DECI 9th Call

Find below the results of DECI-9 (Distributed European Computing Initiative) call.

Projects from the following research areas:


Applied Mathematics (1)


Project Title: New hybrid solver for large sparse linear systems and scalability
Project Leader: Prof. Mustafa Serdar Celebi, Istanbul Technical University, Informatics Institute, Istanbul, Turkey
Resource Awarded: 5 045 759 core hours on EPCC – ICE-Advance and RZG – Hydra

Dr. Ahmet Duran – Istanbul Technical University, Informatics Institute, Istanbul, Turkey
Dr. Figen Oztoprak – Istanbul Technical University, Informatics Institute, Istanbul, Turkey
Mehmet Tuncel – Istanbul Technical University, Informatics Institute, Istanbul, Turkey
We are working on design and implementation of new hybrid algorithm and solver for large sparse linear systems. We focus on scalable direct solvers. We design a new parallel algorithm HybridSuperLU 1.0 for large scale server systems containing fat nodes. HybridSuperLU 1.0 is utilizing the MPI+OpenMP hybrid programming approach. We plan to compare the effectiveness of the hybrid algorithm with the state of the art sparse direct solvers MUMPS, Pardiso and SuperLU_DIST Version 3.1 [1]. Existing version of SuperLU is scalable and tuned for large distributed systems based on MPI. The algorithm is sensitive to tuning and needs further customization for various type of sparse matrices. We plan to modify the SuperLU in order to improve its scalability via several ways. Moreover, we tune the algorithm for various type of large sparse matrices. We plan to make the solver scalable on large server systems (systems having more than 1000 processors). As the system becomes larger, robustness will be a challenging issue because cumulative rounding errors dominate. We believe that robust and scalable direct solvers are important to handle this issue as one of our objectives. [1] X. S. Li, J. W. Demmel, J. R. Gilbert, L. Grigori, M. Shao, I. Yamazaki, SuperLU Users’ Guide, 1999, update: 2012

Astro Sciences (4)


Project Title: Galaxies and Dark- and Luminous-matter Feedback: the origin of cosmic structures and the effects of feedback processes during cosmological evolution
Project Leader: Dr. Umberto Maio, Max Planck Institute for Extraterrestrial Physics, Garching, Germany
Resource Awarded: 1 300 000 core hours on UIO – Abel

Dr. Margarita Petkova – University of Bologna, Department of Astronomy, Italy
Cosmic structure formation is a very crucial issue in modern Astrophysics and is one of the most debated topics of the recent scientific developments. Cosmic structures (stars, galaxies, and clusters of galaxies) arise from primordial matter fluctuations generated in the initial instants of the Universe and representing the original ‘seeds’ which would evolve and grow up into the present-day structures. The general cosmological framework of such process has been extensively studied in the last decades, and this has allowed scientists to establish a standard cosmological model relying on the cosmic expansion of the Universe in a ‘flat’ space-time. In this framework, the key-mechanism leading structure formation is gravitational collapse of the primordial matter perturbations, which takes place inside the over-dense regions, while the external, low-density, cosmic space expands. In the over-dense regions, in fact, gas falls into the growing potential wells of dark-matter fluctuations, cools down and condenses into cold clumps. This is possible because during collapse, gas densities increase, molecule formation is boosted, and molecular cooling becomes more and more efficient. The final result is that gas temperatures drop down dramatically and the gas kinetic energy gets converted into radiation. In this way, first stars can form, and their assembly can lead to the formation of larger objects, like galaxies and galaxy clusters. Their birth and evolution is extremely relevant, as they light up the Universe, heat the surrounding medium, and interact with the ambient gas by a number of physical mechanisms, usually classified as: ‘mechanical feedback’ (like supernova explosions, galaxy interactions, gas stripping), ‘chemical feedback’ (metal spreading from stars and changes in the chemical compositions of cosmic gas), and ‘radiative feedback’ (heating and radiation from stellar sources).

The starting point of any serious investigation on the origins of cosmic structures and feedback mechanisms is the detailed calculation of the most relevant physical processes, i.e. gravity, hydrodynamics, chemistry evolution, and radiative transfer. Then, higher-order complications can derive from molecular evolution at early times, metal production in stellar cores, metal pollution of the cosmic medium, turbulence, establishment of large-scale magnetic fields, cosmic rays, tidal torques, and angular momentum acquisition. All these phenomena should take place and interact with each other during the growth of cosmological structures. The main difficulties in properly following such processes at early and late cosmological times are related to their high non-linearity.

Because of that, advanced studies can be performed only with the help of numerical calculations and detailed cosmological simulations of the cosmic medium, addressing the role of atoms, molecules, and stellar activity, both in intergalactic environments (intergalactic medium, IGM) and in interstellar environments (interstellar medium, ISM).

Due to the lack of self-consistent studies addressing such issues, we propose to run numerical Nbody, hydrodynamic, chemistry, radiative simulations taking into account all the aforementioned processes. This will allow us to provide the scientific community with a set of complete, selfconsistent, but computationally expensive, state-of-the-art simulations, which will be a fundamental tool to better understand the build-up of the observed cosmic structure and the role played by the various physical processes during cosmological evolution.


Project Title: Modelling the molecular content of massive, high redshift galaxies: predictions for ALMA
Project Leader: Prof. Dr. Cristiano Porciani, University of Bonn, Argelander Institute for Astronomy, Bonn, Germany
Resource Awarded: 7 000 000 core hours on EPCC – HeCToR XE12 and UIO – Abel

Dr. Aaron Ludlow – University of Bonn, Argelander Institute for Astronomy, Bonn, Germany
Dr. Emilio Romano-Diaz – University of Bonn, Argelander Institute for Astronomy, Bonn, Germany
Matteo Tomassetti – University of Bonn, Argelander Institute for Astronomy, Bonn, Germany
Spatially resolved observations of nearby galaxies indicate that their star formation rates correlate more tightly with the observed density of molecular hydrogen, H2, than that of the total gas density, as often assumed in standard sub-grid models for star formation. This has motivated several attempts to update the “classic” star formation recipes often implemented in cosmological simulations of galaxy formation for ones in which the star formation law is regulated by the local abundance of H2 rather than the total gas density. This requires either following self-consistently the non-equilibrium chemistry of the formation and destruction of H2, including its formation on dust grains, and the radiative transfer responsible for its ionization and dissociation [a,b,c], or alternatively, relating the abundance of H2 to the HI column density and gas metallicity on the scales of giant molecular clouds (GMCs), provided sufficient resolution is obtained to model their mean internal densities. Recent observational data support the theoretical prejudice (based on the physics of turbulence) that the efficiency of gas conversion into stars per free-fall time in GMCs is roughly universal (~1%), independently of the local gas density [e]. This provides a simple way to compute star formation maps in simulations of galaxy formation provided that they reach spatial resolutions sufficient to resolve GMCs (~100 pc). On the observational front, submillimeter surveys have revealed many more luminous, massive galaxies at high redshift than are typically predicted by galaxy formation models that adopt standard prescriptions for star formation. In addition, the Herschel Multi-tiered Extragalactic Survey and the PACS Evolutionary Probe are currently measuring the bolometric emission from infrared galaxies. This will shed new light on the most active phase of the Universe’s star formation history and on the galaxies responsible for it. In the near future, the Atacama Large Millimeter/sub-millimeter Array (ALMA) will detect massive reservoirs of gas (CO and C+) at high redshifts and will map dust emission back to the epoch of reionization. The interpretation of these observations will require high-resolution simulations capable of modelling the evolution of quiescent galaxies and starbursts, as well as their molecular content. Our goal is to implement an improved treatment of the cold interstellar medium, star formation and feedback in cosmological simulations, improving upon previous work in order to produce self-consistent maps of CO, H2 and C+ for massive, high-redshift galaxies. We will apply new subgrid physics models to two high-resolution simulations of galaxy formation (one quiescent disk, and a merging galaxy pair at z≥2) capable of resolving the spatial scales characteristic of GMCs. We will use these simulations to make predictions for, and interpret results from, forthcoming ALMA observations of massive, high- redshift galaxies. The proposed project is timely given that early science with ALMA has been underway since 2011, and the telescope will be fully operational in 2013, coincident with the proposed project period. Furthermore, since we have direct access to collaborations within the German Regional ALMA Center, our project will also hold major relevance to European and International research initiatives.


