Find below the results of the PRACE 4th Regular Call (in alphabetical order of the name of the project leader).

Direct numerical simulation of reaction fronts in partially premixed charge compression ignition combustion: structures, dynamics

Project leader: Xue-Song Bai, Lund University, Sweden

Collaborators: Yu Rixin, Lund University, Sweden/ Henning Carlsson, Lund University, Sweden/ Fan Zhang, Lund University, Sweden/ Rickard Solsjo, Lund University, Sweden.


Recent public concerns on global warming due to emissions of the green house gas CO2, as well as emission of pollutants (soot, NOx, CO, and unburned hydrocarbons) from fossil fuel combustion, have called for development of improved internal combustion (IC) engines that have high engine efficiency, low emissions of pollutants, and friendly to carbon neutral renewable fuels (e.g. biofuels). The European and world engine industry and research community have spent great effort in developing clean combustion engines using the concept of fuel-lean mixture and low temperature combustion which offers great potential in reducing NOx (due to low temperature) and soot and unburned hydrocarbon (due to excessive air), and meanwhile achieving high engine efficiency. One example is the well-known homogeneous charge compression ignition (HCCI) combustion engine, which operates with excessive air in the cylinder, and produces simultaneously low soot and NOx. However, HCCI combustion is found to be very sensitive to the flow and mixture conditions prior to the onset of auto-ignition. As a result, HCCI engine is rather difficult to control. At high load (with high temperature and high pressure) engine knock may occur with pressure waves in the cylinder interacting with the reaction fronts, causing excessive noise and even mechanical damage. At low load (with lower temperature and pressure) high level emissions of CO and unburned hydrocarbon may occur, which lowers the fuel economy and pollutes the environment. Recently, it has been demonstrated experimentally that with partially premixed charge compression ignition (PCCI) which can be attained by using multiple injections of fuel at different piston positions, smoother combustion can be achieved by managing the local fuel/air ratio (thereby the ignition delay time) in an overall lean charge.

There are several technical barriers in applying the PCCI concept to practical engines running with overall fuel-lean mixture, low temperature combustion. For example, it is not known what the optimized partially premixed charge is for a desirable ignition, while at the same time maintaining low emissions. The main difficulty lies in the non-linear behavior of the dominating phenomena and the interaction among them (e.g. chemistry and turbulence). To develop an applicable PCCI technology for IC-engine industry, improved understanding of the multiple scale physical and chemical process is necessary. Further, there is a need to develop computational models for simulating the process for the design where a large number of control parameters are to be investigated.

The goals of this project are to achieve improved understanding of the physical and chemical processes in overall fuel-lean PCCI processes, and to generate reliable database for validating simulation models for analysis of the class of combustion problems. This shall lead to development of new strategies to achieve controllable low temperature combustion IC-engines, while maintaining high efficiency and low levels of emissions (soot, NOx, CO and unburned hydrocarbons). Direct numerical simulation (DNS) approach that employs detailed chemistry and transport properties will be employed.

Resource awarded: 20.000.000 core hours on CURIE TN (GENCI@CEA, France)

How strange is the nucleon?

Project leader: Gunnar Bali, Universitaet Regensburg – Fakultaet fuer Physik, Germany

Collaborators: Sara Collins, Universitaet Regensburg, Germany/ Vladimir Braun, Universitaet Regensburg, Germany/ Benjamin Glaessle, Universitaet Regensburg, Germany/ Meinulf Goeckeler, Universitaet Regensburg, Germany/ Johannes Najjar, Universitaet Regensburg, Germany/ Paula Perez-Rubio, Universitaet Regensburg, Germany/ Rainer Schiel, Universitaet Regensburg, Germany/ Andre Sternbeck, Universitaet Regensburg, Germany/ Wolfgang Soeldner, Universitaet Regensburg, Germany/ Dirk Pleiter, Forschungszentrum Jülich, Germany/ James Zanotti, University of Adelaide, Australia.

Abstract:The mass of the visible universe is almost exclusively due to nucleons (protons and neutrons). These are composed of quarks and gluons whose interactions are governed by quantum chromodynamics (QCD). Yet not much is known about the individual quark and gluon contributions to the energy-momentum or spin of the proton. These are the lowest moments of so-called unpolarized and polarized parton distribution functions (PDFs). Due to quantum fluctuations, in addition to its up and down valence quarks, the proton contains gluons and sea quarks, including strange quarks. A lot of effort goes into extracting the associated PDFs from global fits to experimental data. The sea quark and gluonic PDFs are a dominant uncertainty in the analysis of the ongoing proton-proton collider experiment LHC at CERN since these are hard to measure experimentally. The main results of this project will be the strangeness contributions to the spin and energy-momentum of the proton, by numerical simulation of QCD on a lattice.

Resource awarded: 21.800.000 core hours on FERMI (CINECA, Italy)

Simulations of global accretion discs: turbulent transport and dynamo action

Project leader: Gianluigi Bodo, INAF – Osservatorio Astronomico di Torino, Italy

Collaborators: Andrea Mignone, Universita di Torino, Iyaly/ Paola Rossi, INAF, Italy/ Claudio Zanni, INAF, Italy.


The process of accretion is of great importance in astrophysics. Many compact objects like planets, stars and massive black holes are formed by the infall of material into a central gravitational well. The release of gravitational energy by the accreting material can be extremely efficient and can power some of the most energetic phenomena in the universe. However many aspects of the accretion process are not yet completely understood.

The accreting material, in general, forms a disc around the central object and, in order for accretion to occur, angular momentum has to be removed from the accreting material and transported outward. There is a general consensus that angular momentum transport is mainly due to turbulence driven by the magnetorotational instability (MRI). This instability requires the combined action of rotation, shear and magnetic field, whose presence is essential, because otherwise the disc would be stable. Turbulent motions must also be able to regenerate through a dynamo process the very magnetic field that is necessary for the MRI to develop. Up to now numerical simulations have been restricted to a local patch of the disk or, in the case of global simulations, to low resolution. The results obtained so f
ar show that it is fundamental for understanding the dynamo and turbulent transport processes to consider both global simulations and high resolution. Only the present availability of computing petascale resources is making now possible to achieve both these aims. We propose here to perform a series of simulations at increasing resolution and with different magnetic field configurations that will allow us to understand in more detail how accretion discs work.

Resource awarded: 10.200.000 core hours on FERMI (CINECA, Italy) and 18.000.000 core hours on JUGENE (GCS@Jülich, Germany)

Star formation in extreme conditions: Shock heating, thermal cooling and turbulent dissipation in Stephan’s Quintet

Project leader: Frederic Bournaud, CEA Saclay – DSM/IRFU, France

Collaborators: Florent Renaud, CEA Saclay, France/ Jared Gabor, CEA Saclay, France/ Pierre-Alain Duc, CEA Saclay, France/ Stephanie Juneau, CEA Saclay, France/ Sebastien Fromang, CEA Saclay, France/ Marc Labadens, CEA Saclay, France/ Katarina Kraljic, CEA Saclay, France/ Romain Teyssier, University of Zurich, Switzerland/ Markus Wetzstein, University of Zurich, Switzerland/ Andreas Bleuler, University of Zurich, Switzerland/ Pierre Guillard, CalTech and IAS Orsay, United States of America/ Phil Appleton, CalTech, United States of America; Eric Emsellem, ESO European Southern Observatory, Germany/ François Boulanger, IAS Institut d’Astrophysique Spatiale, France/ Curt Struck, Iowa State University, United States of America.


Understanding galaxy formation is severely limited by our lack of knowledge about star formation (SF). SF is governed by the chaotic parsec-scale physics of the interstellar medium (ISM), and the link between this small-scale physics and SF on galactic scales is still poorly understood. In the densest gas clouds, on scales of a few parsecs, SF is probably a relatively universal process. But the conversion of low-density galaxy-wide gas reservoirs into molecular clouds and dense substructures may be a non-universal process. The transition between SF physics on small scales and the scaling relations for the rate of SF on large scales is hence a highly debated topic of fundamental importance in galaxy formation.

The connection between the actual formation of individual stars on small scales and the properties of SF in entire molecular clouds and entire galaxies is governed by the structure and properties of the Interstellar Medium (ISM). Numerical simulations of galaxies are a major tool to study the properties of the ISM in various conditions. Only recently, simulations became capable to accurately describe the global structure and dynamics of galaxies and resolve supersonic turbulence in the cold phases of the ISM at the same time. This typically requires a very high resolution (around one parsec) in boxes of tens of kiloparsecs: adaptive-resolution techniques such as AMR hydro-codes are a powerful tool for this.

To date, the properties of ISM turbulence have been explored in details in isolated galaxies, like our Milky Way – our own PRACE project is enabling major progress in this direction, by developing a new predictive model of the interaction between newly formed stars and the surrounding gas (the so-called « feedback » processes) and using it at ultra high resolution (up to 0.1pc) in a global self-consistent model of the Milky Way.

Galaxy formation, however, often proceeds in violent phases during which galaxies accrete mass very rapidly, collide and merge with each other – not just in calm isolation as is the case for the Milky Way. The properties of ISM turbulence and ISM structures are likely different in such systems. Observationally, these systems often undergo strong « bursts » of very rapid star formation. But whether these bursts are only a large-scale property with a universal efficiency of SF on small scales, or whether there could be locally-enhanced of locally-suppressed SF in molecular clouds at some intermediate scale, is still unknown.

Our team has the full knowledge and tools to model ISM physics and SF in such « extreme » events with bursts of star formation. We hence plan to perform simulations of colliding/interacting galaxies with a resolution sufficient to capture the turbulent cascade in the ISM, down to scales of about one parsec and explicitly resolving high gas densities (above 10^5 atoms per cm^3) in which SF can reasonably be supposed to be a locally universal process. A thorough and accurate modeling of stellar feedback, such as the one we have developed, is also required.

We here propose to perform simulation of Stephan’s Quintet at parsec-scale resolution. This is a well-known group of galaxies in mutual interaction/collision. This system is very rich in various types of features that can have different efficiencies of SF (active nuclei, supermassive star clusters, a large-scale shock rich in molecular gas, many tidal tails, etc). Observations indeed suggest that locally-enhanced and locally-suppressed SF are present in this system on intermediate scales, but whether the stabilization of relatively dense gas against SF results from thermal heating in shocks, or from tidally-triggered, slowly-dissipating turbulence, or some other process is now known yet.

With this simulation, we will obtain a new major milestone in understanding the link between star formation and galaxy evolution. In addition to probing the « normal », isolated phases of star formation with our on-going model of a Milky Way-type galaxy, we will now probe the « extreme » modes where the ISM is highly perturbed by tidal interactions, large scale shocks – a process that is best observed and modeled for Nearby galaxies, but may be even more important to understand the early phases of galaxy formation in the distant Universe.

Resource awarded: 8.100.000 core hours on SuperMUC (GCS@LRZ, Germany)

Thousands of trees for 4 billion years of life evolution on Earth.

Project leader: Bastien Boussau, CNRS – Life sciences, France

Collaborators: Gergely Szollosi, CNRS, France/ Vincent Miele, CNRS, France/ Vincent Daubin, CNRS, France/ Eric Tannier, INRIA, France


In the last fifteen years, thousands of genomes have been sequenced from species sampling the entire tree of life. Genomes contain a huge amount of information about how the diversity of life appeared and changed through time, and about how living systems work, from protein functions to ecological communities. However, this information is encrypted. One powerful approach to deciphering this information is to compare the genomes of different species in an evolutionary framework: genomes inform about species evolution, and in return evolution informs about the function of genes. The analysis of genomes requires dedicated algorithms and vast amounts of computation, but methods that are taylored to use all the information included in the genomes are just starting to appear. In particular the reconstruction of gene trees and species trees on a genomic scale is a critical step for evolutionary studies, but is a very complex and costly task. Fortunately, this reconstruction can easily be parallelized. We have developed a program to jointly infer gene trees and species tree by modelling events of gene dupli
cation, loss, and transfer. Simulations and analyses on real genomic data show that our method accurately reconstructs the species tree and gene trees. In addition it can simultaneously infer other key features of the history of the genomes under study, such as ancestral gene contents and speciation times, more accurately than commonly-used methods. This program uses the Message Passing Interface to run on several processors simultaneously. Its scaling properties have been assessed on the supercomputer Jade at the CINES in 2011, and we now plan on using it on 100 to 200 genomes from the three domains of life. This will enable us to reconstruct a species tree based on all the genes present in the genomes under study, and to infer gene trees and events of duplication, transfer and losses for all gene families. In the process, we will obtain precise dates for speciation events billions of years old, at geological times where the lack of interpretable fossils renders commonly used dating methods inaccurate. We will also reconstruct ancestral gene contents over the entire tree of life, thus illuminating metabolism evolution and consequently the history of geochemical conditions on earth for the past 4 billion years. The results of these computations performed at an unprecedented scale will be the core data for a group of leading scientists in Evolution gathered in the context of the French ANR project ANCESTROME (funded from the 2011 call ”Bioinformatique”, ”Investissements d’avenir”). We will study the evolution of genome structure, metabolism, ecological communities, and we will create user-friendly databases to make the result of our computations available to the international scientific community.

