September 2013 - Allocations-admins - mailman3.cels.anl.gov

[LCRC Accounts] Project Request: Anharmonic_systems
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A new project on the LCRC cluster has been requested. Please forward the information on to the LCRC Allocation sub-committee. Applicant's name: Darya Aleinikava Applicant's institution: ANL Applicant's division: CSE Project Name: Anharmonic_systems Project title: Analysis of dynamical complexities of highly anharmonic systems Associated funding: Other Systems: High-performance GPU cluster Golem Science: The goal of this project is to study the dynamical complexities of systems with arbitrary degree of anharmonicity. The first objective of this work is to calculate a statistical descriptor, such as density of states (DOS), and use it to analyze the properties of various structures, such as molecules and clusters, as a function of their composition, size, equilibrium configuration, etc. DOS is the central figure of statistical mechanics since it is the basis for computation of all thermodynamic quantities, and it also plays a major role in a theory of chemical kinetics because it allows a direct evaluation of reaction rates. The non-trivial problem of finding DOS for arbitrary anharmonic systems has been researched extensively over the last 60 years, but no exact solution that would be applicable in all cases was found. We developed an original quantum algorithm with provides an exact solution for arbitrary (including highly anharmonic) systems. This project will involve MD simulations, which are performed on various nanoscale systems with the use of many-body potentials. Such simulations will produce power spectra, which serve as an input into a newly developed algorithm, yielding the exact anharmonic density of states thus eliminating the need to use various approximations. As the second aspect of the project we will address the problem of mixing, where we plan to utilize such descriptors as mixing energy and mixing coefficient, collected during MD simulation runs, to investigate the structural changes of different isomers during high-energy dynamics which is of particular importance in the study of multiple-steps melting transitions. Project description: Efficient codes for original algorithm and MD simulations were developed and extensively tested on a high-performance GPU cluster Golem. We have successfully applied our methodology to Ar, Ne, Al, Ni, and mixed clusters of sizes N=3, 7, 13 and 55. We are planning to use our approach for systems of different compositions with larger sizes, up to N=147 and higher. The number of runs (at different energies) for each system to get a reasonable density of states energy dependence is approximately 15, thus MD simulations time for 70 structures will be 30 core-hours*15 runs*70 structures=31500 core-hours. The new algorithm is implemented in parallel, and its average run will require 8 cores/node * 2 nodes * 25 hours of run time = 400 core-hours, thus the computation of a complete DOS curve for a given system will amount to 400 core-hours *15 runs=6000 core-hours. Including MD simulations time, the total computational time for 70 structures (different in size and composition) will be 31500 core-hours+ 6000 core-hours*70 structures+1000 hours development=452,500 core-hours. The second aspect of the project will be incorporated in MD runs mentioned above. Out of requested 452,500 core-hours, 7% of the time will be allocated for single-core jobs. The number of people to run these simulations is 1. Project URL: Requested allocation: 452500 Q1: 113125 Q2: 113125 Q3: 113125 Q4: 113125 Justification: The requester has used undetermined amount hours of their initial startup project. In addition to approving an initial amount, please specify a Category and Subcategory for this project. For a list of the current selection of approved categories, please see: https://wiki.lcrc.anl.gov/wiki/Processes/Categories Once the Allocation committee has approved the project, please go to the Project Management page to create it: https://accounts.lcrc.anl.gov/projects.php Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Project Request: Bio_catalysis
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A new project on the LCRC cluster has been requested. Please forward the information on to the LCRC Allocation sub-committee. Applicant's name: Cong Liu Applicant's institution: ANL Applicant's division: MSD Project Name: Bio_catalysis Project title: Biomass Catalysis for Energy Associated funding: Director's Fellowship at Argonne Other Systems: Science: A critical challenge for biomass conversion to fuels is to develop efficient multi-functional catalysts that facilitate the complex multi-step processes. The necessary chemical conversions in biomass catalysis often include ring opening (e.g., C-O bond breaking), dehydroxygenation/hydrogenation, and coupling reactions, etc. It is known that most of the metal-containing catalysts are only able to catalyze one specific conversion step effectively, while possibly failing on other steps. This is due to the different reaction mechanisms involved in different chemical conversions. For instance, our previous studies of CO2 activation by 3d transition metal complexes have shown that C-O bond cleavage occurs more easily with earlier metal catalysts while hydrogenation reaction favors with later metal systems. On the other hand, current multi-functional catalysts show lack of efficiency and often require high temperature and pressure. Better