6th Project Access Call – Awarded Projects

Find below the allocations of the PRACE 6th Project Access Call (in alphabetical order of the name of the project leader).

Baryon structure using dynamical QCD simulations with physical values of the light, strange and charm quark masses

Project leader: Constantia Alexandrou, University of Cyprus, Cyprus

Collaborators: Abdou Abdel-Rehim, The Cyprus Institute, Cyprus; Giannis Koutsou, The Cyprus Institute, Cyprus; Vincent Drach, DESY, Germany; Karl Jansen, DESY, Germany; Marc Wagner, Goethe University Frankfurt am Main, Germany; Carsten Urbach, University of Bonn, Germany; Luigi Scorzato, ECT*, Italy; Roberto Frezzotti, Universita di Roma Tor Vergata, Italy; Giancarlo Rossi, Universita di Roma Tor Vergata, Italy; Elisabetta Pallante, University of Groningen, Netherlands; Chris Michael, University of Liverpool, United Kingdom;

Abstract: The current project proposes ab initio calculation of key observables that probe hadron structure. Solution of Quantum Chromodynamics (QCD), the theory that describes the strong force that binds quarks into hadrons and nucleons into nuclei, requires large computational resources. There is a rich experimental program at major accelerator facilities world-wide that aims at measuring such key observables and therefore the input from such calculations is needed in the quest of understanding the properties of the bulk of visible matter in our universe.

In particular, our ability to predict new phenomena beyond the standard model (SM) of elementary particle physics now under search at LHC, relies in part in our reliable calculation of known experimental quantities. This forms the basis of the current project. Namely we aim at calculating well measured quantities that include the nucleon axial charge, charge radius, form factors and moments of generalized parton distributions at the physical point. Resolving the observed discrepancy between lattice calculations and the measured values of these key quantities will open the way for reliable calculation of other difficult to measure quantities such as the axial charges of the hyperons and charm baryons, which can be obtained in an optimal way within the formulation of the current proposal.

Baryon structure calculations in this project will be done using dynamical twisted mass configurations of two light degenerate quarks, a strange and a charm quark all fixed to their physical values. These gauge configurations, that provide the most complete description of the QCD vacuum to date, will be made available in time to be utilized as the project progresses. The generation of these configurations is being done using computer resources beyond the current project. Therefore this project highly leverages resources used for the generation of these configurations providing the required computer time for their analysis and the extraction of important physics results thus enabling the optimal utilization of these gauge configurations. To accomplish our goal of having precise calculations at the physical point requires large computer resources as foreseen by PRACE.

The project is embedded in the European Twisted Mass Collaboration which consists of leading scientists across Europe with expertise in all aspects of lattice QCD calculations.

Resource awarded: 12 million core hours on FERMI @ CINECA, Italy and 12 million core hours on JUQUEEN @ GCS@Jülich, Germany

Solid-solid phase transitions in correlated metals from ab-initio calculations.

Project leader: Bernard Amadon, CEA, France

Collaborators: Jordan Bieder, CEA, France; Agnes Dewaele, CEA, France;

Abstract: The goal of the whole project will help to clarify the fundamental role of electron interaction for the thermodynamics and kinetics of phase transitions in strongly correlated metals. It will help for a better description of strongly correlated systems, such as e.g superconductors, and iron under very high pressure, as found in the Earth core.

When solids are compressed, cooled or heated, their structure and their density can abruptly change. In cerium, the most abundant of the so-called rare-earths metals, the most intriguing feature is a discontinuous collapse of its volume by 15 % when the material is cooled down at a given pressure, or compressed at a given temperature.

The crystal structure, keeps the same symmetry and this so-called “iso-structural” transition is thus purely due to the electrons. Today’s understanding is that at room temperature, due to the large interatomic distance, electrons stay inside localized orbitals instead of hopping from one atom to the other. Under pressure increase or temperature decrease, the electrons start however to delocalize. Several studies have highlighted the important role of local interaction to reproduce electronic properties of the systems, such as photoemission. We propose to use a quantum based ab-initio simulation, using a combination of Density Functional Theory and Dynamical Mean Field Theory with a Continuous Time Quantum Monte Carlo solver to compute the temperature and pressure boundary line for the alpha gamma transition in Cerium.

Resource awarded: 24 million core hours on MareNostrum @ BSC, Spain

TRASFER – integrating Topological order and RAShba-like effects in FERroelectrics

Project leader: Paolo Barone, CNR-SPIN, Italy

Collaborators: Thomas Archer, Trinity College Dublin, Ireland; Ivan Rungger, Trinity College Dublin, Ireland;Silvia Picozzi, CNR-SPIN, Italy; Alessandro Stroppa, CNR-SPIN, Italy; Domenico Di Sante, University of L’Aquila, Italy;

Abstract: In recent years, there has been increasing interest in phenomena emerging from relativistic electrons in a solid. Among other effects, the relativistic spin-orbit interaction lies at the origin of many subtle and interesting effects in the electronic structure of materials such as the emergence of topological insulators and the Rashba effect in the bulk of materials with non-centrosymmetric structures. Materials with large Rashba-like effects are object of intense research due to their potential applications in the field of spintronics, aiming at an electric control of spin transport in new device concepts. On the other hand, it has been recently noticed that topological insulators and noncentrosymmetric materials displaying Rashba effects are characterized by similar spin-polarized states, suggesting a close relationship between the two classes of materials. As a matter of fact, the first known noncentrosymmetric semiconductor showing huge Rashba-like spin-split band, BiTeI, has been predicted to undergo a topological transition under pressure.

On the other hand, an interesting subclass of bulk materials lacking inversion symmetry comprises ferroelectrics, that in principle would allow to bring in a novel functionality through the coexistence and coupling between ferroelectricity (i.e. presence of a permanent and switchable polarization) and Rashba effects (i.e. presence of a k-dependent spin splitting in the band structure). Among these, the semiconductor ferroelectric GeTe may represent a rather simple but instructive playground, where d
ifferent properties relative to possible Rashba effects and/or topological order may be analysed by means of first-principle tools. The potential relevance of these materials has been recently discussed in GeTe, where the reversal of ferroelectric polarization in GeTe has been theoretically predicted to cause a full reversal of the spin polarization; in order to exploit this peculiar property, a Datta-Das spin-FET architecture has been also proposed. On the other hand, a much more interesting class of materials comprises oxide ferroelectrics, where ferroelectric properties are known to be much more robust as opposed to semiconductor ferroelectrics. Furthermore, recent development in the fields of oxide superlattices and oxide electronics allows for the synthesis and characterization of artificial heterostructures where the desired features may be implemented. In principle, these features may be tuned in order to design new topological insulators or new materials displaying the desired coupling between ferroelectricity and spin-polarization effects.

Within this context, we propose a theoretical investigation of candidate materials where ferroelectricity may coexist with large Rashba effects. Our focus will be on known semiconductor ferroelectrics and on new transition-metal oxide heterostructures. Due to the appealing analogy with topological insulators, we will investigate also topological properties of the electronic structures, aiming at a deeper understanding of the possible relationship between topological order and Rashba physics.

Resource awarded: 4.4 million core hours on MareNostrum @ BSC, Spain

(CRYSPHASE) Large scale simulations of the crystallization process of phase change materials for data storage

Project leader: Marco Bernasconi, University of Milano-Bicocca, Italy

Collaborators: Sebastiano Caravati, ETH Zurich, Switzerland; Gabriele Sosso, University of Milano-Bicocca, Italy;

Abstract: Phase-change materials based on chalcogenide alloys are attracting a lot of interest due to their ability to undergo reversible and fast transitions between the amorphous and crystalline phases upon heating. This property is exploited in rewriteable optical media (DVD) and phase change non volatile memories (PCM). Mass production of 45 nm PCM devices has been announced by Micron in July 2012. The key property that makes these materials (typically GeSbTe or doped GeTe alloys) suitable for applications in PCM is the high speed of the transformation which leads to full crystallization on the time scale of 10-100 ns upon Joule heating. What makes some chalcogenide alloys so special in this respect and so different from most amorphous semiconductors is, however, still a matter of debate. Molecular dynamics (MD) techniques hold the promise to shed light on the peculiar features that make some chalcogenide alloys suitable to undergo crystallization in about 10 ns. However, the limitations in size and simulation time prevents a full characterization of the process from the quantum molecular dynamics simulations base on Density Functional Theory (MD) performed so far. On the other hand, for the electrothermal modelling of PCM devices it is crucial to assess the dependence on temperature of nucleation and growth rates. To overcome the limitations of DFT-MD simulations we have recently developed an interatomic potential for the binary phase change material GeTe by fitting a large database of DFT energies by means of a Neural Network (NN) method. The NN potential has an accuracy close to that of the underlying DFT framework at a much reduced computational load that scales linearly with the size of the system. The NN potential allowed us to simulate several thousands of atoms for tens of ns, which is well beyond present-day capabilities of DFT-MD methods.

In this project we plan to use the NN potential to simulate the homogeneous crystallization of 8000-atom models of GeTe aiming at identifying the mechanism of formation and growth of the critical crystalline nucleus. We will consider both the crystallization of the amorphous and of supercooled liquid phases as the programming protocols of PCM devices exploit both possibilities. We will study by standard MD the crystallization at temperatures above the glass transition (Tg) which is expected to be completed in about 2 ns for simulations cell of about 8000 atoms. Crystallization at low temperatures, below Tg, will be addressed as well by means of metadynamics simulations as these conditions are of interest for data retention in the device. The project will provide crucial information to aid the electrothermal modelling of the device which is now particularly needed for the design of PCM at the next technological nodes below 45 nm.

Resource awarded: 12.55 million core hours on FERMI @ CINECA, Italy

SolarFiestaUnderstanding organic solar cells with embedded many-body perturbation theory: the Fiesta initiative.

Project leader: Xavier Blase, CNRS, France

Collaborators: Ivan Duchemin, CEA, France; Claudio Attaccalite, CNRS, France; Carina Faber, CNRS, France; Valerio Olevano, CNRS, France;

Abstract: Due to their low cost and flexibility, in terms of architecture and available materials, organic solar cells, namely solar cells composed of organic molecules or polymers, are promising systems for renewable energy production. These “third generation devices” suffer however to date from a rather low quantum efficiency (<10%) impairing their commercial applications.

Such systems are also a challenging scientific playground since the phenomena that controls the photon-to-electricity transduction rely on a complex interplay between structural and “excited states” electronic properties. In particular, quantum mechanical ab initio approaches based e.g. on the Density Functional Theory (DFT) and its time-dependent extensions fail, in their standard formulation, to properly predict the properties of such systems. While there is much activity to cure DFT from its spurious limitations, other approaches deriving from many-body perturbation theory (MBPT), such as the GW and Bethe-Salpeter (BSE) formalisms, have been shown since the mid 80s to accurately describe the electronic properties of extended semiconductors. However, such techniques are expensive and applications to organic systems are still in their infancy.

A crucial step has been made with the development of a “localized auxiliary basis” MBPT implementation (the Fiesta initiative). Recent work on the parallelization of the Fiesta package demonstrated that full GW calculations on molecular systems up to several hundred atoms could be conducted within a very few hours wall-time on several thousand cores (see Preparatory PRACE project n° 2010PA0855). The accuracy of the GW/BSE approach in the Fiesta implementation was demonstrated with e.g. errors of the order of 0.1-0.2 eV as compared to experiment for the calculation of the charge transfer excitations in typical donor-acceptor complexes.

The ability to perform GW/BSE calculations on systems containing up to a few hundred atoms allows to treat explicitly a molecule surrounded by few shells of neighbors (1D to 3D geometries). We wish to demonstrate that treating explicitly within GW/BSE the interaction of first-nearest neighbors, while describing molecules further away by discrete dynamically pola
rizable models, is an efficient and accurate way to describe, within MBPT approaches, embedded molecular systems. The possibility to compare “embedded” and exact GW/BSE calculations on large molecular clusters offers a unique way to cleanly develop a methodology that we believe to be a very powerful approach to study disordered molecular bulk or solvated organic systems.

As one of the main applications, we wish to apply the Fiesta methodology to realistic bulk donor-acceptor organic interfaces, such as fullerenes/polythiophene derivatives. The accurate calculation of the band gap and excitation energy spectra, as a function of the distance to the donor/acceptor interface, based on the realistic molecular geometries that are starting to appear in the literature, is a crucial step to understand in particular the dissociation process(es) of the photo-generated electron-hole pairs (excitons) that are believed to strongly influence the solar cell quantum efficiency.

Resource awarded: 5.38 million core hours on Curie TN @ GENCI@CEA, France

Ab initio molecular dynamics of lanthanides in protic ionic liquids.

Project leader: Enrico Bodo, University of Rome “Sapienza”, Italy

Abstract: Ionic liquids (ILs) are salts made by complex, sterically mismatched molecular ions which possess a low melting point owing to the fact that the electrostatic interactions are weakened and lattice formation frustrated by geometric and steric effects. In the past few years, the coordination chemistry of metals and in particular of f-element ions in ILs has started to gain high interest due to several potential applications, such as the use of ILs as solvents for the processing of spent nuclear fuel waste and their use a photo-luminescent probes. In contrast to traditional organic solvents, ILs possess negligible flammability and volatility and represent a new class of “green” solvents that are inherently safer and more environmentally friendly than conventional solvents, and their use in solvent extraction processes would be in perfect line with the topics covered in the Horizon 2020 European program[*], as the development of green and sustainable procedures is one of the main objective of it. Another important application of ILs combined with metal ions is their use as optical solvents since the solutions of f-element in ILs represent promising novel soft luminescent materials for photochemistry and spectroscopy.

The microscopic characterization of the coordination properties of the f-element ions in ILs is essential to tailor the chelating properties of these systems and to optimize their specific technological performances in extraction technologies. Besides, the rationalization of the structural features that occur in these systems is of extreme scientific interest because only the understanding of the fundamental issues at the basis of their behavior can allow the planning of new applications. While classical molecular dynamics simulations have been and still are the main choice for the atomistic description of an ionic liquid, ab-initio molecular dynamics on such systems is getting within the grasp of modern supercomputer and is opening unprecedented analysis possibilities on these systems because it allows the control over the electronic degrees of freedom and avoids the bias often introduced by traditional force field methods.

The main focus of this project is a theoretical description of lanthanides in ILs. Our objectives, in particular, is the rationalization of the structural and, possibility dynamical, features of these materials in the presence of an f-elements ion. This complex task will be achieved by using ab-initio, first principle Molecular Dynamics (MD). The joint use of NEXAFS experimental data available to the P.I’s group will allow us to supplement and compare the theoretical findings with experimental structures. The configurations will be explored using small clusters (or droplets) of the ionic liquid with the ion in the center. This allows us to use a smaller system size with respect to a simulation of the bulk fluid by limiting ourselves to 12-24 ionic couples. Two lanthanides (La3+, Lu3+) will be used in combination with a simple but interesting ionic liquid such as ethyl-ammonium nitrate. The latter is able to form a complex h-bonding network which is reminiscent of water and therefore provides a very interesting solvating medium.