Resource awarded: 5.000.000 core hours on CURIE TN (GENCI@CEA, France)

NadiaSpectral DNS of boundary layer transition in channel flows

Project leader: Marc Buffat, Université Lyon 1 – LMFA, France

Collaborators: Julien Montagnier, Université Lyon 1, France/ Anne Cadiou, CNRS, France/ Lionel Le Penven, CNRS, France


Understanding the mechanisms involved in the turbulent transition of boundary layers is crucial for many engineering domains. They are mainly studied by analytical theories supplemented by direct numerical simulations (DNS) of the entire flow dynamics. Such simulations require high-order numerical approximations such as the spectral methods that have been successfully used for fundamental studies on periodic channel flows and free-stream boundary layers since the 70’s.This project focuses on the turbulent transition of the boundary layers developing at the entrance of a plane channel. Many aspects of these flows are not yet fully understood, such as their stability characteristics, their progressive stages towards a well-developed turbulence and the reasons why the turbulence is self-sustaining as it convects downstream. For high enough Reynolds numbers, and if upstream conditions are such that the velocity profile is uniform at the channel entrance, turbulent transition is expected to take place in the boundary layers developing on each wall, and relatively near the entrance section. The transition is essentially similar in that case to what happens to a boundary layer under an accelerated free-stream. At lower Reynolds numbers, one can imagine that the interaction of the boundary layers plays a more complex role on the nature and location of the transition, but this remains to be studied. Since geometries of interest are wide and very elongated, the simulations involve billions of modes, thereby needing massively parallel platforms. In our team, a spectral Galerkin method based on an orthogonal decomposition of the divergence-free velocity field into two orthogonal solenoidal velocity fields has been developed. It has been implemented in an efficient numerical code, NadiaSpectral ( buffa…), dedicated to the spectral simulations of turbulent flows between two parallel plates, with a very high parallel scalability. The new peta-scale DNS that will be achieved in this project will give new insight into the first stages of transition up to the fully developed turbulent state as a function of the Reynolds number. Such DNS data are also vital in order to develop and validate engineering models. The DNS data will therefore be made available to the wider scientific community.

Resource awarded: 10.000.000 core hours on JUGENE (GCS@Jülich, Germany) and 10.200.000 core hours on SuperMUC (GCS@LRZ, Germany)


Project leader: Michel Caffarel, UMR 5626 CNRS-Univers. Paul Sabatier – Lab. Chimie et Physique Quantiques, France

Collaborators: Anthony Scemama, UMR 5626 CNRS-Univers. Paul Sabatier – France/ Angélique Pagès, UMR 5626 CNRS-Univers. Paul Sabatier, France/ Emmanuel Giner, UMR 5626 CNRS-Univers. Paul Sabatier, France/ William Jalby, Université de Versailles Saint-Quentin-en-Yvelines, France/ Emmanuel Oseret, Université de Versailles Saint-Quentin-en-Yvelines, France/ Peter Faller, UPR CNRS 82421, France/ Christelle Hureau, UPR CNRS 82421, France/ Giovanni La Penna, National Research Council (CNR), Italy


In this project we propose to realize large-scale quantum Monte Carlo (QMC) simulations to shed light on some key issues of the chemistry of Alzheimer’s disease (AD). QMC is a powerful approach for solving the Schrödinger equation using random walks. One of the key aspects making QMC approaches very attractive is their ideal adaptation to massively parallel computers and High Performance Computing (HPC). The application proposed here concerns the Alzheimer’s disease (AD) which is the most prevalent cause of dementia in the elderly population affecting more than 20 million people world-wide. The amyloid-beta (Abeta) peptide seems to play a causative role in AD. Abeta is the major constituent of amyloid plaques, a hallmark of AD. A large body of evidence suggests that metallic ions (copper, zinc, and iron) could play a role in the etiology of AD through their influence on this aggregation mechanism and in the case of copper and iron in the production of reactive oxygen species (ROS). As an important consequence understanding and controlling Abeta/metallic ions interactions could lead to promising therapeutic strategies.

In this project we propose to study the energetics implied in the formation of complexes involving metallic ions and Abeta peptide using large-scale Fixed-Node Diffusion Monte Carlo simulations. More precisely, we propose to look at the difference of Cu-Abeta binding in the case of the human and the murine (mouse and rat). Indeed, quite remarkably the murine, whose peptide differs from the human Abeta peptide only by three point mutations, does not show amyloid deposition and does not develop the Alzheimer’s disease. To understand these remarkable facts and thus to eventually propose new therapeutic strategies we propose to study the distinct Cu(II) coordination to the human/murine Abeta and possibly to shed some light on the origin of the mechanisms inhibiting or not the aggregation.

Last year, t
hanks to a PRACE preparatory access we have been able to check the optimal parallel efficiency of our QMC=Chem code up to 10 000 cores on the Curie machine (TGCC, France). Very recently, using the new configuration of Curie this regime has been extended further to its 80 000 compute cores for a first scientific application concerning the Abeta peptides involved in the Alzheimer’s disease.

From a more general perspective, such a project is expected to illustrate the extremely favorable computational aspects of QMC algorithms and the new exciting scientific applications which can be envisioned in the near future on massively parallel platforms running at the petascale level and beyond (exascale horizon).

Resource awarded: 8.809.500 core hours on CURIE FN (GENCI@CEA, France) and 16.913.150 core hours on CURIE TN (GENCI@CEA, France)

Vertical axis wind turbines for future large offshore farms

Project leader: Philippe Chatelain, Université catholique de Louvain – Institute of Mechanics, Materials and Civil Engineering, Belgium

Collaborators: Matthieu Duponcheel, Université catholique de Louvain, Belgium/ Stéphane Backaert, Université catholique de Louvain, Belgium/ Grégoire Winckelmans, Université catholique de Louvain, Belgium/ Stefan Kern, GE Global Research, Germany/ Dominic Von Terzi, GE Global Research, Germany/ Thierry Maeder, GE Global Research, Germany/ Petros Koumoutsakos, ETH Zurich, Switzerland/ Wim van Rees, ETH Zurich, Switzerland/ Alessandro Curioni, IBM Research Division – Zurich Research Laboratory, Switzerland.

Abstract:Wind energy is expected to play a key role in the world’s future energy mix. In that context, the offshore migration of wind turbines is inevitable, given the superior wind resources and much reduced acoustic and visual disturbances. In parallel, economy of scales will cause wind turbines to grow much larger, e.g. up to 20MW. It is recognized today that a simple upscaling of existing designs will not be the answer, hence the interest in a promising alternative, the Vertical Axis Wind Turbines (VAWT). VAWTs can be aerodynamically as efficient as the classical horizontal axis wind turbines (HAWT) and there are indications that they can be significantly more densely packed in farms because of the structure of their wakes.

The wake of a wind turbine is the flow region downstream of the turbine which is characterized by turbulence and lower wind energy. It can extend over large distances and greatly deteriorate the performance of any downstream wind turbine. The region closest to the rotor, the near wake, is dominated by powerful coherent flow structures, i.e. vortices generated by the rotor blades. Further downstream, in the so-called far wake, the velocity deficit has been spread over a greater area and individual large-scale vortices cannot be easily identified. The dynamics of the whole wake, in particular the rate at which it decays, thus impose the separations between wind turbines, and therefore the energy extraction capability of a farm on a given geographical concession.

This project aims at investigating the physics of the wakes of VAWTs at two levels:

In a first step, we study the fundamental physics of the wake of a single wind turbine. The VAWT wake morphology should differ vastly from the classical swirling wake of a horizontal axis wind turbine. It is indeed expected to lose the axisymmetric character of a HAWT wake, and rather exhibit unsteadiness and a flux of transversal linear momentum, just as an aircraft wake. How and how fast the flow relaxes to such a canonical flow are questions that the present project aims to answer.

The second part of this work considers multiple wind turbines, either as an isolated cluster or as a couple of rows in a wind farm. Simulating wake effects on downstream turbines will allow determining optimal wind turbine placement and verifying the above claim of a denser packing in VAWT wind farms.

The contributions of this work are far-reaching. At a fundamental level, the outcome is a better understanding of VAWT wakes and interactions. At a higher level, this could pave the way to large-scale, highly efficient offshore wind farms, required for a paradigm change in offshore energy generation.Resource awarded: 35.000.000 core hours on JUGENE (GCS@Jülich, Germany)

— –

All-atom simulation of the Amyloid-beta peptide interacting with Gold nanoparticles (AmyGo)

Project leader: Stefano Corni, CNR Istituto Nanoscienze – Centro S3, Italy

Collaborators: Luca Bellucci, CNR Istituto Nanoscienze, Italy/ Rosa Di Felice ,CNR Istituto Nanoscienze, Italy/ Giovanni Bussi, Scuola Internazionale Superiore di Studi Avanzati (SISSA), Italy.


The increasing technological importance of nanomaterials naturally raises the concern for possible toxic effects when they accidentally contact living organisms. Such effects will likely involve the interactions of nanomaterials with the protein arsenal of the body. On the other hands, nanoparticles have been proposed as the basis for innovative diagnostic and therapeutic approaches, applications which define the emerging field of theranostic (therapy + diagnostic) nanomedicine. Recent experimental works tackled the effect, in vitro, of nanoparticles on protein fibrillation. Protein fibrillation is involved in many human diseases, such as Alzheimer’s, Parkinson’s, Creutzfeld-Jacob’s, and dialysis-related amyloidosis. The nanoparticles may enhance or inhibit the rate of formation of fibrils, therefore they can potentially lead to novel mechanisms for amyloid diseases as well as to therapeutic opportunities for their treatment. Although these results are of broad relevance, how the nanoparticles interfere with the fibrillation process at the atomistic level is still poorly understood.

The main goal of this project is to unravel microscopic mechanisms that modify the fibrillation propensity of an amyloidogenic peptide when contacting inorganic nanoparticles. To this end, we shall perform all-atom classical Molecular Dynamics simulations in explicit solvent exploiting the highly scalable Hamiltonian-Temperature Replica Exchange enhanced sampling technique. We will focus on an important prototypical system: the Amyloid beta 42 (Abeta42) peptide on gold nanoparticles. The Abeta42 is a peptide of 42 amino acids, whose inclination to aggregation triggers the formation of fibrills and neurotoxic plaques observed in the Alzheimer’s disease. Gold nanoparticles are among the most versatile and easy-to-use ones, and are currently being tested as imaging contrast agents, absorptive heating systems and as dual imaging and therapeutic agents. Our simulations will clarify how and why the interaction with gold modifies the conformational ensemble of Abeta42 and therefore will provide microscopic insights on the nanoparticle role in affecting its fibrillation propensity.Resource awarded: 8.750.000 core hours on SuperMUC (GCS@LRZ, Germany)

Characterization of the bacterial membrane and its interaction with antimicrobial peptides

Project leader: Matteo Dal Peraro, Ecole Polytechnique Federale de Lausanne, EPFL – Institute
of Bioengineering, School of Life Sciences, Switzerland.

Collaborators: Thomas Lemmin, Ecole Polytechnique Federale de Lausanne, EPFL, Switzerland/ Christophe Bovigny, Ecole Polytechnique Federale de Lausanne, EPFL, Switzerland/ Marco Stenta, Ecole Polytechnique Federale de Lausanne, EPFL, Switzerland


Bacterial infections are one of the most frequent diseases in humans, representing the second-leading cause of death worldwide. Antibacterial resistance constitutes nowadays a major concern for human health due to its social and economical implications. One of the main avenues for devising new strategies to overcome bacterial defenses without promoting resistance relies on a deeper understanding of the biophysical and biochemical properties of the bacterial membrane. This knowledge could be used to develop innovative therapies based on a synergic use of drugs inhibiting the activity of intracellular components, and substances increasing drug permeability of bacterial membrane. Great effort has therefore been invested to overcome resistance or to find possible alternative antibiotics. Antimicrobial peptides (AMPs), which are produced by many tissues and cell types as part of their innate immune system, have attracted during the last years considerable interest as possible drug candidates acting as templates for the development of improved remedies. Despite the relevance of their mechanism of action, the fluid nature of the membrane coupled with the current resolution limitations of experimental techniques still prevents a detailed molecular understanding of AMP activity. For this reason, atomistic molecular simulations can play a pivotal role for unveiling their mechanism and can have a large impact to progress towards the development of innovative antibiotic remedies.

We propose to perform a comprehensive set of large-scale simulations aimed at investigating for the first time the biochemical and biophysical properties of a realistic model of the bacterial membrane. Building on established molecular mechanics models of phospholipids along with recent models developed and tested in our lab of key anionic constituents of the bacterial membrane, i.e. cardiolipins, extended molecular simulations will be used to fully characterize the biophysical properties of the bacterial membrane and to dissect its interactions with natural and artificial cationic AMPs. To this purpose we will mainly use molecular dynamics (MD) sampling techniques within the molecular mechanics framework and will investigate model systems with different concentration of phospholipids, cations and AMPs. Importantly, this series of molecular simulations will be designed to consistently complement high-speed atomic force microscopy experiments on similar membrane templates in complex with the same AMPs. Thus, the interplay of state-of-the-art MD and AFM techniques will contribute to compensate their reciprocal spatial and temporal resolution limitations and will provide precious and unique insights into the molecular understanding of the structural and kinetic properties of the bacterial membrane and its response to environmental conditions and AMPs attack. These studies are ultimately intended to pave the way to the rational design of novel antibiotics mimicking the AMP mechanism of action.

Resource awarded: 45.000.000 core hours on JUGENE (GCS@Jülich, Germany)

Structural determination of complex organic/biological molecular assembly from ab-initio NMR

Project leader: Stefano De Gironcoli, International School for Advanced Studies – SISSA – Condensed Matter Theory Sector, Italy.

Collaborators: Davide Ceresoli, CNR, Italy/ Emine Kucukbenli, International School for Advanced Studies – SISSA, Italy/ Riccardo Sabatini, International School for Advanced Studies – SISSA, Italy/ Ngoc Linh Nguyen, International School for Advanced Studies – SISSA, Italy/ Nicola Varini, Irish Centre for High-End Computing (ICHEC), Ireland.


