Find below the results of DECI-4 (Distributed European Computing Initiative) call.
Projects from the following research areas:
Project Title: Large-scale dynamo action in shearing stratified convection
Project Leader: Maarit Korpi, University of Helsinki, Observatory, Finland
Resource Awarded: 900,000 core hours
Axel Brandenburg, AlbaNova University Center, Nordita, Sweden
The mechanism responsible for the observed magnetic activity phenomena of the Sun still remains controversial: the challenge is to understand how ordered magnetic fields can arise in the turbulent solar convection zone. One possible mechanism is a large-scale hydromagnetic dynamo, the basic ingredients of which include convective turbulence and large-scale shear due to solar differential rotation. Whilst simple theoretical arguments suggest that such a dynamo should be capable of efficient large-scale magnetic field generation, numerical simulations of turbulent shearing convection exhibiting large-scale dynamo action have been lacking. As a possible explanation of this it has been suggested that large-scale dynamos cannot co-exist with the small-scale dynamo, which inevitably arises in the high Reynolds number (Rm) regime, and which washes out any large-scale structures. We have, for the first time, produced a set of runs just reaching the desired Rm regime where the small-scale dynamo is operating, which also show clear large-scale dynamo action. Our results indicate that shear is the key ingredient, not only as a generator of magnetic fields via streching, but also as a mediator of magnetic helicity fluxes. The latter is vital for the dynamo for dumping its own garbage, the small-scale magnetic helicity, out of the system. We propose to investigate this issue further with higher resolution runs reaching higher Rm, more realistic shear profile and boundary conditions, to check whether or not the large-scale dynamo action persists also in this regime. The result (positive or negative) will have very important implications for dynamo theory.
Project Title: The Gravitational Billion Body Problem
Project Leader: Dr. Simon Portegies Zwart, University of Amsterdam, Astronomical Institute A. Pannekoek, The Netherlands
Resource Awarded: 3,150,000 core hours on EPCC – HECToR DC, CSC – Louhi QC, SARA – Huygens P6, EPCC – EPCC – HECToR DC QC,CSC – Louhi XT, EPCC – EPCC – HECToR DC EPCC – HECToR QC2
Prof. Andreas Burkert, Ludwig-Maximilians Universität München, Germany
Prof. Douglas Heggie, University of Edinburgh, School of Mathematics, UK
Kei Hiraki, Tokyo University, Japan
Prof. Jun Makino, Center for Computational Astrophysics, Japan
Prof. Steve McMillan, Drexel University, USA
Cees de laat, University of Amsterdam, The Netherlands
Cold dark matter (CDM) is more successful in explaining the structure in the universe than any other cosmological model. Galaxies, according to CDM, accumulate from small clumps of dark matter which grow by subsequent mergers to large galaxies. Each large galaxy is then accompanied by a number of smaller unmerged dark matter clumps in their halos. These small clumps are generally associated with the dwarf galaxies observed in the halos of every major galaxy.
Cosmological simulations of dark-matter halos predict about a thousand dwarf galaxies (see Fig. 2 of ), but the Milky-way galaxy has only 24 dwarf galaxies . This problem is known as the missing dwarf-galaxy problem, and it is illustrated in Fig. 1. We propose to perform a large cosmological dark-matter simulation through which we will effectively address this problem.
We plan on running a cold dark-matter N-body simulation of the small cluster of galaxies at extremely high resolution. This calculation is interesting for two main reasons:
1) we will address the substructure of dark matter halos, in particular in relation to the dwarf galaxy problem and
2) the calculation will computational be extremely challenging, since we plan on running concurrently on five of the largest computers on the planet via one of the fastest internet connections.
The results of our simulation will become publicly available via the web page http://wiki.2048x2048x2048.org. The simulation data will be used for public outreach via the planetarium at Artis (the zoo in Amsterdam), which has expressed interest in using our simulation data for one of their shows.
Project Title: Accelerating Ligand Binding Energetic Calculations and Clay Intercalation Studies
Project Leader: Prof. Peter V. Coveney, University College London, Centre for Computational Science, UK
Resource Awarded: 700,000 core hours on RZG – VIP
Dr. H. Christopher Greenwell, University of Durham, UK
Four research projects will be undertaken building on programmes of work funded through major EU and UK based initiatives. The first is to support a combination of EU funding, in the FP6 funded e-Health project called ViroLab that requires access to DEISA resources to meet our objectives within Europe, and the UK EPSRC/MRC/BBSRC funded life sciences interface doctoral training centre ’CoMPLEX’ programme (EP/F500351/1). This project aims to demonstrate the feasibility of HIV/AIDS related patient-specific drug treatment based on the determination of patient-specific protease and reverse transcriptase drug binding affinities, which must also be computed in a matter of days to impact clinical decisionmaking.
The second project aims to address simulation of the action of novel anticancer therapies and will underpin work funded in the EU funded Contra Cancrum project and Virtual Physiological Human Network of Excellence. The final two projects are materials science related and both are funded UK national priorities; one intends to extend our studies of clay-polymer nanomaterials in order to better understand their materials properties and boosts capability in the UK Technology Strategy Board funded ’New Improved Muds from Environmental Sources’ project (CRD Q2506L) and adds to the EPSRC ’RealityGrid’ platform grant (EP/C536452/1); the other is concerned with lattice-Boltzmann simulations aimed at unravelling the role of defects in the large scale rheological properties of amphiphilic fluids and DEISA time will boost research to the EPSRC ’large-scale lattice Boltzmann simulations of liquid crystals’ project (EP/E045111/1). These applications will exploit highly scalable molecular dynamics codes and associated workflows; these can be managed by the Application Hosting Environment that seamlessly links distributed HPC resources, both within DEISA and to other machines, such as HPCx and HECToR, in the UK.
Project Title: The role of metals in β-amyloid interactions. A proposal for an ab initio study
Project Leader: Paolo Giannozzi, University of Udine, Department of Physics, Italy
Resource Awarded: 400,000 core hours on CSC – Louhi QC,CSC – Louhi XT
Nils Christian, Humboldt University, Germany
Karl Jansen, NIC/DESY Zeuthen, Germany
Giovanni La Penna, National Research Council (CNR), Institut for chemistry of organo-metallic compounds (ICCOM), Italy
Velia Minicozzi, University of Rome Tor Vergata, Italy
Silvia Morante, University of Rome Tor Vergata, Italy
Gian Carlo Rossi, University of Rome Tor Vergata, Italy
Francesco Stellato, University of Rome Tor Vergata, Italy
The interest in elucidating the role of metals in the development of amyloid diseases (to the group of which the Alzheimer’s disease belongs) has significantly increased after noticing that Cu and Zn chelators can be used to solubilize the amyloid-beta aggregates (Aβ-aggregates) which appear to make up the fibrillar material associated with AD.
There exist in the literature somewhat conflicting statements on whether Cu2+ and Zn2+ ions can provide protection against or act as promoters of plaques formation. Recent X-ray Absorption Spectroscopy experiments designed to study the metal atomic environment in Cu-Aβ and Zn-Aβ complexes have been able to detect visible differences between Zn and Cu coordination modes.
Even on the specific geometrical structure of the Cu binding site experimental data are not fully conclusive. On the one hand, in fact, NMR data on copper complexed with Aβ1-28 and Aβ1-40 peptides suggest that three Histidines (His6, His13 and His14) are coordinated to the metal with the fourth ligand being the N-terminal nitrogen. On the other, XAS experiments on various portions of the natural Aβ protein and NMR/EPR experiments on the Aβ1-16 and Aβ1-28 peptides, all point to the conclusion that, besides the above three Histidines, the fourth copper ligand is most probably the oxygen of Tyr10. Clarifying this question is not without interest, because it is expected that the more or less open structure of the peptide can strongly influence its aggregation propensity. Typically one may suspect that a coordination mode where the N-terminal is not bound to Cu will give rise to a more open geometry and thus probably more prone to aggregation structure.