[*] http://ec.europa.eu/research/horizo…

Resource awarded: 6 million core hours on Curie FN @ GENCI@CEA, France

STiMulUs – Lagrangian Space-Time Methods for Multi-Fluid Problems on Unstructured Meshes

Project leader: Walter Boscheri, University of Trento, Italy

Abstract: This project is inserted in the framework of the STiMulUs project, which has begun in 2011, when the PI (Prof. Dr.-Ing. Dumbser) won an ERC Starting Grant of 60 months duration. STiMulUs main task is the development of new robust, efficient and high order accurate numerical algorithms for the solution of time dependent partial differential equations (PDE) in the context of non-ideal magnetized multi-fluid plasma flows with thermal radiation. It will consider both, high order unstructured Eulerian methods on fixed grids as well as high order unstructured Lagrangian schemes on moving meshes, to reduce numerical diffusion at material interfaces. This project and our request of computational resources will focus on the Lagrangian part of STiMulUs, whose growth started last year when the PI et al. were the first who developed a one-dimensional Lagrangian arbitrary high-order one step WENO finite volume scheme for stiff hyperbolic balance laws. The work is currently in progress and we have already ex-tended the previous work to two space dimensions, using unstructured triangular meshes . A very challenging field of application for non-ideal multi-fluid plasma flows with thermal radiation is nuclear fusion, in particular inertial confinement fusion (ICF).). The science is very interested in this topic because of the shortage of energy on the Earth. In fact the world human population is rapidly growing and as a natural consequence also its need for energy. For these reasons and due to their better instabilities resolution, our new Lagrangian schemes would be suitable to study those phenomena occurring during ICF. In this research project we therefore want to carry out very important basic research on that topic and develop complete-ly new, very high order accurate numerical algorithms in space and time for the solution of time dependent partial differential equations (PDE) with stiff source terms on general un-structured meshes that govern non-ideal magnetized multi-fluid plasma flows with thermal radiation, occurring before the onset of the nuclear fusion process. We rely on high order schemes to resolve very well and with only little numerical diffusion also the fine details of the flow that are crucial for this kind of applications. The highly accurate next-generation mathematical tools emerging from the STiMulUs project may lead to completely new fluid-mechanical key insights in ICF flows that can subsequently be used by physicists and engineers to succeed with the next ICF experiments, thus providing modern civilization with clean energy in the future. The impo
rtance of our research topic for our society is underlined by the recent international discussions on the accelerating global climate change, mainly caused by modern civilization and its increasing need for energy.

Resource awarded: 3 million core hours on SuperMUC @ GCS@LRZ, Germany

MSMPS – Multi-scale, Multi-physics Plasma Simulations

Project leader: Francesco Califano, University of Pisa, Department of Physics, Italy

Collaborators: Nicolas Dubuit, CNRS – Aix-Marseille University, France; Matteo Faganello, CNRS – Aix-Marseille University, France; Benkadda Sadruddin, CNRS – Aix-Marseille University, France; Pierre Henri, University of Pisa, Department of Physics, Italy; Claudia Rossi, University of Pisa, Department of Physics, Italy;

Abstract: High temperature laboratory plasmas and, even more, rarefied space plasmas are characterized by a typical collisional time scale much longer than the plasma dynamics time scale. As a consequence, these plasmas can be considered in first approximation as collisionless. The energy, typically injected on the large Hydrodynamics/fluid-scales, spontaneously cascades towards smaller and smaller scales (and higher frequencies) until small scale effects eventually come into play. The micro-scale dynamics is self-consistently coupled to the large scale one at the point that it influences the global evolution of the system itself, as for example during magnetic reconnection events.In this project, we aim at investigating the multi-scale, weakly collisional plasma physics which characterize the magnetized dynamics of space and laboratory plasma systems using a two-fluid approach including the Hall term in the generalized Ohm law. Dissipative effects are provided here by resistivity and/or numerical filters. The long-term challenge of our work will be to provide a 3D magneto-fluid numerical model, including electron inertia effects (to be developed during the project) able to capture the spontaneous transition beyond MHD scales up to kinetic scales (not included in the present proposal).

The goal of the present one year project is the 3D study of multi-scale processes observed by satellites in the Earth’s environment or in the laboratory, e.g. magnetic reconnection and turbulence. These results will also be a fundamental step for the set up a a new fluid-based model where the first important kinetic effects are included. The project is based on a long standing investigation of the same dynamics in a 2D-1/2 configuration.

The results of this project will be a major step in the progress of Space Weather research and laboratory plasma physics, in particular concerning the analysis of global basic plasma processes.

Resource awarded: 3.4 million core hours on SuperMUC @ GCS@LRZ, Germany

AgZnO – Ab initio investigation of multidimensional interface between 0D Ag nanoparticle and 1D ZnO wire for optoelectronic applications

Project leader: Arrigo Calzolari, CNR-NANO Istituto Nanoscienze, Italy

Collaborators: Alessandra Catellani, CNR-NANO, Italy; Alice Ruini, CNR-NANO Istituto Nanoscienze, Italy;

Abstract: The present project is focused on large scale ab initio simulations, based on density functional theory (DFT), of multidimensional interfaces between one-dimensional semiconductors wires (ZnO, Al-doped ZnO) and zero-dimensional Ag nanoparticles. This class of systems is prototypical of nanostructured metal/metal-oxide interfaces: while the mesoscale heterojunctions are wildly used in a large range of optoelectronic devices, such as photovoltaic solar cells [1-2], displays [3], photo-detectors [4], plasmonic devices [3], reproducibility of nanoscaled interfaces is still to be achieved. The implementation of these junctions in nanodevices and their subsequent performances could be properly optimized if we were able to control the structural and electronic properties of the single components at the nanoscale. On the theoretical side this implies a detailed quantum mechanical understanding of ground states of the single systems and of their aggregates. Particularly subtle is the effect of quantum confinement on the electronic structure of the nanoparticle, as well as the effect of surface defect and strain, which make such nanosystems so different from the corresponding bulk materials. The study of effects such as the interplay between the quantum confinement and the band line-up cannot be simply addressed through oversimplified systems (e.g. planar periodic junctions), but it requires the explicit simulation of the overall nanostructures. For instance, the proposed ZnO wire/Ag dot interfaces show a peculiar mixing of dimensionalities – from 3D (core of large-diameter nanowire), to 2D (exposed surfaces), to 1D (edges), to 0D (dot) – that imparts a not obvious charge redistribution and surface atomic relaxation. This is expected to modify the reactivity, the bonding capability and the electronic properties of single nanostructures, and consequently the properties of the resulting interfaces. Here, we will focus on the ground state properties of the interfaces only, which is the prerequisite for any further ab initio investigation.

We will consider undoped (semiconductor) and Al-doped (metallic) ZnO wires in the formation of interfaces with Ag dot. This will realize semiconductor/metal and metal/metal heterojuctions with the same structural configuration. The former is representative of the formation of the contacts between the optical active sites and the external leads in solar cells, the latter of field-enhanced (e.g. plasmonic) transparent conductive oxides (TCO) in photovoltaic devices.

In general, calculations which require the explicit treatment of the electronic structure of the system (such as DFT), represent a tremendous computational task. In particular, the memory request for the simulation of realistic nanostructures (i.e. thousands of atoms) goes beyond the standard capability even of large supercomputers. To overcome this problem it is necessary to use well optimized codes and to have access to extraordinary HPC resources. In this light, PRACE Tier-0 program fulfills such large-scale computational requirements (i.e. memory, CPUs, and CPU time),for the simulation of nanoscale systems otherwise unfeasible. To best exploit the PRACE resources, we previously proposed a PRACE-preliminary access focused to improve the memory request and the scalability of high-performance DFT code (cp.x code included in Quantum-Espresso distribution [6]).

Resource awarded: 7.7 million core hours on Curie FN @ GENCI@CEA, France

TRADELINBOTransition delay in Blasius-like boundary layers by passive control: complementary investigation and numerical support to an ongoing experimental activity

Project leader: Simone Camarri, University of Pisa, Italy

Collaborators: Alessandro Talamelli, Alma Mater Studiorum, Universit di Bologna, Italy; Franco Auteri, Politecnico di Milano, Italy; Flavio Giannetti, Universita’ degli Studi di Salerno, Italy; Andrea Fani, University of Pisa, Italy; Alessandro Mariotti, University of Pisa, Italy; Maria Vittoria Salvetti, University of Pisa, Italy; Jens Fransson, Royal Institute of Technology (KTH), Sweden

Abstract: The main objective of the present project is to provide a numerical support and integration to the experiments of the ERC project AFRODITE (FP7 reference number: 258339). Both projects focus on the investigation of a method to delay transition to turbulence in Blasius-like boundary layers, i.e. the region of the flow adjacent to a flat plate. Transition delay implies a friction drag reduction, which is mainly caused by the turbulent portion of the boundary layer. Thus, the direct implication of transition delay is a drag reduction for aerodynamic bodies, for which the friction drag is an important contribution on the overall drag. The project is divided in two subsequent parts. The first is aimed at validating the DNS setup against the experiments. Once validated, the DNS will be used in the subsequent stage of the work to (i) complete the information available from the experiments on the proposed transition control and (ii) to explore and select potentially interesting alternative configurations for future experiments.

If the project will be successful, we plan to obtain as an output the following points:

1) a part showing the accuracy assessment of the proposed Direct Numerical Simulation (DNS) setup against the experiments of the AFRODITE project. This validation is carried out on three points: (a) validation of the steady flow, (b) validation of the flow with Tollmien-Schlichting waves superposed and (c) validation of the simulation of transition. The three points are subsequent also in terms of difficulty of the DNS involved.
2) A part showing how results from (1) has contributed to add information concerning the transition delay mechanism
3) A part showing set of alternative configurations for the transition control devices, to be tested in future experiments and which could lead, in future, even to a patent
4) A part dedicated to the investigation of global instabilities in the wake past a set of bluff roughness elements.

Resource awarded: 15 million core hours on FERMI @ CINECA, Italy

LGIC – On the Gating Mechanism of Ligand-Gated Ion Channels (LGICs)

Project leader: Marco Cecchini, University of Strasbourg, France

Collaborators: ESQUE Jeremy, University of Strasbourg, France; CALIMET Nicolas, University of Strasbourg, France;

Abstract: Ligand-gated ion channels (LGICs) play a central role in intercellular communication in the central and peripheral nervous systems. Understanding their function at an atomic level of detail will be beneficial for the development of drug therapies against a range of diseases including Alzheimer’s, schizophrenia, pain, and depression. By capitalizing on the increasing availability of high-resolution structures of both pentameric and trimeric ligand-gated ion channels, we aim at the elucidation of the molecular mechanism underlying activation/deactivation by atomistic Molecular Dynamics simulations. The acquired knowledge will boost the development of novel small-molecule modulators of LGICs function.

Resource awarded: 20 million core hours on SuperMUC @ GCS@LRZ, Germany

Ds spectroscopy and decays.

Project leader: Sara Collins, University of Regensburg, Germany

Collaborators: Gunnar Bali, University of Regensburg, Germany; Tommy Burch, University of Regensburg, Germany; Benjamin Glaessle, University of Regensburg, Germany; Issaku Kanamori, University of Regensburg, Germany; Paula Perez-Rubio, University of Regensburg, Germany;

Abstract: In the quark model mesons are thought to contain one quark and one antiquark. Recent experimental results on some charmonia (containing a charm and a charm anti-quark) and some open charm mesons (containing a charm and a light anti-quark) are incompatible with this picture.

Moreover, decays of open charm (D and Ds) mesons have gained prominence in the search for physics beyond the standard model, with existing and new experimental facilities suitable to their study. We investigate 4-quark contributions to Ds mesons and their mixing with quark-antiquark states and weak decays of Ds mesons. Of particular interest are decay form factors of Ds mesons into so-called flavour singlet mesons and electromagnetic corrections to their decay constants that we are studying for the first time.

Resource awarded: 18.35 million core hours on FERMI @ CINECA, Italy

CTHSP90-Investigating conformational transitions of Hsp90 by Bias-Exchange Metadynamics.

Project leader: Giorgio Colombo, ICRM-CNR, Italy

Collaborators: Elisabetta Moroni, ICRM-CNR, Italy; Jacopo Sgrignani, ICRM-CNR, Italy; Gerolamo Vettoretti, ICRM-CNR, Italy;

Abstract: Hsp90 is a molecular chaperone of the heat shock protein family whose functional activity is regulated by ATP binding and hydrolysis. From the structural point of view it is a homo-dimer and each dimer is formed by 3 regions.

Hsp90 plays and essential role in overseeing the folding process of a large number of client proteins and it is involved in the regulation of normal homeostasis and stress response. Given its central role in the control of a number of processes associated to cell life, its dis-regulated activity has been directly related with the onset and progression of neoplastic and neurological disorders.Considering the experimental and theoretical data published so far, it has been hypothesized that the interaction of Hsp90 with its client proteins and its role in the folding process are carried out by progressing through a cycle that encompasses specific conformational changes. In this context, recent results have shown that modulating Hsp90 motions using specific allosteric drugs can lead to develop promising innovative therapeutic agents.

Here we propose to investigate the complex conformational transitions that determine Hsp90 functions using Bias-Exchange Metadynamics (BE-MTD).This methodology permits to overcome the limits, related with computational efficiency, of standard metadynamics calculation where the use of collective variables is limited to a number between 1 and 3, enabling to use a hypothetically infinite number of collective variables (CV) without affecting computational efficiency. Notably, the use of a large set of CVs helps to limit convergence problems that plague standard metadynamics simulations.

Considering recent experimental findings supporting the idea that Hsp90 conformational motions are stochastic in nature and largely determined by thermal fluctuations, in a first phase of our work, we will explore the conformational transitions of the apo form of Hsp90 (no ligand present) and then we will repeat the same calculation considering an ATP or ADP molecules bound inside the two catalytic sites (one for each monomer). The role of ATP and ADP in this model is to select
and stabilize specific, functionally oriented conformations from the ensemble of accessible ones.

Finally the comparison of the conformational free energy profiles will permit us to better understand the influence of the ATP and ADP molecules on the conformational transitions of Hsp90 and also to understand molecular determinants and internal movements essential to induce conformational transitions, and ultimately function.

Finally, the data acquired in this study will be employed in designing new drug-like organic molecules able to modulate Hsp90 activity by interfering with specific protein motions. Meeting these challenges will contribute directly to moving the role of computation in the investigation of complex biomolecular systems to a new level. The overarching perspective is the definition of new sets of rules and strategies for the rational, efficient discovery of chemical entities able to modulate cell functions.

Important international collaborations are established for different aspects of the Hsp90 project including organic chemistry, pharmacology, and computational biophysics. Other important collaborations with structural biologists will be activated in the next months.

Resource awarded: 20 million core hours on FERMI @ CINECA, Italy

Arp2/3 – Mechanistic studies of the Arp2/3 complex activation

Project leader: Zoe Cournia, Biomedical Research Foundation, Academy of Athens, Greece

Collaborators: Georgios Patargias, Biomedical Research Foundation, Academy of Athens, Greece;Bradley Nolen, University of Oregon, United States;

Abstract: The spatial and temporal regulation of actin cytoskeletal assembly and disassembly is essential for many cellular processes that include cytokinesis, membrane trafficking and maintenance of cell morphology. Actin-related-protein 2/3 (Arp2/3) complex is a seven subunit ATP-ase that plays a central role in the regulated actin assembly. The Arp2/3 complex nucleates the polymerization of a new actin filament that emerges from an existing filament and is required for many cellular processes. Importantly, tumor cell migration is thought to require Arp2/3 complex, and Arp2/3 overexpression contributes to pathogenesis, growth, and invasion ofcarcinomas. While inactive on its own, Arp2/3 is activated by interacting with nucleation promoting factor (NPF) proteins that contain the conserved VCA domains. Biochemical evidence suggests that two VCA domains bind Arp2/3 each delivering an actin monomer to the complex. The result of VCA/actin binding to Arp2/3 is the activation of the complex that is accompanied with a large conformational change. It is unclear how the VCA domains bring about the Arp2/3 activation. The aim of the proposed project is to unravel the mechanism that underlies the Arp2/3 complex activation using standard and biased molecular dynamics (MD) simulations. In close collaboration with the experimental group of Dr Brad Nolen in the University of Oregon,our study will investigate the following questions: a) The location of the VCA-binding sites on Arp2/3 complex b) The effect of the bound VCA domains on the conformation of Arp2/3. The results of these studies will help us in the design of new small molecule inhibitors of Arp2/3 complex for basic research and biomedical applications.