Solid State NMR (SSNMR) is a powerful tool to investigate the structure of complex organic systems of biological or medical interest. It is not destructive, does not require crystalline long range order and it is sensitive to local structural and chemical details. Interpretation of spectra is however a complex matter and ab initio theoretical calculations of NMR spectra play an important role in NMR assisted structural determination of complex systems.

Ubiquitous in organic systems, van der Waals (vdW) interactions have long eluded proper treatment by first principle methods. Recently a new class of exchange and correlation functionals in DFT has emerged that can account for dispersion forces at a fundamental level, avoiding semiempirical fitting or ad hoc corrections. The advent of these new theoretical tools has significantly enhanced the reliability of DFT calculations applied to organic systems.

The combination of highly efficient NMR calculations via GIPAW method with the availability of this new class of functionals in the Quantum-ESPRESSO suite of codes offers an unprecedented tool for the study of complex organic systems of great fundamental and medical interest.

In this project we apply this general strategy to i) the search for candidate cholesterol crystal structures associated to different gallbladder diseases by comparing experimental spectra with first principles NMR characterization, and ii) the ab-initio NMR study of proposed structural models of Abeta_40 amyloid fibril, which constitute the senile plaques, found in several neurodegenerative diseases such as Alzheimer and Parkinson disease.

Resource awarded: 13.000.000 core hours on CURIE TN (GENCI@CEA, France)

Massively-parallel molecular simulation studies of nano-scale crystal formation

Project leader: Niall English, UCD – Chemical & Bioprocess Engineering, Ireland

Collaborators: Gilles Civario, Irish Centre for High End Computing, Ireland/ John Tse, University of Saskatchewan, Canada.


Massively-parallel molecular simulation of the formation of nano-scale crystallites will be performed via non-equilibrium molecular dynamics (NEMD) of liquid-crystal systems to investigate crystal formation mechanisms. A particular, historic challenge for molecular simulation of crystallisation in the past decade has been the relatively small molecular system sizes available with most Beowolf-clusters (hundreds of thousands of atoms), necessarily limiting the realism of simulations. This project is dedicated to larger-scale simulation of millions of atoms in non-equilibrium crystal-liquid systems, to reduce substantially the artefact of periodic boundary conditions upon crystallisation. Representative systems to be studied would include ice-liquid water systems, and mechanisms of crystal growth therein based on seeds of differing geometry and size.

Resource awarded: 35.376.211 core hours on JUGENE (GCS@Jülich, Germany)

The domain interplay in pr
otein kinases: free energy calculations on the allosteric effect of the SH2 domain on kinase activation

Project leader: Francesco L. Gervasio, Spanish National Cancer Research Center (CNIO) – Structural Biology and Biocomputing, Spain

Collaborators: Nicole Dölker, Spanish National Cancer Research Center (CNIO), Spain / Silvia Lovera, Spanish National Cancer Research Center (CNIO), Spain


Protein kinases constitute one of the largest protein families and are involved in most cellular pathways. Correspondingly, kinase malfunction is related to an important number of human diseases. Protein kinases share a common canonical catalytic domain, consisting of a number of highly conserved motifs, which have to undergo distinct conformational changes to perform their catalytic activity. Most kinases also require other domains or proteins for full activation. The mechanism by which those protein-protein interactions activate the catalytic domain is not yet fully understood.

The Abelson tyrosine kinase (Abl) is of special interest because of its importance as an anti-cancer drug target. The fusion protein Bcr-Abl leads to over-activation of Abl, causing chronic myeloid leukemia (CML). Currently, Abl inhibitors are the only pharmaceutical treatment for CML, but a significant proportion of patients develop resistance against currently known inhibitors. As most resistance mutants are located around the active site of the catalytic domain, the development of inhibitors that bind to allosteric sites may open the way towards new therapeutic approaches. A more detailed understanding of allosteric effects of other protein domains on the activation of Abl is, therefore, of paramount importance for the rational design of new drugs.

In this project, we focus on the protein-protein interactions of the catalytic domain with the SH2 domain, an allosteric effector involved both in auto-inhibition and activation. As our preliminary calculations indicate, this effect stems mainly from a modification of important functional modes of the catalytic domain, as well as from the stabilization of certain catalytically relevant structural motifs. In order to obtain quantitative data on the thermodynamics and kinetics of Abl activation in the presence of the SH2 domain, we plan to reconstruct the free energy surface corresponding to those important modes. To this end, we plan to use metadynamics with parallel tempering (PTmetaD). PTmetaD is an extremely efficient and intrinsically parallel method for the calculation of the free energy as a function of one or more collective variables (CVs).

In the second part of the project, the effect of the SH2 domain will be tested by introducing specific mutants, for which we will also calculate the corresponding free energy surfaces. These mutants are simultaneously studied experimentally in the group of Prof. Giulio Superti-Furga at CeMM in Vienna. We aim at identifying hot spots of inter-domain interactions, which could serve as putative binding sites for new allosteric Abl inhibitors.

Finally, we plan to expand our findings obtained for the Abl kinase to another, therapeutically relevant system, the FES kinase. FES, which has recently attracted attention as an important tumor suppressor, is also activated by interaction between the catalytic domain and the SH2 domain.

Resource awarded: 25.000.000 core hours on SuperMUC (GCS@LRZ, Germany)

Numerical simulation of air flow in the human large airways

Project leader: Guillaume Houzeaux, Barcelona Supercomputing Center – Computer Applications in Science and Engineering, Spain.

Collaborators: Denis Doorly, Imperial College, United Kingdom/ Alberto Gambaruto, CEMAT/IST, Portugal/ Raul de la Cruz, Barcelona Supercomputing Center, Spain/ Mariano Vazquez, Barcelona Supercomputing Center, Spain/ Hadrien Calmet, Barcelona Supercomputing Center, Spain.


The human upper airways (nasal cavity and trachea) function not merely as a flow conduit but perform essential air conditioning tasks such as temperature equilibration, humidification, filtration and defence. Moreover the airways incorporate the organs or voice production and olfaction. Detailed computational modelling of the airflow is extremely demanding, given factors such as: the complexity of the geometry, the range of length and timescales and the unstable or transitional nature of the flow. Including other processes such as heat and mass exchange, the transport and deposition of dilute particle mixtures, and the motion of the boundary (imposed by the mechanics of breathing or due to flow-structure interaction) increase computational cost and add more unknowns in terms of appropriate models and parameters.

Computational modelling of the three dimensional flow in anatomically accurate models is unsurprisingly limited in scope up to the present time. The state of the art numerical simulations of the upper respiratory airways comprise in a number of assumptions including simplified boundary conditions (often steady state), a rigid wall assumption, and restriction of the spatial extent of the domain modelled (for example by including just the nasal airways, or only the portion from the nasopharynx down). Furthermore, rather than direct numerical simulation (DNS) of the full equations, commonly Reynolds Averaged Navier-Stokes (RANS) and more rarely Large Eddy Simulation (LES) approximations are used.

Depending on the particular biomechanical question to be addressed, some of these restrictions and approximations may prove adequate. However their impact on predictions are as yet unclear. The scope of this project relies on performing DNS large-scale computations of the fluid mechanics for extensive portions of the upper respiratory airways of different subjects (including the exterior of the face, the nasal cavity, the trachea and primary bifurcations of the lungs) for a set of flow boundary conditions to simulate cyclic breathing and sniffing. These DNS simulations of large anatomically accurate models will provide a significant insight into the complex airflow and will help understand the physiological functions.

Numerical simulations of a patient-specific model have already been performed using the Alya software. A study of mesh convergence, boundary conditions and software scalability have all already been performed. See e.g. PRACE project 2010PA0326 and the corresponding whitepapers: “Implementing a XDMF/HDF5 Parallel File System in Alya” and “Parallel Uniform Mesh Subdivision in Alya”. The need for extensive computational resources has arisen from this study, as well as the apparent complex nature of the fluid mechanics in the airways that justifies the need for such computationally intensive studies.

Resource awarded: 20.000.000 core hours on FERMI (CINECA, Italy)

Three-Dimensional Simulations of Core-Collapse-Supernova-Explosions of Massive Stars Applying Neutrino Hydrodynamics

Project leader: Hans-Thomas Janka, Max-Planck-Institut fuer Astrophysik – Hydrodynamics Group, Germany

Collaborators: Florian HANKE, Max-Planck-Institut fuer Astrophysik, Germany/ Bernhard MUELLER, Max-Planck-Institut fuer Astrophysik, Germany/ Andreas MAREK Rechenzentrum der Max-Planck-Gesellschaft, Germany.


Supernova explosions of massive stars are among the most powerful cosmic events. They give birth to neutron stars and stellar black holes, produce strong neutrino and gravitational wave signals, and are prime source candidates of chemical elements from iron to plutonium. The details of the physical mechanism that leads to the final explosion of the stars are not yet fully understood. In this project we perform state-of-the-art simulations of the supernova evolution of massive stars and treat the neutrino-matter interactions in the supernova core with unprecedented accuracy. In this project we plan to move towards three-dimensional models of core collapse supernovae with detailed neutrino-transport.

Due to the limitations of the available computing power neutrino radiation-hydrodynamics simulations have long been forced to assume axi-symmetry (2D), thereby imposing restrictions on the convective flow pattern in the neutrino-heated region behind the supernova shock. While axi-symmetric models capture the interplay of convection and neutrino heating to some extent and thus yield weak explosions in some cases, three-dimensional simulations are indispensable to develop a true understanding of the mechanism and consequences of core-collapse supernovae.

Resource awarded: 97.800.000 core hours on CURIE TN (GENCI@CEA, France) and 48.900.000 core hours on SuperMUC (GCS@LRZ, Germany)

High-accuracy quantum mechanic models for extended molecular systems.

Project leader: Branislav Jansik, Aarhus University – Institute of Chemistry, Denmark

Collaborators: Kasper Kristensen, Aarhus University – Institute of Chemistry, Denmark/ Ida-Marie Hoyvik, Aarhus University – Institute of Chemistry, Denmark/ Patrick Ettenhuber, Aarhus University – Institute of Chemistry, Denmark/ Poul Jorgensen, Aarhus University – Institute of Chemistry, Denmark


Many years of theoretical research together with the on-going revolution in computer technology have made coupled cluster (CC) calculations the state-of-the-art method for small molecules. The CC calculations are able to provide highly accurate molecular energies, nuclear forces, and all parameters needed for describing the interaction of a molecule with its surroundings, including perturbing fields. Molecular properties, such as harmonic frequencies, excitation energies nuclear shieldings, frequency-dependent polarizabilities and hyperpolarizabilities and a plethora of molecular spectra, are all within reach.

Standard CC wave function calculations are expressed in the canonical Hartree-Fock (HF) basis, which is a highly delocalized basis. However, the main task of the correlated calculation is to describe short-range electron-electron interactions, which is a very local phenomenon. The discord in description of a local phenomenon using a nonlocal basis inherently leads to a scaling wall, e.g., the seventh power scaling of CCSD(T) calculations, rendering this highly accurate method limited to small molecules only.

The computational scaling of existing CC methods represents a roadblock to progress. One challenge of the 21st century is to remove this roadblock. The ultimate goal is to obtain CC methods that scale linearly with system size, or even better, where calculations for small and large molecular systems have the same wall time, provided a sufficient number of processors is available.

The key to progress in CC applicability is to express the CC wave function in a basis of local HF orbitals and exploit the inherently short range physics of electron correlation. In a breakthrough, we have recently shown how such a local HF basis may be obtained and shown how to take advantage of the local nature of electron correlation to obtain a linearly scaling CC formulation. We denote the new method DEC-CC (Divide-Expand-Consolidate Coupled Cluster).

Using DEC-CC approach, the CC calculation may be split into number of calculations on small fragments of the molecular orbital space, while retaining both full control of the error in the total energy and the black-box character of the traditional CC methodology. Each of the fragment calculations can be carried out independently, making the method embarrassingly parallel. The number of these calculations scales linearly with molecule size. DEC-CC is entirely on par with standard CC except its scaling is linear with respect to molecule size and therefore DEC-CC is suitable for applications on large molecules

The success of the DEC-CC approach will open a new era of accurate quantum calculations on large molecular systems such as nanoparticles and proteins. It holds potential to accelerate research, not only in chemistry and physics, but in molecular science in general.

Resource awarded: 3.000.000 core hours on CURIE FN (GENCI@CEA, France)

Comprehensive ab initio simulations of turbulence in ITER-relevant fusion plasmas

Project leader: Frank Jenko, Max Planck Institute for Plasma Physics (IPP) – Tokamak Physics, Germany

Collaborators: Tobias Görler Told, Max Planck Institute for Plasma Physics (IPP), Germany/ Daniel Told, Max Planck Institute for Plasma Physics (IPP), Germany/ David Hatch, Max Planck Institute for Plasma Physics (IPP), Germany/ Stephan Brunner, Ecole Polytechnique Federale de Lausanne, Switzwerland/ Tilman Dannert, Rechenzentrum der Max-Planck-Gesellschaft, Germany.