For the purpose of elucidating the structure of the (Cu2+/Zn2+)-Aβ binding site, simulations of the Car-Parrinello type appear to be the most appropriate tool, as they provide a first principle quantum mechanical computation of the atomic force field. In this project we propose to carry out a thorough first principle study of the metal environment in Cu-Aβ and Zn-Aβ complexes.
Project Title: Simulations of bacterial efflux pumps
Project Leader: Prof.Dr. Ulrich Kleinekathöfer, Jacobs University, Bremen, Germany
Resource Awarded: 346,500 core hours on FZJ – JUGENE
Prof.Dr. Paolo Ruggerone, Università degli Studi di Cagliari, Italy
Keeping the upper hand with bacterial resistance against antibiotics will require that Information at molecular level should be available form different techniques, and computational simulations can play an important role thereon. One of the resistance mechanisms implies the overexpression of the efflux pumps, very complex systems involved in the extrusion of different substrates (and also antibiotics and other drugs) from the bacteria. The efflux systems we are interested in belong to the RND family and are responsible for the transport of substrates through the inner as well as the outer membrane of gram-negative bacteria. Since the atomic structure of several efflux pumps were resolved recently or reliable models have been assembled via homology, one now has the possibility to study the structure-function relationship computational via large-scale molecular dynamics (MD) simulations. Both the Italian and the German groups have experience in modelling the influx of ions and antibiotics through cell membranes and now join forces and experience to tackle the more complex problem of efflux pumps. Because of this the system of study consists of several hundred thousand atoms. The MD code NAMD developed by the Schulten group which has excellent parallel scaling properties will be used on the DEISA infrastructure to study these very large systems in an European project which would be otherwise impossible. Studying molecular details of the efflux of different antibiotics in different bacteria, such as E.Coli and P.Aeruginosa, will help to understand why antibiotics are pumped out of the bacterial cell and therefore are unable to fulfil there task of destroying the cell.
Project Title: Microanatomically Accurate and Functionally Realistic Bidoain Models of the Heart
Project Leader: Gernot Plank, University of Oxford, Oxford eScience research Center, UK
Resource Awarded: 1,100,000 core hours on CINECA – CNE-SP6
Computer simulations of the electrical activity of the heart using the bidomain equations have become commonplace during the last decade. However, even whenthe most powerful HPC resources are used, attempts to integrate behaviour from the protein scale of ion channels to the organ scale of cardiac arrhythmias remain enormously challenging and, typically, include significant simplifications to permit the exploration of the parameter space of interest in a tractable fashion. Typically, the geometry of the heart is represented in a stylized fashion or only parts of the heart are modeled. In constraining the degrees of freedom, the choice of the spatial discretization often leads to under-representation of the finer anatomical details and, quite often, ad-hoc adjustments of parameters are required to avoid artificial scaling effects. Many studies resort to using the simpler monodomain formulations which do not account for current flow in the extracellular space and, thus, feedback mechanisms of external current flow on the activation sequence is ignored; The specialised cardiac conduction system is typically not represented, preventing the simulation of the most important control case, the electric activity driven by the sinus node; And finally, myocardial membrane ion transport kinetics is modeled using simplified descriptions which are incapable of accounting for many potentially arrhythmogenic mechanisms.
This study aims to overcome these limitations by finely discretizing the heart to account for anatomical details while employing the latest descriptions of the membrane dynamics based on on Markov state models. After initially optimizing the codes, a micro-anatomically accurate mesh of a rabbit heart will be tuned and validated. Further, simulations will be executed to tackle questions related to arrhythmogenic effects of metabolic sinks. Effects of anatomical structure, functional heterogeneity and pharmacologically induced variations of the myocardial response on the electrocardiogram will also be investigated.
Project Title: Multiscale simulation of membrane-associated multiprotein complexes
Project Leader: Rebecca C. Wade, EML Research gGmbH, Germany
Resource Awarded: 900,284 core hours on CSC – Louhi QC, IDRIS – BABEL
Mark S. P. Sansom, University of Oxford, Dept. of Biochemistry, UK
Membrane-associated multiprotein complexes play a critical role in many biological processes. The modelling and simulation of such complexes is very challenging for computational studies due to their high number of degrees of freedom and their heterogeneity.
Currently, there are established procedures for docking proteins that are membrane-associated are lacking. In this project, we will work towards the goal of establishing a set of tools and workflows for modelling and simulating membrane-associated protein-protein complexes. A multiscale approach will be taken with models using both all-atom and coarse-grain models of the protein, lipid and water molecules, as well as simpler models in which the solvent is treated as a continuum. The most computationally intensive calculations will be performed with molecular dynamics codes that show good scaling to thousands of processors. The results will enable us to establish procedures for modelling multiprotein-lipid complexes.
Moreover, they will provide insights into two systems of biological importance, the cytochrome P450 drug metabolizing enzymes and the Rab GTPase proteins which play a regulatory role in vesicle trafficking.
Project Title: Designing anti-HIV drugs targeting RNA by molecular simulation
Project Leader: Paolo Carloni, SISSA/ISAS, Trieste, Italy
Resource Awarded: 400,000 core hours on CSC – Louhi QC
Michele Parrinello, ETH Zurich, Switzerland
Gabriele Varani, University of Washington, USA
RNA is a potentially very attractive target for pharmaceutical intervention against infectious diseases, yet standard computational approaches are ineffective because docking procedures have limited success in predicting the poses and the relative potency of ligands binding to it. Here we will use advanced and innovative computational approaches, based on the metadynamics method, to address this fundamental issue in the context of anti-HIV drug discovery. Based on our calculations, we will be able to design new ligands targeting viral RNA, which will be then tested experimentally. This work seeks to establish a new and powerful protocol for the computer-aided ligand design targeting RNA.
Project Title: Coupling the Chemistry in Earth System Models on multiple Scales
Project Leader: Dr. Patrick Jöckel, Max Planck Institute for Chemistry, Germany
Resource Awarded: 1,822,500 core hours on SARA – Huygens P6
Dr. Astrid Kerkweg, Institute for Atmospheric Physics, University of Mainz, Germany
Dr. Andrea Pozzer, The Cyprus Institute, Energy, Environment and Water Research Center, Cyprus
In Earth system modelling, the interacting domains (e.g., atmosphere, hydrosphere, cryosphere, etc.) of the environment are simulated in conjunction to study the nature of feedbacks between the different domains and processes and how they influence the properties of the whole system. A challenging task in particular is to represent the constituent cycles of chemically active species, since a wide range of temporal and spatial scales and a large number of species are involved. A global and a regional model are applied to address two different issues which both require tailor made coupling strategies: First, the chemical influence of the ocean on the atmospheric composition includes processes with a strong diurnal cycle (e.g. photochemical production), requiring a frequent exchange of information between the ocean and the atmosphere models. Second, the downscaling from the global to the regional and local scale by nesting a regional model into a global model requires the frequent transfer of boundary conditions from the global model to the regional model and between different instances (in different resolutions) of the regional model. For both model setups the coupling is implemented following different approaches in the Modular Earth Submodel System (MESSy, http://www.messyinterface.org) framework. The infrastructure provided by DEISA provides an ideal environment to systematically test and optimise the new developments.