Project Title: Planck LFI data analysis for second release
Project Leader: Hannu Kurki-Suonio, University of Helsinki, Finland
Resource Awarded: 7 000 000 core hours on CSC – Louhi XT and CSC – Sisu

Enrique Martinez-Gonzalez – Institute of Physics of Cantabria (CSIC-UC), Santander, Spain
Paolo Natoli – University of Ferrara, Italy
Jussi Väliviita – University of Helsinki, Finland
Planck is a European Space Agency satellite mission, whose task is to map the structure of the cosmic microwave background (CMB) in unprecedented detail, surpassing the accuracy of previous missions, like the NASA WMAP. The cosmic microwave background is radiation from the Big Bang, and it shows us the structure of the early universe.

Planck will constrain cosmological models and examine the birth of large-scale structure in the universe. It is thought that this structure originates from quantum fluctuations in the very early universe during a period of accelerated expansion called inflation, but Planck results are needed for a better understanding of this. The main scientific results expected from Planck are cosmological, but as a by-product, Planck will also yield all-sky maps of all the major sources of microwave to far-infrared emission, opening a broad expanse of other astrophysical topics to scrutiny. Planck was launched in May 2009. The first cosmological results will be published in early 2013; there will be a second release in early 2014; and possibly a final third release with the final results from the mission in 2015.

Planck carries two instruments, the Low-Frequency Instrument (LFI) and the High-Frequency Instrument (HFI), utilizing different technologies. It is important to map the sky at many frequencies to be able to separate the cosmic microwave background from the astrophysical foreground radiation. Two Planck data processing centres (DPCs) have been set up, one for each instrument, but the most resource-intensive tasks need to be done on supercomputers.

Because of the weakness of the signal and the high accuracy desired, Planck data analysis is a complicated task, requiring sophisticated statistical methods to separate out the signal from instrument noise and systematic effects.

Simulation work will dominate the computational load of Planck data analysis. Analysis pipelines will indeed be predominantly run on simulated rather than real data since Planck analysis codes either require simulations for self-calibration and validation of results, or depend critically on simulations for the results themselves.

The objective of this project is to carry out two resource-intensive tasks that are needed as part of Planck LFI data analysis, for the second data release and publication of cosmological results in early 2014 (there is an ongoing PRACE DECI project for the first release in early 2013):
1) timeline-to-map Monte Carlo simulation of Planck data: thousands of realizations of instrument noise and cosmic microwave background signal; also simulations of astrophysical foreground radiation signal; the analysis of this data in parallel to the analysis of the real data from the sky
2) cosmological parameter estimation for a number of cosmological models, in particular those related to multi-field inflation.


Project Title: Spin and eccentricity effects on BNS coalescence
Project Leader: Prof. Dr. José Antonio Font, Universidad de Valencia, Relativistic Astrophysics Group, Valencia, Spain
Resource Awarded: 2 500 000 core hours on RZG – Hydra

Filippo Galeazzi – Universidad de Valencia, Relativistic Astrophysics Group, Valencia, Spain
Wolfgang Kastaun – Max Planck Institute for Gravitational Physics, Potsdam, Germany
Vassilios Mewes – Universidad de Valencia, Relativistic Astrophysics Group, Valencia, Spain
Luciano Rezzolla – Max Planck Institute for Gravitational Physics, Potsdam, Germany
Kentaro Takami – Max Planck Institute for Gravitational Physics, Potsdam, Germany
Binary neutron star (BNS) systems coalesce and merge due to the emission of gravitational waves (GW), as predicted by general relativity. Such systems are the most likely candidates for detection with current GW detectors. Matter in the core of neutron stars is in a unique state which is not present in any other system and which cannot be reproduced in laboratories. Observational data of neutron stars are a very important input for nuclear physics. The GW signal produced by NS mergers is very sensitive to the equation of state (EOS) of neutron star matter, and could therefore be used to constrain various theories describing matter at nuclear densities. For this, it is necessary to map out the influence of parameters other than the EOS, namely the NS masses, the eccentricity of the orbit, and the NS spins. Previous studies have focused mainly on the masses and the EOS. Recently, eccentric orbits have been studied as well, using however a strongly simplified model for the EOS. In this project, we will study the influence of eccentricity and spin, using a state of the art EOS considering thermal effects and the influence of nuclear matter composition (electron fraction). For the extreme cases regarding spin or eccentricity, we will consider the influence of the EOS. Besides the GW signal, we will also investigate the spin of the final merged object on the parameters of the initial system, a quantity which might be accessible to standard (E.M.) astronomical observations. Furthermore, we are interested in the amount of matter ejected during the process, which orbits the merged object (most likely a black hole). This hot, dense matter orbiting a black hole is an important ingredient for models explaining observations of so called short gamma ray bursts. To accomplish those goals, we will perform three dimensional, fully relativistic numerical simulations of close binary neutron star systems with different spins and eccentricities. Our code employs a modern high resolution shock capturing (HRSC) scheme for evolving the hydrodynamic part, high order accurate finite differences for evolving the spacetime, and mesh refinement methods allowing to evolve a large domain. The latter is crucial to investigate GWs. An outstanding problem is the current lack of initial data for eccentric or spinning BNS systems fully consistent with general relativity. However, we recently observed that when evolving exact initial data for non-spinning NSs in circular orbits which has been modified to add spin and eccentricity, the violations of general relativity partially dissipate during the first stages of the evolution.

Bio Sciences (10)


Project Title: Interaction of drug delivery liposomes with opsonin proteins (complement activation)
Project Leader: Alex Bunker, University of Helsinki, Finland
Resource Awarded: 5 600 000 core hours on CSCS – Rosa

Seppo Meri – University of Helsinki, Finland
Tomasz Rog – Tampere University of Technology, Finland
Arto Urtti – University of Helsinki, Finland
Tapani Viitala – University of Helsinki, Finland
We are applying for four million core hours on the resources of PRACE to perform a set of molecular dynamics (MD) simulations that will study the interaction between drug delivery liposomes (DDLs) and two proteins found in the bloodstream that are known to coat DDLs.

Once a drug molecule has been designed it must be delivered to the desired location in the body. A very promising new avenue for drug delivery is pharmaceutical nanotechnology (1): the encapsulation of drug molecules into a nanoscale delivery device, that protects the drug molecules as they travel through the bloodstream.