Resource awarded: 11.2 million core hours on Curie TN @ GENCI@CEA, France

Characterization of the cytochrome bc1 complex and its interactions with cardiolipins

Project leader: Matteo Dal Peraro, Ecole Polytechnique Federale de Lausanne, EPFL, Switzerland

Collaborators: Thomas Lemmin, Ecole Polytechnique Federale de Lausanne, EPFL, Switzerland;

Abstract: The cytochrome bc1, also known as complex III, is a large homodimeric complex (0.5 MDa), composed of three catalytic subunits and several cofactors (e.g. heme centers). It forms the central pump for the transfer of protons across the mitochondrial inner-membrane. The interaction with specific phospholipids is essential for its activity. Several X-ray structures of cytochrome bc1 interacting with various phospholipids (in particular cardiolipins) have been in fact solved during the past years. However, little is known about the dynamical properties of this important structure and the role of adjacent cardiolipins for its stability and proton transfer across the membrane. Building on established molecular mechanics models of these negatively charged constituents of the mitochondrial membrane, we propose to perform extended molecular simulations of the cytochrome bc1 complex to dissect the biophysical role of cardiolipins. To this purpose we will mainly use molecular dynamics sampling techniques within the molecular mechanics framework and will investigate model systems with different compositions and protonation states of phospholipids.

Resource awarded: 7.3 million core hours on MareNostrum @ BSC, Spain

External hydrodynamic for automobile

Project leader: Matthieu de Leffe, HydrOcean, France

Collaborators: Julien Candelier, HydrOcean, France; Nicolas Couty, HydrOcean, France; Guibert David, HydrOcean, France; Pierre-Michel Guilcher, HydrOcean, France;

Abstract: The first working steam-powered vehicle was built by Nicolas-Joseph Cugnot in about 1769. Since that first creation automobile has evolved rapidly. To enable this evolution manufacturers have developed methodologies to validate new designs or new technologies. Thus, the first wind tunnel tests were conducted in 1966 for the formula 1. Since the 1980s, in order to improve their models, manufacturers have systematically use wind tunnel tests. But wind tunnel tests are complex to implement and expensive. So the numerical simulation has been introduced over the last ten years in the design process to complete the tests.

More recently, manufacturers have integrated into their processes external hydrodynamic issues. In fact, vehicles are exposed to the elements, especially rain. Water may cause various problems that involve the vehicle safety. These problems are being studied mostly by scale tests and at the end of the vehicle design process. But it is very expensive and complex to work on the car at this stage of the design. In addition, these tests require specific facilities such as climatic wind tunnels, wet tracks or fords. So manufacturers are looking to complete the scale tests by simulation. But if the external aerodynamics of vehicles is now relatively well understood thanks to numerical models, problems of external hydrodynamics for automobile exhibit characteristics that make them very difficult to take into account with traditional numerical methods. In fact, these flows are generally characterized by air / water interface extremely complex and fragmented. In addition they involve issues of moving body like a wheel on the ground, a wiper or a door. These two points are the usual limitations of traditional solvers available on the market. Thanks to an innovative approach by the SPH (Smoothed Particle Hyd
rodynamics) these limitations can now be removed.

The objective of this project is to demonstrate that it is now possible to simulate external hydrodynamic flows in automotive industry and thus accurately predict the risks caused by water. To this end, four issues will be addressed in this project: hydroplaning, river crossing, visibility, sealing.

Resource awarded: 8.2 million core hours on Curie FN @ GENCI@CEA, France

3DMagRoI : High-resolution 3D study of MRI in relativistic rotating stars

Project leader: Roberto De Pietri, Parma University, Italy

Collaborators: Luca Del Zanna, Firenze University, Italy; Niccolo Bucciantini, INAF, Italy; Roberto Alfieri, Parma University, Italy; Alessandra Feo, Parma University, Italy; Luca Franci, Parma University, Italy; Frank Loeffler, Louisiana State University, United States;

Abstract: Many high-energy phenomena such as Gamma-Ray Burst (GRBs) are associated with strong magnetic fields and fast rotating compact objects. In particular, the recently developed millisecond-magnetar model has suggested that such violent events might be the signature of a rapidly rotating and highly magnetized neutron star (NS). Although the origin of this strong and ordered magnetic fields remains poorly understood due, for example, to the fact that these magnetic fields are supposed to arise in an environment characterized by turbulence, instabilities and convection, still simple models of GRB jet launching make strong predictions on the strength and geometry of the field in the engine.

The main open questions are: whether such a field can be produced; whether it can survive instabilities; whether a large dipolar component can be generated or higher multipoles are found to dominate; how much magnetic energy can be stored inside the NS and whether this energy can be released at later times to power the afterglow activity; whether a strong magnetic field introduces deformations in gravitational wave signals; whether these waves can take away a large fraction of the rotational energy of the NS, quenching the engine.

It is clear that the intrinsic multidimensional and non-linear nature of the problem prevents simple analytical treatments from going beyond a general idea or proof of principle. Testable predictions, also in view of the quality of present and upcoming observational data, demand a higher level of accuracy in our models.

This project will involve the high-resolution 3D simulations in full General Relativity to study the effects that magnetic fields may have on the dynamics of magnetized differentially rotating neutron stars, focusing our attention on the role that magnetic instabilities may have in the evolution of newly born neutron star (or proto neutron star, PNS), as a function of the geometry and strength of the initial seed magnetic field. Differentially rotating Star models with initial Poloidal and Toroidal magnetic field configurations will be studied.

In particular, in the case of an initial weak field, our main goal is to follow the onset and growth of the magneto-rotational instability (MRI) that is supposed to redistribute the angular momentum of the star and to amplify the magnetic field itself.

Resource awarded: 25 million core hours on FERMI @ CINECA, Italy

Atomistic Simulation of Human topoisomerase IB in complex with a circular supercoiled DNA substrate.

Project leader: Alessandro Desideri, University of Rome “Tor Vergata”, Italy

Collaborators: Andrea Coletta, University of Rome “Tor Vergata”, Italy; Ilda D’Annessa, University of Rome “Tor Vergata”, Italy; Sarah Harris, University of Leeds, United Kingdom; Thana Sutthibutpong, University of Leeds, United Kingdom;

Abstract: The project is aimed at better understanding the mechanism of action of human topoisomerase IB carrying out different comparative simulations of the enzyme in complex with a supercoiled DNA substrate. Top1 is of significant medical interest since it is the only target of anticancer drugs, such as the camptothecin (CPT) and its water-soluble derivatives, such as irinotecan and topotecan that are used throughout the world for the treatment of various cancers. The full understanding of the enzyme mechanism of action and of its interaction with the DNA is a necessary requirement to develop a new generation of inhibitory drugs with improved efficiency. An important step in this direction has been provided by the resolution of the 3D structure by X-ray diffraction of the enzyme-DNA binary complex and of the enzyme-drug-DNA ternary complex. A further improvement has been provided by a series of study combining biochemical experiments to classical molecular dynamics simulation that have shown the importance for the enzyme functioning of long range communication between far located protein regions. The main limit of these simulations as well of the up to now solved X-ray diffraction structures is the fact that the DNA substrate is represented by a short linear double stranded helix and not by a large supercoiled DNA that is the real natural substrate. A Supercoiled DNA bound to topoisomerase has represented a too large system to be simulated with the up to now available computational resources. In this project we propose for the first time to afford the simulation of the topoisomerase IB in complex with a negatively and a positively supercoiled DNA substrate. In detail comparative simulations will be carried out taking in consideration the covalent and non covalent protein-DNA complex.

These simulations will enable us to identify the role of the different protein regions involved in the relaxation process, allowing us to test the hypothesis that different Top1 domains play key roles in the relaxation mechanism depending on the type of super-coiling. The simulations proposed will permit us to probe the mechanism in more detail than ever before. By describing the interaction of the enzyme with its natural super-coiled substrate we will also be able to test the hypothesis that a second DNA binding site within plectonemic DNA plays a role in the action of the enzyme. Our investigation will improve understanding of the importance of Top1 dynamics and of the roles of the various protein domains, giving useful information for the design of new inhibitors. Human Top1 can be trapped by a variety of compounds, distinct from the CPTs, some of which are in early clinical trials as promising anticancer agents. Therefore, understanding the details of the interaction sites is fundamental to the development of novel strategies to effectively treat human cancers. This information will be shared with other international groups working in this field, and in particular with the group of Prof. Mark Cushman at Purdue University, who is currently developing new drugs with improved efficacy.

Resource awarded: 31 million core hours on FERMI @ CINECA, Italy

CIM128Ki – Enable UpScaling to 128Ki cores and Perform Full Process Simulation at 4Ki cores using CimLib

Project leader: Hugues Digonnet, MINES-ParisTech, France

Collaborators: Luisa Silva, MINES-ParisTech, France; Marc Henri, SCC (Sciences Computers Consultants), France;

Abstract: The CIM128Ki project aims in promoting HPC computation in the field of numerical simulation. To achieve this, two mains objectives will be done: one is to upscale the CimLib library to the range of hundreds of thousands of cores to be able to fully benefit of the power of new supercomputers; the other one is to clearly show the benefit in using several thousands of cores to perform numerical simulations for real industrial processes. We plan to simulate during this project two industrial cases: one is based on scraper mixing process from St-Gobain, the second one is based on a Twin-Screw Extruder (TSE) from Herakles company, a subsidiary of the SAFRAN group.

Performing such simulations needs the adaptation to massively parallel computation of a large number of numerical tools, like the solution of very large linear/nonlinear solver for Stokes/Navier-Stokes problems, but also anisotropic mesh adaptation on unstructured distributed meshes. Theses two main components will be intensively analysed during this project, in a range of 8Ki to 128Ki cores. Soft Speed-Up as well as Hard Speed-Up will be determined and a biggest one shoot resolution will be executed on both the 64Ki cores of Curie Tn and 128Ki cores of JuQUEEN.

Running full simulations need more resources as it requires from hundreds to thousands of increments/iterations to calculate all the process. This kind of simulation on several thousands of cores will be a huge step forward as it deals with large computational domains, but also with complex processes and complex physical phenomena (complex flows, non linear behaviours, transient problems). Doing such realistic simulations will be also a very good vector to promote the use of HPC in industrial configurations in the future.

Resource awarded: 2.9 million core hours on Curie TN @ GENCI@CEA, France and 2.54 million core hours on JUQUEEN @ GCS@Jülich, Germany


Molecular dynamics simulation and experimental characterization of a DNA nanocage family.

Project leader: Mattia Falconi, University of Rome “Tor Vergata”, Italy

Collaborators: Cassio Alves, Universidade de Sao Paulo, BRAZILCristiano L.P. de Oliveira, Universidade de Sao Paulo, BRAZILBirgitta R. Knudsen, Aarhus University, Denmark; Federico Iacovelli, University of Rome “Tor Vergata”, Italy;

Abstract: Understanding and exploiting new, complex functional materials is intrinsically an interdisciplinary effort at the interface between physics, chemistry, biology, material science, and engineering. The unique self-recognition properties of DNA determined by the strict rules of Watson-Crick base pairing makes this material ideal for the creation of self-assembling predesigned nanostructures in a bottom-up approach. The construction of such structures is one of the main focuses of the thriving area of DNA nanotechnology, where several assembly strategies have been employed to build increasingly complex three-dimensional DNA nanostructures. To achieve this goal it is necessary to estimate the thermodynamics of all possible pairings of DNA sequences and select the sequences so that the desired product is by far the most thermodynamically favourable one. Obviously, the complexity of doing so, even when designing rather simple structures involving more than a few DNA strands, by far exceeds the capacity of the human mind. Therefore, the design of DNA sequences for the construction of nanostructures must rely on sophisticated computational tools in order to rule out sequence combinations prone to form unwanted structures.

Common for all DNA nano-structures presented until date is that they rely at least to some extent on synthetic DNA oligonucleotides, which makes their construction rather expensive. This fact, taken together with some of the common analysis techniques, such as Small Angle X-ray Scattering (SAXS) and Cryo-Transmission Electron Microscopy (Cryo-TEM) requiring quite large amounts of material, pose a serious challenge to the validation of the structures. Thus, to counter such obstacles long-time atomistic simulations, which can predict the likelihood of successful assembly as well as structural properties of DNA nano-structures before experiments, are of great value.

Aim of this project is to address the fundamental challenges related to the development of new functionally structured materials based on DNA and to gain a deep understanding of the structure and dynamics of a series of planned nanostructures on multiple length and time scales. To accomplish this task, not achievable with the regular computing resources, we need the large computational facilities offered by the Tier-0 Systems. In detail, an automatic procedure has been implemented to identify the best oligonucleotides sequences that will be assembled to form eight three-dimensional DNA cages, having different regular or irregular geometry. For each nanocage 400 ns of molecular dynamics simulation will be executed. After having screened the oligo sequences using simulative methods, some selected DNA nanocages structures will be experimentally assembled with the help of an extensive toolbox of DNA binding, cutting, ligating, or recombining enzymes, which may all prove valuable for synthesis, manipulation, or functionalization of DNA nanostructures. Finally the structural-dynamical properties of the produced cages will be investigated using spectroscopic experimental techniques, such as SAXS and Cryo-TEM, in conjunction with extensive classical molecular dynamics simulation. The results obtained through the molecular dynamics simulations will help to improve the design and the stability of the studied DNA nanostructures.

Resource awarded: 7 million core hours on Curie FN @ GENCI@CEA, France

Dynamo Action in Compressible Turbulent Plasmas

Project leader: Christoph Federrath, Monash University, Australia

Collaborators: Dominik Schleicher, Georg-August-Universitt Gttingen, Germany; Philipp Girichidis, Max-Planck-Gesellschaft, Germany; Robi Banerjee, Universität Hamburg, Germany; Ralf Klessen, Zentrum für Astronomie der Universität Heidelberg, Germany; Jennifer Schober, Zentrum für Astronomie der Universität Heidelberg, Germany;

Abstract: Magnetic fields are important for virtually all astrophysical objects, ranging from the plasma of the early Universe, over galaxies to our Sun and Earth. The turbulent plasma motions in these astrophysical objects generate magnetic fields in a manner similar to the electromagnetic energy converted from kinetic energy in a bicycle dynamo, which is why this process is called ’turbulent dynamo’. Turbulent astrophysical systems exhibit magnetic Prandtl numbers, Pm=nu/eta — the ratio of viscous to magnetic dissipation — either much smaller or much larger than unity. For instance, interiors of the Earth and stars, as well as liquid metals in terrestrial experiments have Pm<1, while the interstellar and intergalactic medium, as well as the plasma of the early Universe have Pm>1. The problem is that all existing numerical simulations of dynamo action to date had to focus on plasmas with Prandtl numbers of around unity, far away from the realistic
low and high Prandtl-number regimes. This is because extremely high resolution is required to achieve the necessary scale separation to safely reach the desired physical magnetic Prandtl numbers of interest. In addition to the enormous resolution challenge, most of the previous numerical studies concentrated on incompressible turbulence. In contrast, astrophysical systems are often highly compressible, such as the gas in the early Universe from which the first stars formed, and in the interstellar and intergalactic medium, all of which are threaded by shock waves — waves that produce density contrasts of several orders of magnitude. The investigation of dynamo action in such highly compressible plasmas has only recently been touched on. Modeling such flows requires robust and accurate numerical schemes, capable of treating plasma discontinuities.