Project Title: Cloud formation in moist convective turbulence
Project Leader: Prof.Dr. Jörg Schumacher, Technische Universität Ilmenau, Institute for Thermodynamics and Fluid Mechanics, Ilmenau, Germany
Resource Awarded: 759,000 core hours on FZJ – JUGENE
Olivier M. Pauluis, New York University, Center for Atmospheric and Ocean Science, USA
Katepalli R. Sreenivasan, International Centre for Theoretical Physics, Trieste, Italy
Moist convection is an extremely complex physical process involving the turbulent fluid motion and the complicated thermodynamics of phase changes. It is an ubiquitous feature of the Earth’s atmosphere, manifesting itself through the various clouds observed every day around the globe. The combination of turbulence and thermodynamics makes moist convection a particularly difficult physical problem. Its necessary parametrization in global atmospheric models remains a major source of uncertainty for prediction of future climate change. In this project, we will study moist convection in an idealized setting using piecewise linear thermodynamics. Our focus is to simplify the thermodynamics but to fully resolve the turbulence in the presence of phase changes. Our model will compare the behaviour of moist convection with the well-known case of dry convection and analyze the formation, the lifetimes and the statistics of shapes of clouds. The targeted fine spectral resolution allows for a detailed study of the impact of moisture on the turbulent heat transport through the layer and the small-scale structure of turbulence. Our results will suggest new strategies for coarse-grained parametrizations in global climate models.
Project Title: Millennial climate
Project Leader: Prof. Heikki Järvinen, Finnish Meteorological Institute, Finland
Resource Awarded: 1,800,000 core hours on LRZ – HLRB II
The Finnish Meteorological Institute (FMI) will simulate the global climate over the best documented period of Earth’s climate, i.e., from the year 1000 B.P. to present, and a further period of two hundred years into the future. The computational code consists of a coupled atmosphere-ocean general circulation model including the land and ocean biogeochemistry, allowing simulation of the global carbon cycle. The model configuration closely follows the one used in the Millennium experiment (www.mpimet.mpg.de/en → science → Millennium experiment) of the COSMOS network (cosmos.enes.org). Here, the special emphasis is on the soil organic carbon for which a new and more realistic modelling treatment is included.
The main objectives of the simulation are (1) to comprehensively assess the realism of the models used for climate projections in the light of past climate variability and (2) to evaluate the low-frequency ecosystem feedbacks, especially as regards the soil organic carbon. A small ensemble of simulations (an ensemble of 4-6 members) is proposed in order (i) to assess the stability of the model over an extended period of time, and (ii) to obtain some uncertainty estimates of the simulated variability in the carbon fluxes between different carbon pools.
The proposed DEISA experiment, MillCli, is far beyond the computational resources of FMI but it would greatly advance our understanding of using very demanding Earth system models in computational climate research, and improve our readiness in simulating the Earth’s climate beyond the 100 – 200 year time window from present. FMI will be mainly responsible for the experiment itself but the model development and the analysis of the scientific results is shared with the University of Helsinki and the Finnish Environmental Institute.
Project Title: Ground Motion Simulation for Sedimentary Basins
Project Leader: Dr. Martin Käser, Ludwig-Maximilians-Universität München, Department of Earth and Environmental Sciences, Geophysics, Germany
Resource Awarded: 2,313,625 core hours on BSC – MareNostrum, LRZ – HLRB II, RZG – Genius
Dr. Jan Burjánek, Swiss Seismological Service, Switzerland
Dipl.Geophys. Andreas Fichtner, Ludwig-Maximilians-Universität München, Department of Earth and Environmental Sciences, Geophysics, Germany
Dr. Martin Galis, Comenius University Bratislava, FMFI, Slovak Republic
Dr. Jozef Kristek, Comenius University Bratislava, FMFI, Slovak Republic
Prof.Dr. Peter Moczo, Comenius University Bratislava, FMFI, Slovak Republic
Dr. Josep de la Puente, Instituto de Ciencias del Espacio, CSIC-IEEC, Barcelona, Spain
Computational Seismology is becoming an increasingly important tool in order to understand the effects of complex 3D subsurface structures on the propagating seismic wave field. Therefore, highly accurate computer simulations are necessary to resolve the exact waveforms caused by the geometrical, rheological and site-specific properties of the material and by the spatial and temporal description of the seismic source. The creation of such reliable synthetic data sets in form of seismograms and their comparison to real observational data allows us in a further step to produce sharper tomographic images of the subsurface using full wave form inversion techniques. However, this approach relies heavily on the iterative re-calculation of the complete 3D forward problem after appropriate model adjustments and therefore increases the computational cost easily by orders of magnitude. Today, this approach still is hardly followed, as the required computational resources basically have been out of reach, and has led to a number of simplified and therefore much less accurate methodologies in the past. However, for a detailed and reliable simulation-driven seismic hazard assessment the available 3D model information has to be included into cutting-edge ground motion modelling. This is particularly important in specific areas, such as densely populated and industrialized sedimentary basins in order to generate accurate shake maps and well-defined geological subsurface models.
Project Title: Numerical simulation of turbulent flow past a circular cylinder at Re=20,000-200,000
Project Leader: Roel Verstappen, University of Groningen, Institute of Mathematics and Computing Science, The Netherlands
Resource Awarded: 650,000 core hours on BSC – MareNostrum
Xavier Trias Miquel, Univerisitat Politècnica de Catalunya, CTTC, Spain
Jaap de Wilde, MARIN (Maritime Research Institute Netherlands), The Netherlands
For all but the lowest Reynolds number the flow around a cylinder separates and vortices are formed in the wake. The asymmetric shedding of vortices into the wake induces forces on the cylinder. These forces can cause the cylinder to vibrate. Such a vibration is termed a Vortex-Induced Vibration (VIV). Pipelines linking the seabed to the offshore platform for oil production, for example, exhibit VIV. Marine riser pipes are often exposed to high Reynolds number currents, Re>100,000 where the Reynolds number is based on the free stream flow speed, the diameter of the pipe and the viscosity of water. The fluid excitation of marine riser pipes forms a potent source of fatigue. Studying the potentially destructive consequences of VIV requires a coupled fluid and structural dynamics model. Since the numerical simulation of turbulent flow around a rigid cylinder at Re>100,000 forms a grand challenge to current Computational Fluid Dynamics (CFD) models we will concentrate on solving the flow problem in this project. To that end, simulations based on a symmetry-preserving regularization model of turbulent flow will be performed in the range Re=20,000-200,000. At Re=20,000, the results will be compared to that of Direct Numerical Simulation (DNS).
Project Title: Direct Aeroacoustics by using Optimal System Architectures and Domain Decomposition
Project Leader: Prof.Dr. Claus-Dieter Munz, Universität Stuttgart, Institut für Aerodynamik und Gasdynamik, Germany
Resource Awarded: 442,368 core hours on NEC SX-9, HLRS – Laki
The costs for the direct simulation of aeroacoustic problems can become very high if not insurmountable due to a multi-scale problem: Noise producing small-scale structures such as vortices as well as the propagating acoustics with large wavelengths and small amplitudes need to be resolved highly accurate at the same time. On the other hand, direct simulations require the least modelling and include the retroaction of the acoustics to the flow field, a feature that other hybrid models do not offer.
In order to reduce the computational effort of direct simulations, the calculation domain is divided into subdomains, where the grids, methods, time steps, orders of accuracy, etc. are adapted to the local physical requirements. By doing so, the optimal solver is used for each domain. To give an example, an unstructured mesh is only employed in the vicinity of a complicated geometry (e.g. a nozzle), while Cartesian grids are used farther away from an obstacle, delivering faster and more accurate results.
Moreover, each solver should run on the most suitable processor: High-order methods for unstructured grids, such as the Discontinuous Galerkin (DG) and Finite Volume (FV) methods are efficient on cache-based machines with scalar CPUs. On the other hand, Finite Volume and Finite Difference schemes for structured grids show good vectorization properties and can therefore especially exploit vector architectures such as the NEC-SX8.
The IAG in-house code KOP3D is a framework which provides different numerical methods and a well-tested coupling procedure. While being a completely stand-alone code, it uses PACX-MPI to organize the data distribution between different system architectures.
In the past, KOP3D has successfully demonstrated the efficiency of the domain decomposition approach for a sphere scattering benchmark example and shall now prove its applicability to real life 3D calculations. For this purpose, a supersonic nozzle flow with a free-jet (Ma=1.4, Re=30000) is calculated in 3D.