understanding the reaction mechanisms of biomass conversions using computational tools is essentially crucial to new catalysts development. Our group has considerable experience in the area of computational modeling of catalytic pathways as well as molecular simulations on accurately predicting thermochemical properties of chemical conversions of sugar molecules. These studies show, among other findings, that reaction mechanisms of biomass catalysis can provide valuable insights and guidance for experimental discovery of novel catalysts. Our work in this area will be using first principle-based calculations to develop better understandings of the possible reaction pathways for converting anhydrosugars and furans to diesel fuels. Anhydrosugars and furans are major primary products from biomass pyrolysis. These C5 or C6 compounds are potential source to synthesize diesel fuels that are a mixture of C8-C21 hydrocarbon compounds. In order to convert the C5 or C6 chemicals into diesel fuels, an important step is to activate a C=C bond to carry out C-C coupling reaction. On the other hand, ring opening and hydrogenation will be used to either saturate a C=C bond or to remove the oxygens. Recently, nano-scale materials have received increased attention on biomass catalysis, because of the ability of these materials to activation inert chemical bonds. For instance, Atesin et al has reported a selective etheric C-O bond hydrogenolysis catalyzed by Ln(Otf)3/Pd-nanoparticles in ionic liquid. Therefore, designing appropr iate nano-scale heterogeneous catalysts that can facilitate C-C coupling reaction, and multi-functional catalysts for both ring-opening and hydrogenation, is a key scientific challenge that we are going to address in this project. Quantum mechanical calculations will be carried out as follows: a. Identifying suitable computational methods The overall reactions involved in the investigations will be identified and the thermodynamic data for these reactions will be analyzed with a series of candidate computational methods, including both DFT and higher level ab initio methods. Based on our experience, B3LYP (DFT) will be utilized to carry out geometry optimizations and frequency calculations. The energy calculations will be performed with also G4MP2, CCSD(T) and MP2 methods. The highly accurate G4MP2 method that was developed by our group will be used as the bench mark method. The reaction energy and barrier will be calculated for these overall reactions. The most suitable method will be chosen based on efficient computational time and acceptable accuracy. b. Reaction pathways The themochemistry of individual reactions with detailed reaction pathways including important intermediates and transition states is the essential first step towards the understanding of the chemical transformations. This will be investigated using the suitable computational method from step a. Since the chemical transformations such as C-C coupling reactions are rather complicated and will generate many possible intermediate and product species, detailed reaction pathways and all the possible species will be calculated. In addition to gas phase calculations, non-aqueous solvent effect will be considered for reaction energies and barriers. c. The catalysis of C-C coupling reactions The C-C coupling reaction is the key step to convert C5-C6 compounds from the biomass pyrolysis to C8-C21 diesel fuel compounds. Diels-Alder (DA) reactions and aldol condensation are available reactions to achieve this type of transformations. Both homogeneous and heterogeneous catalysts will be studied, in which transition metal complexes (will be used for homogeneous catalysis and metal/metal oxide/alloy nanoparticles will be used for heterogeneous catalysis. For the homogeneous catalysis of the DA reaction, for instance, Lewis acid metal complexes will be considered as potential molecular catalysts. The catalytic properties of Lewis acid metal for DA reaction are due to the coordination of the metal to the functional groups of 1,3-diene, causing the decrease of the LUMO energy of the diene. Thus in the studies of the homogeneous catalysts the effect of the metal center will be the focus of interest. Examples of potential catalysts include 3d metal complexes such as titaniu m dichloride diisopropoxide (Ti(IV)Cl2(iPrO)2), Cr(III), Ni(II) and Cu(II) pincer complexes, and Pd (II)(acac)2 complex. These fundamental studies of molecular models will provide constructive guidance for choosing heterogeneous catalysts, and a set of nanoparticle catalysts will be selected for further investigations, and examples include BaTiO3, Pd and Cu nanoparticles. The correlations and differences between the homogeneous and heterogeneous catalysis will be analyzed, in terms of the catalytic properties of the metal, the impact of the ligands, and the steric/structural effects. d. Multi-functional catalysts for ring opening and hydrogenation reaction An important step of getting the diesel fuel products is to remove the unsaturated bonds and oxygens in the longer chain carbon compounds. In many cases, this step involves both ring opening (breaking C-O bond in a ring) and hydrogenation (removing oxygens). To facilitate both reactions, efficient multi-function catalysts need to be developed. An ideal catalyst would be a bimetallic nanoparticle catalyst supported on a high surface area material, in which one metal acts as an oxophilic acid to reduce the polar bond and the other metal is being a base to help produce dissociated hydrogen. Potential