One of the most promising forms of nanoscale drug delivery device is the DDL (2): a phospholipid membrane that encapsulates the drugs to be delivered in an internal cavity, and expresses a protective polymer sheath on the exterior. The protective polymer sheath inhibits uptake by the bodies defence mechanism, the reticuloendothilian system (RES), in a process known as complement activation (3). A particularly effective coating is the polymer poly(ethylene)-glycol (PEG). While the efficacy of PEG is impressive, it extends bloodstream lifetime of the DDL from ~1 h to 1 – 2 days there is still significant room for improvement; red blood cells and antibodies circulate in the bloodstream for up to 1 – 2 months. Thus, the search for a superior protecitve polymer to PEG is an extremely active field of research (4).

The PEG polymer is particularly effective as a protective polymer coating, with better performance than several superficially similar alternate polymers, however the mechanism through which PEG achieves its superior performance is not entirely clear. In order to follow a rational design approach to finding a superior alternative polymer to PEG we must first understand the protective mechanism of PEG. One proposal is that PEGylation inhibits the first step of RES uptake: opsonisation (5), where a series of bloodstream proteins stick to the outer surface of the DDL. The experimental data on this question is however inconclusive, and our proposed project is to used computational molecular modelling with an all atom model to address this question.

The current work is part of a larger research stream to study the PEGylated liposomes. Our initial work (6,7,8), to simulate a PEGylated bilayer at physiological conditions has yielded important and unanticipated results, that shed light on how the properties of PEG effect its protective efficacy. The current study continues this investigation. We propose to simulate the membrane of the DDL interacting with two bloodstream proteins that have been found to play an important role in opsonization: Human Serum Albumin (HSA) and Complement component 3.

We will perform MD simulation of four systems: membrane both with and without PEGylation with the protein situated next to it. This is to be followed by force biased simulation (9) for each system to further probe the membrane protein interaction. Each system is to be simulated for a total of 1 μs (~ one million core hours each). Our results have the potential to directly assist the development of future generations of DDLs, and is supported by the experimental research program of Prof. Arto Urtti, Dr. Tapani Viitala, and Prof. Seppo Meri.


Project Title: Molecular Dynamics simulations of mixed DOPC/DOPE based membrane bilayer
Project Leader: Dr Roberta Galeazzi, Università Politecnica delle Marche, Ancona, Italy
Resource Awarded: 3 000 000 core hours on CINECA – PLX

Luca Massaccesi – Università Politecnica delle Marche, Ancona, Italy
Recent progress in nanotechnology has triggered the site specific drug/gene delivery research and gained wide acknowledgment in contemporary DNA therapeutics. Thus, Gene therapy is considered a promising approach for the treatment of a wide range of diseases such as cancer, AIDS, and neurodegenerative and cardiovascular pathologies and is expected to be of paramount importance in the treatment of genetic disorders. It is accepted today by people working in this field that the main problem to be settled in order to realize a full practice of gene therapy consists on the availability of vectors and/or methodologies able to transport DNA inside the cells efficiently, selectively and safe for patients. For these reasons, many researchers are today trying to synthesize new vectors or at least to optimize the ones already existing. Cationic liposomes are the most studied, although some inherent cytotoxicity that causes negative effects on cells and the low stability of their complexes with plasmid DNA in serum are serious drawbaks and still limit their application. In order to increase this efficiency, the search for more stable and more efficient vectors is therefore of great importance. Neutral liposomes formed by zwitterionic phospholipids DOPC and DOPE are non toxic and more stable in serum but only a limited number of studies have been made so far because of a supposed instability of their complexes with DNA. Our group started few years ago studying neutral liposomes, i.e. nanovectors that we demonstrated in some in vitro experiments being able either to form stable complexes with plasmid DNA in the presence of bivalent metal cations (Ca, Mg, Mn), either to transfect this material to cells. The main project objective is to give insights in the biophysical properties of liposomal gene delivery systems containing new synthetic lipids lacking in positive charge but acting as effective cationic lipids. For this purpose we are studying neutral synthetic vectors containing lipids functionalized with groups able to coordinate bivalent metals and to form stable complex with plasmidic DNA. With the aim to optimize the structure of the chelating agent lipids with different polar heads have been synthesized: Lipids functionalized with crown ethers, Lipids functionalized with polydentate ligand containing nitrogen donor atoms, Anionic lipids derived from malonic acid. All the synthetic amphipatic lipids have been mixed with commercial zwitterionic lipids (DOPC o DOPE) in different percentage and employed in the preparation of multilamellar liposome. Then, these mixed DOPC/DOPE based membranes containing these new synthesized functionalized lipids will be studied by Molecular dynamics simulations in order to elucidate the molecular organization of synthetic neutral lipids into liposomes, their size, and their rigidity influence interactions with cells. All these simulations of the lipid bilayer will be carried at the atomistic level since this is the only way to understand the parameters that influence such an organization. the results will be used to reproduce biophysical experimental data, such as the single lipid molecular areas, and the influence of the salt nature and concentration into the bilayer organization and structure, and others which can deeply influence the”synthetic” membrane capability to organize in liposomes and thus strongly interact with DNA, an exential requisite to be a good genetic vector


Project Title: Large scale molecular dynamics simulations of GPCR proteins in membrane environments: fine tuning the 3D structure for optimal virtual screening campaigns aimed at developing new drugs
Project Leader: Salvatore Guccione, University of Catania, Catania, Italy
Resource Awarded: 660 000 core hours on EPCC – HeCToR XE8

Danilo Milardi – University of Catania, Catania, Italy
Matteo Pappalardo – University of Catania, Catania, Italy
G-protein coupled receptors (GCPR). Membrane proteins are attached to or associated with either the cell membrane or an organelle. Membrane function is mediated in a large extent by integral membrane proteins, which are often organized as assemblies of polypeptide segments interacting with the lipid bilayer and serving as channels, receptors and energy transducers. Accordingly, they constitute biological machines involved in essential cellular processes like ion and molecular transport across the membrane, cell communication and signaling. Therefore, their study is a field of enormous interest. Despite their importance, most of their structural and functional properties still need to be unraveled. In particular, the difficulties in isolating purifying and crystallizing membrane proteins resulted in a extremely low number of available crystal structures of membrane proteins when compared to that of globular proteins. In eukaryotes, many of these proteins are coupled to G proteins and therefore, they are also called G-protein coupled receptors (GPCR). These receptors play a key role in transmission of transduction cell signals responding to hormones and neurotransmitters, regulating basic physiologic- alprocesses, being of great pharmacological interest. The 5-HT7 receptor. 5-HT-7 receptor agonists might be useful for the treatment of pain and the symptoms of pain, especially certain subtypes of pain like neuropathic pain and inflammatory pain and symptoms involving allodynia and hyperalgesia, the prevention or the prophylaxis of pain. Serotonin 5-HT7 receptor antagonists are potential candidates for migraine therapies, sleeping disorders, depression,schizophrenia, anxiety, obsessive compulsive disorders, circadian rhythm disorders, ocular disorders. Details and differences between agonism and antagonism related to 5-HT-7 receptor which share a similar affinity for many ligands and have high sequence homology with the 5HT-1A subtype in the putative binding sites might help to establish a basic foundation for drug discovery aimed at differentiating between these receptors and modulating the pharmacological activity in the antidepressants area by more selective ligands. Molecular Dynamics simulations in explicit membranes. The use of MD simulations provided the atomistic detail necessary for the study of protein motions in membrane. The increasing power of computers in the last years hasprovided the opportunity to consider the lipid bilayer explicitly in the simulations. The results of these studies suggest that the explicit inclusion of lipids is crucial to describing protein dynamics and to provide realistic simulations ranging over several tens or even hundreds of nano-seconds. Modeling GPCRs in their lipidic environment requires a careful selection of a computational protocol to avoid possible artifacts in the simulations.