Project Title: Direct Numerical Simulation of Nanofluids
Project Leader: Alfredo Soldati, University of Udine, Italy
Resource Awarded: 1,200,000 core hours on RZG – VIP
Nanofluids are dilute liquid suspensions of nanoparticulate solids, including nanoparticles, nanofibers and nanotubes. Experiments have demonstrated that nanofluids can be used to design new-concept heat transfer fluids with properties given by those of the base fluid modulated up to the target, desired amount by the presence of dynamically-interacting, suitably-chosen, discrete nanoparticles. As of today, a clearcut understanding of the modifications of the physical heat transfer mechanisms occurring in nanofluids is still lacking. A possible way to improve the knowledge of these mechanisms is to use accurate and reliable numerical tools such as direct numerical simulation (DNS) and Lagrangian particle tracking (LPT), which may complement complex and costly experiments.
In this project, we propose to perform a comprehensive numerical analysis accounting for mass, momentum and heat transfer mechanisms all together, tailored for the specific case of nanodispersed fluids. Nanoparticles will be modeled as active heat transfer agents interacting both with the temperature field and the velocity field to study heat transfer modifications. To this aim, necessary energy and momentum coupling terms must be incorporated in the governing equations of both phases.
The main objective of the present proposal is thus to investigate microscale heat transfer enhancement mechanisms occurring in nanofluids by means of pseudospectral DNS at high resolution (17 to 135 million grid points) and LPT of large swarms of nanometer-size particles (1 billion), focusing on turbulent channel flow at different values of the Reynolds and Prandtl numbers. Such parametric study is currently unavailable, first because of the challening non-trivial modeling issues, which of course reflect upon the complex interactions between the two phases; and second because of the (unprecedented) huge computational cost required by the high grid resolution combined with the large number of individual nanoparticles to be tracked.
Project Title: Direct Numerical Simulation of turbulent flames at high Reynolds numbers
Project Leader: Dominique Thevenin, University of Magdeburg Otto von Guericke, Germany
Resource Awarded: 831,600 core hours on EPCC – HPCx, IDRIS – BABEL, EPCC – EPCC – HECToR DC X2
Bénédicte Cuenot, CERFACS, France
Alain Laverdant, ONERA, France
Orlando Rivera, Leibniz Supercomputing Centre, Germany
Combustion phenomena are of essential scientific and technological interest, in particular for the improvement of energy generation and transportation systems. For many decades to come an overwhelming part of human energy needs will still be covered by combustion of fossil fuels. Obviously, combustion-driven devices (Internal Combustion Engines, gas turbines, furnaces…) must be optimized to reduce pollutant emissions and fuel consumption. Direct Numerical Simulations (DNS) have become an essential and well-established research tool to investigate turbulent combustion, since they do not rely on any approximate turbulence model. In this project some of the most powerful supercomputers available in Europe will be used to carry out DNS of turbulent flames at realistic (i.e. high) values of the Reynolds number. These simulations will be used in particular to check the importance of two modelling issues concerning possible extensions or modifications of the Navier-Stokes equations: a) volume viscosity; b) additional terms for non-isothermal flows. For both issues, controversial information can be found in the literature. All computations will employ the reactive Navier-Stokes equations including accurate models for chemistry (complex reaction mechanisms) and molecular transport (mixture-averaged diffusion coefficients). DNS are ideally suited for this purpose since all scales down to the Kolmogorov scale are resolved in the simulation, leading to a ’numerical experiment’. For these investigations the ignition and propagation of turbulent hydrogen and n-heptane flames, modelled using up to 30 chemical species, will be considered. This allows a suitable description of most chemical systems used in practical combustion applications, from simple fuels up to complex hydrocarbon molecules.
Project Title: Fluid Dynamics of Film Cooling Investigated by Large-Eddy Simulation
Project Leader: Prof. L. Kleiser, ETH Zurich, Institute of Fluid Dynamics, Switzerland
Resource Awarded: 300,000 core hours on NEC SX-8, NEC SX-9
Prof. U. Rist, Universität Stuttgart, Institut für Aerodynamik und Gasdynamik, Germany
High blade temperatures limit the performance of gas turbines. Film cooling is used to lower the blade temperature by ejecting cold gas through holes in the blade surface directly into the boundary layer. The resulting flow configuration is very complex with respect to the geometry and the involved physical flow phenomena. Numerical simulations of film cooling are mostly done using Reynolds-averaged Navier-Stokes (RANS) simulations which by their nature are incapable of capturing the full physics of this problem. Therefore, we investigate this flow by large-eddy simulations (LES) which are able to resolve the essential flow structures and the large-scale turbulence in space and time. For modelling the effect of subgrid scales we use our approximate deconvolution model (ADM).
We aim at simulating large film cooling configurations with an accurately represented geometry, increasing the level of complexity step by step. This will allow us to investigate in detail the influence of separate effects on the film cooling efficiency (e.g., the inflow disturbance level or the arrangement of cooling holes). Although LES need only about one percent of the computational time of corresponding direct numerical simulations (DNS), they still require exceptionally large computing resources.
Project Title: Global instability Analysis of TUrbulent Separated flows
Project Leader: Vassilis Theofilis, School of Aeronautics, UPM, Spain
Resource Awarded: 2,240,000 core hours on FZJ – JUGENE
The instability properties of the turbulent separated flow over aircraft airfoils at high angles of attack are studied through the solution of partial-derivative-based eigenvalue problems. The complex geometry and involved physics make it impossible to use simplified analysis, within the possibilities of personal computers. Massive parallelization is required then to obtain physical meaningful results, useful on the industrial fiel
Project Title: Numerical Simulation of complex hydrodynamic and MHD flow phenomena
Project Leader: Prof.Dr. Claus-Dieter Munz, Universität Stuttgart, Institut für Aerodynamik und Gasdynamik, Germany
Resource Awarded: 198,000 core hours on FZJ – JUGENE
The goal of this project is the numerical simulation of complex large scale flow phenomena with an explicit high order accurate Discontinuous Galerkin (DG) scheme.
Project Title: The Langevin equation for Large Reynolds number turbulent flow
Project Leader: J.G.M. Kuerten, Technical University of Eindhoven, The Netherlands
Resource Awarded: 1,000,000 core hours on SARA – Huygens P6, NEC SX-9
In many examples in industry and the environment particles are transported in a turbulent flow. A well-known example is dispersion of pollution in the atmosphere or in a river. Particle motion is influenced by all scales of motion of the turbulent flow. Due to the enormous computational requirements calculation of particle dispersion by solving the complete flow field by means of direct numerical simulation (DNS) is only possible at relatively low Reynolds numbers. For larger Reynolds numbers a statistical description of the particle behaviour might offer a solution. Recently, such a method has been proposed based on a stochastic Langevin equation for particle velocity. In order to determine the coefficients in this model Lagrangian velocity correlation functions are required. The aim of this proposal is the determination of these velocity correlation functions at intermediate values of the Reynolds number by means of DNS and passive particle tracking in turbulent channel flow. In this way, the recently developed model can be validated and tested and an efficient model for the calculation of particle dispersion in turbulent flow at large Reynolds number has been obtained.