In this PRACE project, we run and analyze the first high-resolution simulations of magnetohydrodynamical turbulence in the compressible regime with much more realistic, low and high magnetic Prandtl numbers than ever achieved before. The aim of this project is to gain theoretical insight into fundamental properties of turbulent magnetic field amplification, by directly testing theoretical predictions. Confirming or falsifying those theories is of interest to a broad scientific community in physics and astrophysics dealing with magnetic fields in turbulent systems, and sets the basis for applications of structure formation and magnetic field growth ranging from the time right after the Big Bang, over the present-day Universe, to terrestrial applications such as the study of nuclear fusion.

Resource awarded: 6.5 million core hours on SuperMUC @ GCS@LRZ, Germany

EPIGRAPH – Electronic and optical properties of graphene nanoribbons on metal surfaces

Project leader: Andrea Ferretti, CNR, Italy

Collaborators: Deborah Prezzi, CNR, Italy; Alice Ruini, University of Modena and Reggio Emilia, Italy;

Abstract: Some of the most intriguing properties of graphene are predicted for specifically designed nanostructures such as nanoribbons. Functionalities far beyond those known from extended graphene systems include electronic and optical gap variations related to quantum confinement and edge effects. The possibility to produce graphene nanostructures with the needed atomic precision was recently demonstrated by exploiting on-surface synthesis techniques.

On the theory side, most of the studies on excited-state properties rely so far on calculations for ideal isolated systems. However, the presence of a substrate is often required, e.g. for its catalytic role or to support further device applications, and has demonstrated to give rise to significant modifications of the intrinsic optoelectronic properties of the systems. Our main goal within the present project is the accurate first-principles investigation of edge- and quantum-confinement-effects for realistic graphene nanoribbons (GNRs) as altered by the coupling to a substrate.

To this end, we will employ different levels of theory, from density-functional theory (DFT), for the study of ground-state properties, to many-body perturbation theory (MBPT) approaches, for an accurate investigation of excited-state properties. In close collaboration with our experimental partners, we have identified a number of relevant target systems to address, which comprise: a) graphene nanoribbons on metals; b) GNRs on metals intercaled with insulating buffer layers; c) aromatic molecular precursors used for the catalytic growth of GNRs, both in gas phase and on substrate. Our results will be directly compared with top-quality experimental data, resulting from STS, ARPES and optical characterization of the above mentioned systems.

The innovation potential of the project is related to the possibility of investigating realistic graphene-based nanostructures, that can now be produced and characterized with unprecedented accuracy. This possibility is strictly related to the presently available HPC resources, which make now possible the application of sophisticated, predictive-power computational tools to the study of complex systems.

The here-proposed project is part of a joint theoretical/experimental collaboration, where cutting-edge experiments are planned to be coupled to state-of-the-art first-principles simulations of realistic systems. In fact, the group led by Prof. Roman Fasel [“nanotech@surfaces” Laboratory at EMPA-ETH (Zurich)] was the first to prepare atomically defined ultrathin graphene nanoribbons on metal substrates [see Cai et al, Nature 466,470 (2010) and in Figure 1], and further experiments are ongoing on the preparation of similar structures on different substrates, possibly including semiconducting ones, and on the investigation of their excited-state properties.

In this project, we plan to match these experimental efforts with highly predictive ab-initio simulations, which -in terms of cell dimension and number of atoms and electrons- are only possible through the massive use of HPC resources and well-optimized codes. In particular, our objectives were out-of-reach before the availability of tier-0 machines (such as Fermi): a class A project on this machine makes now realistic this kind of first-principles calculations also on excited-state properties, which allows us not only to explain experimental data, but also to predict further graphene-based nanostructures with desired functions.

Resource awarded: 17 million core hours on FERMI @ CINECA, Italy

QCDpQED – QCD plus QED and the stability of matter

Project leader: Zoltan Fodor, Bergische Universitaet Wuppertal, Germany

Collaborators: Christian Hoelbling, Bergische Universitaet Wuppertal, Germany; Stefan Krieg, Bergische Universitaet Wuppertal, Germany; Kalman Szabo, Bergische Universitaet Wuppertal, Germany; Laurent Lellouch, CNRS (Institut de Physique) and Univ. Aix-Marseille II, France; Alfonso Sastre, CNRS (Institut de Physique) and Univ. Aix-Marseille II, France;

Abstract: The stability and the very existence of atoms and ordinary matter relies heavily on the fact that neutrons are slightly more massive than protons. If the neutron were not more massive than the hydrogen atom, the latter would not exist because the proton and electron that compose it would undergo reverse beta decay and transform into a neutron. Moreover, if the neutron were substantially more massive than the proton, it would decay into a proton so rapidly that heavier nuclei could not have formed during Big Bang nucleosynthesis, in the first twenty minutes of the universes life.

The neutron-proton mass difference is known experimentally to very high precision and is very small: approximately 0.14%. We have strong reasons to believe that this tiny difference results from two small competing effects, which break the near isospin symmetry observed in nature, between up and down quarks. On the one hand, the mass of the electrically charged proton is augmented with respect to that of the neutral neutron by the energy carried in the electric field surrounding it. On the other hand, the mass of the neutron is enhanced because the sum of the masses of its constituents (one up and two down quarks) is larger than it is for the proton (composed of one down and two up quarks). While experiment clearl
y demonstrates that the latter effect wins if no unanticipated effects play a role, the resulting mass difference has never been shown from first principles, to be significantly in the statistical sense larger than the mass of the electron, which would be required to establish the stability of hydrogen.

The natural tool to compute the properties of hadrons are large scale simulations in lattice quantum chromodynamics (QCD). Here, we propose to perform such simulations which incorporate electromagnetic and isospin breaking effects for valence and sea up, down, strange and charm sea quarks. We further propose to perform these computations with pion masses down to around their physical value, on lattices with spatial sizes of up to more than 6 fm and for multiple lattice spacings down to about 0.054 fm. This is what is required for a precise and fully controlled of the neutron proton mass difference from first principles.

Resource awarded: 91 million core hours on JUQUEEN @ GCS@Jülich, Germany

Massively Parallel Navier-Stokes Solver for Breaking Waves (MAPAW)

Project leader: Stephane Glockner, Institut de Mecanique et d’Ingnierie de Bordeaux, France

Collaborators: Pierre Lubin, Institut de Mecanique et d’Ingnierie de Bordeaux, France;

Abstract: In the last three decades, significant attention has been devoted to improving the knowledge of the hydrodynamics and the general processes occurring in the surf zone, widely affected by the breaking of the waves. Nevertheless, the wave breaking phenomenon remains a very challenging fluid mechanics problem, turbulence and aeration interactions making it more complex to investigate.

Performing numerical simulations of breaking waves requires a large number of mesh grid nodes, robust and accurate numerical methods and long CPU time calculations to compute the hydrodynamics from the largest to the smallest length and time scales (Lubin et al., 2011). Recent progress in computational capacities allowed us to run fine three-dimensional simulations giving us the opportunity to observe for the first time fine vortex filaments generated during the early stage of the wave breaking phenomenon (Lubin et al. 2013, Lubin et al. to be submitted). To date, no experimental observations of these structures have been made in laboratories. They have only been visualized in rare surfing footages (BBC, 2009). These fine coherent structures are three-dimensional vortical tubes, like vortex filaments, connecting the splash-up and the main tube of air, elongated in the main flow direction. Some of the vortex filaments have been observed to entrain air, while some structures interacted with each other, forming larger vortex filaments and sometimes coiling. This project will help to lead a parametric numerical study to confirm these first observations by varying the values of the initial steepness and dispersion parameter. Several breaker intensities will also be investigated to detail the conditions of occurrence of the vortex filaments, their generation and their development (sizes and spacing between vortex filaments, interactions, air entrainment, etc.).

The project is based on Thetis CFD code which has been parallelized a few years ago thanks to domain decomposition and MPI library (see http://thetis.enscbp.fr and Ahusborde et al. 2011). It is linked to the massively parallel solver and preconditioner Hypre library. Weak scalability has been tested up to 1 billion mesh points and 16 384 cores. It proved that the code run efficiently as the number of processors (and so the total amount of mesh nodes) increased. A PRACE preparatory access allowed us to overcome some bottleneck parts of the code such as I/O, to improve the partitioning stage and to optimize some part of the code and the placement of the MPI process on the Curie supercomputer. First simulations of 3D breaking waves were done with 80 millions nodes on 576 processors. The goal of this project is to achieve a better description and understanding of the small filament structures and to reach a typical mesh size of 640 millions nodes on 4096 processors. Few simulations will run on 4 billions nodes on 16 384 cores.

Resource awarded: 13 million core hours on Curie TN @ GENCI@CEA, France

DENSITIESDirEct Numerical SImulation of TurbInE flowS

Project leader: Nicolas Gourdain, CERFACS, France

Collaborators: Tony ARTS, Von Karman Institute for Fluid Dynamics, Belgium; Frederic Sicot, CERFACS, France; Marcello MANNA, Universita’ degli Studi di Napoli “Federico II”, Italy;

Abstract: The transfer of thermal energy between a flow and a wall occurs in a lot of industrial applications, including gas turbine. While the life duration of a turbine, as well as the whole system performance, relies on the capability of designers to correctly estimate the blade temperature, its prediction remains very difficult due to the complex physical phenomena occuring in these components (e.g. turbulence).

The DENSITIES (DirEct Numerical SImulation of TurbInE flowS) project proposes thus to run a DNS in a high-pressure turbine guide vane, at flow conditions close to industrial applications (compressible transonic flow, high Reynolds number, etc.). The objectives of DENSITIES are very ambitious:
1) Build a reference database available to the community for turbulence modelling validation,
2) Provide some insights on the correctness of commonly used assumptions in numerical methods (for instance constant turbulent Prandtl number in heat transfer problems),
3) Advance in the understanding of bypass transition and its effect on heat transfer in a real turbine blade geometry.

The configuration is a well-documented test case extensively used for CFD validation (the so-called LS89 test case) that operates at transonic conditions and a Reynolds number close to one million. Due to the existence of very different scales (Hairpin vortices, inlet turbulence, Von Karman vortices, etc.), the required mesh is very large ( 5 billion points). The simulation is performed thanks to a multi-block structured flow solver, which already shows its capability to run on massively parallel architectures and to simulate turbulent flows, including aerodynamics and aero-acoustics problems. To run the simulation in a reasonable amount of time, the minimum number of computing cores used is 8,192.

In order to achieve the ambitious objectives of DENSITIES, the project proposes to associate three complementary research centres, each internationally recognised in its field of expertise: the Von Karman Institute for Fluid Dynamics (turbomachinery measurements), the University of Naples (simulation and analysis of turbulence) and CERFACS (high-performance computing of gas turbine flows).

Resource awarded: 25 million core hours on Curie TN @ GENCI@CEA, France

RHODQMC – Energy storage in the first step of vision explored by Quantum Monte Carlo / Molecular Mechanics calculations

Project leader: Leonardo Guidoni, University of L’Aquila, Italy

Collaborators: Daniele Bovi, La Sapienza, Universit di Roma, Italy; Daniele Varsano, La Sapienza, Universit di Roma, Italy; Andrea Zen, La Sapienza, Universit di Roma, Italy; Sandro Sorella, SISSA, Italy; Matteo Barborini, University of L’Aquila, Italy; Emanuele Coccia, University of L’Aquila, Italy; Maria Montagna, University of L’Aquila, Italy; Daniele Narzi, University of L’Aquila, Italy; Fabio Pitari, University of L’Aquila, Italy;

Abstract: Rhodopsin is the light-detecting membrane protein responsible for the dim light vision of vertebrates. Its chromophore, the Retinal Protonated Schiff Base, is covalently bound to the seven-helix transmembrane protein and upon photon absorption undergoes a cis/trans isomerization from 11-cis to all-trans isomer [1]. During this process a certain amount of energy is stored in the distorted chromophore, which subsequently relaxes triggering the protein structural changes that activate the signal cascade to vision. In absence of light the 11-cis to all trans isomerisation can still occur by thermal activation in the ground state, a process that is limiting visual sensitivity [2]. The accurate determination of activation energy and energy storage along the ground state path moving from the dark state to the batho state of Rhodopsin is a long-standing open question. Experimental values seem to be contradictory since thermal activation energy was estimated to be smaller then the energy storage.

To unravel this puzzling data, quantum mechanical calculations based on the available crystal structures might play an important role, estimating both the storage and activation energies. Electronic structure calculations on these systems are still a big challenge because of the necessity to use correlated quantum chemistry methods to obtain an accurate description of the electronic structure and of the total energies in the complex protein environment. Recently, Quantum Monte Carlo calculations are emerging as a correlated, accurate and scalable electronic structure tool to investigate the electronic and geometrical properties of protein chromophores and conjugated polymers [3-5]. Thanks to the use of high performance computing resources QMC calculations on systems up to 150 electrons are affordable both in gas phase and in a Quantum Mechanics / Molecular Mechanics scheme.

In the present proposal we will use Quantum Monte Carlo / Molecular Mechanics calculations to investigate the ground state minimum energy path for retinal isomerization in Rhodopsin. The comparison of our QMC results with the experimental and other theoretical data in literature will shed new light to the energetics of the thermal isomerization path of rhodopsin.

[1] Palczewski, K. Annuv. Rev. Biochem., 75, 743 (2006).
[2] Gozem, S., Shapiro, I., Ferr, N. and Olivucci, M. Science, 337, 1125 (2012).
[3] Coccia, E., Varsano, D. and Guidoni, L., J. Chem. Theory Comput., submitted (2012).
[4] Coccia, E. and Guidoni, L., J. Comput. Chem., 33, 2332 (2012).
[5] Valsson, O. and Filippi, C. J. Chem. Theory Comput., 6, 1275 (2010).

Resource awarded: 38.99 million core hours on JUQUEEN @ GCS@Jülich, Germany


Project leader: Bruno Guillaume, ARIA Technologies, France

Collaborators: Armand Albergel, ARIA Technologies, France; Claude Derognat, ARIA Technologies, France; Christophe Olry, ARIA Technologies, France; Qijie Zhang, ARIA Technologies, France;

Abstract: Until recently, extreme event databases used by the reinsurers have been built largely on collection of information on real past extreme events.

Due to the lack in many cases of large amounts of information data for such severe events, and due to the difficulty to perform sensitivity tests only on the basis of such real case information, recent work by insurance-reinsurance R&D teams have started in order to provide a new generation of extreme event databases by using “event generators” (ex. “weather generators”). Use was made for example of the validated dynamical meteorological model MM5 at fine scale (downto several square kilometer resolution) to build such an database of realistic extra-tropical storms (Keller et al., 2004). Starting from the large scale meteorological conditions which enable the model to reproduce satisfactorily real past events (Lothar, Martin…), perturbations were introduced to simulate new events, having physically realistic footprints and intensities, and enabling to provide insight in the potential damages of slight variations from the real extreme events.

In this project, we further extend this approach, by assessing in a more general way the perturbations, by using perturbations at a global scale from the newly produced ensemble global NOAA20Century Reanalysis, in order to generate new realistic events using the meteorological model WRF over the current climate time period (20 last years). In a first step, we perform in this time period a selection of the days, that are favorable conditions to generate a local extreme event following a method developed in collaboration with IPSL (Vrac and Yiou, 2010).

The studied region is the Netherlands country, and the perils simulated are the extreme rainfall events causing pluvial flooding. We concentrate in this project on 3-nest WRF simulations, forced by NOAA 20CR (Compo et al. 2006, Compo et al., 2011) fields and with resolutions of 27kmx27km hourly, of 9kmx9km hourly and 3kmx3km hourly.

This production of a new methodology and associated database of rainfall extreme vents does interest at high level both the reinsurance companies and the research partners involved in the Climate-KIC OASIS project (http://www.climate-kic.org/fileadmi…), leaded by Imperial College, with IPSL, TU Delft and Deltares, ARIA technologies being also a partner.