Project Title: Effect of mutations distant from the active site in the nickase activity of I-CreI and related homing endonucleases
Project Leader: Dr. Francesco Luigi Gervasio and Dr. Antonio Sánchez-Torralba Spanish National Cancer Research Centre, Madrid, Spain
Resource Awarded: 1 100 000 core hours on EPCC – HeCToR XE9

Dr. Antonio Sánchez-Torralba – Spanish National Cancer Research Centre, Madrid, Spain
Dr. David R. Bowler – University College London, London Centre for Nanotechnology, UK
Dr. Tsuyoshi Miyazaki – National Institute for Materials Science (NIMS), Computational Materials Science Unit, Tsukuba, Japan
Meganucleases (or Homing Endonucleases, HEs) are DNA hydrolases that recognize, very specifically, long DNA sequences (14-40 bp) and produce double strand breaks (DSBs) on DNA. Since DSBs lead to DNA repair by homologous recombination with an intact allele, HEs have been used in vivo for therapeutical purposes, for instance in the monogenic disease Xeroderma Pigmentosum, and have strong potential as general tools for gene therapy. Unfortunately, DSBs can also lead to non-homologous end joining (NHEJ) and genome instability. In order to avoid this undesirable side effect, the use of nickases, i.e. ssDNA endonucleases, has been proposed. HEs of the LAGLIDADG family, including I-CreI and I-DmoI, possess two pseudo-symmetric active sites and hence are particularly suitable for nickase design by selectively inactivating one of them. However, before their potential can be fulfilled the determinants of catalysis must be understood and distinguished from those of DNA binding. Recent work shows that several I-CreI mutants present reduced activity without significant loss of DNA-binding affinity, but the reasons why some of the mutations are far from the active site are poorly understood. In this project, we investigate the effect of several mutations on the activity of dimeric I-CreI. We plan to use linear scaling Density Functional Theory (O(N) DFT) to obtain total quantum mechanical (QM) energies of most of the enzyme. The method is suitable to avoid the artificial interfaces found in hybrid approaches. We will find minimized structures for the wild type enzyme, I-CreI mutants, among others K7E, E8K and G19S, and the homologous enzymes I-DmoI and I-DreI (a chimera between I-CreI and I-DmoI). The effect of these mutations will be assessed using the Electron Localization Function (ELF), which characterizes chemical bonding from static pictures of the electronic density matrix. The study of this variety of homologous structures is expected to clarify the allosteric interactions between distant points in the structure. In the second part of the project, we will explore the QM dynamics of I-CreI, in comparison with those of I-DmoI, aiming at characterizing the DNA cleavage reaction. In particular, our preliminary work shows that even thought only two binding sites for metals were assigned to the crystal structure of I-DmoI, as opposed to three in I-CreI, it might still use three ions during in vivo operation. The implication is that the different activity and mechanism must be due to more subtle effects than a simple difference in metal binding.


Project Title: In silico Mechanical Investigation of Glycosaminoglycans
Project Leader: Alfonso Gautieri, Politecnico di Milano, Italy
Resource Awarded: 2 500 000 core hours on BSC – MinoTauro

Annalisa Manenti – Politecnico di Milano, Italy
Andrea Mezzanzanica – Politecnico di Milano, Italy
Federica Rigoldi – Politecnico di Milano, Italy
Simone Vesentini – Politecnico di Milano, Italy
Project objectives. The aim of the present project is apply molecular dynamics (MD) techniques to investigate how the different features of chondroitin sulphate (CS) and dermatan sulphate (DS) drive their mechanical properties and in particular their resistance to compression. Scientific rationale. Elucidating the construction rules of living matter will offer the possibility to create new materials. The strategy envisioned is to fish among the existing natural molecular building blocks for assembling new materials and to understand their “language of shape”. However a real breakthrough requires an understanding of the basic building principles of living matters and a study of their physical properties, to control the form, size, and function of these new systems. Innovation potential. The key impact of the present project is to provide theoretical insights on cartilage tissue mechanical behavior. This will be useful providing design criteria for the development of biomimetic extracellular matrix substituents for cartilage tissue engineering. These new molecule are intended to be used in cell cultures for cartilage regeneration as an effective substituent of the natural proteoglycan present in the healthy cartilage. Indeed, despite recent advances in cell biology and material sciences have contributed to tissue engineering becoming a promising therapeutic modality for the treatment of osteoarticular disorders, available scaffolds are still far from being able to reproduce the in vivo structural and mechanical behavior and thus to generate a tissue that is comparable to native cartilage with respect to quality, stability, and integration. Outcomes and high-impact scientific advances expected. Our computational-aided molecular design approach will provide knowledge-based criteria to rationally design new biomimetic extracellular matrix substituents which will provide structural and mechanical properties similar to those of the naturally occurring proteoglycan. This would eventually be important for cell culture and cartilage regeneration.


Project Title: On the gating mechanism of ligand-gated ion channels
Project Leader: Dr. Marco CECCHINI, Université Louis Pasteur, Strasbourg, France
Resource Awarded: 1 200 000 core hours on UIO – Abel

Ligand-gated ion channels (LGIC) play a central role in intercellular communications in the brain and are involved in fundamental cognitive processes, such as attention, learning and memory. Understanding their function at an atomic level of detail will be beneficial for the development of therapies against severe diseases including autism, schizophrenia, anxiety, Parkinson and Alzheimer. By capitalizing on the increasing availability of high-resolution structures of both pentameric and trimeric LGICs we aim at shedding light onto the ion-gating mechanism by atomistic Molecular Dynamics (MD) simulations with an explicit treatment of the solvent and the membrane environment. We pursue the long-term goal to characterize the functional dynamics of these molecular machines and describe the transition paths between their open and closed forms at high resolution in order to provide a structural basis for allostery and ultimately rationalize the chemical design of potent allosteric modulators.


Project Title: Local Control of Retinal in Rhodopsin
Project Leader: Prof Ursula Roethlisberger, EPFL, Lausanne, Switzerland
Resource Awarded: 9 900 000 core hours on EPCC – ICE-Advance and PDC – Lindgren