Project Title: Atomistic Coulomb Explosion Simulations at X-rays
Project Leader: Dr. Andrea Fratalocchi, Universita di Roma la Sapienza, Department of Physics, Italy
Resource Awarded: 400,000 core hours on RZG – VIP
Dr. Claudio Conti, National Research Council (CNR), Institute for the Physics of Matter (INFM), Italy
Prof. G. Ruocco, National Research Council (CNR), Institute for the Physics of Matter (INFM), Italy
The early days of 21st century are experiencing a revolution in synchrotron source intensities driven by the new generation of X-ray Free Electron Lasers (XFEL). Such sources will be able to deliver femtosecond pulses of peak powers above 100GW, characterized by atomic-scaled wavelengths and a high degree of spatial and temporal coherence. This opens in the near future the possibility to perform single molecule experiments, involving the real-time reconstruction of molecular images from diffraction patterns recorded at x-rays. However, the physics behind such phenomena, at the intensities reached by XFEL sources, is in many respect unknown. In particular, the most problematic issue concerns the problem of radiation damage of samples exposed to intense XFEL radiation. The process is triggered by the XFEL photoionization of core electron shells that, in turn, either escape or decay through Auger relaxation, forcing the system to accumulate a strong positive charge around the nucleus. This results into an abrupt explosion of the atom as due to the unbalanced Coulomb forces mutually exerted by the ions. A quantitatively analysis of this process, which is fundamental for the next generation of imaging experiments, is still lacking.
The ACES-X project employs a massively parallel numerical approach to deal with this problem. Our ab-initio theory makes use of molecular dynamics for ions evolution, of time-dependent density function theory for electrons dynamics and of nonlinear finite-difference time-domain methods for the exact propagation of electromagnetic waves. This approach will allow for the real-time simulation of Coulomb explosion with reference to realistic samples, as well as the theoretical determination of the ionization intensity and optimal pulse length for the next generation of molecular imaging experiments.
Project Title: Addressing Molecular Nanomagnets On Surfaces
Project Leader: Valerio Bellini, INFM-CNR S3, Modena, Italy
Resource Awarded: 500,000 core hours on FZJ – JUMP P6, CINECA – CNE-SP6
Alessandro Bencini, University di Firenzeo, Dipartimento di Chimica, Italy
Franca Manghi, INFM-CNR S3, Modena, Italy
Federico Totti, University di Firenzeo, Dipartimento di Chimica, Italy
Quantum computing, ultra-high density magnetic storage, and molecular electronics are some of most appealing and promising fields of novel and important technological breakthroughs. Single Molecule Magnets (SMM) can find important application in such relevant fields since in many of them the requirement of small magnetic ’units’ is mandatory. Nevertheless, to be technological appealing, they must be organized on solid surfaces or wired to metal electrodes in a controlled way. This represents the nowadays challenge and a key achievement is represented by the functionalization of the magnetic ’core’ of the SMM with appropriate chemical groups with the aim to preserve both the uncommon magnetic properties and to make it suitable for being attached to a surface.
SMM can be grafted on surfaces mainly through two different class depending on the type of interactions, i.e. covalent interaction through functionalized groups (bond formation) or direct π-π interaction (Van der Waals stacking); in addition, they can experience direct magnetic interaction in presence of a magnetic surface or their native magnetic properties can be preserved or even enhanced. Understanding and predicting, with the appropriate theoretical tools, how such molecular magnets arrange and organize themselves onto the surfaces, and how much their magnetic properties can be varied, is something that experimentalists are strongly asking in order to rationalize the new experimental data available through STM, XCMD, EPR, micro-EPR, etc.
We individuated two optimal SMM candidates, each representing one of the class of ystems described above; the Fe4(C11H19O2)6[(OCH2)3C(CH2)9SCOCH3]2 grafted on Au(111) and the Co phthalocyanine molecule (CoC32H16N8), adsorbed on Ni(100). The programs suited for handling such big systems are CP2k and Wien2k.
Project Title: Accurate Quantum Understanding of Adsorption: water on carbon
Project Leader: Prof. Dario Alfe, University College London, Department of Earth Sciences, UK and Prof. Angelos Michaelides, University College London, London Centre for Nanotechnology, UK
Resource Awarded: 1,228,800 core hours on CSC – Louhi QC
Enge Wang, Chinese Academy of Sciences, Institute of Physics, China
The interaction of water with carbon is central to an almost endless list of scientific areas. For example, the adsorption of water on carbon nanotubes is relevant to the properties and function of nanotubes in biological media and the water-graphite interaction is key to lubrication and to chemical reactions in the interstellar medium. Despite the ubiquity and importance of water-covered carbon, it is remarkable that the most fundamental and important question of how strong the bond between water and carbon is, is neither well-established experimentally nor theoretically. With this project we plan to tackle this question with a novel set of ’first principles’ (parameter free) quantum Monte Carlo simulations for water on grapheme and on a carbon nanotube. The simulations will be carried out with the highly efficient parallel code CASINO; leading to the first reliable theoretical estimates of the strength of the bond between water and two forms of carbon. The data to come out of this joint UK-China project will be of tremendous interest to the large and active international scientific community focussed on understanding the properties of different forms of carbon. In addition, it will facilitate future innovation in the development of cheaper computational approaches (e.g. density functional theory and parameterised potentials) for treating weakly interacting physisorption systems.
This project, which is impossible without the computational capacity offered by the DEISA extreme computing initiative, will push the first principles calculation of surface processes to a new level of sophistication and in so doing propel European science to the forefront of capability computing in materials science.
Project Title: Quantum-Mechanical Prediction of Gas Phase Biomolecular Secondary Structure
Project Leader: Dr. Volker Blum, Max Planck Society, Fritz-Haber-Institute, Berlin, Germany and Prof. Matthias Scheffler, Max Planck Society, Fritz-Haber-Institute, Berlin, Germany
Resource Awarded: 3,072,000 core hours on LRZ – HLRB II, CSC – Louhi QC, RZG – VIP, SARA – Huygens P6, EPCC – EPCC – HECToR DC QC, EPCC – EPCC – HECToR DC EPCC – HECToR QC2
Prof. Ville Havu, Helsinki University of Technology, Institute of Mathematics, Finland
The accurate computational prediction of peptide / protein behaviour is among the most competitive research areas today, but fully QM predictions (here: van der Waals corrected density functional theory) for such systems are still hampered by the computational demands for the necessary molecular dynamics timescales. DEISA will allow us to advance these limits significantly, by predicting conformational equilibria and ensembles of two benchmark peptides, Ac-Ala19-Lys-H+ and (Ac-Lys-Ala19- H+)2 (220 and 440 atoms, respectively). Key parts of the results can be directly compared to accurate gas-phase experimental data (vibrational spectroscopy), and will provide a key check and corroboration of the current ideas of secondary structure formation and its driving force in such molecules.
Project Title: Phase Transitions in discotic semiconductor LC
Project Leader: Prof. Claudio Zannoni, Università di Bologna e INSTM, Italy
Resource Awarded: 1,200,000 on EPCC – EPCC – HECToR DC QC, EPCC – EPCC – HECToR DC EPCC – HECToR QC2
We propose to perform the first predictive computer simulation of the phase organisation and transitions of a discotic liquid crystal [1,2] as employed in organic electronic devices, using all atom force fields and massive parallel computer simulations. While some studies exist on the subject [3,4,5], they failed to provide in-silico realistic phase transition temperatures, as well as a quantitative agreement with experimental data for properties, due to the short time scales reached and the limited system sizes afforded by the computational power available.
Our aim is to exploit the computational power behind the DEISA infrastructure to accomplish the above tasks, relying in our previous experience in simulations of liquid-crystalline cinnamates  and cyanobiphenyls . These studies represent the first successful predictions of transition temperatures for rod-like nematics ; here we aim at extending the approach to the more complex columnar systems, focusing on the hexa-hexyl-thio-triphenylene discotic molecule, particularly interesting both for its charge transport properties  and its mesomorphism .
