, and stronger light-weight materials that can result in energy savings. Additionally, the study of this area of nanoscience continues to provide advances for the design of improved materials for electronic devices and sensors as well as to address future challenges in integrating materials with exceptional properties into other application areas.

This work was performed at the Oak Ridge National Laboratory, the Center for Nanophase Material Science and the Oak Ridge Leadership Computing Facility.

Full Document: http://pubs.acs.org/doi/abs/10.1021/ar500180h

Conformational Electroresistance and Hysteresis in Nanoclusters

Xiang-Guo Li, X.-G. Zhang, and Hai-Ping Cheng

The existence of multiple thermodynamically stable isomer states is one of the most fundamental properties of small clusters. This work shows that the conformational dependence of the Coulomb charging energy of a nanocluster leads to a giant electroresistance, where charging induced conformational distortion changes the blockade voltage. The intricate interplay between charging and conformation change is demonstrated in a Zn3O4 nanocluster by combining a first-principles calculation with a temperature-dependent transport model. The predicted hysteretic Coulomb blockade staircase in the current−voltage curve adds another dimension to the rich phenomena of tunneling electroresistance. The new mechanism provides a better controlled and repeatable platform to study conformational electroresistance.

Work was performed at University of Florida and the Center for Nanophase Materials Sciences, Materials Sciences and Engineering Division. This work was supported by the US Department of Energy (DOE), Office of Basic Energy Sciences (BES), under Contract No. DE-FG02-02ER45995. A portion of this research was conducted at the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National Laboratory by the Division of Scientific User Facilities (X.-G.Z.). The computation was done using the utilities of the National Energy Research Scientific Computing Center (NERSC).

Full Document: http://pubs.acs.org/doi/abs/10.1021/nl5014458

Almost Perfect Diffraction Limited Beam Combination from a Commercial Quality Non-Identical Array of High-Power Broad Area Laser Diodes

Bo Liu and Yehuda Braiman

Researchers demonstrated coherent beam combination from an array of high-power broad area laser diodes with almost perfect beam quality (1.5-1.6 diffraction limit) and 95-99% visibility, high power conversion efficiency (in the range of 20%), and narrow line-width (0.1 nm).

Free-running broad area laser diodes support multiple spatial modes and subsequently emit very poor beam quality. Demonstration of almost perfect phase synchronization and coherence from commercial quality, non-identical, broad area array is of fundamental importance to our understanding of how large, heterogeneous, multiple spatial mode cavities phase-synchronize. This demonstration has direct impact on variety of applications including beam combining of high power lasers for (a) directed energy, (b) laser communication source, and (c) laser pump of fiber and solid state lasers.

Funding for this work was provided by the Office of Naval Research.

Full Document: http://www.csm.ornl.gov/newsite/documents/OptExp_21_31218_2013.pdf

Stochastic Parameterization to Represent Variability and Extremes in Climate Modeling

R. Langan, R. Archibald, M. Plumlee, S. Mahajan, D. Ricciuto, C. Yang, R. Mei, J. Mao, X. Shi, and J. S. Fu

Unresolved sub-grid processes, those that are too small or dissipate too quickly to be captured within a model's spatial resolution, are not adequately parameterized by conventional numerical climate models. Sub-grid heterogeneity is lost in parameterizations that quantify only the 'bulk effect' of sub-grid dynamics on the resolved scales. A unique solution, one unreliant on increased grid resolution, is the employment of stochastic parameterization of the sub-grid to reintroduce variability. The researchers administered this approach in a coupled land-atmosphere model, one that combines the single-column Community Atmosphere Model (CAM-SC) and the single-point Community Land Model (CLM-SP), by incorporating a stochastic representation of sub-grid latent heat flux to force the distribution of precipitation. Sub-grid differences in surface latent heat flux arise from the mosaic of Plant Functional Types (PFT) that describe terrestrial land cover. With the introduction of a stochastic parameterization framework to affect the distribution of sub-grid PFT's, the researchers alter the distribution of convective precipitation over regions with high PFT variability. The stochastically forced precipitation probability density functions (pdf) show lengthened tails, demonstrating the retrieval of rare events. Through model data analysis they show that the stochastic model increases both the frequency and intensity of rare events in comparison to conventional deterministic parameterization.

Coupled land-atmosphere climate calculations were run using Oak Ridge Leadership Computing Facility's (OLCF's) Titan supercomputer. Funding for this work was provided by the US Department of Energy through ORNL's LDRD program.