Fusion energy offers a promising option for serving the future world-wide energy demand in a sustainable and carbon-free way. Therefore, much effort is invested to remove the last remaining obstacles on the way to efficient power plants. On the physics side, one of the key issues to be addressed is the understanding of plasma microturbulence as it causes anomalous transport which significantly reduces the characteristical energy confinement time. Over the last two decades, numerical implementations of the appropriate theoretical framework given by gyrokinetic theory as well as the high-end supercomputing hardware equipment have evolved in a stunning manner. Hence, computer resources as made available by PRACE allow for a so far unmatched realism of gyrokinetic simulations of anomalous transport in tokamaks. Using the state-of-the-art plasma turbulence code GENE (, comprehensive global computations of actual ASDEX Upgrade and JET discharges involving realistic MHD equilibria, finite-beta effects, collisions, etc. will be performed, using experimental data as input – and exploiting the full potential of the new HERMIT supercomputer. In a second step, the input parameters will be extrapolated to ITER—the upcoming flagship of the nuclear fusion experiments—therefore representing simulations unprecedented in their complexity. The present project will thus constitute a significant step (“milestone”) towards a “numerical tokamak,” demonstrating the rather impressive physical and computational capabilities of present-day gyrokinet
ic codes on modern state-of-the-art hardware.

At the same time, the targeted GENE runs will shed light on key tokamak physics issues, including a better understanding of the multi-scale nature of plasma turbulence in actual high-performance discharges of ASDEX Upgrade and JET. Thanks to the possibility to run GENE also in the local limit, the global simulations can be systematically compared to flux-tube simulations, allowing to assess the degree of nonlocality under realistic experimental conditions. Finally, targeted extended-scale flux-tube simulations will help tackle the important question if and in which cases sub-ion-scale turbulence must be retained to match the experimentally observed fluxes. This is clearly a watershed issue for future modelling efforts.

Resource awarded: 23.300.000 core hours on HERMIT (GCS@HLRS, Germany) and 26.700.000 core hours on SuperMUC (GCS@LRZ, Germany)

Discrete Logarithms on Elliptic Curves over composite extension fields in characteristic 2

Project leader: Antoine JOUX, Univ. Versailles St-Quentin-en-Yvelines – PRISM, France

Collaborators: Vanessa Vitse, Univ. Versailles St-Quentin-en-Yvelines – PRISM, France


The goal of this proposal is to study the state of the art algorithm for computing discrete logarithms on elliptic curves over composite extension field in characteristic 2 from a computational point of view. This includes 4 main sub-topics:

  1. Studying the applicability of the advances recently published
  2. Extending the recent results obtained on Curie for odd characteristic to even characteristic. These results have been accepted to Eurocrypt 2012: and this paper has been announced to be the recipient of the best paper award at this conference.
  3. Improving the computation of the Semaev’s polynomials which are a key tools for these attacks in characteristic 2. In particular, it is important to study the practical impact for these polynomials of the symmetrization technique.
  4. Improving our iterative linear algebra implementation to gain more from the MPI parallelism.

The results of this proposal should include at least on record-breaking computation for discrete logarithm on elliptic curves in characteristic 2, with a crypto-significant size of around 160 bits. These record computations have two main (computation intensive) phases:
a) Sieving (embarrassingly parallel step) [Structured gaussian elimination between the phases — cheap but essential] b) Linear algebra, Lanczos style (Both MPI message passing between nodes and local threads within a node)

Resource awarded: 750.000 core hours on CURIE FN (GENCI@CEA, France) and 750.000 core hours on CURIE TN (GENCI@CEA, France) and 750.000 core hours on JUGENE (GCS@Jülich, Germany)

— –

Full-f gyrokinetic simulation of edge pedestal in Textor

Project leader: Timo Kiviniemi, Aalto University – School of Science, Finland

Collaborators: Jukka HeikkinenVTT, Finland

Abstract:Although research on magnetically confined fusion experiments already started several decades ago, perfectly controlled nuclear fusion has not yet been achieved. A major obstacle is that the transport of the fuel particles perpendicular to the magnetic field lines cannot be predicted by only neoclassical transport theory. Turbulence induced transport, a still unsolved problem in physics in general, has been attributed to be the main reason. It is causing fuel particles to drift from the magnetic field lines, disturbing the plasma confinement and thereby the fusion process. Turbulence induced transport is a transport regime characterized by chaotic properties. The ELMFIRE code, developed at Aalto University in Finland investigates this regime with a so-called first principal computer model which tracks individual particles. The code solves the coupled problem of Boltzmann’s gas kinetic equation for the total particle distribution function and Maxwell’s equations in a complex magnetic field, providing information on the complex interplay between the magnetic field, the electric field and the particle trajectories.

In the present proposal, the Elmfire code will be used to study turbulence and plasma rotation in the low and and high confinement regimes of Textor. The steady state experimental profiles are obtained in the simulation by the balance of a heating model, particle/energy transport, radiation losses and energy transported to the limiter surface. The turbulence levels in the two regimes will be compared to the experimentally ’silent stage’ in the inter ELM period. Furthermore the contribution of the phase velocity and the ExB back ground flow to the poloidal rotation of density fluctuations will be analyzed and investigations into decorrelation rates and probability distribution function of the turbulent fluctuations will be performed. The radial structure of the rotation shear and the radial electric field will be compared to the experimental obtained values.

Resource awarded: 30.000.000 core hours on SuperMUC (GCS@LRZ, Germany)

Eulerian and Lagrangian Turbulence over a reduced fractal skeleton

Project leader: Alessandra Sabina Lanotte, Consiglio Nazionale delle Ricerche – Istituto di Scienze dell’Atmosfera e del Clima, Italy.

Collaborators: Luca Biferale, University of Rome “Tor Vergata”, Italy/ Federico Toschi, Eindhoven University of Technology, Netherlands/ Prasad Perlekar, Eindhoven University of Technology, Netherlands/ Stefano Musacchio, Centre national de la recherche scientifique (CNRS), France.


Turbulence is everywhere around us. It arises whenever a fluid is stirred by some external mechanism (mechanical, thermal, magnetic,..) and there is a large separation of spatial scales between the typical scale of the forcing and the scale at which kinetic energy is transformed into heat, by molecular viscosity. The most ideal case is a statistically stationary, Homogeneous and Isotropic Turbulent (HIT) flow. Yet this is the central problem in turbulence, whose understanding would impact a large variety of applications, from geophysics to engineering. The main deadlock of HIT is unanimously recognized to be the strong non-Gaussian statistics (intermittent) developing at smaller and smaller scales. Still, we do not know what are the physical, kinematical and topological ingredients leading to intermittency in HIT. Moreover, we do not know what is the role played by the inviscid quadratic invariants, Energy and Helicity; we ignore what are the necessary –and sufficient—degrees of freedom needed to develop (or to kill) intermittency; we do not understand if the entangled population of small-scales vortex filaments is important or not to determine turbulent statistical properties in the bulk volume. Turbulence is also considered one of the most cha
llenging problems for extreme computations, a paradigmatic problem for peta-, exa- and even yotta-scale HPC.

This project aims to implement numerically a novel theoretical tool to investigate the key questions raised above. In particular, we intend to apply a decimation of the Navier-Stokes equations dynamics on a Fractal-Fourier set, in order to understand the dependence of small-scales intermittency on the number of degrees of freedom involved in the energy cascade. Such question has never been asked before using this promising methodology.

What is absolutely remarkable is that such questions can be asked only in-silico, by performing state-of-the-art direct numerical simulations at high Reynolds and at changing the embedding Fractal dimension in the Fourier space. We are going to use numerical simulations in a most innovative way, investigating nature with tools unavailable in the labs.

Resource awarded: 22.000.000 core hours on FERMI (CINECA, Italy)

Identification of redox competent species in biological processes

Project leader: Giovanni La Penna, National research council of Italy – Institute for chemistry of organo-metallic compounds, Italy

Collaborators: Peter Faller, CNRS, France/ Christelle Hureau, CNRS, France


In many biological systems, oxido-reductive (redox) reactions efficiently occur when complexes of metal ions access configurations displaying extraordinary high rates for electron transfer. The occurrence of these configurations is modulated by the molecular environment and crowding.

As an emblematic example, the relationship between aging and neurodegenerative diseases (dementia, Parkinson, amyotrophic lateral sclerosis, etc.) is due to the impairing of the biological mechanisms involved in metal ions transport within cells and in intercellular space (like in synapsis). Metal ions like Zn and Cu are particularly abundant in the central nervous system of mammals, where both ionic currents and redox reactions occur more frequently than in other compartments. The same metal ions are found coprecipitated and associated in various forms in the protein aggregates that are hallmarks of neurodegeneration. The mostly investigated example is the structural role of Zn and Cu in the amyloid-beta peptides that form the stable fibrils characterizing the Alzheimer’s disease (dementia).

The Zn and Cu pathways in neurons and in synapsis are coupled by two main processes.

i) Zn and Cu ions, this latter in two possible oxidation states, compete in the weak interactions with the 42 aminoacids amyloid-beta peptide (Abeta(1-42)), that is the major component of the fibrils in Alzheimer’s disease;
ii) when Cu is associated with the Abeta peptide, it may behave as a catalyst for the oxidation of other molecules.

Accurate measures, via cyclic voltammetry, of the redox parameters (both thermodynamic, like the reduction potential, and kinetic) for the Cu2+/Cu+ reduction (usually coupled with oxidation of ascorbic acid, both in vitro and in vivo) show a strong dependence on the peptide environment, that is by the chance of pre-organizing the Cu2+ site into a reduction-prone (or redox-competent) center. This pre-organization puts Cu2+ and Cu+ in metastable states, that may be otherwise silenced in the other conformations available to the Cu-Abeta complex. Oligomeric forms play active role in modulating the chance of these metastable Cu-Abeta forms.

The structures of monomeric complexes of Abeta (the region 1-16 is that involved in metal binding) with Zn2+ and Cu+ have been studied via first-principle molecular dynamics simulations in explicit water models. These simulations showed the importance of competition between protons and metal ions for the same peptide atoms, with metal ion coordination assisting deprotonation of groups with usually high pKa (like the backbone amidic N atoms). Simulations of the Cu2+ interactions with Abeta(1-16) is still in progress, but the collected data well fit with recent experimental results.

The combination of empirical simulations of monomeric Abeta(1-16), Cu and a sample of water with short first-principle molecular dynamics simulations (these latter performed in the Car-Parrinello scheme) provided affordable tests for hypothesis about the chance of reaching states possibly promoting or silencing redox activity. Nevertheless, resonable estimates of the redox parameters for these states were not possible because of the limited statistics collected.

With this project we aim at combining several advances in first-principle simulations for identifying redox-competent forms:

i) massively parallel simulations of peptide models in water samples within the density-functional theory approximation for interatomic forces;
ii) extended statistical sampling induced by external potentials.

The sampling of different coordinations for the Cu ion (in both oxidation states) allows, in theory, the calculation of redox parameters averaging over the many possibilities offered by the peptide flexibility and in a single multiple replica simulation. The PRACE experiment performed for a single Cu ion in a small water sample showed that a simple empirical definition of coordination number greatly helps in describing the most efficient pathways for the reduction of the aqua complex of Cu. Structures for Cu-Abeta(1-16) complex have been already simulated (in both oxidation states) and will be used as starting conditions for this project. Even though a quantitative simulation of redox parameters requires a huge amount of computational resources, the identification of redox-competent states may be performed with less resources. This possibility will be exploited by this project.

Resource awarded: 6.500.000 core hours on JUGENE (GCS@Jülich, Germany)

The Life-cycle of Energy in Space Weather

Project leader: Giovanni Lapenta, KU Leuven, Belgium

Collaborators: Stefano Markidis, KU Leuven, Belgium/ Jan Deca, KU Leuven, Belgium/ Anna Lisa Restante, KU Leuven, Belgium/ Alexander Vapirev, KU Leuven, Belgium/ Vycheslav Olshevsky, KU Leuven, Belgium


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 a variety of socioeconomic losses. The conditions in space are also linked to the Earth climate. The activity of the Sun affects the total amount of heat and light reaching the Earth and the amount of cosmic rays arriving in the atmosphere, a phenomenon linked with the amount of cloud cover and precipitation. Given these great impacts on society, space weather is attracting 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 imp
act 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. Modelling space weather is a daunting task: daunting because the system is enormous and because it includes a wide variety of physical processes and of time and space scales.

A computer model of space weather must face this challenge by using state of the art mathematical techniques to deal with multiscale systems. Our simulation code, iPic3D relies on the implicit moment method. The fields and the particles are studied together in a coupled manner. The word implicit refers to the ability of the method to advance both fields and particles together without any lag between the two (the time lag is a typical aspect of the explicit methods, instead). The word moment refers to the use of moments of the particle statistical distribution. The moments are local statistical averages that characterize the particle properties.

The implicit moment method allows us to select the local level of resolution according to the scales of the local processes. This feature allows to model space weather events with the minimum effort, increasing the resolution only where absolutely needed.

Resource awarded: 5.000.000 core hours on CURIE FN (GENCI@CEA, France) and 5.000.000 core hours on CURIE TN (GENCI@CEA, France)

Towards an atomistic understanding of the selective oxidation of alcohols at the Au/TiO2(110) solid-liquid interface


Collaborators: Matteo Farnesi Camellone, RUHR-UNIVERSITAET BOCHUM, Germany


The selective oxidation of alcohols represents one of the most challenging reaction in green chemistry.

Although a number of efficient methods have been already developed to oxidize alcohols; the search for new, cost-effective and environmentally begin procedures attracts substantial interest in the scientific community.

In this respect, the gold-supported TiO$_2$ catalyst in liquid water offers a sustainable, environmentally benign alternative to traditional processes that use expensive inorganic oxidants and harmful organic solvents.

Experimental studies have indeed shown that gold nanoclusters supported on various metal oxides (MgO, CeO$_2$, TiO$_2$) are able to catalyze a number of reactions.

The outstanding catalytic activities of the supported metal are related to the shape and size of the deposited nanoparticles, and the interplay between the gold and the oxide support. However, there is little consensus on the atomistic details of the reaction mechanisms and on the nature and the role played by the active sites. Computer simulations can offer detailed insight into alcohol oxidation on the gold-supported TiO$_2$ catalyst in liquid water. We propose to perform an extensive theoretical investigation of the oxidation of methanol on the gold-supported TiO$_2$ (110) liquid interface.