Project Title: Extremely Versatile Inorganic Nanotubes
Project Leader: Dr. Stefano Leoni, Max Planck Institute für Chemische Physik fester Stoffe, Dresden, Germany
Resource Awarded: 675,000 core hours on SARA – Huygens P6
Dr. Francesco Mercuri, National Research Council (CNR), Institute of Molecular Sciences and Technologies, Italy
Prof.Dr. Gotthard Seifert , Technische Universität Dresden, Germany
The eVerINT project meets the challenge of simulating mechanical and electronic properties of inorganic nanotubes and nanoparticles. In doing so, it intends to strongly contribute to the field of computational nanotechnology, and support the development of applications related to the field of electronic structure and tribology computation on complex nanostructured materials. These targets aim at designing and realizing novel nanostructures with interesting electronic properties, which can find application as photovoltaic and/or energy storage devices or as coatings or additives as a new generation of solid lubricants. Indeed, the outstanding lubrication properties and extremely long life times of these new materials indicate inorganic nanostructures as a new and promising class of material. In the simulation strategy, the project is going to interrogate the electronic and mechanical response of chemical systems like TiO2, MoS2 or WS2, on system dimensions ranging from few hundred to order of 105 atoms, for a level of theory based on DFT.
The highly scalable CPMD, CP2K and CP- DFTB codes are used in connection with advanced molecular dynamics simulation schemes to elucidate atomistic details of critical process of material deformation. The implementation of the codes fully benefits from the use of aggregate memories, needed for the treatment of large systems with first-principles techniques, with a performance speed-up. Additionally, the use of DFTB drivers will empower the project to simulate systems of millions of atoms.
The eVerINT project is presented in the framework of the collaboration between the Max-Planck-Institute for the Chemical Physics of Solids Dresden, Germany, Technische UniversitÃ€t, Dresden and the CNR Institute for Molecular Sciences and Technologies Perugia, Italy
Project Title: Ab Initio Simulation of Hydrous Silicates
Project Leader: Jürgen Horbach, DLR (German Aerospace Center), Institute of Materials Physics in Space, Germany
Resource Awarded: 1,800,000 core hours on RZG – VIP
Dr. Matthias Krack, Paul Scherrer Institut (PSI), Switzerland
A crucial step for the understanding of geological processes (e.g. volcanic eruptions) as well as the materials properties of technological glasses is the understanding of microscopic mass transport in silica (SiO2) and mixtures of silica with other oxides. Of particular interest are silicates containing water. Even small amounts of water reduce the viscosity of a silicate melt by orders of magnitude. Despite its importance for glass technology and geosciences, many aspects of the structural arrangement and the diffusive transport of water in silicates are largely unknown. In this project, mixtures of SiO2 and water (hydrous silica) will be studied by ab-initio simulations. Born-Oppenheimer molecular dynamics (BO-MD) simulations based on density functional theory (DFT) will be performed with the Quickstep code which employs a hybrid basis set of Gaussian functions and plane waves. This scheme is particularly suitable for the simulation of large systems and thus we will be able to study systems of about 500 particles on a time scale of about 50 ps. Melts and glasses with different water content ranging from 1 wt% to 10 wt% H2O will be considered. The first aim is to analyze how the water molecules are built into the Si-O network. Here, the accuracy of the simulation can be directly tested by comparing the static structure factor to recent neutron scattering data on hydrous silica. The second aim is to elucidate the mechanism how OH or other water-based units diffuse through the Si-O network. Also in this case, our simulation results can be compared to recent quasi-elastic neutron scattering experiments on the proton dynamics in hydrous silica. On the other hand, the detailed information provided by the simulation helps to better understand the results from the neutron scattering experiment.
Project Title: Mechanisms of material cracking induced by hypervelocity nanoparticle impacts
Project Leader: Prof. Kai Nordlund, University of Helsinki, Department of Physics, Finland
Resource Awarded: 570,000 core hours on EPCC – HECToR DC QC
The aim of the study will be to detect cracking and shock wave induced strengthening/weakening of materials upon hypervelocity nanoparticle impact using large-scale molecular dynamics simulations. Nanoparticle impacts occur, for instance, in space where small meteoroids hit surfaces of planets, moons, and spacecraft. They are also important in industrial processing of surfaces when cluster ion beams are used. In our recent study it was shown that transition to macroscopic cratering behaviour occurs already at very small impactor sizes (impactor diameters less than 20 nm). The next step that will be taken in this study, is to find out whether also the macroscopic cracking mechanisms occur at this size regime.
Project Title: Parallel Simulation of Electron Transport in Nanostructures
Project Leader: Boris N. Chetverushkin, Institute for Mathematical Modeling of RAS, Russia
Resource Awarded: 196,608 core hours on BSC – MareNostrum
Sergey V. Polyakov, Institute for Mathematical Modeling of RAS, Russia
The aim of the project is to investigate the electron transport processes in the nanostructures for the purpose of creation of the next generation of electronic devices, using quantum effects. We intend to consider electron processes in the semiconductor nanostructures with two-dimensional electron gas and the electron auto-emission processes in silicon cathodes with nano sizes. The suggested problems are of great innovative potential. Most of the program support for the project is already elaborated and adapted to high-performance computing. Use of DEISA systems can significantly put forward the numerical simulations and to increase the accuracy of computational results.
Project Title: Ab-initio study of GeSbTe phase change alloys
Project Leader: Michele Parrinello, ETH Zurich, Switzerland
Resource Awarded: 1,050,000 core hours on CSC – Louhi QC
Marco Bernasconi, University of Milano-Bicocca, Dept. Materials Science, Italy
Dr. Matthias Krack, Paul Scherrer Institut (PSI), Switzerland
Phase change materials are widely used in optical information storage (DVD) and are the active part of the most promising non-volatile memories of new concept, the Phase Change Memory (PCM) device which exploits the large change in conductivity between the crystalline metallic phase and the insulating amorphous phase of a chalcogenide film. Although Ge2Sb2Te5 (GST) is presently the material of choice for PCM, doping and changes in composition of GST are under scrutiny to improve PCM performances and scalability. In spite of their great technological importance, the physical insight in chalcogenide materials and their reversible crystal-to-amorphous phase-change mechanisms is still far from being satisfactory. By making use of a novel ab-initio molecular dynamics scheme (Kühne et al., Phys. Rev. Lett. 98, 66401 (2007)), developed in the group of the principal investigator, we have recently provided the first reliable atomistic model of amorphous GST by quenching from the melt. In the present project we plan to extend the ab-initio simulations to other ternary systems along and off the pseudo-binary line (GeTe)x(Sb2Te3)y in order to establish useful relationships among the microscopic structure, the stability of the amorphous phase and the crystallization speed measured experimentally. The simulation results would provide critical insight for the search of new better performing materials in this class for PCM applications.
Project Title: Quantum Monte Carlo simulation of High-Temperature superconducting materials
Project Leader: Sandro Sorella, SISSA/ISAS, Trieste, Italy
Resource Awarded: 800,000 core hours on SARA – Huygens P6
Claudio Attaccalite, Universidad del País Vasco, Vitoria, Spain
Michele Casula, Ecole Polytechnique, France
Leonardo Spanu, University of California Davis, USA
In this project we study two important high temperature (HTc) superconducting materials, La2CuO4 that is the prototype of copper oxide superconductors discovered in 1988, and LaFeAsO, representative of the novel iron based rare-earth oxypnictide superconductors.
By means of state-of-the-art quantum Monte Carlo (QMC) techniques, our main target is to understand whether the peculiar magnetic properties of the two materials are related to the pairing mechanism that leads to superconductivity. The superconducting regime in the iron-based and copper-oxide families occurs in close proximity to a long-range ordered antiferromagnetic state. This suggests a strong interplay between magnetism and HTc superconductivity, which is probably one of the most challenging problems in condensed matter physics.
In the past, different lattice models, such as the Hubbard and the t-J Hamiltonians, have been introduced for a simplified theoretical description of the HTc superconductors. Within these models a superconducting phase arising only from electron correlation has been found by several QMC techniques.