The request of capacity on the order of a Tier-0 (precise need detailed hereafter) for this project is motivated by the numerous simulations caused by simultaneous use of the ensemble reanalysis (up to 56 members), of a large simulation periods (a climate time slice of 20 years) and a detailed hourly-resolution meteorological modelling downto 3kmx3km over full Netherlands with WRF model.

Resource awarded: 6 million core hours on Curie TN @ GENCI@CEA, France

StarLife – Protostars: From Molecular Clouds to Disc Microphysics

Project leader: Troels Haugboelle, University of Copenhagen, Denmark

Collaborators: Christian Brinch, University of Copenhagen, Denmark; Soeren Frimann, University of Copenhagen, Denmark; Aake Nordlund, University of Copenhagen, Denmark; Jacob Trier Fredriksen, University of Copenhagen, Denmark; Remo Collet, University of Copenhagen, Denmark; Colin McNally, University of Copenhagen, Denmark; Gareth Murphy, University of Copenhagen, Denmark; Martin Pessah, University of Copenhagen, Denmark; Paolo Padoan, University of Barcelona, Spain;

Abstract: Understanding the physical processes that determine the rate at which gravitating bodies accrete mass and radiate energy is vital for
unraveling the formation, evolution, and fate of almost every type of object in the Universe. This is particularly true in the context of star formation, where there is an intricate interplay between large-scale environmental factors, which regulate the supply of mass and angular momentum for the ensemble of forming stars, and small-scale plasma physics processes, which control the ultimate rate at which energy is dissipated. These two fundamental processes operate over a wide range of spatial and temporal scales and it is thus impossible to study all of them at once in a single numerical simulation. This proposal addresses this limitation by employing two code families, which have been specifically developed to deal with the challenges at each end of the associated physical scales. The assembled team has a broad combination of scientific and technical expertise, and is focused on bridging the gap between molecular cloud scales, which determine the initial and boundary conditions for protostellar disc accretion, and the microphysics scales, which determine the ultimate fate of the accretion energy.

Herschel and ALMA are revolutionizing observational star formation, resolving for the first time the intricate structure of filaments that feed the stellar envelopes. Interpreting the observations will require modeling large statistical samples of protostars at high enough resolution, so that they can be used as synthetic templates for interpreting real observations. We will perform groundbreaking global simulations that will follow the formation and evolution of a thousand protostellar systems down to the opacity limit, with resolved discs and outflows embedded in a large scale model of a molecular cloud. We will employ sophisticated post processing techniques to build a large database with more than a hundred thousand synthetic images of the protostars from different viewpoints, evolutionary stages, and wavelengths, which can be used for non-parametric modeling and interpretation of observations of protostars, in particular from ALMA.

The simulations will also provide an unprecedented global view of the formation of protostellar discs, in the presence of selfconsistently evolved ‘external’ magnetic fields. Previous disc studies have used idealized initial conditions and relied on the fluid-MHD framework, which breaks down at the very low levels of ionization in cold discs, and at the very low densities in disc coronae. The details of how magnetic fields reconnect and energy is dissipated in protostellar discs is of fundamental importance, and needs to be understood on solid physical grounds by going beyond the fluid regime. We will carry out the first ever 3D particle-in-cell simulation that follows the magneto-rotational instability and its non-linear evolution to saturation. By self-consistently calculating the neutral-ion-electron interactions, as well as the photon spectrum emitted by accelerated charged particles, we will be able to compute both the thermal energy dissipation and the emergent non-thermal spectrum from first principles. This will open up a completely new window into one of the most important instabilities in modern astrophysics.

Resource awarded: 75 million core hours on JUQUEEN @ GCS@Jülich, Germany

Simulating Dark Matter on the Lattice

Project leader: Ari Hietanen, University of Southern Denmark, Denmark

Collaborators: Claudio Pica, University of Southern Denmark, Denmark; Francesco Sannino, University of Southern Denmark, Denmark; Ulrik Sondergaard, University of Southern Denmark, Denmark;

Abstract: Most of the matter in the universe is in the form of dark matter. It does not emit or absorb light, and its existence has only been observed from its gravitational effect on the visible matter. However, little is known about the elementary constituents of dark matter, and the study of those, is one of the most topical problems in experimental and theoretical particle physics. In this project, we study a model where a dark matter candidate is a composite particle, which dynamics is described by yet to be found technicolor interactions. The particle would be many ways similar to a proton, except it does not have a charge, and only interacts weakly with the visible matter. In addition, we will study if a light Higgs particle could emerge from the Technicolor theory. Because the theory includes strong dynamics, no analytic solution to them is known, and large scale lattice calculations are the only first principle method known to gain knowledge about the model.

Resource awarded: 7.3 million core hours on MareNostrum @ BSC, Spain

The CASINO code used to help design industrial catalysts: the first Quantum Monte Carlo (QMC) benchmark for adsorbed reactions with reliable experimental data.

Project leader: Philip Hoggan, Institut Pascal, France

Collaborators: Daniel Claves, Clermont University, France; Geert-Jan Kroes, Leiden University, The Netherlands; Philip Thomas, Leiden University, The Netherlands; Neil Drummond, Lancaster University, United Kingdom;

Abstract: This proposal aims to bridge the gap between model and industrial catalysts for the hydrogen dissociation reaction on Cu(111). This is chosen since accurate experimental data exists.

Previous studies of model catalysts for the water gas shift reaction by Quantum Monte Carlo (QMC) techniques used an award of 1.5 million hours on Bluegene/P running the CASINO code in 2011. The surfaces studied were copper (100) with and without oxide impurities.

The same partners involved have been joined by G-J Kroes (and P Thomas from his group). They suggest benchmarking the hydrogen dissociation reaction to chemical accuracy using QMC. This would be accurate and useful ab intio determination of key points on the potential energy surface, for subsequent use in molecular dynamics.

We suggest setting up a data base of such QMC results. The results of this proposal could be the inaugural entry in this data base.

The method of choice for evaluating salient positions on the potential energy surface (containing the adsorbed reaction path) is Quantum Monte Carlo. This approach is well-known for controllable accuracy and being appropriate to Tier-0 resources.

The CASINO code will be used, as it is suitable for solid state applications. The systems to be studied are close-packed metal surfaces at which molecular hydrogen is dissociated.

It is too time-consuming to check more than a few key points on the dissociation pathway, e.g. the transition state and reactants, thus estimating activation barriers.

Resource awarded: 2 million core hours on MareNostrum @ BSC, Spain

QCDOVERLAP – QCD Thermodynamics with three flavors of dynamical overlap fermions

Project leader: Sandor Katz, Eotvos University, Hungary

Collaborators: Daniel Nogradi, Eotvos University, Hungary; Attila Pasztor, Eotvos University, Hungary; Norbert Trombitas, Eotvos University, Hungary; Tamas Kovacs, Institute of Nuclear Research, Hungary; Ferenc Pittler, Institute of Nuclear Research, Hungary;

Abstract: At very high temperatures in the early universe and at the Large Hadron Collider (CERN) & Relativistic Heavy Ion Collider (Brookhaven) ordinary hadronic matter undergoes a transition to a quark-gluon plasma. This transition is the remnant of a chiral phase transition, which would happen in an exactly chiral theory. All continuum extrapolated results, which studied this transition, used formulations of the theory with non-chiral fermions (Wilson or staggered case, where the latter has another symmetry than the continuum theory). Therefore it is absolutely necessary to study the transition with a chiral fermion formulation. One can confirm earlier finding (nature of the transition and its characteristic scale) and even more importantly study features, which have been non-accessible until now with non-chiral actions. One prominent example is the chiral endpoint in the light quark mass versus strange quark mass plane. In the future this endpoint may be continued into the non-vanishing chemical potential plane and give rise to the conjectured critical endpoint searched in the energy scan program of RHIC. Through these efforts large scale simulations can provide a point of direct comparison between fundamental theory, model and experiment.

Resource awarded: 74.4 million core hours on FERMI @ CINECA, Italy

Towards chemical accuracy for methane reacting on metal surfaces

Project leader: Geert-Jan Kroes, Leiden University, The Netherlands

Collaborators: Francesco Nattino, Leiden University, The Netherlands; Bret Jackson, University of Amherst at Massachusetts, United States;

Abstract: In the allocated project, research will be performed that will enable modeling of the interaction of methane molecules with metal surfaces with chemical accuracy. This is important to heterogeneous catalysis in general and hydrogen production in particular: Ni surfaces catalyse the so-called steam reforming reaction in which methane reacts with water to form hydrogen, in the currently commercial process for making hydrogen. Hydrogen is needed to make ammonia as a raw material for fertiliser, and is potentially useful as a fuel replacing fossile fuels. In heterogeneously catalysed processes, surfaces catalyse reactions of molecules with each other, and these processes are important because the production of most man-made chemicals involves heterogeneous catalysis at some point.

The accurate modeling of reactions on metal surfaces is hampered by the lack of accuracy of so-called electronic structure methods for large systems. So far, the only method of practical use for molecules interacting with metals is density functional theory (DFT). In the versions of the theory presently applicable in dynamics calculations on molecule-surface reactions, reaction barrier heights can be computed with an accuracy of no better than 2.2 kcal/mol. However, “chemical accuracy” (errors less than 1 kcal/mol) is required for accurate simulations of heterogeneously catalysed proccesses and of the elementary surface reactions involved in these processes. Fortunately, as demonstrated previously for the elementary surface reaction in which hydrogen dissociatively chemisorbs on a copper surface (i.e., the H-H bond is broken and the H-H atoms form new bonds to the surface), elementary surface reactions can now be simulated with chemical accuracy through a semi-empirical procedure. In this procedure, the so-called density functional used in the DFT calculations is fitted to a set of experiments that are sensitive to the height of the reaction barrier for the specific system studied.

We will apply this semi-emperical procedure to the dissociative chemisorption of methane on two metal surfaces, to develop the functionals with the required accuracy. The calculations will be done with the so-called ab initio molecular dynamics (AIMD) method, in which the forces exerted by the surface on the impinging molecule are computed “on the fly” from first principles. The use of the AIMD method is required because it allows the motion of all the atoms in the molecule and of the atoms in the top layers of the metal surface to be simulated, as is required to achieve the accuracy aimed for. Because AIMD calculations need to be computed with high enough accuracy for many trajectories involving individual molecule-surface collisions, a high amount of computer time is required on a “Tier-0” machine, which is a massively parallel computer.

Resource awarded: 25 million core hours on MareNostrum @ BSC, Spain

SWEET: Space Weather Effects on the Earth environmenT

Project leader: Giovanni Lapenta, KU Leuven, Belgium

Collaborators: Jorge Amaya, KU Leuven, Belgium; Arnaud Beck, KU Leuven, Belgium; Jan Deca, KU Leuven, Belgium; Maria Elena Innocenti, KU Leuven, Belgium; Viacheslav Olshevskyi, KU Leuven, Belgium; Anna Lisa Restante, KU Leuven, Belgium; Alexander Vapirev, KU Leuven, Belgium;

Abstract: We plan to study with our massively parallel code iPic3D the processes involved in the connection of the Sun with the Earth, known collectively as Space Weather.

Space Weather refers to conditions on the Sun, in the interplanetary space and in the Earth space environment that can influence the performance and reliability of space-borne and ground-based technological systems and can endanger human life or health. Adverse conditions in the space environment can cause disruption of satellite operations, communications, navigation, and electric power distribution grids, leading to strong socioeconomic impacts. This is the main reason why space weather is currently receiving a growing attention and is the subject of international efforts worldwide.

We focus here on achieving a true physics-based ability to predict the arrival and consequences of major space weather storms. Great disturbances in the space environment are common but their precise arrival and impact on human activities varies greatly. Simulating such a system is a grand-challenge, requiring state of the art computing resources at the limit of what is possible. Modeling space weather requires the best computational resources available, as it demands the resolution of a wide variety of physical processes and of different time and space scales.

Our simulation code, iPic3D relies on the implicit moment method. The fields and the particles are studied together in a coupled manner. The code has been developed in the last three years based on the implicit moment method that previously available only on single processor computers. iPic3D implements an advanced mathematical method to couple small and large scale models. This project for the first time allows us to study large astrophysical systems with small scale physical processes included self-consistently.

Resource awarded: 10 million core hours on Curie TN @ GENCI@CEA, France and 10 million core hours on FERMI @ CINECA, Italy

High Performance Computing with Generic Solvers for Partial Differential Equations 2 (HPC-PDE 2)

Project leader: Frederic Nataf, Universite Pierre et Marie Curie, France

Collaborators: Victorita Dolean, Universite Nice Sophia-Antipolis, France; Frederic Hecht, Universite Pierre et Marie Curie, France; Pierre Jolivet, Universite Pierre et Marie Curie, France; Christophe Prud’homme, Universite de Strasbourg, France;

Abstract: A partial differential equation is a relation between a function of several variables and its (partial) derivatives. It is used in physics, mathematics and engineering to model most of the natural phenomena such as propagation of sound, electrodynamics, fluid flow, elasticity or more complex problems. From the point of view of applied mathematicians, one of the different ways to solve such equations is by using the so called finite element method.

FreeFem++ and Feel++ are open source software projects used to solve these equations numerically on arbitrary finite element spaces on arbitrary unstructured and adapted two- and three-dimensional meshes. They are based on C++ and runs on most computer architectures, from low-end desktop personal computer to high performance clusters. It enables the end-user to solve PDE easily, without having to take care of some difficult implementation problems in the finite element method framework such as: automatic mesh generation, mesh adaptation, matrix assembling, quadrature approximations, MPI bindings.

Our goal is to use those generic frameworks to develop and assess the efficiency of parallel numerical algorithms. Among others, one of the most promising paradigm in this area is the field of domain decomposition methods, where – roughly speaking – the initial workload that would result from solving a PDE on a single computer is spread among a group of processes. As is, one-level domain decomposition methods are not scalable, meaning that by increasing the number of processes, i.e. the number of subdomains in our original problem, the solving of the parallel problem becomes harder. However, it is possible to address this issue by solving an auxiliary problem known as coarse space problem, that can be build using geometrical consideration, or by solving local problems. In contrast with one-level methods, these are known as two-level methods.

Using this technique, we have achieved linear speedup in the solving of various classes of PDE: the Darcy equation with highly heterogeneous coefficients and the system of linear elasticity with highly heterogeneous materials. We have also successfully solved heterogeneous large linear systems preconditioned by two-level methods, in both 2D (circa 3 billion unknowns) and 3D (circa 300 million unknowns). These results were obtained on Curie (French Tier-0 system), thanks to last year HPC-PDE grant. From a mathematical point of view, we are currently investigating harder problems by using our domain decomposition framework to solve nonlinear problems. It should be noted that we plan on porting our framework to the MPI/OpenMP hybrid parallelization paradigm. This clearly shows why the access to a Tier-0 system would represent an invaluable asset.

Resource awarded: 2 million core hours on Curie TN @ GENCI@CEA, France

QUASINO – QUAntum SImulation of ultimate NanO-devices

Project leader: Yann-Michel Niquet, CEA, France

Collaborators: Ivan Duchemin, CEA, France; Viet-Hung Nguyen, CEA, France; Fabio Perreira, CEA, France; Francois Triozon, CEA, France;

Abstract: The characteristic dimensions of the transistors on a processor chip have steadily decreased over the last fifty years, allowing for ever more performances and functionalities. The International Technology Roadmap for Semiconductors (ITRS), which sets targets for the microelectronics community, is now discussing options beyond the “16 nm node”. At this scale, transistors will likely resemble, in a way or the other, to nanowires with diameters around or below 10 nm. The physics of such devices is, however, very complex and goes beyond semi-classical understanding. Modeling and simulation are therefore expected to play an increasing role in the exploration of original and innovative designs.