Resource awarded: 35.333.333 core hours on FERMI (CINECA, Italy)

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

Project leader: Frederic Nataf, Universite Pierre et Marie Curie – Laboratoire Jacques-Louis Lions, France

Collaborators: Frederic Hecht, Université Pierre et Marie Curie, France/ Pierre Jolivet, Université Pierre et Marie Curie, France/ Christophe Prud’homme, Université Joseph Fourier, France/ Victoria Dolean, Université Nice Sophia-Antipolis, France


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++ is an open source software project that started more than 15 years ago to solve these equations numerically on arbitrary finite element spaces on arbitrary unstructured and adapted two- and three-dimensional meshes. It is a high level programming language based on C++ that 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.. Advanced users can also generate their own C++ piece of software and then load it within FreeFem++ (assuming it has been compiled with dynamic loading support).

Our goal is now to use FreeFem++ 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 super linear speedup in the solving of a Darcy equation with highly heterogeneous coefficients, which describes the flow of a fluid through a porous medium on CEA supercomputer Titane and on IDRIS supercomputer Babel (two GENCI national centres). We have also successfully solved heterogeneous large linear systems (circa 200,000,000 unknowns) preconditioned by two-level methods. However, there are still theoretical work to conduct for the construction of the coarse space problem who could lead to more efficient methods. Moreover, we would now like to investigate harder problem such as linear elasticity or the Navier-Stokes equations on even bigger decompositions using multi-level methods, i.e. by adding even coarser subspaces, and the PRACE supercomputers represent an unforeseen opportunity to do research in that direction.

Resource awarded: 1.000.000 core hours on CURIE TN (GENCI@CEA, France)

Ab Initio Modeling of Solar Active Regions

Project leader: Aake Nordlund, University of Copenhagen – Niels Bohr Institute, Denmark

Collaborators: Klaus Galsgaard, University of Copenhagen, Denmark/ Troels Haugboelle, University of Copenhagen,
Denmark/ Jacob Trier Frederiksen, University of Copenhagen, Denmark/ Remo Collet, University of Copenhagen, Denmark/ Gisela Baumann, University of Copenhagen, Denmark/ Damian Fabbian, Instituto de Astrofísica de Canarias, Spain/ Fernando Moreno-Insertis, Instituto de Astrofísica de Canarias, Spain/ Robert Stein, Michigan State University, USA


The Sun is a fascinating astrophysical object, given its proximity and relevance to us, and the intricate physics of its readily observable surface layers, which provide a unique and comprehensive test bench for non-thermal astro- and plasma-physics. Studies of the Sun’s atmosphere and heliosphere thus help us understand basic physical processes, and are also of direct importance for understanding the environment in which the Earth moves and the perturbations to which its magnetosphere is subjected (space weather).

The overarching scientific aim of this proposal is to produce the first ab initio models of the dynamics of solar active regions – sunspots and their neighborhoods – and their interaction with the sub-surface solar convection zone and the overlying solar corona, on scales that range from 10 km to 50,000 km, using both 3-D MHD simulations and 3-D relativistic charged particle (particle-in-cell code) simulations. The interaction of magnetic flux emerging through the solar surface with the overlying coronal magnetic field creates violent events – solar flares – where charged particles are accelerated to relativistic energies, and with the combination of MHD simulations that can model the large scale dynamics and charged particle simulations that can model the non-thermal particle acceleration we can study both the cause and the effects of these violent events.

This project was initiated with a grant from the 2nd PRACE call and produced already in the first 6 months of JUGENE access outstanding results, which are now in the process of being published (for details see the separately emailed progress report). The results from this project are being compared with observations from the current satellite observatories SDO (Solar Dynamics Observatory) and RHESSI (Reuven Ramaty High Energy Solar Spectroscopic Imager). Predictions are also being made for the future IRIS (Interface Region Imaging Spectrograph) satellite observatory. European scientists have a heavy involvement in all of these three satellite observatories, and the PI of the current proposal is a collaborator on the SDO and IRIS missions.

The project uses mainly two well-proven MPI-codes that parallelize well to hundreds of thousands of cores; a staggered mesh magneto-hydrodynamics code (the Copenhagen Stagger Code), and the PhotonPlasma Code — a relativistic particle-in-cell (PIC) code with modular provisions for particle-particle interactions (Coulomb collisions, Compton scattering, etc.), which also has built-in facilities for high cadence particle tracing, and on-line computation of non-thermal radiation diagnostics.

Resource awarded: 45.000.000 core hours on JUGENE (GCS@Jülich, Germany)

Extreme Star-Formation Modeling: From the Galactic Fountain to Single Stars in One Run

Project leader: Paolo Padoan, Catalan Institute for Research and Advanced Studies (ICREA) and University of Barcelona – Institute of Cosmos Sciences (ICC), Spain

Collaborators: Per-Ake Nordlund, University of Copenhagen, Denmark/ Christian Brinch, University of Copenhagen, Denmark/ Jes Kristian Joergensen, University of Copenhagen, Denmark/ Troels Haugboelle, University of Copenhagen, Denmark/ Remo Collet, University of Copenhagen, Denmark/ Mika Juvela, University of Helsinki, Finland/ Sami Dib, Imperial College, London, UK/ Eugenio Schisano, Istituto Nazionale di Astrofisica, Italy/ Sergio Molinari, Istituto Nazionale di Astrofisica, Italy


The process of star formation is crucial to cosmology and astrophysics. The first massive stars may have been responsible for re-ionizing the Universe. Stars also provide a dominant energy source to the interstellar medium (ISM) of galaxies and control their chemical enrichment. Because of the complexity of this process, involving the mutual interaction of magnetic fields, gravity and supersonic turbulence, no complete theory or numerical model of star formation is available to date. Further progress in this field requires a new class of numerical simulations that can address both the main physical processes and the huge range of scales involved. These challenging simulations are timely because of the wealth of new observational data from ongoing international campaigns with Planck, Herschel, and ALMA, in which our team is involved.

The objective of this PRACE proposal is to study star formation as a multi-scale process, using simulations with an unprecedented dynamic range (seven orders of magnitude in linear scale).

The ISM goes through a complex life cycle known as the Galactic fountain: Massive stars explode as SNe, sending hot gas out of the disk into the halo, where it cools, condenses, and then falls down on the disk again. Because of computational limitations, current star formation simulations are based on unphysical initial conditions and artificial driving forces that can only mimic this energy injection from large-scale processes. We propose to overcome these limitations using numerical simulations of unprecedented size and complexity, which can resolve the collapse of individual protostellar cores while simultaneously modelling the Galactic fountain. We are already carrying out the largest star formation simulations to date, thanks to a series of large allocations on the NASA/Ames Pleiades system, awarded to the PI of this proposal. These simulations cover a range of scales from 100 AU to 100 pc. Using the European supercomputing resources provided by this PRACE call, we propose to jump to the next level reaching 10 kpc, where we can simulate the whole Galactic fountain, while retaining the ability of resolving the formation and evolution of every individual star in the volume. These simulations are now possible, thanks to the specific combination of our unique OpenMP/MPI hybrid version of the public adaptive-mesh-refinement code Ramses and the large-node supercomputers accessible in this PRACE call. Over the last year we have added several new physics modules to Ramses, and we have improved its performance on many levels, particularly through the development of a new hybrid version. We have proved excellent scalability and performance of our hybrid version of Ramses, up to 2,048 nodes (24,576 cores) on Pleiades, demonstrating that we are able to carry out our extreme simulations running on the whole Hermit machine.

Resource awarded: 16.875.000 core hours on SuperMUC (GCS@LRZ, Germany)

Strong interactions beyond QCD

Project leader: Claudio Pica, CP3-Origins, University of Southern Denmark – IMADA, Denmark

Collaborators: Francesco Sannino, CP3-Origins, University of Southern Denmark, Denmark/ Luigi Del Debbio, University of Edinburgh, UK/ Biagio Lucini, Swansea University, UK/ Antonio Rago, University of Plymouth, UK/ Agostino Patella, CERN, Switzerland/ Stefan Sint, Trinity College Dublin, Ireland


The Strong Nuclear Force, described by the theor
y of Quantum ChromoDynamics (QCD), is one of the three fundamental forces of Nature contained in the Standard Model (SM) of particle physics. QCD is responsible for binding the quarks together to form nuclei and for the origin of about 90% of the mass of all ordinary matter.

Recently people started to realize that new strong interactions could contain the key to understand the new physics beyond the SM, which may soon be discovered by the Large Hadron Collider experiments at CERN. These new kinds of strong interactions are, in fact, becoming rapidly more relevant and popular in the scientific community.

Nonetheless, so far QCD is the only strong interaction which has been studied in great detail. We have acquired a precise quantitative understanding of QCD thanks to Lattice simulations during the last 30 years. QCD dynamics is however not suitable for beyond SM physics.

It is only in the last few years that reliable Lattice simulations of new strong forces have started. In particular it was demonstrated that a particular model, known as Minimal Walking Technicolor (MWT), has a dynamics which is fundamentally different from QCD, namely it is infrared (IR) conformal. Models such as MWT could indeed be the starting point to build novel extensions of the SM which satisfy the experimental constraints. Much however remains to be done to understand MWT in detail.

With this project we aim to push our initial investigation of MWT to a new level of accuracy. We aim to understand, in a precise quantitative way, some of the most important aspects of the novel dynamics of MWT: the running of its couplings and its critical behavior in the IR. Such knowledge is not only fundamentally interesting from a theoretical perspective, but also extremely relevant for building new models based on the novel dynamics.

Achieving the level of accuracy we aim for requires the large computational resources available to PRACE. Local resources available to members of the team in the UK and Denmark will also be used for this project, to complement the larger numerical simulations performed at PRACE facilities.

Resource awarded: 22.133.333 core hours on FERMI (CINECA, Italy)

Quantum Monte Carlo simulation of hydrogen at high pressure

Project leader: Carlo Pierleoni, University of L’Aquila – Physics, Italy

Collaborators: Elisa Liberatore, University of Roma “Sapienza”, Italy/ Miguel Morales, Lawrence Livermore National Laboratory, USA/ Jeffrey Mc Mahon, University of Illinois at Urbana-Champain, USA/ David Ceperley, University of Illinois at Urbana-Champain, USA/ Sandro Sorella, SISSA/ISAS, Italy/ Gugliemo Mazzola, SISSA/ISAS, Italy/ Markus Holzmann, CNRS/LPTMC, France


We propose to apply first principles simulation methods based on quantum Monte Carlo and density functional theory to elucidate the equilibrium properties and to explore superconductivity of hydrogen at high pressure.

The properties of dense hydrogen and of hydrogen-helium mixture are important in astrophysics, for example in the understanding of the giant planets and the recently observed exo-planets. The equation of state of dense hydrogen and its heavier isotopes is important for technological reasons such as for experiments conducted at the National Ignition Facility at the Lawrence Livermore National Laboratory.

What phase transitions dense hydrogen undergoes is a long standing problem in condensed matter physics since Wigner in 1935. However, for pressures greater than about 300GPa, its properties and even the underlying structures are not well understood. Recently, we have suggested new crystal structures of hydrogen and estimated the density and pressures of an unusual liquid-liquid transition of hydrogen. Using the recently developed Coupled Electron Ion Quantum Monte Carlo method, we propose to determine the ordering of the various crystal structures and their melting temperatures, and investigate whether liquid hydrogen could be stable at low temperatures. If so, it would have unusual properties due to the strong quantum effects of the ions. To definitively determine the ordering of the various phases we will perform precise free energy calculations, evaluate finite size effects and the quantum effects of both electrons and protons. These techniques require significant computational resources.

In the same interesting region hydrogen has been predicted to be a superconductor with particularly high critical temperature (around or above room temperature). We will investigate the superconducting behavior of the detected structures. In particular we wish to study superconductivity by optimizing an highly correlated wave functions that allows singlet pairing between electrons just like in a conventional superconductor. Although we will not consider electron-phonon coupling, unconventional superconductivity can show up due to strong electron correlation with a mechanism explained within the RVB theory of High-Temperature superconductors. Using Path Integral Monte Carlo techniques we will also investigate the effect of quantum proton statistics on the detected superconducting behavior within the free-energy Born-Oppenheimer approximation.

Our calculations will establish benchmarks useful for approximate density functional theory simulations as well as develop a powerful new tool for investigating material properties.

Resource awarded: 27.000.000 core hours on FERMI (CINECA, Italy) and 24.000.000 core hours on HERMIT (GCS@HLRS, Germany)

Lattice Study of Baryon Resonances

Project leader: Paul Rakow, University of Liverpool – Mathematical Sciences, UK

Collaborators: Gerrit Schierholz, Deutches Elektronen-Synchrotron DESY, Germany/ Roger Horsley, University of Edinburgh, UK/ James Zanotti, University of Adelaide, Australia/ Ross Young, University of Adelaide, Australia/ Yoshifumi Nakamura, Riken, Japan/ Raffaele Millo, University of Liverpool, UK/ Holger Perlt, University of Leipzig, Germany


Explaining the spectrum of hadrons is core to our understanding of QCD in the low-energy regime. The spectrum of stable particles has been successfully reproduced by lattice simulations. Most interesting are resonances and excited states, as they reveal a great amount of information on hadron structure at the confinement scale.

The experimental investigation of the excited baryon spectrum, in particular, has been a long-standing element of hadron physics. We wish to bring the theoretical understanding of these excited states to a higher level.