catalysts include Au/Fe3O4 and Au/ZnO nanoparticles, as well as Pt or Pd metal nanoparticles-zeolite catalysts. The catalytic properties will be measured by reaction energetic. Project description: Both methodological and chemical investigations will be carried out. Methodological studies will be focused on identifying suitable computational methods for calculating thermochemical properties, which include reaction energies and barriers. For chemical reaction investigations, initial studies on transformations from C5-C6 to C8-C21 will be to map out the reaction pathways. Based on the thermodynamics, dominant reaction pathways will be chosen to explore catalytic reactions. A series of metal nanoparticle catalysts will be screened to determine their catalytic properties. The factors that will be analyzed for the catalytic properties include particle sizes and surface structures. These potential catalysts will be initially chosen from available catalysts for C-C coupling reaction and hydrogenation. This work will be in collaboration with the experimental groups at the National Renewable Energy Laboratory (NREL) and Argonne National Lab (ANL). Project URL: Requested allocation: 480000 Q1: 120000 Q2: 120000 Q3: 120000 Q4: 120000 Justification: The requester has used undetermined amount hours of their initial startup project. In addition to approving an initial amount, please specify a Category and Subcategory for this project. For a list of the current selection of approved categories, please see: https://wiki.lcrc.anl.gov/wiki/Processes/Categories Once the Allocation committee has approved the project, please go to the Project Management page to create it: https://accounts.lcrc.anl.gov/projects.php Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from Li_air_battery
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Kah Chun Lau Project Name: Li_air_battery Division: MSD Project title: Computational studies of Li-Air battery components Associated funding: DOE-BES Other Systems: CNM Science: Unlocking the true energy capabilities of the lithium metal (negative) electrode in a lithium battery has been mainly limited by the low capacity in intercalation and conversion reaction at the positive electrodes. However, it can be overcome by removing these electrodes and allowing lithium to react directly with oxygen in the amosphere, forming the lithium-air battery as being a hybrid battery-fuel cell system. Compared to other metal-air batteries, Zn-air and Al-air cells have aqueous electrolytes and operate at a relatively low voltage of ~1.4 V and ~1.2 V, respectively, whereas for the non-aqueous Li-air cells, it provides ~3 V and therefore yields a significantly higher specific energy. During the electrochemically reduction of oxygen, the small yet highly reactive Li cations tend to form ionic bonds in terms of lithium (per)oxides (i.e. Li2O2 and Li2O) leading to their precipitation on the electrode surfaces. These surface coverage species by the O2 reduction products passivate the electrode, shut down the reduction, and render the reaction irreversible. To advance this new technology, a systematic theoretical study of these system are therefore critically important. To date, today's battery manufacturers and developers have not yet been able to unlock the theoretical potential of Li-Air battery due to : (1) the unstable non-aqueous electrolytes that easily oxidized and decomposed during discharge/charge, thereby seriously limiting cycle life; (2) during discharge, the solid and insoluble Li2O2 and/or Li2O products are deposited on the surface or within the pores of the carbon cathode, thereby clogging the pores , restricting oxygen flow and shut down the electrical conductivity; (3) poisoning of the lithium electrode due to oxygen crossover destroys the integrity and functioning of the cell; and (4) commonly used cathode catalysts, such as metals, metal complexes, and metal oxides, do not access the full capacity of the oxygen electrode or enable sufficiently high rates. Thus in the computational studies, these problems will be addressed and explored based on state-of-the-art ab initio atomistic modeling techniques. This is a project in its 1st year that was awarded under the DOE-BES JCESR led by Argonne. Project description: Project Renewal Request: In the coming fiscal year (FY2014), we will need 1200 000 core-hour to complete our project. We will complete our current on-going work and explore different problems in Li-air batteries based on experimental input from our colleagues at ANL. To complete the current ongoing study (i.e. Li2O2 interaction on Ag_n nanoclusters catalyst/Al2O3 coated electrode), our experience indicates that 100, 000 hours is required. For our new study at FY2014, the computational methods will involve largely AIMD (CPMD, VASP) and DFT method (Gaussian09, VASP). For Gaussian09, the DFT calculation scales well up to 8 processors and for 4-8 nodes is the optimal number of nodes for geometry optimization for system size below 100 atoms. We will require about 200,000 core hours to study the Electron Paramagnetic Resonance (EPR) of O-rich Li2O2, Li2O2 and LiO2 nanoparticles, electrochemical stability and reaction paths of Li2O2 nanoparticles with DMSO and hydrophobic ionic liquids solvents against different Li salts in gas phase and solution phase. To extend our previous DFT study on Li2O2 (crystalline bulk, surfaces and clusters) system, more extensive solid phases of Li2O2/LiO2 and Li2O/Li2O2/LiO2 hetero-interfaces, grain boundaries, and amorphous