Coherent control offers a way of manipulating chemical processes by modulating the amplitudes and phases of the external perturbation applied. Experimentally, this is usually realised using a closed loop-learning algorithm [1,2], and therefore control can often be achieved without a close understanding of the under-lying dynamics involved. The aim of the proposal is to apply local control theory (LCT) [3-5] to simulate the manipulation of the photopigment rhodopsin, a transmembrane receptor that is part of the protein family called G protein-coupled receptors (GPCRs) [6]. Important for this proposal is that this protein plays a central role in vision, for which the early dynamics are dominated by the cis-trans isomerisation of the retinal chromophore [7]. The coherent control of a related sysyte, bacteriorhodopsin, has been investigated experimentally by Prokhorenko et al [8]. They studied the transition of the retinal chromophore from the all-trans to the cis form, under weak field excitation, and demonstrated that by modulating the phases and amplitudes of the spectral components in the photoexcitation pulse, the absolute quantity of the cis form could be enhanced or suppressed by approximately 20%. Simulating such experiments is important to assist in understanding the underlying dynamics of these large and complex systems. Theoretical, control experiments are usually simulated using optimal control theory. Here a pulse is optimized using a learning algorithm yielding an approach, which is conceptually similar to the experimental optimisation procedure. However the computational expense of performing the dynamics multiple times to achieve convergence of the optimum pulse limits its application to small or model systems. Alternatively, we have recently coupled local control theory (LCT) with nonadiabatic ab initio molecular dynamics [3]. In this approach, a target is defined and a control pulse is calculated on-the-fly, depending on only the instantaneous properties of the dynamics, ensuring an increase (or decrease) in the expectation value of the target at each time step. When coupled with nonadiabatic ab initio molecular dynamics this approach obtains the potential, dynamics and the control all from one nuclear dynamics propagation and therefore a full dimensional study of medium to large systems becomes tractable. In this proposal we will apply LCT implemented within the framework of nonadiabatic ab initio molecular dynamics to study the control of the cis-trans photoisomerisation of the chromophore in the visual photoreceptor rhodopsin. The excited state dynamics will be performed using the Tully’s Surface Hopping approach for which the potential of the important region will be calculated on-the-fly within LR-TDDFT [9-13]. The rest of the protein and its membrane environment will be described using a classical force field, using a QM/MM approach [14,15].


Project Title: Describing conformational changes in proteins using simulation and NMR
Project Leader: Kresten Lindorff-Larsen, University of Copenhagen, Denmark
Resource Awarded: 2 080 000 core hours on CSCS – Rosa

Dr. Francesco Luigi Gervasio – Spanish National Cancer Research Centre, Madrid, Spain


Project Title: Identifying Proton Pathways in a Multi-Drug-Resistance Membrane Transporter by EVB-MD Simulations
Project Leader: Dr. Jose Faraldo-Gomez, Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt, Germany
Resource Awarded: 4 650 000 core hours on ICHEC – Fionn-thin

Dr. Claudio Anselmi – Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt, Germany
Prof. K. Martin Pos – University of Frankfurt, Institute of Biochemistry, Germany
Prof. Gregory Voth – University of Chicago, Department of Chemistry, USA
The emergence of multi-drug resistance in pathogenic bacteria is a threat to human health on a global scale. Resistance is in part conferred by an array of proteins in the bacterial membranes whose function is to remove a variety of toxic compounds from the cell interior – for example, man-made antibiotics. Novel strategies to inhibit these drug-efflux systems are therefore of great interest from a biomedical standpoint. To be able to do so, however, much remains to be understood about their molecular mechanisms – and in particular about how drug-transport is energized.

Multi-drug resistance transporters have been identified in several membrane protein subfamilies. For example, the main drug-efflux pump in Escherichia coli, termed AcrB, is a member of the so-called RND family. Like other RND pumps, AcrB is a trimeric protein, and functions through an intricate conformational mechanism that entails remote allosteric effects within each protomer as well as cross-talk between them. Importantly, to power this complex mechanism AcrB harvests the energy stored in the membrane in the form of an electrochemical gradient of H+. That is, AcrB facilitates the downhill passage of H+ across the membrane, through a structural mechanism that is tightly coupled to its pumping activity.

Previous large-scale molecular dynamics simulations in our group, based on a novel high-resolution crystal structure of AcrB, have revealed a number of water channels within its transmembrane domain. Importantly, these channels are redefined in each of the states adopted by the transporter during its conformational cycle, and are consistent with the concept of alternating-access, which is key for tight conformational coupling. However, whether or not these channels are actual proton pathways remain to be demonstrated.

We now plan to elucidate this question, by examining the energetics of H+ passage along each of the water channels identified thus far. To do so, we will calculate the potential of mean force associated with H+ movement, via a systematic series of umbrella-sampling molecular dynamics simulations. To be able to simulate the dynamics of the permeating H+, we will use the most up-to-date implementation of the Multi-State Empirical Valence Bond method, in combination with the CHARMM molecular-mechanics force field.

In sum, the proposed calculations will employ state-of-the-art methods to reveal a level of mechanistic insight not yet attained for any H+-driven membrane transporter, and in particular the multi-drug resistance efflux pump AcrB. These investigations will be carried out in parallel with novel experimental work, both at the functional and structural levels.


Project Title: In-silico screening of new drugs to treat tuberculosis in the dormant phase
Project Leader: Prof. Dr. Marcelo A. Marti, University of Buenos Aires / CONICET, Structural Biology – Biological-Chemistry / INQUIMAE, Buenos Aires, Argentinean
Resource Awarded: 736 000 core hours on CINECA – PLX

Dr. Luciana Capece – German Research School for Simulation Science, ICGEB, Jülich, Germany
Prof. Dr. Dario A. Estrin – University of Buenos Aires / CONICET, Structural Biology – Biological-Chemistry / INQUIMAE, Buenos Aires, Argentinean
Tuberculosis (TB) is one of the most important infectious diseases in the world, for which every year approximately 2 million people die, especially in developing countries. TB treatment presents nowadays several complications, due to the extended length of the conventional treatment and the increasing amount of multi-resistant (MRS) and extreme resistant (XRS) strains. Novel anti-TB drug design is also hampered by the limited knowledge on the bacillus physiology, particularly under nitrosative stress conditions as those found inside the host, where the bacillus enters the dormant or latent state, where conventional drugs are ineffective. Considering that approximately 20% of healthy people harbouring a latent bacillus develop the disease, finding effective drugs against TB during the latency phase, the main aim of this project is of paramount relevance. In previous phase of the project we focused on nitrosative stress related genes using two criteria: I) Relevance criteria, as determined by the increased expression profile of the corresponding protein under stress conditions, as determined by micro-array experiments; and ii) Sensitivity criteria, determined by the likeliness of the target protein to be inactivated by reactive oxygen and nitrogen species. The resulting stress related targets (proteins) druggability was then assessed using recently developed method by our collaborators and the most promising targets are selected. Currently we have already identified five possible targets (while others are still being analyzed): i) N-acetyl-glutamate semialdehyde dehydrogenase (P63562), ii) Uncharacterized protein Q7D606. iii) Possible Mycolic Acid Synthase UMA-A. iv) Cytochrome p450- Cyp121. v) MycothiolA (MshA). Once the protein targets are selected, key point is to perform the Virtual Screening (VS) and further Structure Based Drug Design based on the analysis of predicted protein-drug complex, for this part of the general project, which involves performing the mentioned Molecular Dynamics and high throughput Docking calculations to perform the VS, we request the computational time from PRACE. The requested time is expected to allow the parallel evaluation of the multiple selected drug targets and to perform faster and larger screening of possible compounds. Besides the main goal of obtaining possible new lead compounds a TB, the present project is also expected to allow us to evaluate the performance of several innovations related to the VS and evaluation procedure (See below) that due to its general nature could be translated to other targets. In this context the access to massive computer resources will allow us to increase the performance of control simulations (i.e inclusion in the VS and evaluation protocol of known target-inhibitor pairs) to test and refine the proposed innovations. We expect that the successful outcome of the project will provide i) a general strategy for the performance of massive VS projects and more important ii) a short-list of 10-20 small drug like molecules to be tested in-vitro (and in-vivo) for their potential to inhibit the target enzymes (and if possible mycobacterial growth) to de used as new lead compounds for fighting this disease.