Full Document: http://www.sciencedirect.com/science/article/pii/S1877050914002804

Understanding How to Manipulate the Nanoscale Assembly of Organic Photovoltaic Donor/Acceptor Films

M. Shao, J. K. Keum, R. Kumar, J. Chen

Researchers reported the results of a comprehensive investigation of the effects of the processing additive diiodooctane (DIO) on the morphology of the established blend of PBDTTT-C-T polymer and the fullerene derivative PC71BM used in organic photovoltaics (OPVs), starting from the casting solution and tracing the effects to spun-cast thin films by using neutron/x-ray scattering, neutron reflectometry and other characterization techniques corroborated by theory and modeling. The results reveal that DIO has no observable effect on the structures of PBDTTT-C-T and PC71BM in solution, however in the spun-cast films, it significantly promotes their molecular ordering and phase segregation, resulting in improved power conversion efficiency (PCE). Thermodynamic analysis based on Flory-Huggins theory provides a rationale for the effects of DIO on different characteristics of phase segregation due to changes in concentration resulting from evaporation of the solvent and additive during film formation. In summary, a comprehensive suite of characterization techniques and theoretical analyses revealed both the lateral and vertical morphological effects of the processing additive diiodooctane, DIO, on the formation of bulk-heterojunctions and the resulting organic photovoltaic device parameters starting from a donor/acceptor polymer blend PBDTTT-C-T:PC71BM in solution, to the spin-cast films.

This research was conducted at the Center for Nanophase Materials Sciences (CNMS), High flux Isotope Reactor (HFIR) and Spallation Neutron Source (SNS) at Oak Ridge National Laboratory. This research was conducted at the Center for Nanophase Materials Sciences (CNMS), High flux Isotope Reactor (HFIR) and Spallation Neutron Source (SNS) that are sponsored at Oak Ridge National Laboratory by the Scientific User Facilities Division, U.S. Department of Energy. KX and DBG acknowledge the support provided by a Laboratory Directed Research and Development award from the Oak Ridge National Laboratory (ORNL). M. Shao and J. K. Keum contributed equally for this work.

Full Document: http://onlinelibrary.wiley.com/doi/10.1002/adfm.201401547/abstract

A Computer Program for Uncertainty Analysis Integrating Regression and Bayesian Methods

D. Lu, M. Ye, M. C. Hill, E. P. Poeter, and G. P. Curtis

This work develops a new functionality in UCODE_2014 to evaluate Bayesian credible intervals using the Markov Chain Monte Carlo (MCMC) method. The MCMC capability in UCODE_2014 is based on the FORTRAN version of the differential evolution adaptive Metropolis (DREAM) algorithm, which estimates the posterior probability density function of model parameters in high dimensional and multimodal sampling problems. This paper tests and demonstrates the MCMC capability using a 10-dimensional multimodal mathematical function, a 100-dimensional Gaussian function, and a groundwater reactive transport model. The use of the MCMC capability is made straightforward and flexible. With the new MCMC capability, UCODE_2014 can be used to calculate three types of uncertainty intervals, linear confidence intervals, nonlinear confidence intervals, and MCMC Bayesian credible intervals. Ready access allows users to select methods best suited to their work, and to compare methods in many circumstances.

Full Document: http://www.sciencedirect.com/science/article/pii/S1364815214001662

Rational Design of Nanostructured Polymeric Materials

R. Kumar, S. W. Sides, M. Goswami, B. G. Sumpter, K. Hong, X. Wu, T. P. Russell, S. P. Gido, S. Rangou, K. Misichronis, A. Avgeropoulos, T. Tsoukatos, N. Hadjichristidis, F. L. Beyer, and J. W. Mays

Researchers developed an understanding of the role of conformational asymmetry in self-assembly of ordered multi-block copolymer morphologies. This understanding enhances their ability to effectively utilize self-assembly to generate nanoscale structures (morphologies) over large 3D volumes that are important for improving multiscale functional materials.

This work was performed at the Oak Ridge National Laboratory, the Center for Nanophase Material Science and the Oak Ridge Leadership Computing Facility.

Full Document: http://pubs.acs.org/doi/abs/10.1021/la304576c

Hierarchical acceleration of a stochastic collocation method for partial differential equations with random input data

M. Gunzburger, P. Jantsch, A. Teckentrup, and C.Webster

Drawing inspiration from recent work in multilevel Monte Carlo methods, this work proposed a multilevel stochastic collocation method, based on a hierarchy of spatial and stochastic approximations. A detailed computational cost analysis showed, in all cases, a sufficient improvement in costs compared to single-level methods. Furthermore, this work provided a framework for the analysis of a multilevel version of any method for SPDEs in which the spatial and stochastic degrees of freedom are decoupled. The numerical results practically demonstrated this significant decrease in complexity versus single level methods for each of the problems considered. Likewise, the results for the model problem showed multilevel SC to be superior to multilevel MC even up to N = 20 dimensions (see right).