Lattice gauge theory is a Euclidean formulation of Field Theory. It is easy to find the properties of ground states, but finding the masses and decay rates of excited states is much more difficult, and is only recently becoming feasible.

Resource awarded: 31.300.000 core hours on JUGENE (GCS@Jülich, Germany)

The way to heating the solar corona: Finely-resolved twisting of magnetic loops

Project leader: Fabio Reale, University of Palermo – Department of Physics, Italy

Collaborators: Massimiliano Guarrasi, University of Palermo, Italy/ Marco Miceli, University of Palermo, Italy/ Salvatore Orlando, Istituto Nazionale di Astrofisica (INAF), Italy


The question of what heats the solar and stellar coronae to million degrees is very important in Astrophysics and Plasma Physics. The Sun is a unique laboratory, since it is the only star that we can resolve in detail and it is also a place where one can find extreme ambient conditions. Coronal heating concerns mechanisms of magnetic energy release and confinement, with very important implications for research on energy production. At the same time its investigation is very challenging. The energy from photospheric plasma motions is carried by the magnetic field that confines plasma in a multitude of arch-like structures anchored in the photosphere, the coronal loops.

The aim of this project is to study the twisting of coronal loops with unprecedented model (ten-fold) resolution and completeness, necessary to answer reliably and for the first time important questions on coronal heating, e.g., how is reconnection distributed along and across the loop? How does it compete with kink instability? Does it reach a steady state? The results will produce high-impact scientific publications. The modelled region includes a single coronal loop, described as a straight magnetic flux tube linking two chromospheric layers. The loop model accounts for the reduction of beta and the consequent expansion of the magnetic flux tubes from the chromosphere to the corona. The plasma and magnetic field evolution is described by solving the full 3-D MHD plasma equations including gravity, ohmic and thin plasma radiative losses, thermal conduction. The thin chromosphere/corona transition region is accurately described with resolution down to 20 km, that leads to a very small time integration step, and that has made this task prohibitive so far.

Our project is based on a single large-scale 3D-MHD simulation. We extend the initial 2D configuration to 3D by a simple 360deg rotation. Then we keep one boundary surface fixed and apply a rotation motion to the inner part of the opposite boundary, thus forcing the magnetic field lines to twist around the loop central axis. The magnetic field will be stressed and eventually become unstable, reconnecting or rearranging to less stressed configuration. The geometric domain is a cube with 768x384x384 grid cells, enclosing a loop with total length of 20000km. We follow the loop evolution for about two 360deg footpoint rotations (5000s). We use the PLUTO 3D-MHD code, with the super-time-stepping technique to optimize the demanding thermal conduction. PLUTO is already well-tested on our problem in 2D, and optimized on PRACE Tier-0 systems. In particular, since it is already tested on the BLUE-Gene/P up to 32,768 processors on a problem with very similar size, the CINECA/FERMI machine is highly suitable for the simulation we propose. The 3D simulation scales linearly with respect to the 2D well-tested ones, requiring a computing time of 28.8Mhours on the FERMI HPC system. Using 40000 cores per step the simulation can be performed in 1 month of continuous wall clock time, and, including the warm-up, a total of 3 months and 30.4Mhours CPU-time.

Resource awarded: 30.432.000 core hours on FERMI (CINECA, Italy)

Understanding the bacterial efflux systems: insights into structure-function relationship from all-atom simulations

Project leader: Paolo Ruggerone, University of Cagliari – Department of Physics, Italy

Collaborators: Attilio Vargiu, CNR, Italy/ Francesca Collu, University of Bern, Switzerland/ Michele Cascella, University of Bern, Switzerland


Multidrug resistance (MDR), i.e., resistance towards many chemically and structurally unrelated antibiotics, is of particular concern in Gram-negative bacteria, as they include most life-threatening MDR (and increasingly, pan-resistant) pathogens responsible for extreme resistant and even untreatable infections. Mainly responsible for this appearance are efflux systems belonging to the Resistance-Nodulation-Division (RND) superfamily, which are the object of the present project. This emergency calls for new strategies to design new antibiotics and/or inhibitors, and an efficient path leading to the next generation of compounds will pass through the knowledge of the molecular processes governing resistance mechanisms.

Our aim is to provide microscopically well-funded hints on the intrinsic structural and dynamical complexity associated with MDR that might be helpful for a more efficient drug design. With the same protocol (flexible docking, standard MD, and TMD) tested in previous studies, we will continue our investigation on the effects of point mutations in the binding pocket of AcrB, i.e., the transporter of E. coli RND efflux pump, in strong connection with experimental groups. Other techniques will be also used, such as metadynamics and SMD. The design of inhibitors and/or antibiotics able to escape efflux pumps requires gaining insights on the role of the residues in the crucial regions of the transporter (binding pocket, gate, etc.) and on the interaction patterns at the basis of an efficient extrusion. Thus, we will apply our protocol to unveil the determinants behind the molecular recognition of AcrB by a dozen of antibiotic substrates. It is a challenging task because of the sizes of the systems, the probable presence of multiple binding sites and cooperativity effects, along with the scarcity of data on binding and kinetics. A second application of our computational strategy is related to the functioning of MexB, the transporter of P. aeruginosa. Although it bears high similarity to AcrB, MexB is very elusive, since only a crystal structure without a substrate is available. A comparative study will allow the identification of the common features and differences when compared with AcrB, possibly paving the way to a better understanding of the functionality of these systems.

A further issue associated with the comparative study of the two transporters is the identification of microscopic determinants of the action of efflux inhibitors. Some of them are known to be active against AcrB but never tested on P. aeruginosa. The comparative study will be extended to the investigation of the combination of different conformations of the transporter proposed as steps of the extrusion. They will be simulated for AcrB and MexB and quantitatively evaluated in terms of the free energy costs of the transitions among the configurations. Additionally, examples of good and poor substrate for MexB will be investigated aiming at identifying the still largely unknown key aspects of an efficient and inefficient extrusion associated with structure-dynamics relationship and interactions with the solvent. This information might make feasible a preliminary screening of the compounds based on a microscopic picture.

Resource awarded: 27.000.000 core hours on FERMI (CINECA, Italy)


Project leader: Yanick Sarazin, Commissariat à l’Energie Atomique et aux Energies Alternatives (CEA) – Institut de Recherche sur la Fusion par confinement Magnétique (IRFM), France

Collaborators: Virginie Grandgirard, CEA/IRFM, France/ Guillaume Latu, CEA/IRFM, France/ Chantal Passeron, CEA/IRFM, France/ Guilhem Dif-Pradalier, CEA/IRFM, France/ Xavier Garbet, CEA/IRFM, France/ Philippe Ghendrih, CEA/IRFM, France/ Jérémie Abiteboul, CEA/IRFM, France/ Thomas Cartier-Michaud, CEA/IRFM, France/ Antoine Strugarek, CEA/IRFM, France/ David Zarzoso, CEA/IRFM, France/ Olivier Thomine, CEA/IRFM, France


Energy and particle confinement is among the most critical research topics in magnetic controlled fusion. Energy and particles losses in tokamak plasmas are mainly governed by micro scale turbulence. So will be the case in ITER, the international thermonuclear experimental reactor in construction at Cadarache, France. Recent works have shown that transport coefficients were systematically overestimated by fluid simulations. Five dimensional gyro-kinetic simulations are therefore mandatory to understand such turbulence in these weakly collisional plasmas. The gyrokinetic equation derives from the Vlasov equation by averaging out the fast gyromotion of charged particles immerged in a strong magnetic field. Phase space reduction from 6D to 5D is possible since cyclotron frequency exceeds typical turbulence frequencies by orders of magnitudes. The system is closed by Maxwell’s equations. In practice, for turbulence develops at scales larger than the Debye length, Maxwell-Gauss is safely replaced by quasi-neutrality.

We have developed such a 5D gyrokinetic code, named GYSELA (for GYrokinetic SEmi-LAgrangian) for it uses the semi-Lagrangian numerical scheme: it takes advantage of both Eulerian (fixed grid) and Lagrangian (f constant along trajectories) methods. It is global, in the sense that a large portion of the torus is modelled, and full-f, such that no scale separation between equilibrium and fluctuations is assumed a priori. In this framework, the interplay between small scale fluctuations and large scale structures, such as zonal flows, or events, such as avalanche-like transport and profile relaxation, appears as an important characteristic of turbulence dynamics. So far, the code studies the electrostatic limit, and the electron response is assumed adiabatic. A versatile volume source is implemented, allowing one to inject separately heat, momentum or vorticity. Collisions are taken into account via a Fokker-Planck like operator, acting on the parallel velocity only, and preserving particle, momentum and energy, as it should.

The aim of the project is twofold. First of all, it aims at exploring turbulence saturation mechanisms. These can be either intrinsic, if self-generated by turbulence, or extrinsic. We will focus on the latter mechanism, by exploring the possibility to control turbulence by means of fast particles. Recent simulations with GYSELA have already shown that fast particles can excite so-called energetic geodesic acoustic modes or EGAMs, which exhibit large scale oscillations of the radial electric field at about the sound frequency. They also correspond to a radially sheared poloidal velocity. In turn, such modes are potentially susceptible of reducing turbulent transport by shearing apart turbulent convective cells, as suggested by recent experimental observations. Secondly, we will address particle and heat turbulent transport, since GYSELA can now evolve two distribution functions. The proportionality between density and potential fluctuations is then relaxed. The second specie can be trace impurities, a second ion species (e.g. D-T), or electrons. Several gyrokinetic codes have already addressed some of these issues. The critical novelty resides here in the ability to explore the so-called flux-driven regime, where constant sources of both particles and heat are simultaneously injected in the system. Especially, such regimes allow one to self-consistently address plasma self-organization.

Resource awarded: 4.000.000 core hours on CURIE FN (GENCI@CEA, France) and 8.000.000 core hours on CURIE TN (GENCI@CEA, France)

Direct Numerical Simulation of the Flow in an Internal Combustion Engine

Project leader: Wolfgang Schröder, RWTH Aachen University – Institute of Aerodynamics, Germany

Collaborators: Heinz Pitsch, RWTH Aachen University, Germany/ Stefan Pischinger, RWTH Aachen University, Germany/ Matthias Meinke, RWTH Aachen University, Germany/ Georg Eitel-Amor, RWTH Aachen University, Germany/ Claudia Günther, RWTH Aachen University, Germany/Andreas Lintermann, RWTH Aachen University, Germany/ Pascal Meysonnat, RWTH Aachen University, Germany/ Stephan Schlimpert, RWTH Aachen University, Germany/ Micheal Schlottke, RWTH Aachen University, Germany/ Lennart Schneiders, RWTH Aachen University, Germany/ Christoph Siewert, RWTH Aachen University, Germany/ Andreas Henze, RWTH Aachen University, Germany


The increasing energy demand and therewith increasing carbon dioxide (CO2) emissions, alongside the limited availability of fossil energy resources, represent one of today’s greatest societal challenges. The ’Cluster of Excellence ’’Tailor-Made Fuels from Biomass”(TMFB)’ funded by the German federal and state governments adopts an interdisciplinary approach to research in the area of energy utilization from renewable resources: Through the application of optimized synthesis processes, new biomass-based synthetic fuels are created while modern combustion technologies are developed simultaneously. The tailor-made characteristics of the new fuels and their potential for efficient and clean combustion procedures for combustion engines are investigated.

The central objective of this research project within the TMFB cluster is to analyze the flow field under real engine conditions. The small and large scales of the turbulent flow affect the mixing significantly and interact with the flame front, thus influencing flame area, propagation speed, and the stability of the flame.

During the granted period of calculation resources the focus of the simulation lies on the cold flow field during the intake and compression strokes of a four-valve single-cylinder, optical demonstrator engine at realistic engine speeds. At first, a high resolution large eddy simulations (LES) with will be performed. To analyze cyclic variability 100 – 200 consecutive cycles will be computed. Based on these results, a detailed analysis of the turbulence structures, scales, and intensities will be carried out and compared to already obtained experimental results. Since a thorough analysis of the turbulent length scales in engine flows has not been given in the literature, such an analysis will be carried out by means of proper orthogonal decomposition methods. These results will allow a high fidelity analysis of the impact of the velocity fluctuations on turbulent flame fronts and on the stability properties of the flames and thus the extinction likelihood will be determined. The LES calculation should also serve as a baseline for a subsequent direct numerical simulation (DNS).

To remove the necessity of any modelling a full DNS with an adequate mesh resolution to resolve all turbulent scales will be conducted next. To obtain statistical results, 5 to 10 cycles will be computed, which is a compromise between statistical requirements and the high computational costs of the DNS computations. Such DNS simulations in internal combustion engines at realistic operating conditions have not been reported in the literature yet. The results will provide meaningful initial conditions for the subsequent computations including combustion and will allow to analyze t
he small- and large-scales of the flow in great detail. In addition, from this analysis, the strengths and weaknesses of the LES can be identified.

DNS computations of the flow with premixed combustion at full load operation and at partial load, including direct injection, spray formation and combustion will be conducted later stages of the TMFB.