This calls for the development of advanced simulators able to account for the quantum effects, such as tunneling and confinement, prevailing at the nanometer scale. The Non-Equilibrium Green’s Functions (NEGF) method is one of the most versatile approach to quantum transport. It describes all important scattering mechanisms (scattering of electrons by impurities, lattice vibrations, etc…) in an unified framework. We have recently developed a new NEGF solver based on effective mass and atomistic tight-binding models of the electronic structure of the materials, carefully optimized for high-performance computing infrastructures and hybrid CPU/GPU machines. This code is able to simulate devices with realistic geometries and sizes in the 10 nm range, representative of the next generation transistors being developed in the laboratories at present.

The objective of this project is to model nanowire transistors in close connection with experimental teams of the CEA/LETI fabricating and characterizing these devices for fundamental and technological purposes. We will, in particular, investigate the effects of the gate architecture (“tri-gate”, “gate-all-around”…) and of the mechanical strains on the performances of the transistors, in order to help their design and optimization. We endeavour to understand the intrinsic physics of these devices, and to refine the description of the scattering mechanisms at the nano scale. We shall that way be able to assess clearly the strengths and weaknesses of nanowire devices with respect to traditional planar architectures. We will also investigate localization by surface roughness and impurities within the channels of ultimate nanowire devices with diameters < 5 nm, to understand puzzling transport spectroscopy experiments showing signatures of localized states in the current up to room temperature.

Resource awarded: 0.155 million core hours on Curie H and 6 million core hours on Curie TN @ GENCI@CEA, France

DRAGON – Understanding the DRAG crisis: ON the flow past a circular cylinder from critical to trans-critical Reynolds numbers

Project leader: Assensio Oliva, Tecnical University of Catalonia, Spain

Collaborators: Ricard Borrell, Tecnical University of Catalonia, Spain;Jorge Chiva, Tecnical University of Catalonia, Spain;Oriol Lehmkuhl, Tecnical University of Catalonia, Spain;Ivette Rodriguez, Tecnical University of Catalonia, Spain;

Abstract: This project is in the context of the numerical simulation of the turbulent flow past bluff bodies which is of importance due to its presence in many engineering applications (e.g. flows past an airplane, a submarine and an automobile, in turbomachines, etc.). The flow past a circular cylinder is associated with various instabilities which involve the wake, separated shear layer and boundary layer. When Reynolds number increases the transition to turbulence in the separated shear-layer (SL) moves forward the cylinder surface and at about Re=2e5, this point is located too close to that of the separation of the SL. As a results, SL eddies cause mixing of the flow and undergoes transition from laminar t
o turbulent. The transition causes the delaying of the separation point and an important reduction of the drag force on the cylinder surface. This phenomena is known as Drag Crisis. In the critical Reynolds number range, flow transition is accompanied of different regimes such as the presence of asymmetric forces in the cylinder surface which causes average lift forces greater than zero. At larger Re, these asymmetries disappear and the flow is again symmetric, but now with turbulent separation.

Considering the advances in computational fluid dynamics (CFD) together with the increasing capacity of parallel computers, which have made possible to tackle complex problems, we propose to perform well solved large-eddy simulations of the flow from critical to supercritical Reynolds numbers in the range 6e5-4e6. In a previous project at Marenostrum Supercomputer we have carried out LES computations at critical Re=1.4e5-5.3e5 and a first attempt to compute the flow at Re=1e6 (FI-2012-1-0011 and FI-2012-2-0010). Due to the large computational resources required, only the flows up to Re=5.3e5 were accurately computed, considering the importance of solve well the boundary layer in the cylinder surface. Thus, the main goals of this project is to advance in the understanding of the physics of turbulent flows and in particular to gain insight in the mechanism of the shear-layer transition and its influence in the wake characteristics and in the unsteady forces on the bluff body surface. In the range of Re numbers to be study, different flow features are to be investigated, such as the presence of separation bubble which has been suggested to be a characteristic of the transition to the trans-critical regime, the existence of vortex shedding during at its characteristic frequency in the whole range (there is a large scattering in the measurements and controversies about the suppression of vortex shedding at some Reynolds numbers), near wake characteristics, among others. The computed data and the results can also be used as reference data for comparison with results from different turbulence modeling. In this sense it is our aim to make available for the research community the results obtained through the creation of a database with specific data from these computations.

Resource awarded: 23 million core hours on MareNostrum @ BSC, Spain


Project leader: Ignacio Pagonabarraga, University of Barcelona, Spain

Collaborators: Francisco Alarcon, University of Barcelona, Spain;Paolo Malgaretti, University of Barcelona, Spain;Andrea Scagliarini, University of Reykjavk, Iceland

Abstract: ESFSAP focuses on the study of the collecive properties of suspensions composed by particles that can selfpropell at the microscale. Microorganism suspensions and microbots propel through reactions which are catalyzed at their surfaces. The internal activity of these systems confer them with unique properties due to the fact that they are intrinsically out-of-equilibrium. As a result, emergent structures develop and promote new structures and morphologies that are not controlled by the environment parameters in which these systems are embedded. Hence, their sensitivity and mechanical properties are qualitatively different from the ones known for their passive counterparts.

This project analyses the competition between the hydrodynamic stresses induced by the internal activity and the direct (directional) interactions between self-propelling particles. We will clarify the capability of such competition to promote cluster formation and the possibility that these clusters subsequently merge into large spanning morphologies. We will characterize if these structures can be a basis for active gels. The intrinsic motion of active particles generate fluid flow. Hence, we will also consider the features of the fluid flow induced by these structures and thei rpotential applications for fludi stirring at small sclaes and use in microfluidic devices.

An alternative way to induce larger structures from active suspensions is by bonding them permanently to form filamentous structures, analogous to polymer chains. These active polymrs move spontaneously in a solvent and can lead to a novel kind of polymer solutions. We will address for the first time the collective response of ensembles of such polymers and the emerging structures they develop.ESFAP will also consider the fundamental question of the fluid flow generated by active swimmer suspensions. The flow is produced at short scales due to their intrinsic activity but give rise to irregular large scale fluid flows. Although the experimental observations ofa time-dependent, disorder flows has been refereed to as “bacterial turulence”, we will address the specifity of this new type of irregular motion and its similarities and differences with inertial turbulence.

Resource awarded: 17 million core hours on MareNostrum @ BSC, Spain

Lyman-alpha forests and neutrino mass

Project leader: Nathalie Palanque-Delabrouille, CEA, France

Collaborators: Julien Lesgourgues, CERN, Switzerland; Arnaud Borde, CEA, France; Jean-Marc Le Goff, CEA, France; Graziano Rossi, CEA, France; Christophe Yche, CEA, France; Matteo Viel, INAF/OATS, Italy; James Bolton, University of Nottingham, United Kingdom;

Abstract: A major breakthrough in particle physics over the last decade is the confirmation that neutrinos are massive. Their mass, however, remains unknown and is the source of intense research activity. In the near future, cosmological data are expected to offer the best sensitivity to the neutrino mass, better than what can be achieved with laboratory experiments. The objective of this project is to use the signature left in quasar spectra by the presence of neutral hydrogen in the Universe to measure, or constrain, the sum of the masses of the three neutrino flavors at the 0.1 eV level.

We propose to use data collected by the BOSS experiment between 2009 and 2014. By the end of the survey, up to 160,000 quasar spectra should be measured, exceeding previously available data sets by over an order of magnitude. To complete the project we will, on the one hand, analyse the measured quasar spectra to derive the flux power spectrum, and simulate, on the other hand, the distribution of matter in the Universe in the presence of massive neutrinos, through N-body simulations that we develop from an already existing code. Given the significant improvement in data quality a similar technological upgrade is needed in the simulation part required to interpret the results. The N-body approach presented in this proposal is the only one that provides the required precision.

Resource awarded: 7 million core hours on Curie TN @ GENCI@CEA, France

FMOC – Fast Multi-physics Optimization of a Car

Project leader: Marc Pariente, Renault SAS, France

Collaborators: Argiris Kamoulakos, ESI Group, France; Raymond Ni, ESI Group, France; Antoine Petitet, ESI Group, France; Guillaume Pierrot, ESI Group, France; Torsten Queckboerner, ESI Group, France; Derrick FONGANG FONGANG, Ecole des Mines de Saint-Etienne, France; Rodolphe Le Riche, Ecole des Mines de Saint-Etienne, France; Stphane Dugardin, Renault SAS, France; Frederic Mercier, Renault SAS, France; Yves Tourbier, Renault SAS, France; Pascal Vaumoron, Renault SAS, France;

Abstract: The modeling of a car crash is an important task for Renault engineering. Regarding our policies for the security of our customers and the environment (other occupants, driver, pedestrians, etc…), we have been working on our way to model and to improve numerical simulations.

Today we realize several optimization phases in our standard design process of a car. These optimizations use standard model of 3 million of finite elements. During this, we vary around 50 parameters (like thicknesses, material…).

This method is efficient but can be improved: our numerical simulations cannot represent the entire physical phenomenon; the final validation with real test is different to simulation. This difference can increase with optimization: optimization uses the numerical simulations in a very different context as usual because we vary a great number of design parameters (the final solution can be far away from the initial point which has been validated by some real crash-tests).

We demonstrate that increasing the number of parameters will improve the quality of the solutions in numerical phase. The total number of design parameters for a parametric optimization of the car body is greater than 200. Assessing for the first time such parametric optimisation is currently unreachable with our current R&D facilities. The conclusion was we need to improve the crash model accuracy and the performance of our optimization algorithms.

We are building a big car model (more than 20 million finite elements, x50 on CPU) including many physics phenomena not taken into account today:

  • Iron sheet cracking
  • Welding point cracking
  • Failure of screw body
  • Wheel rim deformation
  • Airbag and passenger modeling
  • Windshield cracking
  • Stamping coupling

We want to do a breakthrough on our optimization tools, for example use all available data from crash simulation. We will use BIG DATA technology to extract more information and improve the accuracy of our statistical models

We want to take benefit from the PRACE facilities for demostrating the feasability and the efficiency of such numerical approaches in order to model a full car optimisation analysis with more than 200 parameters on a 20 millions elements mesh. Moreover, we want to solve optimization problems including combinatorial aspects, for example reach specifications with minimal cost and weight and re-use 80% of the parts of an old car to design the new one (choose the parts to modify, then choose parameters like thickness, material properties, shape… for each chosen part).

Resource awarded: 42 million core hours on Curie TN @ GENCI@CEA, France

Structure and evolution of an active region on the Sun

Project leader: Hardi Peter, Max-Planck-Institut fr Sonnensystemforschung, Germany

Collaborators: Sven Bingert, Max-Planck-Institut fr Sonnensystemforschung, Germany; Philippe Bourdin, Max-Planck-Institut fr Sonnensystemforschung, Germany;

Abstract: Cool stars are surrounded by hot coronae, which are heated to some million degrees Kelvin. The heating processes, widely proposed to be related to the stellar magnetic field, lead not only to temperatures in the outer atmosphere well in excess of the stellar surface, but result also in a highly dynamic response of the plasma, inducing flows and waves. To study these processes the Sun is of pivotal interest because here we can spatially resolve individual structures in the corona.

Utilizing increasing computing power, we plan to run new numerical experiments which will allow to study the structure and evolution of an active region in the solar corona. To this end we describe part of the solar corona, i.e. an active region, in a box using magneto-hydrodynamics. The system is driven by fluid motions on the solar surface driven by convection which carry around the magnetic fieldlines. This leads to currents in the upper layers which are dissipated, subsequently heat the atmosphere and thus create the corona.

So far this modelling has been possible only in simplified setups. The proposed simulation will allow for the first time to model the structure and evolution of a solar active region with a high spatial resolution (230 km) to resolve most of the driving motions on the surface and at the same time to describe the full extend of the active region (235×235 Mm).

The results from these simulations will be compared to real observations of the Sun by deriving synthesised observations, i.e. emission line spectra, from the numerical experiments (which is not part of this proposal, as it can be done on a normal computer).

Resource awarded: 5.94 million core hours on Curie TN @ GENCI@CEA, France

Phase transition based control of friction at the nanoscale

Project leader: Carlo Antonio Pignedoli, Empa, Switzerland

Collaborators: Andrea Benassi, Empa, Switzerland; Daniele Passerone, Empa, Switzerland;

Abstract: The possibility to exploit some kind of phase transition in order to control the friction coefficient of two sliding bodies has been theoretically demonstrated and experimental evidence is already available. However, a suitable and experimentally accessible phase transition to actuate the friction control has not yet been identified. Through this PRACE project we will set up a series of simulations whose aim is to demonstrate that the rotational melting phase transition occurring in molecular crystals such as C60 fullerite, is a good candidate to actuate friction control at their surface. Through classical molecular dynamics we will simulate a model atomic force microscopy (AFM) tip sliding on a fullerite slab and we will show how its frictional properties can be modified by the occurrence of the rotational melting transition. Our results will serve as a basis to design new friction control experiments in collaboration with our experimental partners.

Resource awarded: 18.6 million core hours on SuperMUC @ GCS@LRZ, Germany


High Performance Finite Element Embedded Library (or Language) in C++

Project leader: Christophe Prud’homme, Universite de Strasbourg, France

Collaborators: Trophime Christophe, Laboratoire National des Champs Magnetiques Intenses, France; Cecile Daversin, Laboratoire National des Champs Magnetiques Intenses, France; Vincent Doyeux, Universite Joseph Fourier Grenoble 1, France; Mourad Ismail, Universite Joseph Fourier Grenoble 1, France; Pierre Jolivet, Universite Joseph Fourier Grenoble 1, France; Abdoulaye Samake, Universit
e Joseph Fourier Grenoble 1, France; Stephane Veys, Universite Joseph Fourier Grenoble 1, France; Chabannes Vincent, Universite Joseph Fourier Grenoble 1, France; Yannick Hoaraux, Universite de Strasbourg, France; Marcela Szopos, Universite de Strasbourg, France; Ranine Tarabay, Universite de Strasbourg, France; Goncalo Pena, University of Coimbra, Portugal;

Abstract: The HP-Feel++ project aims at developing two research applications that require now access of TIER-0 computing resources: blood flow rheology and high field magnets.

Although these domains are quite different they have been thoroughly developed for the past few years within the Feel++ project (http://www.feelpp.org). They share the same mathematical kernel that encompasses a large range of numerical methods to solve partial differential equations such as (i) arbitrary order continuous and discontinuous Galerkin methods in 1D, 2D and 3D, (ii) domain decomposition methods, (iii) fictitious domain methods, (iv) level-set methods or (iv) certified reduced basis methods. These methods are developed and used easily thanks to a domain specific language embedded in C++ mimicking the mathematical language associated to Galerkin methods. This language allows physicists, engineers and mathematicians to focus on the numerical methods as well the physics whilst it hides the computer science details (e.g. parallelism) or algebraic solvers and enables the user to ramp up very quickly from rapid prototyping numerical methods to large scale computations. Within this context, blood flow rheology and high field magnets are the two domains driving Feel++ developments.

In blood flow rheology, we are interested in simulating suspensions of red blood cells (RBC) in arteries and veins and in studying the fluid properties (i.e. the fluid apparent viscosity) either in healthy contexts (our current focus) or pathological contexts (in the longer term). Not only the RBC are deformable entities, arteries and veins deform also during blood pulse; in both cases fluid structure interaction modeling and simulations are required. We have developed two main alternatives to tackle these problems: (i) fluid structure interaction within the so-called Arbitrary Lagrangian Eulerian framework coupled with a fictitious domain method to handle the RBC and (ii) fluid structure interaction using level-set methods. In both cases the fluid-structure coupling is strongly nonlinear and the computational and storage costs for realistic simulations require the TIER-0 infrastructures.