Project Title: Simulation of Nuclear Fuels
Project Leader: Dr. Matthias Krack, Paul Scherrer Institut (PSI), Switzerland
Resource Awarded: 900,000 core hours on LRZ – HLRB II
Marcella Iannuzzi , Paul Scherrer Institut (PSI), Switzerland
The world-wide growing demand for energy and in particular electric energy has revived the interest in nuclear energy. Many countries have already extended the scheduled time of operation for their nuclear power plants in consideration of the expected energy supply gap. Even the construction of new and more advanced (Generation III and III+) nuclear power plants is planned, because longer operation times of old plants alone are insufficient to bridge the gap and might also raise safety concerns. Moreover, due to the global warming low CO2 emission has become an important issue for the design of new power plants and nuclear power plants fulfill this condition very well. Uranium dioxide (UO2) is the main component of the currently employed nuclear fuels. An in-depth knowledge about the fuel material and its behaviour under various conditions is crucial to ensure the safe operation of nuclear power plants. However experiments with hot materials are difficult and very costly. Alternatively, computer simulations can be employed nowadays to investigate the nanostructure and the structural dynamics of fuel materials like UO2. In the proposed project we plan to investigate UO2 using an electronic structure method based on density functional theory (DFT). Finite temperature molecular dynamics (MD) simulations will be performed at DFT level to deliver insights from first principles into the structural and dynamical behaviour of pristine and defective UO2 at the atomic level. The CP2K program package (http://cp2k.berlios.de) will be employed for the study which has already proven to run on parallel supercomputers with high efficiency. The proposed activity will be performed in the framework of the EU FP-7 project F-BRIDGE.
Project Title: Simulation of reactivity at interfaces
Project Leader: Philippe Sautet, Ecole Normale Supérieure de Lyon, France
Resource Awarded: 1,260,000 core hours on CINECA – CNE-SP6
Jean-Sébastien Filhol, University of Montpellier II, France
Tony Lelièvre, Ecole Nationale des Ponts et Chaussées, France
Hervé Toulhoat, Institut Français du Pétrole, France
The breaking and formation of chemical bonds is at the crossroad of a large number of natural or industrial processes, not only in chemistry but also in life, materials, earth and environment sciences. Reactions at the solid/gas or solid/liquid interface have a special importance. However, chemical reactivity is a quantum phenomenon, which is difficult to model. Ab initio methods are precise but limited to a small number of atoms. Reactive force fields have started to be developed, but their accuracy is still insufficient. Reactivity of complex molecules on surfaces also requires the exploration of potential energy surfaces for a large number of degrees of freedom.
The SIRE project has been financed by the French ANR agency for 3 years (project ANR-06-CIS6-014, 2007-2009) and it aims, by a combination of mathematics and theoretical chemistry, at developing new simulation methods for reactive processes at surfaces: search of reaction paths, including environment, temperature and pressure effects, new reactive force fields, methods for electrochemistry. These codes are designed for highly parallel computers. At half way in the project, new innovative modules have been developed for reaction pathway searches (CARTE), for electronic structure calculations at the electrochemical interface (Metelec), and are coupled to various academic electronic structure codes (VASP, PWSCF, CPMD).
In the second half of the project (September 2008 – December 2009) these methods will be applied to selected challenges in heterogeneous catalysis and in electrochemistry, with strong societal or industrial implications. These applications are very innovative for the realism of the description of the catalytic system, but very computationally intensive. ANR did not finance the CPU time for these applications but recommended us to submit applications to European facilities. This is clearly the purpose of this proposal, which gathers 4 institutions and 24 researchers, including 6 post-doc or students dedicated to the project. We wish to underline that this application to DECI is especially timely, since last year the methods were not ready, and since in 2010 the SIRE project will be finished.
Project Title: Slater Type Orbital Project for Quantum monte carlo large molecule simulations
Project Leader: Philip Hoggan, LASMEA-UMR6602, CNRS, Clermont Universit, France
Resource Awarded: 1,440,000 core hours on FZJ – JUGENE, RZG – Genius
R. Assaraf, LCT-UMR7616, CNRS, Paris VI University, France
Peter Reinhardt, LCT-UMR7616, CNRS, Paris VI University, France
Slater-type orbitals (STO) are rarely used as atomic basis sets for molecular structure and property calculations, since integrals are expensive to evaluate, reliable basis sets are scarce and exact properties such as Kato’s cusp condition and the correct exponential decay of the electron density are not significantly better described numerically than with commonly used Gaussian basis sets.
We propose the systematic parallelized development of integration routines for multi-centre integrals, and high-quality basis sets over STOs, useful for modern electron correlation calculations via compact low-variance trial wave-functions for QMC (Quantum Monte Carlo).
Molecular QMC applications are also rare, because the method is comparatively complicated to use, however it is extremely precise and can be made to include nearly all the correlation energy. It also scales well for large numbers of processors (1000s at nearly 100% efficiency).
Applications need to be carried out on a large scale, to determine electronic structure and properties of large (~100 atoms) molecules of chemical interest, including intermolecular interactions, best described using Slater trial wave-functions for QMC. Such functions combined as hydrogen-like atomic orbitals possess the correct nodal structure for the high precision FN-MC (Fixed Node Monte Carlo) methods, which include more than 95% of the electron correlation energy.
Project Title: Global electromagnetic gyrokinetic simulation in 3D equilibria
Project Leader: Prof. P. Helander, Max Planck Institute of Plasma Physics (IPP), Greifswald, Germany
Resource Awarded: 700,000 core hours on CINECA – CNE-SP6
Dr. F. Castejón, CIEMAT para Fusión, Spain
Prof. Laurent Villard, CRPP-EPFL, Lausanne, Switzerland
Simulations of plasma microinstabilities and related turbulence are a necessary complement to stellarator experiments as e.g. Wendelstein 7-X. Especially important are full torus simulations for three-dimensional stellarator configurations. Gyrokinetics as a first principle based theory is well suited to describe the relevant physics. An established and flexible method for solving the gyrokinetic system of equations is the simulation via the particle-in-cell (PIC) Monte-Carlo method. For this purpose the EUTERPE code has been established which originally solved the electrostatic gyrokinetic equation globally in arbitrary stellarator geometry. With the availability of up-to-date high performance computing hardware it is possible to further enhanced the physical model in two aspects. The full kinetic treatment of the electrons will include trapped particle effects and the inclusion of electromagnetic effects will extend the scope of applicability to the magnetohydrodynamic (MHD) regime. These developments will make EUTERPE the first code worldwide that is able to simulate global gyrokinetic electromagnetic instabilities in three dimensions.
Project Title: Fully Nonlinear Gyrokinetic Computation of Tokamak Turbulence and Transport
Project Leader: Dr.habil. Bruce D. Scott, Max Planck Institute of Plasma Physics (IPP), Garching, Germany
Resource Awarded: 700,000 core hours on SARA – Huygens P6
Priv.-Doz.Dr. Alexander Kendl, University of Innsbruck, Austria
Prof. Fernando Serra, Centro de Fusão Nuclear (CFN), Instituto Superior Técnico, Portugal
The FEFI (full electrons, full ions) code is at present the only fully nonlinear ’total-f’ electromagnetic gyrokinetic model which can address the steep gradient edge region of tokamak plasmas without any separation between background and fluctuating components of the dependent variables. Scale separation is to be achieved within the model on a high resolution spatial grid, with a spatially dependent velocity space grid for treatment of large temperature contrasts. The electromagnetic field responses to the dynamics of electron and ion gyrocenters include the self consistent magnetic equilibrium. The code is currently parallelised up to 512 CPUs and can be run for medium sized domains. The goal is to have it fully capable for global computation of turbulence and transport for ITER. Within DECI it is planned to simulate the entire pedestal region (edge-to-core transition) of the ITER tokamak.