phases together with the equation of states, mechanical, electronic and thermodynamic properties will be explored using DFT planewave VASP code. The VASP code is well-parallelized for 64-72 cores for 100-200 atoms system for a finite k-point calculation. For a single Γ-point calculation with a total atom ~ 300 atoms, a parallel efficiency of ~ 60% -75% can be achieved using 256 cores. Thus to explore a moderately large systems (~ 200-300 atoms), 200 000 core-hour is needed to perform the tasks. To correlate the Li-air cell with other new metal-air system (e.g. Na-air, K-air cell), the crystalline bulk and surfaces of Na-O2 and K-O2 (e.g. Na2O2, NaO2, K2O2, KO2) based on DFT calculations will be carried out using VASP code. Thus an extra 200 000 core-hour is needed. In addition, the larger system of combined electrolyte/electrode (electrolyte/Li2O2/electrode with and without metal catalyst) interfaces will be computed using CPMD and VASP code. For system involves metal surfaces and metal-catalyst (e.g. Au, Cu, Pt) that involves d-orbital electrons, the constrained-AIMD and nudge-elastic-band method implemented in VASP code and metadynamics approach based on CPMD code will be used to explore the chemical reactions. In particular, the atomistic features of electric double-layer at the metallic cathode interfaces in Li-Air battery will be systematically studied in close collaborations with experimentalist (e.g. Nenad Markovic’s group and K. Amine’s group) at Argonne. In order to have a reliable statistics and long enough simulation time, we will require about 500, 000 core hours to complete these computational intensive simulations. Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 1200000 Q1: 300000 Q2: 300000 Q3: 300000 Q4: 300000 Justification: For the details and evidence how to use the hours effectively, please see the project description and project report (FY2013). Further details of the scaling and the performance of the codes can be found as follows: (1) VASP: http://cms.mpi.univie.ac.at/vasp/vasp/Performance_parallel_code_on_various_… (2) CPMD: http://cpmd.org/documentation (3) Gaussian09: http://www.gaussian.com/g_prod/g09_glance.htm Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from CosmoInfererences
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Rahul Biswas Project Name: CosmoInfererences Division: HEP Project title: Estimation of Cosmological Parameters from Large Surveys Associated funding: Other Systems: Science: The science objective of this project will be to develop codes to estimate cosmological parameters from simulations of future cosmological surveys. Constraining cosmological parameters from multiple probes simultaneously is the best way of improving our knowledge about dark energy, the cause of the late time acceleration of the universe. The focus of interest will be the supernova component of the Dark Energy Survey, but will also be working on the same problem for the Large Synoptic Survey Telescope. The project will explore different methods of performing the parameter estimation, taking into account systematics of these surveys to determine which methods afford the best recovery of information within a practical time scales. A code that prove well suited to the task, might ultimately be used for analyzing the data from these surveys. Project description: The main computational methods involved in the project are Markov Chain Monte Carlo Methods which are mostly embarrassingly parallel. We do aim to try non-Markov methods which can use several processors to reach convergence using multiple walkers. We will also use programs based on ROOT which attempt to minimize large number of parameters. The software requirements for this project are recent intel compilers (C and Fortran with openMP and MPI 8 implementations) v13, Python version 2.7+ , Numpy 1.4+ , Scipy, R, Root . We have tried most of these codes on fusion, and will have to try them on Blues. The number of project members is expected to be less than 10. Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 12000 Q1: 3000 Q2: 3000 Q3: 3000 Q4: 3000 Justification: Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from Comp_Soft_Catalysts
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Juan Lopez-Encarnacion Project Name: Comp_Soft_Catalysts Division: CSE Project title: Soft Catalysts for Green Chemistry and Energy applications Associated funding: Strategic Initiative on Materials for Energy (LDRD-2012-060-R2) Other Systems: None Science: The objective of this project is to perform extensive modeling and computational studies that will aid in design and synthesis of new catalytic materials for efficient transformation of abundant C1 sources (carbon dioxide (CO2) and methane (CH4), in particular) into chemical feedstocks and fuels. The planned theoretical/computational work will be performed in close contact with the ongoing experimental effort on the subject at CSE. The goal is to design and synthesize a new type of catalytic materials that will combine the best characteristics of both homogeneous catalysts (high selectivity at low temperature, tunability) and heterogeneous catalysts (stability, ease of separation). Such catalytic systems are not available at present. Project description: In the FY2014, the computational plan will focus on thermal stability, pore and aperture sizes molecular dynamics at different temperatures of amorphous porous organic polymers (POPs) by using molecular structures of POPs generated in our recent developed code on FY2013. The POPs we are