As to high field magnets (i.e. magnetic intensity greater than 24T), they are being developed by a large scale equipment laboratory (Laboratoire national des champs magntiques intenses) and they are accessible to the international scientific community through project calls. Studies range from solid physics to applied supra-conductivity and magneto-science. The design and optimisation of these high field magnets require the solution of large scale multi-physics (and mildly multi-scale) non-linear partial differential equations. Moreover to ensure a robust design, we need to assess uncertainties through quantile estimations and sensitivity analysis. The latter is built on the former as it requires hundred or thousands evaluations of the former. We have developed the so-called certified reduced basis in this context to reduce the computational cost within the uncertainty quantification and optimisation processes from millions of degrees of freedom to a few tens or hundreds. This huge computational gain requires however the acceptance of an intensive offline stage allowing to get the independence with respect to the costly (typically finite element) underlying models and which demands now the access to TIER-0 infrastructures.

Resource awarded: 60 million core hours on SuperMUC @ GCS@LRZ, Germany

Investigating Thrust Performance of Flettner Rotors with Thom Discs

Project leader: Alistair Revell, The University of Manchester, United Kingdom

Collaborators: Juan Uribe, EDF R&D UK Centre, United Kingdom; Charles Moulinec, STFC Daresbury Laboratory, United Kingdom; Timothy Craft, The University of Manchester, United Kingdom; Brian Launder, The University of Manchester, United Kingdom;

Abstract: A combination of the ever-growing threat of dangerous changes to the Earth’s climate from the mounting levels of atmospheric CO2 and the acceptance that, in any event, reserves of fossil fuels are rapidly dwindling (while their cost rises inexorably) has led to the Flettner rotor again being considered for marine propulsion. Originally motivated by work with Prandtl, Flettner successfully demonstrated that large vertical-axis rotating cylinders mounted on a ship’s deck could be employed to generate a propulsive force via the Magnus effect. Today, however, several ship-design companies propose Flettner rotors as an auxiliary power source while, since autumn 2010, Enercon’s E-Ship-1, has been delivering the company’s wind turbines to its customers. A fleet of ships powered by Flettner rotors has been recently postulated as a means of tackle climate change via geoengineering. It was suggested by Thom in the 1930s that the addition of spanwise discs along the length of the cylinder would act to improved performance at high spin rates, and while this has been shown to be the case for moderate Reynolds numbers, as yet it has not been fully verified at the operating conditions of a Flettner rotor. Indeed, many questions about Flettner rotors remain unanswered and there is a desperate need for accurate reference data at relevant flow conditions. Experimental data at such high rotation rates and Reynolds number is fraught with challenges, and thus moving forwards, numerical investigation is the clear candidate; increased activity in the area is needed in order to make these proposals a reality. The primary objective of the current work is to perform a series of wall-resolved Large Eddy Simulations of the flow around a rotating cylinder with different disc configurations, at the representative value of alpha=10 and Re=5×10^5; which approaches the operating conditions of a Flettner rotor (Re=1×10^6). At this rate of spin, the flow velocity relative to wall on one side is increased by a factor of 10; implying a local maximum Reynolds of 5×10^6. This represents the first such study undertaken and would provide valuable and detailed insight into 1) the principle flow mechanism in operation at these flow conditions, and 2) a detailed set of data for the purpose of assessment and development of the more efficient turbulence modelling and simulation methods; whose informed usage would then become invaluable as a design tool, so as to fully explore the parameter space. The debate as to whether or not the addition of discs actually improves the performance remains unanswered for the relevant flow conditions, and thus the work proposed here would go quite some way to addressing this question. Flow at moderate and high rotating rates will be computed for a smooth cylinder, as well as for a cylinder modified with first spanwise discs, and then ‘blended’ discs.

Resource awarded: 20.8 million core hours on Hermit @ GCS@HLRS, Germany

Longitudinal and Transverse Electronic Transport in Atomically Doped Graphene from First Principles

Project leader: Stephan Roche, Catalan Institute of Nanotechnology, Spain

Collaborators: Jean-Christophe Charlier, Universit catholique de Louvain (UCL), Belgium; Nicolas Leconte, Universite catholique de Louvain (UCL), Belgium; Frank Ortmann, Catalan Institute of Nanotechnology, Spain;

Abstract: By combining linear scaling transport approaches with first principles parameters, we study charge transport in disordered graphene which is displaying remarkable properties and stand as among the most promising innovative materials for future applications.

Nowadays graphene research is entering a new phase focusing on the microscopic understanding of doping and realistic material disorder owing to the need of understanding physical properties of real samples including polycrystalline graphene and doped or chemically modified graphene matrix. Chemical or micro-mechanical routes of processing graphene impact on its transport properties and have to be understood to realize this potential. Transport and magneto-transport are used to characterize this material and assess its potential for electronics applications. Therefore, also (magneto-) transport simulations are strongly required to assist and deepen this understanding microscopically which evades experimental access. Such theoretical characterization includes longitudinal as well as transverse (Hall) transport under magnetic fields.

We recently overcome severe limitations of Hall transport simulations in developing an efficient real space and order N simulation code that copes with macroscopic samples of micron size and realistic disorders as usually observed in experiment. One main goal of this project is to unravel microscopic mechanisms that can lead to transport gaps in disordered graphene or unconventional additional Hall plateaus by comparing model disorders and first-principles calculations of the effect of epoxy defects which are currently under experimental examination. The outcomes of our computational study could for the first time connect the phase diagram of the quantum Hall effect to the nature of localization effects in realistic models.

The computational strategy is based on a real space implementation of transport coefficients, and on the Lanczos approach to compute off-diagonal elements of the Green´s function matrix. This enables to simulate disordered materials of macroscopic sizes, thus reaching a quantitative and predictive power for comparison with experimental data achieved in real materials. The developed algorithms are highly parallelized and will run optimally on massively parallel machines such as the present generation of tier-0 HPC machines.

Resource awarded: 14.43 million core hours on Curie TN @ GENCI@CEA, France

CWIN – Mapping the conformal window

Project leader: Kari Rummukainen, University of Helsinki, Finland

Collaborators: Marco Panero, Helsinki Institute of Physics, Finland; David Weir, Helsinki Institute of Physics, Finland; Tuomas Karavirta, Jyväskylä University, Finland; Kimmo Tuominen, Jyväskylä University, Finland; Anne Mykknen, University of Helsinki, Finland; Teemu Rantalaiho, University of Helsinki, Finland; Jarno Rantaharju, Riken advanced institute for computational science, Japan;

Abstract: The LHC particle accelerator at CERN, together with other upcoming experiments and astrophysical observations, is on the course of completing our picture of the Standard Model of particle physics in the near future. Even more exciting are the hints that these experiments may open the window on new physics, physics beyond the Standard Model. At the same time, modern numerical lattice simulation methods of quantum fieldtheories have matured to very precise and reliable calculational tools, as shown by the remarkable success of lattice QCD simulations.

In this project we apply simulation methods to a novel direction: we study theories which are conformal or almost conformal at long distances. This means that the long distance physics becomes scale invariant. Conformal theories are ingredients for a candidate theory for new physics possibly found at LHC, technicolor. Technicolor is among the most popular extensions of the Standard Model. The physics is strongly interacting; thus, non-perturbative lattice simulations are needed to study it.

We study two candidate theories for conformal or almost conformal physics: SU(2) gauge field theory with either six fundamental representation fermions or two adjoint representation fermions. The initial lattice studies of these theories started a few years ago in a flurry of activity, including by our group. However, definite results are still lacking. In this project we aim at precision results of the two theories under study, using large volumes and statistics enabled by PRACE and highly developed simulation methods and tools.

Resource awarded: 50 million core hours on FERMI @ CINECA, Italy

DynSupercap – Modelling the charge/discharge dynamics of supercapacitors

Project leader: Mathieu Salanne, Universite Pierre et Marie Curie, France

Collaborators: Celine Merlet, Universite Pierre et Marie Curie, France; Clarisse Pean, Universite Pierre et Marie Curie, France; Benjamin Rotenberg, Universite Pierre et Marie Curie, France; Paul Madden, University of Oxford, United Kingdom;

Abstract: How much energy can I store in a device? How fast can it be charged? These two questions are at the heart of the research on electricity storage. We will focus on supercapacitors, a family of devices in which the charge is stored at the electrode/electrolyte interface through reversible ion adsorption at high-surface-area porous carbon electrode. The processes which govern their performance depend strongly on the transport of the ions inside the pores or at the interfaces, which is difficult to probe experimentally. The objective of this project is to simulate the dynamic processes occuring during the charge/discharge cycles at the atomic scale, in order to understand how the power density can be maximized. We are focusing our efforts on two types of electrolytes, i.e. neat ionic liquids and salts dissolved in organic solvents (e.g. acetonitrile). We use the molecular dynamics simulation technique, and we have a major technical advantage in our ability to simulate realistic electrode geometries through using a method to allow us to treat the electrode as polarizable, that is one that develops charge in response to the local electrical potential. Our theoretical work is done in strong collaboration, including regular meetings, with Patrice Simons team (University of Toulouse) which is systematically conducting a major experimental efforts on supercapacitance, involving synthesis and characterization, electrochemistry, electron microscopy, NMR. There are several key issues: How does the pore size distribution influence the charge/discharge rate? Which parameters control the charge/discharge times? How do the sizes of the anions and cations impact these characteristic times? The investigation of these questions by molecular dynamics simulations is challenging because of i) the necessity to include large carbon structures, which reproduce at least partially the heterogeneity observed in experimental nanoporous carbons, and polarization of the electrodes, and
ii) the time scales needed to observe complete charges/discharges. With rapid access to appropriate HPC facilities through this proposal, combined with in-place personnel, unique simulation capability, and collaboration with world-leading experimentalists we will be able to make a major contribution to this very hot topic.

Resource awarded: 16.78 million core hours on Curie TN @ GENCI@CEA, France

EAGLE – Evolution and Assembly of GaLaxies and their Environments

Project leader: Joop Schaye, Leiden University, The Netherlands

Collaborators: Simon White, Max Planck Institute for Astrophysics, Germany; Claudio Dalla Vecchia, Max Planck Institute for Extraterrestrial Physics, Germany; Robert Crain, Leiden University, Netherlands; Joop Schaye, Leiden University, Netherlands; Ian McCarthy, University of Birmingham / University of Liverpool, United Kingdom; Richard Bower, University of Durham, United Kingdom; Carlos Frenk, University of Durham, United Kingdom; Michelle Furlong, University of Durham, United Kingdom; John Helly, University of Durham, United Kingdom; Adrian Jenkins, University of Durham, United Kingdom; Yetli Rosas-Guevara, University of Durham, United Kingdom; Matthieu Schaller, University of Durham, United Kingdom; Tom Theuns, University of Durham, United Kingdom; Craig Booth, University of Chicago, United States;

Abstract: The history of the universe is rich and includes a variety of milestones, from the emergence of the first stars, the formation of star-bursting galaxies that are observed to drive supersonic shock waves into intergalactic space, to the assembly of the galaxies that we see around us today.

Our understanding of this rich cosmic history is limited by the complexity of the physics governing the ordinary matter, i.e. matter that interacts electromagnetically. On the other hand, dark matter, which is thought to dominate the matter content of the universe, is subject only to gravity and in that sense simpler to model. Numerical simulations, in which initial conditions that are well posed by the physics of the early universe are evolved to late cosmic epochs, have therefore emerged as the leading technique for i) obtaining predictions from theoretical models that can be confronted with observations and ii) gaining insight into the relative importance of the physical processes that control the formation and evolution of galaxies.

We propose to carry out the EAGLE project, which consists of gas-dynamical simulations of the evolution of galaxies and the gas around them. EAGLE will use realistic initial conditions to follow the evolution of both ordinary and dark matter in a universe dominated by dark energy, as required by the latest cosmological observations. The volume will be large enough to sample the galaxy mass function over the range observed in current wide-field surveys. The main EAGLE run will contain an order of magnitude more resolution element than the largest existing runs of this kind, which is sufficient to resolve the physics down to the length scales relevant for the warm interstellar medium.

Guided by the results of our previous work (in particular the OWLS project) and extensive parameter testing and code development programs that we have already completed, we will be able to obtain a remarkably good match to the observed galaxy mass function. This success will allow us to use EAGLE to address a variety of questions, ranging from the origin of galaxy shapes and masses to the impact of gas flows on galaxy evolution and on the ability of upcoming missions to provide precision measurements of the cosmological parameters.

Resource awarded: 39.8 million core hours on Curie TN @ GENCI@CEA, France

Kaon semi-leptonic form factor

Project leader: Enno Scholz, University of Regensburg, Germany

Collaborators: Gunnar Bali, University of Regensburg, Germany; Vladimir Braun, University of Regensburg, Germany; Alessio Burello, University of Regensburg, Germany; Benjamin Glaessle, University of Regensburg, Germany;

Abstract: The CKM-matrix describes the mixing between the mass eigenstates and electro-weak eigenstates of the different quark flavors in the Standard Model of Particle Physics. Tests of the unitarity of the CKM matrix might reveal deviations from the Standard Model and consequently will be a sign of new physics beyond the Standard Model. The CKM matrix elements can be obtained, e.g., from experimental measurements of certain decay modes like the kaon decaying into a pion and a lepton plus anti-neutrino pair (Kl3 decay) if the form factors for such a decay are known. Our calculation will address the kaon semi-leptonic form factor needed for the determination of the matrix element |V_us| from the Kl3 decay rates. The form factor will be calculated non-perturbatively from lattice QCD simulations carried out with 2+1 flavors of dynamical SLiNC-fermions, where pion masses down to 210 MeV have been reached.

Resource awarded: 11.7 million core hours on SuperMUC @ GCS@LRZ, Germany

Daemmerung Simulations — Exploring the cosmic dawn with large galaxy surveys of the Universe

Project leader: Robert Smith, Max Planck fur Astrophysik, Germany

Collaborators: Stefan Hilbert, Max Planck fur Astrophysik, Germany; Laura Marian, Max Planck fur Astrophysik, Germany; Simon White, Max Planck fur Astrophysik, Germany; Raul Angulo, KIPAC Stanford University, United States;

Abstract: We propose to perform four new simulations to explore the impact that different inflationary models would imprint on the late time distribution of galaxies and on weak lensing maps of the Universe. The first simulation we will perform will be the control, a standard Gaussian initial condition; the remainingthree will imprint three different generic bispectra into the initial conditions, consistent with different inflationary scenarios. Each simulation will evolve roughly 300 billion particles from the initial time to the final end state in a comoving cube of size L=3000 Mpc/h. These simulations will be unique in that they will havesufficient mass resolution to be able to accurately follow the hierarchical build up of galaxy populations, but will whilst also being large enough to explore the observational signatures of primordial non-Gaussianity on the largest scales currently observable. The “Dammerung’’ series of simulations will therefore provide the crucial missing link between the observable universe and the primordial initial conditions. Thus allowing us the potential to robustly explore the generation of perturbations at the dawn of our Universe.

Resource awarded: 10 million core hours on SuperMUC @ GCS@LRZ, Germany

Simulation of seismic wave propagation in multi-scale media

Project leader: Vladimir Tcheverda, Institute of Petroleum Geology and Geophysics SB RAS, Russia

Abstract: In order to simulate the interaction of seismic waves with cavernous/fractured reservoirs, a finite-difference technique based on locally refined time-and-space grids is used. The need to use these grids is due primarily to the differing scale of heterogeneities in the reference medium and the reservoir. Domain Decomposition methods allow for theseparation of the target area into subdomains containing the reference medium (coarse grid) and reservoir (fine grid). Computations for each subdomain can be carried out in parallel. The data exchange between each subdomain within a group is done using MPI through nonblocking iSend/iReceive commands. The data exchange between the two groups is done simultaneously by coupling the coarse and fine grids.