Project Title: Full f gyrokinetic simulation of plasma edge
Project Leader: Jukka Heikkinen, VTT (Euratom-Tekes association), Espoo, Finland
Resource Awarded: 786,240 core hours on EPCC – HECToR DC, FZJ – JUGENE, EPCC – EPCC – HECToR DC QC
Timo Kiviniemi, Aalto University (Euraton-Tekes association), Espoo, Finland
Francisco Ogando, Spanish National University for Distance Education (UNED), Spain
Understanding the plasma turbulence is of major importance for success of ITER, which is in turn the most important step in the development of fusion energy. ELMFIRE is a gyrokinetic plasma simulation code developed to study the dynamics of turbulence and its influence on plasma global behaviour. ELMFIRE can be used, and has already been used, to understand the physics underlying the formation and development of plasma turbulence and its undesirable effects on plasma confinement. Of particular importance in the study of plasma confinement is the transition between L and H states, referring to Low and High confinement states. Understanding the way to induce an L-H transition in a plasma would lead to an immediate increase of plasma confinement applicable to different plasma devices. The DECI resources obtained for this year have supported the plasma analysis with calculations that showed first indications of the mentioned L-H transition. Further understanding of pedestal transport and its control has been acquired. The diagnostics of turbulent structures has been improved by wavelets. These calculations have been so far promising and are now proposed to be continued with more memory-efficient code version (using Domain Decomposition) for longer (and heavier) simulations to collisional time scale.
Project Title: Perturbation theory for lattice QCD
Project Leader: Dr. Alistair Hart, University of Edinburgh, UK
Resource Awarded: 768,000 core hours on EPCC – HPCx, LRZ – HLRB II, CSC – Louhi QC
Prof. Ronald Horgan, Cambridge University, UK
Dr. Georg von Hippel, DESY-Zeuthen, Germany
Lattice QCD calculations are crucial to understanding and interpreting the results from particle collider experiments such as at the upcoming LHC in CERN. The simulations are numerically very expensive and require high-capability HPC resources. Theoretical calculations can be carried out to improve the accuracy and efficiency of the simulations. These supporting, ’improvement’ calculations can be carried out using capacity HPC resources.
In this project we use DEISA resources to carry out such improvement calculations to enable lattice QCD simulations to uncover signals of new, ’Beyond the Standard Model’ physics in experimental data. We principally focus on improvements for simulations describing the physically interesting decays of heavy B-mesons. In doing so, we will develop optimised numerical integration tools for use more widely in scientific calculations on a variety of supercomputer architectures.
Project Title: Gauge theories with Nf =1 flavours of dynamical fermions
Project Leader: Prof.Dr. Gernot Münster, Westfälische Wilhelms-Universität, Germany
Resource Awarded: 5,280,000 core hours on FZJ – JUMP P6, FZJ – JUGENE, IDRIS – BABEL
Prof.Dr. István Montvay, DESY-Zeuthen, Germany
This project belongs to the theory of elementary particles. We would like to perform numerical Monte Carlo simulations in two related theories, which pose open questions concerning fundamental issues in quantum field theory: Supersymmetric YangMills (SYM) theory with SU(2) gauge group and Quantum Chromodynamics (QCD) with a single quark flavour. Besides being theoretically related to each other in some specific limits, the Monte Carlo simulations on the lattice have the same basic problem, namely the nonpositive definiteness of the fermion determinant. We deal with this problem by applying the Polynomial Hybrid Monte Carlo (PHMC) update algorithm and by monitoring the sign of the determinant, which is then taken into account by reweighting in the statistical averages. As we showed in previous studies, for nonnegative fermion masses, this is a powerful method producing valuable results on the particle spectrum and on other until now unknown properties of these theories. With the computer time we apply for in this DECI project, our main aim is to extend previous investigations towards larger (32^3*64) lattices, allowing to reduce finite volume effects and to go closer to the continuum limit.
Project Title: Laser-Plasma Accelerators Towards the Energy Frontier: 10 GeV and beyond
Project Leader: Luís O. Silva, Instituto Superior Técnico, Portugal
Resource Awarded: 835,584 core hours on FZJ – JUGENE, IDRIS – BABEL
Warren Mori, University of California, Los Angeles, USA
Dr. Raoul G.M. Trines, STFC/ Rutherford Appleton Laboratory, UK
Ultra intense lasers have opened some of the most exciting new fields and avenues nowadays for research. Extreme laser intensities, associated with intensities in excess of 1022 W/cm2 and pulse durations shorter than 1 picosecond, where the electron quiver motion in the laser field becomes relativistic, are an extraordinary tool for new physics and new applications (which has already been called relativistic engineering) and are becoming widely available in many laboratories around the World. One of the most exciting applications of these systems are compact plasma-based accelerators; recently near 1 GeV electron beams have been produced in experiments at the Rutherford Appleton Laboratory (RAL) and the Lawrence Berkeley Laboratory (LBL). Near future laser systems, such as the 10 PW laser at RAL, open the way to electron beams very close to the energy frontier, with our theoretical models indicating the possibility to generate up to 50 GeV electron beams in laser-guided configurations. In this proposal, and using a combination of massively parallel simulations in Lorentz boosted reference frames, we aim to address this new regime of laser-plasma accelerators by performing transformative one-to-one three dimensional simulations including all the relevant microphysics, with the particle-in-cell code OSIRIS, capable of demonstrating the potential of using plasma-based accelerators to develop compact particle accelerators to use at the energy frontier, in medicine, in novel light sources, and in probing new materials.
Project Title: Solving the Mysteries of Quarks
Project Leader: Prof. Jonathan Flynn, University of Southampton, UK
Resource Awarded: 1,200,000 core hours on CSC – Louhi QC, FZJ – JUGENE
Dr. Peter Boyle, University of Edinburgh, School of Physics, UK
Prof. Richard Kenway, University of Edinburgh, School of Physics, UK
Quarks are the fundamental particles that make up 99.9 per cent of ordinary matter, such as protons and neutrons in atomic nuclei. Quarks are bound together by the strong nuclear force, mediated by the exchange of gluons. The theory describing the strong force is called Quantum Chromodynamics, or QCD. The strong force is actually weak when the quarks are close together but increases steadily as you try to separate them, making it impossible to isolate a single quark, a property known as ’confinement’. This means that in experiments we detect not quarks and gluons but particles which are complicated bound states. The basic properties of the six types or flavours of quarks, such as their masses or the strengths of the interactions which turn one type of quark into another, are thus very hard to determine. The interactions changing one flavour to another are related to the small difference between matter and antimatter, called CP violation, that may help to explain why our Universe is dominated by matter (and hence why we can exist at all).
Supercomputer simulations are needed to discover whether our current theories can explain this or if there is some new physics at work. The simulations are the vital link between fundamental theories and the observed particle interactions seen in high energy physics experiments. The calculations enable scientists to ’look inside’ quark and gluon bound states, such as the proton and a plethora of other states known collectively as hadrons. The calculation is performed by constructing a discrete four dimensional space-time grid (the lattice) and then solving the QCD equations of motion on this grid. Such lattice QCD simulations are the only known first-principles method for studying hadronic interactions.
Project Title: Technicolor on the lattice
Project Leader: Kari Rummukainen, University of Oulu, Finland
Resource Awarded: 1,800,000 core hours on IDRIS – BABEL
Kimmo Tuominen, University of Jyväskylä, Finland
The Standard Model of particle physics is a very successful description of hitherto known physics. However, there are theoretical reasons to expect that the new LHC particle accelerator at CERN finds signs of ’new physics’, physics beyond the Standard Model. Technicolor is one of the most popular alternatives for the new physics. It is very important to know the detailed physical properties of the theory in order to be able to compare the predictions of the theory with the experiments.
Technicolor models rely on so-called strong coupling phenomena, making their study with analytical methods difficult and often impossible. Only large-scale numerical simulations can yield reliable quantitative results; however, so far only very prelimnary studies have been made.
We perform detailed analysis of two technicolor theories, where fermions belong to the symmetric representation of SU(2) or SU(3) gauge group. These are candidates for simplest technicolor models. We study the spectrum of particles predicted by the structure of the theory and the evolution of the coupling constant with the energy scale.