planning to generate are those with the building blocks monomers relevant to experiments in our project. The dynamics/kinetics and elucidation of the paradigmatic reaction mechanism of the catalytic chemical reactions that will be explored experimentally. Specifically, we will continue/complete the exploration of alkane metathesis reaction mechanisms promoted by different type of Ta-based/POP catalysts. One of the targeted processes is conversion of ethane into methane and propane. Based on our preliminary results, we expect the Ta-based reaction centers embedded on POP supports to reduce the relevant reaction barriers as compared to those of the more common supports, e.g ., silica or alumina based. In addition, we are planning to study energetics of elementary reaction steps relevant to alkane, alkene metathesis and ammonia synthesis targeting those critical steps which required more energy, i.e. large energy barriers. The structural and electronic aspects of reactants, intermediates, transition states, and products will be explored and characterized using state-of-the-art density functional theory. The number of atoms treated at a time will vary from a few to a few tens (e.g., in organometallic reaction centers), from tens to about a hundred (e.g., in struts), and to a few hundreds (e.g., in POP networks). In the case of generated 3D amorphous POP networks, the studies in will require inclusion of thousands of atoms, and they will be performed using efficient and at the same time adequate semiempirical potentials and/or force fields to characterize their low-energy structures and dynamical aspects. Then single-point DFT treatments will be performed on these structures to characterize their electronic features using SIESTA or CP2K which allow to performed calculations for thousands of atoms. The dynamical and thermal behavior will be explored using both first principles based and semiempirical potentials/force fields. The kinetic aspects will be evaluated and characterized using transition state theory. In addition to the SIESTA and CP2K codes we will also use other pa ckages (e.g., LAMMPS, NWChem, and Gaussian) as appropriate. The requested time is based on estimates performed on Fusion (see below). The methods used to study the dynamical/kinetics aspects of our systems require considerably computational power/time than those methods used during the FY2013 and the test show below (e.g., crystal, atomic, and electronic structure computations). The expected number of project members is three (3) or more. Project URL: http://www.cse.anl.gov/catalysis_energy_conversion/homogeneous/index.html Current FY Hours Used: undetermined amount New FY Requested allocation: 980000 Q1: 245000 Q2: 245000 Q3: 245000 Q4: 245000 Justification: The scaling efficiency on Fusion was tested using the code SIESTA in the full optimization of the atomic structure and cell volume of a crystalline porous organic polymer, which contain a total of 584 atoms: 192 atoms of H (5 numerical basis function per each H), 360 atoms of C (13 numerical basis function per each C), and 32 atoms of O (13 numerical basis function per each O) in its unit cell; a total of 6,056 basis functions. Substantial speed-up is achieved with the increase in the number of cores: Cores Average elapsed time per optimization step (min) 16 4.38 32 2.70 64 2.34 128 1.68 Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from lith_ox_md
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: John Russell Project Name: lith_ox_md Division: MSD Project title: Molecular Dynamics Calculations of Lithium Oxides Associated funding: Strategic LDRD project entitled Multiscale Modeling of Advanced Electrical Energy Storage Systems Other Systems: Carbon, CNM Science: We will perform molecular dynamics (MD) simulations on nanocrystalline lithium-oxide compounds (including Li2O and Li2O2) of relevance to lithium-air batteries. These simulations include determination of the structures, energies and some properties (such as the diffusivities) of the free surfaces and grain boundaries in these microstructures. A key feature of the simulation approach involves the use of a method to evaluate the Coulomb energy, forces and stresses by direct 1/r summation with spherical truncation (Wolf et al. 1999). This order-N method, which has recently been incorporated into the LAMMPS code that will be used for this work, is computationally superior to both the widely used Ewald method and fast-multipole methods while at the same time producing numerically identical and physically more transparent results. Using this approach, we are able to consider ionic systems consisting of a very large number of ions, thus enabling elucidation of microstructu ral phenomena and physical behavior that can subsequently be used as input to parameterize a physically realistic mesoscale battery model of microstructure evolution during the charge/discharge cycle. Project description: The goal of this project is calculate structural and thermodynamic properties of lithium-oxide materials and to develop parameters for collaborators studying mesoscale models where experimental values may be unavailable. Our initial approach will be to use a Buckingham potential to model covalent interactions and a Wolf summation potential for elecrostatic interactions in a fixed charge Li2O model as implemented in the LAMMPS molecular dynamics package. Our model systems will include free particles, surfaces and nanocrystalline materials. This modeling approach is well suited for large systems due to the favorable scaling of the potentials and software. LAMMPS is massively parallel with nearly linear scaling with system size which can allow for 100k+ atom models for tens of nanoseconds. We are also developing a variable charge lithium-oxide molecular dynamics potential suitable for modeling lithium superoxide, lithium peroxide, and Li2O. The computational cost of this method is estimated as 25x as expensive as a Buckingham pair potential, but with much greater flexibility to model different lithium-oxide stoichiometries. Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 500000 Q1: 125000 Q2: 125000 Q3: 125000 Q4: 125000 Justification: LAMMPS is a linear scaling molecular dynamics code. We will be using the Wolf summation potential to model electrostatics, which is very important due to the fact that FFT methods for electrostatics such as PPPM and PME are NlogN scaling. The Wolf summation is directly incoporated into our variable charge model. The variable charge itself has costs due to charge equilibration, which should put it on par with MEAM and AIREBO. In LAMMPS these methods are compared to a Lennard-Jones (LJ) potential as a benchmark reference (LJ = 1.0, Buckingham ~ 1.0). MEAM is rated at 16.5x and AIREBO at 40.6x. This is primarily due to loops over second nearest neighbors. Therefore, we expect the potential to be in the range of 25x scaling. Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from OSCon
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Dmitry Karpeyev Project Name: OSCon Division: MCS, MSD Project title: Optimizing Superconductor Transport Properties Through Large-scale Simulation Associated funding: DOE ASCR/BES SciDAC-3 project OSCon Other Systems: N/A Science: Most energy applications of superconductivity, such as power transmission over superconducting cables, are based on achieving low energy dissipation in high-temperature superconductors. Dissipation is minimized by restricting the mobility of the vortices carrying magnetic field in the superconducting material by pinning them with admixed inclusions. Understanding the interaction of vortices with general configurations of inclusions is a major outstanding challenge both for fundamental science and energy applications. The task is complicated by the high density of the vortices, their mutual long-range interaction, and the dependence of their behavior on external parameters, such as temperature and the applied magnetic field. These features in general preclude analytical description of vortex dynamics and, until recently, made numerical simulation prohibitively expensive. The new capabilities of DOE's leadership-‐class computing hardware and the development of scalable algorithms and software now put detailed numerical investigation of vortex pinning within reach. Capitalizing on these developments, this project aims to advance the fundamental understanding of vortex dynamics in superconductors and to determine the optimal size, shape and concentration of the admixed particles required to achieve optimal power transmission properties. Project description: This year OSCon will focus on two different computational goals: I. Implementation of the efficient handling of quasiperiodic boundary conditions. II. Building an adjoint-based sensitivity calculation code. I. The implementation of quasiperiodic boundary conditions is a crucial piece in the investigation of the behavior of vortex lattices. Quasiperiodicity places constraints on the grid degrees of freedom that must be correctly reflected in the residual and the Jacobian of the nonlinear system being solved. In particular, this leads to an iterative correction to the sparsity structure of the Jacobian. This in turn incurs substantial computation and communication overhead into the Jacobian assembly. We are investigating efficient methods of circumventing these problems and still producing scalable forward-problem solvers. Analogous issues can be expected in the periodic boundary conditions case on fully unstructured meshes required by OSCon calculations. II. The implementation of the sensitivity code poses a completely different set of challenges. Adjoint calculations requires the integration of the linearized system *backward* in time. This necessitates (a) complete or periodic checkpointing of the forward problem state, incurring substantial I/O overhead, (b) initialization of the linear backward problem at the final time, (c) retrieval or recomputation of the forward problem states at and between the checkpoints, (d) solution of the backward linearized system equations. Most of the development problems are associated to the I/O and/or storage management for checkpointing and are completely new to our project. Project URL: http://www.oscon-scidac.org Current FY Hours Used: undetermined amount New FY Requested allocation: 150000 Q1: 25000 Q2: 25000 Q3: 50000 Q4: 50000 Justification: Thank You, The LCRC Accounts System

1 0

New LCRC report submitted for OSCon
by support＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Full project report at https://accounts.lcrc.anl.gov/reporting/project/2013/?project=OSCon

1 0

[LCRC Accounts] Yearly Allocation Request from ESMF