Resource awarded: 72.700.000 core hours on HERMIT (GCS@HLRS, Germany)

Shock Acceleration in the Laboratory with Ultraintense Lasers: from astrophysics to ion acceleration for medical applications

Project leader: Luis Silva, Instituto Superior Técnico – Instituto de Plasmas e Fusão Nuclear – Laboratório Associado, Portugal

Collaborators: Warren Mori, University of California Los Angeles, USA/ Raoul Trines, Rutherford Appleton Laboratory, UK/ Federico Fiuza, Instituto Superior Técnico, Portugal/ Ricardo Fonseca, Instituto Superior Técnico, Portugal/ Jorge Vieira, Instituto Superior Técnico, Portugal/ Joana Martins, Instituto Superior Técnico, Portugal/ Marija Vranic, Instituto Superior Técnico, Portugal/ Paulo Alves, Instituto Superior Técnico, Portugal


Can collisionless shocks be used to accelerate ions for proton therapy, what is the origin of cosmic rays, what are the dominant acceleration mechanisms in relativistic shocks, how are relativistic collisionless shocks formed are longstanding scientific questions, closely tied to extreme plasma physics processes, and where a close interplay between the micro-instabilities and the global dynamics is critical to fully capture the full nonlinear dynamics.

Relativistic collisionless shocks are closely connected with the propagation of intense streams of particles pervasive in many astrophysical scenarios and are at the core of these questions. The possibility of exciting shocks in the laboratory to study the set of complex and nonlinear phenomena involved in these scenarios, such as magnetic field generation and particle acceleration, will lead to transformative results and will be available very soon with multi-PW lasers coming online, such as the ARC beam in the National Ignition Facility and the ESFRI roadmap project HiPER (High Power Laser Energy Research); moreover, the possibility to excite these shocks to accelerate protons with energies close to 200 MeV would represent a transformative step towards producing more efficient and economic particle sources for proton therapy. However, the conditions required for collisionless shock formation and particle acceleration in laboratory are not yet understood. Computational modelling is critical to understand the physical mechanisms behind shock formation and particle acceleration, and to establish the conditions where these shock waves can be excited in the laboratory, by enabling the fully kinetic modelling of these systems for the first time.

In this proposal, we aim to perform, for the first time with realistic mass ratios and laboratory conditions, self-consistent ab initio massively parallel simulations to study the physics of relativistic shocks driven by ultra-intense lasers, bridging the gap between the multidimensional microphysics of shock onset and the global system dynamics. Leveraging on the state-of-the-art relativistic massively parallel particle-in-cell code OSIRIS, and its recently incorporated hybrid model, we will address the challenge of understanding and establishing a full picture for shock formation and particle acceleration, relevant to astrophysical scenarios, in laboratory conditions using ultra-intense lasers, with the goal of solving some of the central questions in plasma/relativistic phenomena in astrophysics and in the laboratory, and opening new avenues between theoretical/massive computational studies, laboratory experiments and astrophysical observations, as well to the identification of the conditions that can lead to proton acceleration in the 200+ MeV range.

Resource awarded: 37.000.000 core hours on JUGENE (GCS@Jülich, Germany)

Numerical Simulation of 3D Unsteady Reactive Dense Particle Laden Flows

Project leader: Olivier Simonin, Institut National Polytechnique de Toulouse – Institut de Mécanique des Fluides de Toulouse, France

Collaborators: Pascal Fede, Institut National Polytechnique de Toulouse, France/ Hervé Neau, Institut National Polytechnique de Toulouse, France/ Ali Ozel, Institut National Polytechnique de Toulouse, France


The purpose of the project is the development of mathematical models for the 3D numerical simulation of dense particle laden reactive flows from laboratory to industrial scales. Basically, the main practical applications are: solid fuel combustion (circulating fluidized bed boiler, biomass pyrolysis, chemical looping combustion) and gas-particle chemical reactor (fluid catalytic cracking, polymerization reactor, uranium fluoration). An important challenge of such numerical simulations comes from the huge range of the involved scales. Indeed, the diameter of the industrial reactors is of the order of several meters, typically, whereas the solid particle diameter is less than 100 microns. In addition, in dense gas-particle flows, the particle-particle and fluid-particle interactions lead to particle accumulation under the form of unsteady meso-structures of a few centimetres which may have a drastic effect on the global flow behavior.

The numerical simulation of dense reactive fluid-particle flows is generally based on the computation of separate Eulerian transport equations governing the gas and dispersed phase local averaged variables (volume fraction, velocity, temperature …) which are coupled through interphase transfer terms. The computed particle variables represent the first order moments of the velocity probability density function and the corresponding equations are derived directly from the Boltzmann-like kinetic equation in the frame of the kinetic theory of granular media. However, for most of the practical cases, sensitivity analysis shows that affordable industrial meshes (with a few millions of cells) are too coarse to predict accurately the effect of the meso-structures. In order to overcome this problem, by analogy with Large Eddy Simulation approach in single-phase turbulent flows, we have proposed to develop the modelling of subgrid models accounting for the meso-scale effects in the resolved transport equations.

Since 2008, we have been performing numerical simulations on very simplified configurations (periodic boxes) in order to investigate and to model the effect of the subgrid contributions. The mesh is refined up to compute accurately the meso-scale structures, about 20 000 000 cells (using 512 cores). Then, by applying some spatial filtering, the subgrid contributions may be extracted and analyzed in details. Such an approach allowed us to propose effective models for the subgrid velocity variance and drag contributions which are found in the resolved filtered momentum equation. The implementation of the proposed models in an industrial multiphase flow CFD software (NEPTUNE_CFD) and the application to realistic flow configurations lead to satisfactory predictions. However, the proposed models are restricted to isothermal gas-solid flow mixture with a unique diameter for the dispersed phase. In 2012, we plan to ex
tend the model to polydispersed solid mixture and to study the effect of the meso-structures on the heat and mass interphase transfers in reactive flows. So the mesh size should not change drastically in contrast with the needed number of cores due to the increasing number of equations to compute.

Resource awarded: 4.000.000 core hours on CURIE FN (GENCI@CEA, France)

First Lattice QCD study of B-physics with four flavors of dynamical quarks

Project leader: Silvano Simula, INFN – Sezione di Roma Tre, Italy

Collaborators: Vittorio Lubicz, University of Rome III, Italy/ Cecilia Tarantino, University of Rome III, Italy / Giancarlo Rossi, University of Rome “Tor Vergata”, Italy/ Roberto Frezzotti, University of Rome “Tor Vergata”, Italy/ Petros Dimopoulos, Univerisity of Rome “La Sapienza”, Italy/ Damir Becirevic, Universite de Paris XI, France/ Benoit Blossier, Universite de Paris XI, France/ Francesco Sanfilippo, Universite de Paris XI, France/ Vicent Gimenez Gomez, Universitat de Valencia, Spain/ Nuria Carrasco Vela, Universitat de Valencia, Spain/ Carsten Urbach, Universitaet Bonn, Germany/ Andrea Shindler, Humboldt Universitaet zu Berlin, Germany/ Siebren Reker, University of Groningen, The Netherlands


The aim of the present project is an extensive study of B-physics based on Lattice QCD, addressing issues which are relevant for the phenomenology of the Standard Model (SM) as well as for its most general New Physics (NP) extensions, that are currently under investigation by the LHCb experiment and will be studied also at the planned super B-factories

Using the first-principle, non-perturbative technique of Lattice QCD simulations, we plan to compute a number of hadronic B-physics observables with unprecedented accuracy, namely:
i) the b-quark mass, which is a fundamental parameter of the SM;
ii) the decay constants of B, Bs and Bc mesons (fB, fBs and fBc), which are basic hadronic parameters entering the theoretical predictions of several B-physics processes. In particular, from the planned experiments on the B -> tau nu decay and the accurately computed value of fB, one can determine the CKM coupling |Vub|, or get an information on the potential presence of the specific NP scalar states, such as the charged Higgs boson. Current studies of Bs -> mu+ mu- decay at LHCb will provide one of the most stringent flavor physics constraints on the NP model building and in that respect fBs will play a decisive role. As for the double-heavy meson Bc, its mass has been measured recently and only few lattice determinations exist to date; moreover its decay constant, fBc, has not yet been determined experimentally;
iii) the bag parameters of the B-barB and Bs-barBs mixings relevant for the study of the CP violation in the SM as well as in NP scenarios, in particular for the determination of the “xi” parameter, which is crucial for the analysis of CKM unitarity triangle;
iv) the form factors describing the semileptonic weak decays of the B-meson, whose precise knowledge is necessary to extract accurate information on the matrix elements of the CKM matrix, which describes the weak quark mixing.

This goal will be reached thanks to the following main features of the calculation we propose:
i) we account, for the first time in lattice studies of B-physics, the effects of 4 flavours of quarks in the sea, namely the quantum loop effects of the up, down, strange and charm quarks;
ii) we extend to the determination of various B-physics observables the method for studying the b-quark on the lattice to evaluate the b-quark mass and the B-meson decay constant;
iii) through the use of suitable smearing techniques we reduce the coupling of the source and sink interpolating fields to excited states and enhance at the same time their coupling to the ground state, making it possible to extract the relevant signals at smaller time separations. This is relevant when a heavy meson occur in the initial and/or final states, i.e. for the extraction of the B-meson bag parameters and semileptonic form factors. Moreover smearing techniques allow to extend the study of the observables of interest to larger values of the heavy quark mass, up to about three times the physical charm quark mass.

Resource awarded: 30.000.000 core hours on FERMI (CINECA, Italy) and 35.000.000 core hours on JUGENE (GCS@Jülich, Germany)

Global SENSitivity analysis of the MEDiterranean sea biogeochemical model

Project leader: Cosimo Solidoro, Istituto Nazionale di Oceanografia e di Geofisica Sperimentale – OGS – Dept. of Oceanography, Italy

Collaborators: Paolo Lazzari, Istituto Nazionale di Oceanografia e di Geofisica Sperimentale, Italy/ Gianpiero Cossarini, Istituto Nazionale di Oceanografia e di Geofisica Sperimentale, Italy/ Giorgio Bolzon, Istituto Nazionale di Oceanografia e di Geofisica Sperimentale, Italy


Marine ecosystem models are currently employed for a wide range of applications (spanning from short-term forecast to climate change studies), and their results are used in support of management strategies and decision making policies.

Growing complexity of ecosystem marine models increases their realism and usefulness, however uncertainty in model formulation and parameterization should be evaluated in order to increase the confidence in model results.

Objective of the present project is the implementation of a global sensitivity analysis of a state-of-the-art 3D biogeochemical model of the Mediterranean Sea (OPATM-BFM) currently applied in the framework of both climatological and operational oceanographic applications. The OPATM-BFM biogeochemical model is a complex ecosystem model with more than 50 state variables, spanning from the microbial and inorganic compartments to the large zooplanktons.

Kinetics of chemical and biological processes and relationships among variables are non-linear and controlled by more than 200 parameters. The model has been successfully used so far for short-term forecast ( and for hindcast and future climate related scenarios (Lazzari et al., 2010; Lazzari et al., 2011).

The global sensitivity analysis uses the Morris’ method, which is based on a once-at-a-time design of independent runs, which enable for a high performance in parallelization.

The method is based on the evaluation of a number of elementary effects of the model output. An elementary effect is defined as the difference of two consecutive runs that differ for an elementary variation of one parameter. A sequence is done by running elementary variations for all parameters. Multiple sequences are performed optimizing the efficiency in exploring the multidimensional parameters space. The sensitivity is given by computing statistics on the elementary effects. The analysis will require the computation of ad hoc synthetic output indexes subjected to the sensitivity analysis and the statistical analysis (PCA) of temporal-spatial distribution of sensitivity indexes. A relevant by-product of the analysis is a me
asure of the uncertainty of the model forecasts.

This sensitivity method has been successfully used for zero-dimensional complex ecosystem model (Cossarini and Solidoro, 2008), but the computational burden for its implementation on a 3D model has represented the limiting factor till now.

With the world class computational resources provided within the PRACE initiative, we intend to perform the sensitivity analysis of the 3D model, exploring the 200 parameters along 50 trajectories in the parameters space for a total of 10000 simulations. Results of the present project will allow to identify the most important model parameters and ecosystem processes.

The results will provide also an estimation of the model uncertainty, which can be used to constrain future projections simulated by OPATM-BFM ecosystem model.

Moreover, we expect that the results will have an impact on the understanding of the functioning of 3D complex marine ecosystem models. Results will be published into international journals and used in other EU projects.

Resource awarded: 21.760.000 core hours on FERMI (CINECA, Italy)

LUCIDUS – Cosmological simulations of galaxy formation on a moving mesh

Project leader: Volker Springel, University of Heidelberg – Heidelberg Institute for Theoretical Studies, Germany

Collaborators: Federico Marinacci, University of Heidelberg , Germany/ Ewald Puchwein, University of Heidelberg , Germany/ Ruediger Pakmor, University of Heidelberg , Germany/ Lars Hernquist, Harvard University, USA/ Mark Vogelsberger, Harvard University, USA/ Debora Sijacki, Harvard University, USA/ Shy Genel, Harvard University, USA


Galaxy formation is one of the most important unsolved problems in cosmology today. It poses a challenging multi-scale and multi-physics problem that can be fruitfully studied through simulations that probe far into the non-linear regime of cosmic structure formation. In the LUCIDUS project, we perform the first cosmological hydrodynamic simulations with several billion resolution elements and full adaptivity, allowing us to resolve galaxy formation and the cosmic web with an unprecedented combination of spatial resolution, volume, and included physics. Our calculations are based on a novel numerical technique that employs an accurate finite-volume Godunov scheme on an unstructured, fully dynamic and adaptive mesh, combined with particle-based N-body methods (collisionless components), using adaptive time-stepping and a tree- and grid-based Poisson solver for self-gravity. Our physics models account for metal-dependent radiative cooling, star formation, galactic winds, metal enrichment, as well as the growth of supermassive black holes and their energy feedback. The LUCIDUS project will help to understand the connection between the physics of galaxy formation and the large-scale distribution of gas in the Universe. In particular, our simulations will be crucial for interpreting forthcoming galaxy surveys and for testing the prevailing LCDM model for galaxy formation. We will also gain important insights about the reliability of numerical methods presently used in cosmology.