Resource awarded: 32 million core hours on Hermit @ GCS@HLRS, Germany

Towards an improved cell based mesoscopic model for large scale blood flow simulations

Project leader: Federico Toschi, Eindhoven University of Technology, The Netherlands

Collaborators: Stefan Frijters, Eindhoven University of Technology,The Netherlands; Jens Harting, Eindhoven University of Technology, The Netherlands; Dennis Hessling, Eindhoven University of Technology,The Netherlands; Florian Janoschek, Eindhoven University of Technology,The Netherlands; Badr Kaoui, Eindhoven University of Technology,The Netherlands; Prasad Perlekar, Eindhoven University of Technology,The Netherlands;

Abstract: Simulation of human blood flow is a multi-scale problem: in firstapproximation, blood may be treated as a dense suspension ofhighly deformable red blood cells (RBCs) in plasma. Thisparticulate nature requires to resolve individual cells. On theother hand, in realistic vessel geometries typical length scalesvary over several orders of magnitude. Most current models eithertreat blood as a homogeneous fluid and neglect particulateeffects or describe relatively few RBCs with high resolution butfail to reach relevant time and length scales. We developed ahighly efficient particulate model that fills the gap betweenthese extremes and allows to simulate millions of cells whilefully resolving the plasma flow field. The model utilizes alattice Boltzmann solver for the plasma and describes the RBCs assoft ellipsoids interacting via empirical intercellularpotentials. The method has been successfully applied to reproduceimportant rheological features of blood flow. In a previous PRACEproject, a fully-deformable RBC model based on the immersedboundary method was added to our code. It can be applied at equalor higher spatial resolutions than the coarse-grained model ifconvergence is to be studied. Having both models available in thesame massively parallel code offers a unique chance for aone-to-one comparison of different levels of coarse graining. Wewill use the opportunity now to assess the range of validity ofthe coarse-grained model and fine-tune it where necessary.

Further, we will enhance the way our coarse grained modelaccounts for tank-treading and cell deformation in closecomparison with the results of the fully deformable model. Thenew method will be completely in line with our idea of describingeach of the microscopic effects observed in hemodynamics in aminimal but still cell-based way and additionally will enable usto account for the full dissipative coupling of nearby cells dueto lubrication interactions which play an important role for thediffusive motion present in flowing blood. With this enhancedcoarse-grained model we will quantify the local RBC and plasmadiffusivity in a set of large-scale geometries representative formacroscopic blood flows. Together with our collaborators from theproject “Blood in Motion” at TU Eindhoven, we will analyse theresults. If we succeed in developing a predictive model fordiffusive transport based only on the shear rate, hematocrit, andgeometrical confinement, all of which are available locally in acontinuous Navier-Stokes/Advection-diffusion simulation of bloodflow, this could improve the quality of macroscopic blood modelssignificantly, having a major impact on problems to which suchmodels are applied, for example targeted drug delivery, theoptimization of artificial heart valves, or gauging the risk ofrupture impending from an aneurism in the brain.

Resource awarded: 41.16 million core hours on JUQUEEN @ GCS@Jülich, Germany

H2O@RUNG5 – Enabling the next level of accuracy in first principles simulations at finite temperature: double hybrid DFT and RPA simulation of bulk liquid water.

Project leader: Joost VandeVondele, ETH Zurich, Switzerland

Collaborators: Mauro Del Ben, University of Zurich, Switzerland; Juerg Hutter, University of Zurich, Switzerland; Mandes Schoenherr, University of Zurich, Switzerland; Iain Bethune, University of Edinburgh, United Kingdom;

Abstract: Predictivity and hence usefulness of simulation strongly depends on the quality of the underlying model. In first principles simulation, the model of the electronic structure is crucial to obtain a reliable description of the interactions between atoms and molecules in a wide range of systems ranging e.g. from simple liquids, catalysts, chromophores, enzymes, solids to interfaces and nanoparticles. With accurate models, insight in the chemistry and physics of materials can be gained, and new and improved molecules can be designed. This impacts our daily lives, for example through improved drugs, cheaper polymers, cleaner cars, better phone batteries, or brighter displays.

The vast majority of first principles simulations, in which intermolecular interactions must be computed thousands of times, are performed based on simple models of electron correlation, the successful GGA approximation to density functional theory that provides fair accuracy at low computational cost. However, this approximation has notorious failures when it comes to describing for example weak interactions (dispersion) or the localization of electronic states (the self-interaction problem). This impacts the quality of predictions made with these models. Recently, two significantly improved models of electron correlation have been proposed, together with an recipe for their efficient computation. These models are known as double hybrid functionals and the random phase approximation, and represent a fifth rung on Jakob’s ladder towards chemical accuracy. Using Laplace transforms or frequency integration together with resolution of identity (RI) techniques, both models can now be computed with a favorable O(N**4) scaling and on parallel architectures. The prefactor is such that petascale resources and a highly efficient implementation are needed to enable first principles simulations with these models. We have recently finished such a highly efficient and scalable implementation, and are now ready to demonstrate and establish first principles simulations on this level. Through this proposal we seek to obtain access to the necessary petascale resources. Our simulation program, CP2K, is freely provided and widely used be researchers in Europe and world-wide. This enhancing the impact of our research and multiplies the benefits of these developments.

The system we
have selected for demonstration is deceptively simple: bulk liquid water. Surprisingly, the structure of the liquid is still highly controversial, both experimentally and theoretically. For example, we recently demonstrated that simple GGAs underestimate the density significantly (800 g/l) and that dispersion corrections are essential to obtain a proper density. Recent experiments have challenged the view of tetrahedral coordination. With the current proposal, we aim to provide the first simulation results at the 5th rung of DFT, which might confirm or refute the traditional view and earlier simulations. The simulation challenge is significant. This is related to the delicate nature of hydrogen bonding and the fact that ambient temperature and pressure is close to the boiling and freezing point of water. Only with a high quality water description can more complex aqueous systems (e.g. proteins or electrochemical cells) be attempted. The proposed simulations will thus provide a valuable reference for future work.

Resource awarded: 40 million core hours on Hermit @ GCS@HLRS, Germany

Unlocking the role of lipids in the activation mechanism of the EpidermalGrowth Factor Receptor (LIPIDS-EGFR)

Project leader: Ilpo Vattulainen, Tampere University of Technology, Finland

Collaborators: Karol Kaszuba, Tampere University of Technology, Finland; Adam Orlowski, Tampere University of Technology, Finland; Tomasz Rog, Tampere University of Technology, Finland;

Abstract: Epidermal growth factor receptor (EGFR) is a membrane glycoprotein composed of an extracellular ligand binding domain, a single helical transmembrane segment, and an intracellular domain with kinase activity. It is considered as one of the most important membrane receptors, since a major fraction of drug development is targeted to EGFR, with an objective to alter its activity. This is largely due to the fact that EGFR-mediated signaling pathways regulate, e.g., cell proliferation and differentiation, which implies that uncontrolled activation of EGFR is often linked to emergence of diseases such as breast and lung tumors. In essence, EGFR is one of the important targets for cancer therapies.

Yet, despite about 40000 articles published about EGFR, the understanding of how it is activated and stimulated is still quite limited. There is clearly a need for new ideas to unravel how the function of EGFR is modulated by its environment.

In this project, we approach this issue from a new perspective using very recent findings that suggest the role of lipids to be important in regulating EGFR activity. For instance, depletion of cholesterol from plasma membranes has been shown to lead to hyperactivation of EGFR, whereas increasing cellular GM3 levels have been highlighted to inhibit its action. The recent biochemical observations raise an intriguing question as to how the conformational changes in EGFR are induced by the lipid environment surrounding the receptor. As unraveling this issue is very difficult through experiments, here we resort to extensive atomistic molecular dynamics simulations. Using this approach we consider the behavior of EGFR in different lipid environments that are chosen to match the compositions used in previous and on-going biochemical studies. The research will be carried out in close collaboration with experimental partners, the focus being on the effect of GM3 on EGFR activation, and the objective being to unravel the inhibitory mechanism induced by GM3.

Resource awarded: 60.036 million core hours on Hermit @ GCS@HLRS, Germany

Ultra-Relativistic Beam Plasma Interactions: From Miniaturized Plasma Based Accelerators to Extreme Astrophysical Conditions

Project leader: Jorge Vieira, Instituto Superior Tecnico, Portugal

Collaborators: Eduardo Alves, Instituto Superior Tecnico, Portugal; Ligia Amorim, Instituto Superior Tecnico, Portugal; Ricardo Fonseca, Instituto Superior Tecnico, Portugal; Joana Martins, Instituto Superior Tecnico, Portugal; Luis Silva, Instituto Superior Tecnico, Portugal; Anne Stockem, Instituto Superior Tecnico, Portugal; Marija Vranic, Instituto Superior Tecnico, Portugal; Warren Mori, University of California Los Angeles, United States;

Abstract: Can plasma based accelerators be scaled to the energy frontier and be used for high-energy physics experiments? What are the physical mechanisms at the origin of the formation of collisionless shocks and cosmic ray acceleration? These are prominent scientific questions where ultra-relativistic beam plasma interactions could play a decisive role. This proposal aims to exploit the unique computing facilities provided by PRACE and to address these exciting challenges resorting to high fidelity, fully relativistic, and massively parallel particle-in-cell plasma simulations.

The main scientific route to address some these outstanding challenges consists in exploring in the laboratory beam-plasma instabilities of ultra relativistic beams and mimicking extreme astrophysical conditions resorting to complex and large acceleration devices.

Novel and exciting research possibilities, with potential to provide miniaturized plasma-based accelerators capable to accelerate particles beyond the energy frontier, and to recreate extreme astrophysical conditions, could be explored with available ultra-relativistic hadron and lepton bunches at several laboratories (e.g. CERN, SLAC, DESY, RAL). Upcoming experiments using ultra-relativistic protons, electrons and positron bunches at CERN and SLAC are a first step to go beyond the energy frontier using plasma accelerators. This proposal aims to address several outstanding challenges associated with the competition between relevant beam plasma instabilities in this configuration. The work associated with this research program may lead to optimal acceleration regimes for upcoming experiments.

Another key goal of this research is to design first-principle simulations of extreme astrophysical conditions using currently available ultra-relativistic hadron and lepton (proton, electron, and positron) bunches in the laboratory. These simulations may enable fundamental research on large-scale magnetic field amplification mechanisms through Weibel and Kelvin-Helmoltz instabilities in the laboratory.

The relevant temporal and spatial scales associated with these scenarios are very disparate. In addition, the relevant microphysics that rules plasma and ultra-relativistic beam particle trajectories is highly non-linear, and the associated collective phenomena very complex. Thus, full-scale, massively parallel particle-in-cell simulations are critical to address the challenges outlined in this proposal.

This research proposal will take advantage of the available outstanding numerical and visualization frameworks in our group, and from the use of advanced algorithms that significantly reduce computational requirements, in order to model extreme beam-plasma interaction scenarios. The unique computational infrastructures provided by PRACE will be critical to explore some of the most exciting fundamental physics questions at the forefront of science identified in this proposal.

Resource awarded: 45 million core hours on SuperMUC @ GCS@LRZ, Germany

Modeling large scale protein conformat
ional change and folding

Project leader: Wolfgang Wenzel, Karlsruhe Institute of Technology, Germany

Collaborators: Zoe Cournia, Bioacademy GR, Greece;

Abstract: Proteins are the ubiquitous nanoscale machines in all cellular life and perform a multitude of functions. Protein function and dysfunction are frequently related to diseased and models of protein structure and functional analysis are important to understand biological function and to develop novel therapeutic strategies. For this reason modeling large scale conformational change in biomolecules, including protein folding, has been an eminent challenge to the simulation sciences for many decades]. Experimental methods still yield mostly snapshots of interesting biological molecules, while simulation methods promise to elucidate the entire process with atomic resolution. The increasing focus of life science research towards molecular mechanisms, including computer aided drug discovery, generates an overwhelming number of simulation challenge. These include protein folding and misfolding, modeling protein function, understanding protein-protein and protein-ligand associations and modeling transcriptional and translational control. In this project we will investigate protein folding and conformational change to better understand protein function. Recently, all atom MD simulations of the Shaw group have demonstrated reversible folding of several small proteins using proprietary specialized hardware. Presently PRACE HPC resources cannot be used to perform such simulations. Here we use a novel massively parallel Monte-Carlo based simulation approach that enables such simulations on ‘standard’ HPC architectures. In the first part we will use this approach to completely characterize folding equilibriums of small proteins and their stability under mutation. In a second part of the project we will focus on the conformational equilibriums of large proteins, which alternate between two (kinases) or many (ubiquitin) conformational substates for their biological function. Due to computational limitations, these processes have never been observed in unbiased atomistic simulations using transferable biophysical forcefields. Ubiquitin is important in regulating cellular behaviour, specifically protein degradation, while kinases constitute 30% of all targets for the presently available drugs, this investigation will help understand and control biological function, also in regard to pharmaceutical applications. This project will pioneer a novel methodology to exploit the investment in European HPC infrastructures for all-atom simulations of protein structure and function and pave the way to elucidate complex biological phenomena.

Resource awarded: 8 million core hours on Hermit @ GCS@HLRS, Germany

Multiscale modelling of electronic processes in thin organic films

Project leader: Wolfgang Wenzel, Karlsruhe Institute of Technology, Germany

Collaborators: David Beljonne, University of Mons, Belgium; Thierry Deutsch, CEA, France; Ilian Todorov, STFC, United Kingdom;

Abstract: In the EU-funded FP7 ICT e-infrastructures project MMM@HPC we develop a novel, work-flow based approach to develop multiscale materials modeling methods (www.multiscale-modelling.eu). The application area pursued in this project will be the simulation of charge transport through organic materials. Organic materials are suited to optoelectronics applica-tions: the 3rd generation solar cells (organic photovoltaics, OPV, and solid-state dye-sensitized solar cells, SS-DSCs, that used doped organic materials as hole transporters) and organic light emitting devices, OLEDs for displays and lighting. In many applications the cells are stacked to absorb more of the solar spectrum (OPV, SS-DSC) or emit red, green, blue creating white light (OLED). Simulation workflows in MMM@HPC are realized on the basis of GRIDBEANS, which integrate into the UNICORE workflow system, which is implementing the workflow concept as a container of integrated services. In this project we will address three specific challenges: Modelling electron and hole transport in amorphous materials for small molecule OLED, Guest-Host Systems and Processes at organic-organic interfaces which are crucial for computer aided materials design.

Resource awarded: 14.7 million core hours on FERMI @ CINECA, Italy

Molecular Dynamics Simulation of Protein-Protein Complex Formation in a Crowed Environment

Project leader: Martin Zacharias, Technical University Munich, Germany

Abstract: The process of protein-protein complex formation is of fundamental importance for a better understanding of a variety of biological processes. In a cellular environment the high concentration of surrounding proteins can influence the association process between proteins. Large scale molecular dynamics (MD) simulations will be used to study protein-protein binding in the presence of explicit solvent molecules and ions. As a model system the dynamics of up to 50 copies of the protein colicin E9 and 50 copies of the inhibitor Immunity protein 9 (Im9) will be studied during several 100 ns simulation time at high protein concentration which are feasible using Tier-0-HPC resources. The mechanism of protein-protein association at specific and non-specific sites will be studied at atomic resolution. The large-scale simulations will give fundamental new insight into the molecular mechanism of complex formation, the role of surrounding water molecules and the influence of a crowed (high-protein concentration) environment similar to the in vivo situation inside a biological cell.

Resource awarded: 27 million core hours on SuperMUC @ GCS@LRZ, Germany