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Silverio Vasquez Project Name: ESMF Division: MCS Project title: Earth System Modeling Framework Associated funding: DOD, NASA, NSF, NOAA Other Systems: external: NERSC, NCAR, ORNL, NCSA, NASA Science: The Earth System Modeling Framework (ESMF) enables the coupling of model components in high performance Earth system modeling codes. ESMF component interfaces have been implemented in most of the prominent climate and weather codes in the U.S.: CCSM, WRF, NASA GEOS-5, EMC GFS/NEMS, NRL COAMPS and the GFDL MOM4 ocean. In general, componentization with ESMF has been implemented at the level of major physical domains, where simulated interactions require inter-component data communications (e.g. atmosphere, ocean). This implementation is compatible with codes such as WRF that impose their own interoperability conventions at the level of individual physics parameterizations. The codes listed above, plus the full GFDL modeling system under FMS, can now operate within a component-based, hierarchical architecture in which large-scale components can be called as subroutines. This was not the case when ESMF began. Over the years ESMF has evolved and now can be used in multiple ways: 1) to create interoperable component-based modeling systems; 2) as a source of remapping, time management, metadata handling, and other functions that can be used independently of the component constructs; and 3) as a file-based offline generator of interpolation weights for many different kinds of grids. This tool reads in two grid files in NetCDF format and outputs weights for interpolation between the two grids. The recent technical focus has been on continued development of ESMF grid classes and a regridding capability. These are built on a base of two flexible data representations that were constructed in previous years: an unstructured finite element mesh library and a general array class designed to support fields on logically rectangular grids. At this point ESMF can produce conservative and non-conservative interpolation weights, in parallel or serial, in 2D for logically rectangular grids or meshes composed of triangles or quadrilaterals and in 3D for logically rectangular grids or meshes composed of tetrahedra or hexahedra. Project description: The goal of the Earth System Modeling Framework project is to build high performance, flexible software infrastructure to increase ease of use, performance portability, interoperability, and reuse in climate, numerical weather prediction, data assimilation, and other Earth science applications. The framework is intended for use by individual researchers as well as major operational and research centers. ESMF products include an API specification, a reference implementation accompanied by extensive documentation, and an exhaustive regression test suite that includes 4600+ unit tests, system tests, and examples. The ESMF software consists of a superstructure for coupling and exchanging data between models components (e.g., atmosphere, ocean) and an infrastructure consisting of 1) data structures for representing grids and fields and 2) an optimized, portable set of low-level utilities. The data constructs and low-level utilities can be used by the coupling superstructure and may also be used separately to compose scientific applications. As computing architectures evolve, optimizations of the ESMF software is ongoing. Future changes include updates for codes running on computers with general purpose GPUs or accelerators, where data movement in and out of device memory is the source of the major performance bottleneck. Accelerators provide increased computational power which can be leveraged by ESMF in the intensive portion of the code, such as the grid remapping algorithms. However, the main focus in ongoing performance optimization of ESMF is to ensure that the software scales readily and does not introduce overhead as codes start to routinely run on hundreds of thousands of processors. In addition to ESMF full Fortran 90 and partial C/C++ interfaces, ESMPy, a Python package for high-performance grid remapping using ESMF, was released during spring 2013. ESMPy improves on the previous ESMF Python interface, which was called ESMP, by making it more objected oriented. ESMPy is a Python interface that promotes fast prototyping of coupled Earth system modeling codes by providing access to a high-performance and scalable grid remapping package. The main feature is that it allows for high-performance and scalable regridding between 2D and 3D logically rectangular and unstructured grids using a variety of methods and options. At the present time, the ESMPy supports bilinear, finite element patch recovery and first-order conservative methods, for 2D and 3D logically rectangular and unstructured grids. There is also an option to ignore unmapped destination points and mask out points on either the source or destination. Regridding on the sphere takes place in 3D Cartesian space, so the pole problem is not an issue as it can be with other Earth system grid remapping software. Current users of the ESMF Python interface include individuals as well as UV-CDAT (PCMDI) - Ultrascale Visualization Climate Data Analysis Tools; PyFerret (NOAA) – a Python-based interactive visualization and analysis environment; and the Community Surface Dynamics Modeling System (CSDMS) – tools for hydrological and other surface modeling processes. Project URL: http://www.earthsystemmodeling.org Current FY Hours Used: undetermined amount New FY Requested allocation: 10000 Q1: 2500 Q2: 2500 Q3: 2500 Q4: 2500 Justification: Thank You, The LCRC Accounts System

1 0

[LCRC Accounts] Yearly Allocation Request from ESMF
by accounts＠lcrc.anl.gov 04 Sep '13

04 Sep '13

Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Silverio Vasquez Project Name: ESMF Division: MCS Project title: Earth System Modeling Framework Associated funding: DOD, NASA, NSF, NOAA Other Systems: external: NERSC, NCAR, ORNL, NCSA, NASA Science: The Earth System Modeling Framework (ESMF) enables the coupling of model components in high performance Earth system modeling codes. ESMF component interfaces have been implemented in most of the prominent climate and weather codes in the U.S.: CCSM, WRF, NASA GEOS-5, EMC GFS/NEMS, NRL COAMPS and the GFDL MOM4 ocean. In general, componentization with ESMF has been implemented at the level of major physical domains, where simulated interactions require inter-component data communications (e.g. atmosphere, ocean). This implementation is compatible with codes such as WRF that impose their own interoperability conventions at the level of individual physics parameterizations. The codes listed above, plus the full GFDL modeling system under FMS, can now operate within a component-based, hierarchical architecture in which large-scale components can be called as subroutines. This was not the case when ESMF began. Over the years ESMF has evolved and now can be used in multiple ways: 1) to create interoperable component-based modeling systems; 2) as a source of remapping, time management, metadata handling, and other functions that can be used independently of the component constructs; and 3) as a file-based offline generator of interpolation weights for many different kinds of grids. This tool reads in two grid files in NetCDF format and outputs weights for interpolation between the two grids. The recent technical focus has been on continued development of ESMF grid classes and a regridding capability. These are built on a base of two flexible data representations that were constructed in previous years: an unstructured finite element mesh library and a general array class designed to support fields on logically rectangular grids. At this point ESMF can produce conservative and non-conservative interpolation weights, in parallel or serial, in 2D for logically rectangular grids or meshes composed of triangles or quadrilaterals and in 3D for logically rectangular grids or meshes composed of tetrahedra or hexahedra. Project description: The goal of the Earth System Modeling Framework project is to build high performance, flexible software infrastructure to increase ease of use, performance portability, interoperability, and reuse in climate, numerical weather prediction, data assimilation, and other Earth science applications. The framework is intended for use by individual researchers as well as major operational and research centers. ESMF products include an API specification, a reference implementation accompanied by extensive documentation, and an exhaustive regression test suite that includes 4600+ unit tests, system tests, and examples. The ESMF software consists of a superstructure for coupling and exchanging data between models components (e.g., atmosphere, ocean) and an infrastructure consisting of 1) data structures for representing grids and fields and 2) an optimized, portable set of low-level utilities. The data constructs and low-level utilities can be used by the coupling superstructure and may also be used separately to compose scientific applications. As computing architectures evolve, optimizations of the ESMF software is ongoing. Future changes include updates for codes running on computers with general purpose GPUs or accelerators, where data movement in and out of device memory is the source of the major performance bottleneck. Accelerators provide increased computational power which can be leveraged by ESMF in the intensive portion of the code, such as the grid remapping algorithms. However, the main focus in ongoing performance optimization of ESMF is to ensure that the software scales readily and does not introduce overhead as codes start to routinely run on hundreds of thousands of processors. In addition to ESMF full Fortran 90 and partial C/C++ interfaces, ESMPy, a Python package for high-performance grid remapping using ESMF, was released during spring 2013. ESMPy improves on the previous ESMF Python interface, which was called ESMP, by making it more objected oriented. ESMPy is a Python interface that promotes fast prototyping of coupled Earth system modeling codes by providing access to a high-performance and scalable grid remapping package. The main feature is that it allows for high-performance and scalable regridding between 2D and 3D logically rectangular and unstructured grids using a variety of methods and options. At the present time, the ESMPy supports bilinear, finite element patch recovery and first-order conservative methods, for 2D and 3D logically rectangular and unstructured grids. There is also an option to ignore unmapped destination points and mask out points on either the source or destination. Regridding on the sphere takes place in 3D Cartesian space, so the pole problem is not an issue as it can be with other Earth system grid remapping software. Current users of the ESMF Python interface include individuals as well as UV-CDAT (PCMDI) - Ultrascale Visualization Climate Data Analysis Tools; PyFerret (NOAA) – a Python-based interactive visualization and analysis environment; and the Community Surface Dynamics Modeling System (CSDMS) – tools for hydrological and other surface modeling processes. Project URL: http://www.earthsystemmodeling.org Current FY Hours Used: undetermined amount New FY Requested allocation: 10000 Q1: 2500 Q2: 2500 Q3: 2500 Q4: 2500 Justification: Thank You, The LCRC Accounts System

1 0