This work is sponsored by the Department of Energy's Advanced Scientific Computing Research program.

For more information about this work, please go to www.equinox.ornl.gov.

The Eclipse Integrated Computational Environment and its Readiness for High Performance Computing

Jay Jay Billings

The Eclipse Integrated Computational Environment (ICE) provides a common, extensible platform that improves productivity and streamlines the workflow for computational scientists. It has successfully integrated tools and simulation suites from across the DOE complex into a single, unified, cross-platform workbench. It works well on everything from standalone machines to large clusters, including running and managing remote parallel jobs as well as connecting to remote visualization engines and retrieving remote data. Recent work is extending the platform to work on Titan and other Leadership-class resources where launching jobs in the queue and rendering large scale visualizations are a priority.

ICE enhances the productivity of computational scientists by streamlining their workflow. It automates many difficult and tedious tasks and encapsulates confusing details. It increases the accessibility of sophisticated modeling and simulation tools and high-performance computer systems for those with limited experience in such environments.

The development of ICE has been sponsored by DOE Office of Nuclear Energy, Advanced Modeling and Simulation (NEAMS) Program and the DOE Office of Energy Efficiency and Renewable Energy, Computer-Aided Engineering for Batteries (CAEBAT) project.

A Web-based Visual Analytic System for Understanding the Structure of Community Land Model

D. Wang, Y. Xu,, T. Janjusic, and X. Xu

Researchers developed a Fortran CLM specific code syntax parser to extract application data-structure flow. This parser enables a web-based visual system to understand CESM/CLM code structures.

This work was performed by Oak Ridge National Laboratory, Climate Change Science Institute, and Oak Ridge Leadership Computing Facility.

Publication: Y. Xu, D. Wang, T. Janjusic, and X. Xu, "A Web-based Visual Analytic System for Understanding the Structure of Community Land Model", The 2014 International Conference on Software Engineering Research and Practice, July 21, 2014, Las Vegas, Nevada.

HERCULES: Strong Patterns Towards More Intelligent Predictive Modeling

Eun Jung (EJ) Park, Christos Kartsaklis, and John Cavazos

Researchers introduced a static technique to characterize a program using the pattern-driven system HERCULES. This characterization technique not only helps a user to understand programs by searching pattern-of-interests, but also can be used for a predictive model that effectively selects the proper compiler optimizations. They formulated 35 loop patterns, then evaluated their characterization technique by comparing the predictive models constructed using HERCULES to three other state-of-the-art characterization methods.

The researchers showed that their models outperform three state-of-the-art program characterization techniques on two multicore systems in selecting the best optimization combination from a given loop transformation space. The researchers achieved up to 67% of the best possible speedup achievable with the optimization search space they evaluated.

Publication: EunJung Park, Christos Kartsaklis, John Cavazos, "HERCULES: Strong Patterns Towards More Intelligent Predictive Modeling", 43rd International Conference on Parallel Processing (ICPP 2014).

This work was performed at the Oak Ridge Leadership Computing Facility.

Stochastic Finite Element Methods for Partial Differential Equations with Random Input Data

M. Gunzburger, C. Webster, and G. Zhang

Together with Max Gunzburger (Florida State University), Clayton Webster and Guannan Zhang published a review article in the premiere applied mathematics book series ACTA Numerica, Cambridge University Press, Volume 23, pp. 521650, 2014. The review article is entitled "Stochastic finite element methods for partial differential equations with random input data." Acta Numerica is the top-cited mathematics journal for the last two years in MathSciNet. It was established in 1992 to publish widely accessible summaries of recent advances in the field. Its annual volume of review articles (5-8) is by "invitation-only" and includes survey papers by leading researchers in numerical analysis and scientific computing. The papers present overviews of recent advances and provide state-of-the-art techniques and analysis. Covering the breadth of numerical analysis, articles are written in a style accessible to researchers at all levels and can serve as advanced teaching aids.

This work is sponsored by the Department of Energy's Advanced Scientific Computing Research program.

For more information about this work, please go to equinox.ornl.gov

Events

January 6, 2015 - David M. Weiss: Industrial Strength Software Measurement

ASCR Workshop on Quantum Computing for Science (February 17-18, 2015)

Travis Humble

At the request of the Department of Energy's (DOE) Office of Advanced Scientific Computing Research (ASCR), this program committee has been tasked with organizing a workshop to assess the viability of quantum computing technologies to meet the computational requirements in support of DOE's science and energy mission and to identify the potential impact of these technologies. As part of the process, the program committee is soliciting community input in the form of position papers. The program committee will review these position papers and, based on the fit of their area of expertise and interest, selected contributors will have the opportunity to participate in the workshop currently planned for February 17-18th, 2015 in Bethesda, MD.

Visit the site [here].