Earth Sciences (1)


Project Title: Ensemble modelling for probabilistic forecast of squall lines.
Project Leader: Dr. Armand ALBERGEL, Université de Toulouse , IRSAMC, Toulouse, France
Resource Awarded: 960 000 core hours on EPCC – HeCToR XE7

Regions near the equator are located in the Inter Tropical Convergence Zone (ITCZ) and are susceptible to squall lines, which are special convective weather phenomena accompanied by strong, highly varying winds and intense precipitation. Squall lines can cause severe damage to property and structures, affect navigation and offshore operations, as well as posing a threat to human life. The essence of this project is to explore the feasibility of implementing a probabilistic forecast system based on a regional ensemble modelling system and state of the art data assimilation techniques that use satellite observations. This approach ties wind, temperature and moisture observations to a stochastic ensemble modelling system for short term (0~24hrs) squall line probabilistic forecasts. The proposed data assimilation system should improve any individual forecast, and recent studies strongly suggest that an ensemble prediction of extreme weather such as hurricanes performs better than traditional deterministic ones. In addition, the ensemble also provides valuable uncertainty and risk information to operations managers and for use in decision aid systems. The project will focus on the Gulf of Western Africa. The community Weather Research and Forecasting model (WRF: with its satellite data assimilation capability (GSI: will be used.

Engineering (2)


Project Title: Massively parallel implementation of Total-FETI algorithms
Project Leader: Dr. David Horak, Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
Resource Awarded: 2 656 250 core hours on EPCC – HeCToR XE11

Vaclav Hapla – Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
Martin Mensik – Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
Michal Merta – Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
FETI methods are very successful for the solution of large scale engineering problems. The reason is that duality reduces large primal problem to smaller dual, relatively well conditioned strictly convex iteratively solved quadratic programming problem. Total-FETI-1 (TFETI-1) simplifies the inversion of stiffness matrices of subdomains by using Lagrange multipliers not only for gluing the subdomains along the auxiliary interfaces, but also for the enforcement of the Dirichlet boundary conditions. Our research deals with parallel implementations of TFETI-1 algorithm for problems in non-linear and contact mechanics using PETSc and Trilinos frameworks. The main goal of our research is to develop new generation of algorithms for efficient solution of very large and complex problems in engineering on upcoming multi-petascale systems. Current experiments show that our codes scale up to thousands of cores and the main goal of the proposed project is their performance analysis, comparison and optimization on large scale parallel architectures to further improve the scalability up to tens of thousands of cores and hundreds of millions of unknowns. The performance of the codes will be tested on the solution of huge real world or model problems. For the performance analysis and optimization we are going to use some established performance analysis tools like Scalasca. Some of the bottlenecks of the TFETI-1 codes are already known. The natural effort to reduce the computational and memory requirements for subdomain problems’ solution decomposing domain into large number of subdomains (thus reducing the local primal dimension) leads to increase of the dual and null space dimension. This affects the time needed for the coarse problem solution and natural coarse space matrix-vector multiplication appearing in the application of orthogonal projectors, the reconstruction of the primal solution from the dual one etc. However there is a lot of space for an optimization. For example the coarse problem can be solved in many ways – directly using the Cholesky factorization, iteratively using the conjugate gradient method, using the explicit inverse computation, or this problem can be even eliminated through the natural coarse space matrix orthonormalization using Gram-Schmidt process. All these approaches can be performed either sequentially by master process, or in parallel on each of processes. Thus the objectives of the proposed project are to analyze an impact of various data distribution on the algorithm performance, to implement various parallelization of TFETI-1 coarse problem solution and to compare these approaches as well as to compare the implementations in PETSc and Trilinos frameworks. These frameworks should be good choice for interfacing of implemented codes to existing engineering codes for their application to solution of real world problems.


Project Title: High end computational modelling of wave energy converters
Project Leader: Prof. Frederic Dias, University College Dublin, Dublin, Ireland
Resource Awarded: 1 620 000 core hours on CSCS – Rosa and NCSA – EA ECNIS

Dr. Ashkan Rafiee – University College Dublin, Dublin, Ireland
It has been known that bottom hinged Oscillating Wave Surge Converters (OWSCs) are an efficient way of extracting power from ocean waves [1]. OSWCs are in general large buoyant flaps, hinged at the bottom of ocean and oscillating back and forth under the action of incoming incident waves [2,3]. The oscillating motion is converted into energy by pumping high-pressure water to drive a hydro–electric turbine [4]. This project deals with numerical studies of wave loading and wave impact on an OWSC using a three dimensional two-phase Smoothed Particle Hydrodynamics (SPH) code. In the implemented SPH approach, accurate estimation of impact loads on an OWSC is achieved by utilizing a particle pair-wise solution of the Riemann problem. Furthermore, to capture the turbulent features of the flow the Lagrangian form of the RANS k − ε model is included in the SPH equations. In order to better understand the reliability of OWSCs as an efficient way of harnessing wave energy, the SPH simulations will be performed in both two–dimension and three–dimension and results will be compared for flow pattern and impact loads. Besides, the simulations will be per- formed at two different scales, namely full scale and model scale, and results from excitation torque, wave loads and efficiency of OWSC will be compared in order to understand the scaling effects. Furthermore, a two-dimensional two-phase compressible SPH code coupled with an in- compressible potential flow solver (FSID code [5]) will be used to model wave impact on a rigid fixed and moving wall. This will shed lights on the effect of entrapped air pocket on the impact pressure during the wave impact.

Materials Science (11)


Project Title: Ab initio MD simulation for design of polyaromatic framework for hydrogen storage
Resource Awarded: 231 000 core hours on EPCC – ICE-Advance

The biggest problem facing the practical use of hydrogen in fuel-cell powered automobiles is its storage. Hydrogen can be stored in many ways: in gas cylinders, in cryogenic tanks as a liquid or in solid materials – in the form of metal hydrides or adsorbed on high surface area sorbents. The first two methods are conventional technologies with several limitations the most important of which is their low energy efficiencies. Nowadays most of the scientific studies are focused on high surface area sorbents in which hydrogen can be stored by physisorption. The relative weakness of this kind of physical interactions between sorbent and hydrogen can provide reversibility of the process, which is necessary for mobile applications. The goal in hydrogen storage is gravimetric capacity of more than 6% by weight at ambient temperature and pressure, which can be estimated by the surface area and binding energy to hydrogen. For the past several years carbon has been considered as a potential hydrogen storage material for being light and able to form high surface area nanomaterials. Unfortunately, a lot of theoretical and experimental studies indicated that storage capacity is very low and limited in all forms of carbon nanostructures. This is due to the usually very low binding energies to hydrogen. There are two different approaches for increasing the chemical activity of carbon structures. One is substitution of heteroatom in carbon materials and the other is using metal doping. Both of these techniques aim to cause polarization in hydrogen molecules. H2 are moderately polarizable and a dipole moment can be induced in the presence of other charges. Polarization of the hydrogen molecule leads to occurrence of electrostatic charge-quadrupole and charge-induced dipole interactions and hence stronger binding. Searching for a suitable material for hydrogen storage we used both of the techniques for increasing of enthalpy of adsorption and modelled a system in which two of the carbon atoms in the benzene ring were substituted with boron (Li2C4H6) and added two Li atoms preserving the aromatic nature of the system. Our quantum chemical calculations showed relatively strong binding of H2 molecules to the Li atoms. Here we suggest a modifyed known porous material (PAF-1) suitable for hydrogen storage. It has significant advantage over recently investigated MOF and COF systems, as it is light and has large surface area. The modeled system will be stable on hydrogenation and aggregation of Li atoms. Furthermore, there are no ab-initio calculations for such systems in the literature. That is why we propose performing first principle calculations on the periodic 3D structure of the tunned PAF-1in order to obtain more reliable and actual understanding on the adsorption of hydrogen on porous structures and the stability of the proposed material at different temperatures and pressures.