Resource awarded: 20.000.000 core hours on CURIE TN (GENCI@CEA, France)

Fluid turbulence: self and passive scalar diffusion. Application to stably stratified flows

Project leader: Daniela Tordella, Politecnico di Torino – Ingegneria Meccanica e Aerospaziale, Italy

Collaborators: Michele Iovieno, Politecnico di Torino, Italy/ Francesca De Santi, Politecnico di Torino, Italy/ Silvio Di Savino, Politecnico di Torino, Italy/ Luca Gallana, Politecnico di Torino, Italy


Experiments show that turbulent diffusion is complex and that discrete structures or processes, spatially localized within the system, may exist. To obtain a better handling of fundamental issues, we propose an approach where the turbulence self-diffusion is modelled by the interaction between two different isotropic turbulent fields. This simplifies the main mechanisms. In fact, it does not include the nonlinear production of turbulent energy. However, it retains two of the most important features present in real flows: inhomogeneity and anisotropy.

Recent simulations in our group revealed the generation of small-scale anisotropy in turbulence self-diffusion. A long-term interaction must be active to transfer to small scale the information on the anisotropy of the initial and boundary conditions (PRL 2011). Data from direct numerical simulations show that there is a departure of the longitudinal velocity derivative moments from the values found in HIT and that the anisotropy induced by the presence of a kinetic energy gradient has a different pattern from the one generated by an homogeneous shear. Other results concern the relationship between the correlation length and intermittency. A variation of the correlation length is not necessary to depart from Gaussianity (PHYSD 2012, PRE 2008). However, if the correlation length variation is concurrent with that of the energy, the mixing is enhanced, if is opposite, the mixing is decreased (JFM 2006). The transport of a passive scalar or a stable stratification added to the system highlight other phenomenology. The dimensionality of the system is in particular of great relevance for some aspects (temporal mixing growth and vorticity suppression).

We propose here to carry out a number of simulations to account for: a) the variation over two orders of magnitude of the parameters associated to the presence of a stable density stratification, b) the passive scalar transport across an interface, and c) an increase of the amount of statistics for the turbulence self-diffusion already studied. The results, including raw data, will be made available to the community by giving access to the web disks of our group and by posting the data to HPC repositories (e.g. i-cfd at CINECA, or web sites of international cooperations, like ICTR

The impact of this research is expected in all the fields where turbulence does matter and where a manifestation of the nonuniversal behavior of small scales is closely related to small-scale anisotropy. In this concern, the project is fully interdisciplinary. Potential applications of this research span from atmospheric science to astrophysics, but are ubiquitous also in engineering systems.

Resource awarded: 2.440.500 core hours on CURIE FN (GENCI@CEA, France) and 536.850 core hours on CURIE TN (GENCI@CEA, France)

Mesoporous silica for drug delivery: a quantum mechanical simulation

Project leader: Piero Ugliengo, Università di Torino – Dip. Chimica, Italy

Collaborators: Massimo Delle Piane, Università di Torino Italy/ Marta Corno, Università di Torino Italy/ Fabio Chiatti, Università di Torino Italy/ Bartolomeo Civalleri, Università di Torino Italy
/ Roberto Dovesi, Università di Torino Italy/ Roberto Orlando, Università di Torino Italy/ Matteo Ferrabone, Università di Torino Italy/ Lorenzo Maschio, Università di Torino Italy/ Alfonso Pedone, Università di Modena e Reggio Emilia, Italy


The contact between surfaces of inorganic materials and fluids containing biologically relevant molecules is of extraordinary relevance in medical devices, biotechnology and also in proteomics. Among the most important inorganic materials, silica based ones play a key role, because they are chemically inert, structurally robust, easy to prepare and characterize. Considering this class of materials, ordered mesoporous structures have attracted a lot of interest in the last decades. They have found many applications in separation, catalysis, sensors and devices. Generally, they show a regular arrangement of either cylindrical pores or cages of mesoporous size (2-10 nm) with high surface area. As drug delivery systems they are compatible with the fundamental requirements of biodegradability and biocompatibility. MCM-41, a member of this family of materials, was first proposed for this use in 2001. Since then, a lot of investigations have been done in this area. Among drugs whose transportation in the body can be assisted by mesoporous silica, the most studied one is ibuprofen, a common non-steroidal anti-inflammatory drug.

The adsorption of drugs into the pores is a surface phenomenon that is governed mainly by interactions between silanol (SiOH) groups and the functional groups of the guest molecules, but also London type interactions can play an important role considering the hydrophobic nature of ibuprofen. An interesting point of debate in the study of ibuprofen confinement in mesoporous silica concerns the physical state of this molecule in the system. In particular, it is still not clear if the majority of the drug population is in interaction with pore walls or in a free state and, in this case, if it is as a free molecule or in a dimeric form.

The aim of this project is to simulate by quantum mechanical methods the features of MCM-41 mesoporous silica material with respect to adsorption of ibuprofen drug as a single molecule and in its dimeric form. The adopted approach is based on well know quantum mechanical methods based on DFT plus a posteriori dispersion correction. The atomistic details provided by these accurate calculations would allow, for the first time in the literature, to understand the interaction energy of ibuprofen with a mesoporous silica material. This study will also try to compute the relative stability between ibuprofen adsorbed as a monomer vs. the dimer, a key point to understand its release in the body. We believe that the expected results, in particular the calculated interaction energies, may be useful to design new functionalized MCM-41 materials apt to modulate the strength of interaction with ibuprofen and similar drugs.

Resource awarded: 20.000.000 core hours on SuperMUC (GCS@LRZ, Germany)

Electronic structure of wet systems

Project leader: Paolo Umari, Università degli studi di Padova – Dipartimento di Fisica ed Astronomia, Italy

Collaborators: Filippo De Angelis, CNR, Italy


A large number of chemical processes which are important either for technology or for science take place in solution. Indeed, not only most of biology happens in a wet environment but also nowadays chemicals production strongly relies on wet chemistry and chemical reactions in solutions are at the base of devices such as catalysts.

The presence of a solvent induces important effects on the physical properties of a system with respect to the case when it is in the gas-phase. In particular, this happens when the solvent is water, which is a prototypical case due to its relevance for biological systems and its wide frequency in technological realizations.

One of the first steps for investigating, from an atomistic point of view, the physical properties of a solvated molecular system is the assessment of its electronic properties. Indeed, the relative position of energy levels determines the reactivity of simple chemical processes such as electron charge transfer.

The theoretical study of electronic properties requires the use of first-principles methods. In contrast to the case of small molecules, for which very accurate quantum-chemistry schemes can be easily used, larger molecules comprising from tens to hundreds of atoms call for methods showing better scaling of the computational cost with respect to the size of the system. Indeed, most of the studies of solvated molecules have been performed till now using approaches based on density functional theory (DFT). However, the accuracy of these methods is far to be completely satisfactory. Much more accurate methods based on the treatment of the many-body perturbation theory formalism with the GW approximation have proved successfully in describing electronic properties of large organic molecules and of bulk aqueous systems. Here, we propose to use a recently developed method for performing accurate GW calculations in large systems for addressing the electronic properties of solvated molecules keeping an explicit quantum description of the solvent. We want to address three important cases: simple molecules solvated in water as the simplest prototypical realization, solvated DNA fragments, adsorbed organic dyes in water solution. In all these cases, we will first performed molecular dynamics runs at fixed temperature. Then, we will select a set of configurations for which we will perform GW calculations assuring in this way a good statistical average.

The GW approach yields nice results for the electronic structure of isolated DNA bases and it is interesting now to include in the simulation also the phosphate backbone together with the surrounding counter-ions and water molecules as their relevance has been highlighted in previous DFT studies.

In the final part of our proposed work, we will address the case of a solvated organic dye adsorbed on a titania surface. Systems of this kind are strongly investigated nowadays for building electrochemical solar-cell avoiding the use of Ru based dyes. In such systems, the inclusion of the modifications due to the presence of the water solvent has been shown to be the key for reproducing optical excitations.

Resource awarded: 6.500.000 core hours on FERMI (CINECA, Italy)

3D-3V Vlasov simulations of plasma turbulence

Project leader: Francesco Valentini, Universita’ della Calabria – Physics, Italy

Collaborators: Francesco Califano, University of Pisa, Italy/ Pierluigi Veltri, Universita’ della Calabria , Italy/ Denise Perrone, Universita’ della Calabria , Italy


The interplanetary medium, the bubble of plasma that is generated by the Sun and that fills the Heliosphere, is known to be hotter than expected for an expanding, almost collisionless plasma. Understanding how energy from the Sun can be dissipated into heat in such a collision-free system represents a top priority in space physics. The Sun injects energy into the Heliosphere through large wavelength fluctuations, mainly in the form of Alfvén waves. This energy is then channelled towards short scales through a turbulent cas
cade until it can be transferred to plasma particles in the form of heat. Turbulence is therefore thought to be responsible for the local heating of the solar wind and kinetic effects at short wavelengths are considered as the best candidate in replacing collisional processes and in “dissipating” the energy coming from the large scales.

Nowadays, due to the impressive growth in computational resources, it is possible to investigate numerically the role of kinetic effects in the evolution of turbulence and in other basic processes, e.g. magnetic reconnection. In this context, the Particle In Cell (PIC) semi-Lagrangian approach is the most used numerical model that retains kinetic effects. However, in 2007 we introduced a new Eulerian hybrid-Vlasov model to analyze the ion kinetic dynamics of a collisionless magnetized plasma. Through this algorithm we succeeded in describing the evolution of the turbulent cascade in the solar wind going from large scales (somewhat larger than the ion skin depth) towards short wavelengths, in a phase space of reduced dimensionality and in providing important insights into the nature of the local and fast reconnection events occurring as the results of the turbulent cascade.

The main advantage in using Eulerian (Vlasov) instead of Lagrangian (PIC) algorithms consists in the fact that Eulerian schemes do not suffer from statistical noise. This can be crucial, for example, in the analysis of the short-scale region of the turbulent spectra, where the statistical noise introduced by PIC can mask the physical informations.

Within the present project, we propose to extend our study of turbulence and magnetic reconnection in space plasmas to the full dimensional 3D-3V phase space configuration. In particular, we are interested in the analysis of the role of kinetic processes at play in the turbulent cascade in the solar wind, its relationship with explosive magnetic reconnection events and the evolution of the distribution of particle velocities. We expect that the results of these massive simulations will provide crucial informations in the interpretation of in situ measurements from spacecraft. The 3D-3V Vlasov simulations we plan to run will represent the first and unique (to date) attempt to provide a realistic interpretation of experimental measurements in space.

Resource awarded: 25.000.000 core hours on FERMI (CINECA, Italy)

High-fidelity simulations of multiscale-generated turbulence

Project leader: John Christos Vassilicos, Imperial College London – Aeronautics, UK

Collaborators: Sylvain Laizet, Imperial College London, UK/ Eric Lamballais, P’ Institute, France/ Veronique Fortune, P’ Institute, France/ Ning Li, Numerical Algorithms Group (NAG), UK


After more than a century of exhaustive research on the aerodynamics and hydrodynamics of geometrically simple shapes, whether streamlined as in wings or bluff as in spheres/cylinders, it is blindingly natural to expect much of the future in fluid mechanics to lie in the aerodynamics and hydrodynamics of geometrically complex, and thereby multiscale, shapes. There has of course been work over the past decades on how to model and simulate complex turbulent flows, but the emphasis here is on working out rules for the design of multiscale objects so as to obtain desired flow effects beneficial for particular applications. The simplest cases of multiscale shapes are fractal, which is why they have been a good start.

Multiscale/fractal generation/design is about using multiscale/fractal objects (grids, fences, profilers etc) to shape the nature of the resulting turbulent flow over a broad range of scales for a broad range of applications, such as:

  1. Fractal mixers: fractal grids can be used to design turbulent flows with low power losses and high turbulence intensities for intense yet economic mixing over a region of designed length and location.
  2. Fractal combustors: the fractal design of a long region of high turbulence intensity and its location are of great interest for premixed combustion and may pave the way for future fractal combustors particularly adept at operating at the lean premixed combustion regime where NOx emissions are the lowest. In fact results recently obtained at Imperial College ’s Mechanical Engineering Department suggest that fractal design seems to generate turbulent flame speeds which increase by even more than the increase in turbulence intensities!
  3. Fractal spoilers and airbrakes can have significantly reduced sound pressure levels without degrading the lift and drag characteristics of the wind system.
  4. Fractal wind breakers and fractal fences: a fractal fence, for example, can have increased resistance, yet be an effective fence by modifying the momentum profiles in its lee and thereby forcing deposition of particulates, snow etc where desired.

The next unavoidable step towards progress in this new field of research is to perform high-fidelity Direct Numerical Simulations (DNS) (combined with a new raft of experiments currently underway at Imperial College London and elsewhere) to gain fundamental understanding of the spatial structure of these flows. This is necessary to understand the origins of the above-mentioned unique properties and find strategies to utilise these properties to create new industrial applications. Such multiscale-generated turbulent flows cannot be simulated without High Performance Computing (HPC) on Tier-0 systems. They pose a formidable challenge to researchers due to complex geometries and the presence of a multitude of imposed length scales in space and in time. This requires state-of-the art top-end parallel computing for high fidelity simulations to understand the origins of the unique properties of multiscale objects and propose new strategies for industrial applications.

Resource awarded: 5.475.000 core hours on SuperMUC (GCS@LRZ, Germany)

Share: Share on LinkedInTweet about this on TwitterShare on FacebookShare on Google+Email this to someone