Project Title: AuPd-Seg
Project Leader: Dr. Hazar Guesmi, Université Pierre et Marie Curie, Paris, France
Resource Awarded: 640 000 core hours on RZG – Hydra

The knowledge of the composition and surface structure of nanoalloy particles is crucial to explain their catalytic performance. In addition, the bonding of adsorbates may, in some cases, induce modifications in local atomic composition and surface structure, changing the activity and selectivity of the catalyst. These facts were observed for Au-Pd nanoparticles. Indeed, in the presence of reactive gas (for example CO and O2) the segregation of the Pd to the surface was reported.
Because the extent to which segregation occurs can have implications for the performance and the lifetime of the catalyst, it is important to understand whether the particular configuration is stable under the operating environment for a specific application. For this purpose the present project aims to study the structure, chemical order and the reactivity of Au-Pd surface nanoparticles under vacuum and in the presence of adsorbates.
Theoretical studies of catalytic properties are often investigated on model systems such as surfaces or model aggregates and rarely on “realistic” systems similar to those observed experimentally. In addition, one feature of the vast majority of these studies is that no account is taken for the possibility that the surface composition can be modified after the gas exposure. These are serious drawbacks that may prevent reliable description of the nanoparticle reactivity who mainly depend on the structure configuration of the surface. In order to address this challenging task, we intend to combine original theoretical approaches (DFT based Ising model and semi-empirical based simulations) to model Au-Pd nanoparticles (FCC based symmetry cluster and icosahedron) and to study segregation and transition state properties under vacuum and reaction conditions. In addition DFT cluster optimisations (taking in to account adsorbate-induced surface change) will be devoted to the study of the reactivity of these nanoalloys in CO oxidation reaction and the comparison to the experimental data. This project will be developed in collaboration with physicists and chemists experts which are very active in the domain of nanoparticles and nanoalloys.


Project Title: Computational Studies of Advanced Functional Materials
Project Leader: Olle Eriksson, Uppsala University, Department of Physics and Materials Science, Sweden
Resource Awarded: 9 687 608 core hours on RZG – Hydra and UIO – Abel

Dr. Biplab Sanyal – Uppsala University, Department of Physics and Astronomy, Uppsala, Sweden
This proposal deals with state-of-the-art computational materials science methods applied to advanced functional materials. We aim to study the following topics, (i) correlated electron system, (ii) graphene and molecular-interface systems, (iii) lattice dynamics and (iv) core shell structures and nanoparticles. All these topics will be studied by ab initio density functional theory based methods implemented in codes like VASP and SIESTA. For some applications, in-house codes such as SCAILD for lattice dynamics, Rspt for dynamical mean field theory will be used.


Project Title: Diffusion in multicomponent nitrides and vibrational thermodynamics from first-principles
Project Leader: Prof. Igor Abrikosov, Linköping University, Department of Physics, Chemistry and Biology (IFM), Linköping, Sweden
Resource Awarded: 6 250 000 core hours on EPCC – ICE-Advance and PDC – Lindgren

Dr. Björn Alling – Linköping University, Department of Physics, Chemistry and Biology (IFM), Linköping, Sweden
Prof. Leonid Dubrovinsky – Universität Bayreuth, Bayerisches Geoinstitut, Bayreuth, Germany
Prof. Lars Hultman – Linköping University, Department of Physics, Chemistry and Biology (IFM), Linköping, Sweden
Transition metal nitrides (TMN), such as TiN and ZrN, are important technological materials owing to their outstanding mechanical, electrical, and corrosive-resistant properties, in combination with applicability in industrial-scale thin film deposition systems. Inline with our application from last year we are using PRACE allocation to conduct pioneering theoretical work on the diffusion of adatoms on different crystal surfaces of TiAlN. In continuation of this project we are aiming to study aspects of diffusion in nitrides, focusing on TiAlN, ZrAlN and HfAlN. Due to complexity of configurational disorder all our tasks require a huge amounts of computation resources. To investigate atomic diffusion in these materials we are using electronic structure codes to calculate the energy barriers needed to be overcome by diffusing species, both inside bulk materials and on top of crystal surfaces. The second part of our project related to Earth’s core structure and is based on the new technique we developed in our group to study vibrational thermodynamics from first-principles at high pressure and high temperatures. The project targets completely disordered alloy phases of Fe-Ni and Fe-Ni-Si using special quasirandom structure method in first-principles simulation. Although the technique is developed from the theoretical point of view, the code of realisation is yet to be finished and we are counting on help of PRACE experts in order to obtain greater scalability in parallelization of our codes.


Project Title: Efficient Screening Methods for Self-assembled Monolayers in Organic Electronic Devices
Project Leader: Dr. Robert Send, BASF SE, Quantum Chemistry Group, Ludwigshafen, Germany
Resource Awarded: 3 000 000 core hours on EPCC – HeCToR XE6

Dr. Manuel Hamburger – Universität Heidelberg, Organisch-Chemisches Institut, Heidelberg, Germany
Self-assembled monolayers (SAMs) are organic molecules chemically attached to an inorganic substrate that arrange in two-dimensional periodic patterns. These monolayers are mechanically and chemically stable and can be used to adjust several surface properties like surface-tension, wettability or most importantly, the energy necessary to inject or extract charges from the inorganic substrate: the workfunction.

The necessity to understand and develop SAMs comes along with the development of organic electronic devices such as organic light emitting devices, organic photovoltaics and organic field effect transistors. These devices make use of the unique properties of organic materials such as transparency, flexibility and simple processability, but all these devices are used within conventional inorganic electronic frameworks. Thus all organic electronic devices have organic-inorganic interfaces.

SAMs are a versatile tool to design these organic-inorganic interfaces and allow adjusting the workfunction of the inorganic material to ensure efficient charge flow through the interface. They therefore have the potential to become a key building block in organic electronic devices. For well-directed synthesis of SAMs, computational screening methods are necessary in order to search for organic molecules with given properties in the SAM. The prediction of SAM-properties is possible with reasonable accuracy using plane-wave density functional theory calculations. However, the costs of these methods are too high to allow systematic in silico searches for new materials.

Within this project, we will develop a computational protocol that reduces the costs for property prediction of SAMs by evaluating the minimal accuracy required for each property. We will further investigate the possibility to use non-periodic methods for improved efficiency. Our aim is to develop a validated screening protocol that makes standard procedures more efficient, permits prescreening and allows assessing an order of magnitude more molecules than currently possible.


Project Title: Hydrogen storage materials for energy applications
Project Leader: Prof. Rajeev Ahuja, Uppsala University, Department of Physics and Astronomy, Uppsala, Sweden
Resource Awarded: 2 501 125 core hours on UHEM – Karadeniz

Dr. Carlos Moyses Araujo – Yale University, Department of Chemistry, New Heaven, USA
Dr. Börje Johansson – KTH, Applied Materials Physics, Stockholm, Sweden
Dr. Biswarup Pathak – IIT Indore, Department of Chemistry, Indore, India
Our energy-hungry world has become increasingly depending on new methods to store and convert energy for new, environmentally friendly modes of transportation and electrical energy generation as well as for portable electronics . Mobility — the transport of people and goods — is a socioeconomic reality that will surely increase in the coming years. It should be safe, economic and reasonably clean. Little energy needs to be expended to overcome potential energy changes, but a great deal is lost through friction (for cars about 10 kWh per 100 km) and low-efficiency energy conversion. Vehicles can be run either by connecting them to a continuous supply of energy or by storing energy on board. Hydrogen would be ideal as a synthetic fuel because it is lightweight, highly abundant and its oxidation product (water) is environmentally benign. However, the storage remains a problem for this highly desirable development. Present proposal deals with studying the interaction of hydrogen with novel materials having multiple length scales such as clusters, nano-particles, nano-tubes, multi-layers, and crystalline bulk. The materials include light metal hydrides such as alkali-alanates and boro-hydrides and Metal organic frameworks. Due to the light weight of these materials, the gravimetric density of hydrogen is higher than that in the inter-metallic hydrides. Although these materials are regarded as potential candidates for a new generation of hydrogen storage materials and are critical to a new hydrogen economy, very little fundamental understanding is available about the nature and strength of hydrogen bonding, the influence of catalysts on the uptake and release of hydrogen and the effect of nanostructuring on the thermodynamics of hydrogen. This proposal is aimed at providing this fundamental understanding by carrying out first principles calculations based on density functional theory. We seek answers to the following questions: (1) What is the preferential site of hydrogen? (2) Does hydrogen bind atomically or molecularly? What is the nature of hydrogen bonding – ionic, covalent, metallic, or weak van der Waals? (3) What roles do surface morphology and defects play in hydrogen absorption and desorption? (5) In what way is it beneficial to store hydrogen in nanostructured and porous materials? (6) How do catalysts help in improving the thermodynamics of hydrogen absorption and desorption?


Project Title: Lattice Boltzmann Simulation of Soft Condensed Matter
Project Leader: Dr Oliver Henrich, University College London, Centre for Computational Science, UK
Resource Awarded: 3 200 000 core hours on CSC – Louhi XT, CSC – Sisu and EPCC – HeCToR XE10

Prof, FRS Michael Cates – University of Edinburgh, School of Physics, UK
Prof. Peter V. Coveney – University College London, Centre for Computational Science, UK
Dr Davide Marenduzzo – University of Edinburgh, School of Physics, UK
Ignacio Pagonabarraga – Universidad de Barcelona, Spain
Kevin Stratford – University of Edinburgh, EPCC, Edinburgh, UK
Soft matter systems of biological, medical or technological relevance are part of our every day world. Liquid crystals (LCs) for example are states of soft matter in which anisotropic molecules are locally fluid but have short- or long-range orientational or positional order. They have caused a major technological revolution, are now widely used in a large range of displays and photonic devices and become increasingly important for functional materials. Also, many naturally occurring substances such as DNA, chitin, cellulose and collagen show liquid crystalline behaviour and exhibit a hierarchy of self-assembled structures, which is typical for multiscale materials. Therefore LCs can be regarded as models of the precise morphogenetic processes in biological systems. Blue phases (BPs) are a specific class of equilibrium phases in thermotropic LCs close to the isotropic-chiral nematic transition. BPs have recently been stabilised over a wide temperature range including room temperature and have huge potential for possible future applications. They may be used in a range of photonic switchgear and tunable switches for light, photonic integrated circuits and novel designs for displays. Dispersions of colloids in LC show a multitude of self-assembled phases and are most interesting candidates for new nanomaterials such as colloidal crystals and photonic crystals. All these systems have complex equilibrium and flow properties in common, particularly when electromagnetic as well as mechanical forces are acting. Flow-structure interactions are crucial in applications but can rarely be predicted in advance. Physical processes occur across multiple length scales and need to be accurately resolved: these include hydrodynamic interactions, which play a dominant role in all non-equilibrium processes and strongly alter relaxation times close to steady states. In order to realise the potential of possible technological and biomedical applications our understanding of these systems needs to be fundamental and detailed. High-performance computing enabled large-scale simulations constitute an important pillar for current research on this subject as they can go beyond the reach of analytical theories and provide valuable guidance for experimentalists. We propose to make use of state-of-the-art computing technologies to study kinetic aspects as well as equilibrium and non-equilibrium behaviour of liquid crystals and colloid-liquid crystal-dispersions. We plan to investigate how colloids behave dynamically in blue phases, how colloidal crystals form by using the defect topology of blue phases as a template structure. A question of particular relevance is, how the growth of these structures depends on thermodynamic parameters and how it can be directed by external fields. We plan to study the rheology of pure blue phases, colloid-blue phase and colloid-liquid crystal suspensions and the influence of external electric fields on their flow behaviour. We plan to investigate how charges influence the phase behaviour of binary fluids and colloidal suspensions and how polyelectrolytes change the order structure of liquid crystals. These projects are embedded into a wider collaboration between the Centre of Computational Science at University College London (UCL), the School of Physics and Astronomy (SPA) and Edinburgh Parallel Computing Centre (EPCC), both University of Edinburgh (UE) and the Departament de Fisica Fonamental, Universitat de Barcelona (UB).


Project Title: Molecular clusters – non zero pressure Monte Carlo simulations
Project Leader: Dr. Ales Vitek, Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
Resource Awarded: 2 352 000 core hours on UHEM – Karadeniz and UIO – Abel

Assoc. Prof. Dr. Rene Kalus – Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
Martin Stachon – Vysoka Skola Banska – Technical University of Ostrava, Centre of Excellence IT4Innovations, Ostrava, Czech Republic
The project is focused on the modelling of thermodynamic and structural properties of small clusters consisting of several through tens of water molecules. More specifically, classical Monte Carlo methods enhanced with the parallel-tempering scheme and two-dimensional multiple histograms will be used to calculate enthalpies, heat capacities, and selected structural parameters of water clusters containing 17 – 22 molecules (a size region within which a transition from the all-molecules-on-the-surface structures to one-molecule-in-center structures takes place) with the main emphasis on the phase changes in this clusters. A full pressure-temperature phase diagrams will be obtained. In addition, selected larger clusters (up to 50 molecules) will also be studied to get a deeper insight into temperature and pressure induced structural transformations and their evolution with cluster size. In last two decades, a lot of theoretical studies on thermodynamic properties have been published. However, the studies almost exclusively focus on constant volume – constant energy (NVE) or constant volume – constant temperature (NVT) calculations, and only a few papers describe molecular clusters at non-zero pressures. The main goal of the present project is to fill this gap. For the modelling of water clusters under non-zero pressures, we will use the parallel tempering Monte Carlo algorithm, originally developed for the NVT and NVE statistical ensembles and later modified for the NPT ensemble. A classical thermodynamic Monte Carlo simulation performed for a given temperature/energy and volume/pressure employs Markov’s chains and is, thus, inherently a serial process. Calculations performed for different temperatures/energies and volumes/pressures can be, on the other hand, straightforwardly parallelized. For example, a parallel tempering approach has been recently proposed for the NVT ensemble, which significantly accelerates convergence by simulating a bunch of systems at different temperatures in parallel and occasionally exchanging information (configurations) between different systems. Later on, parallel tempering approach has been extended to the NPT ensemble. In that case, hundreds of systems are simulated at different temperatures and pressures in parallel and exchanges of configurations and volumes between randomly selected systems take place periodically. This computationally demanding algorithm is predestined for an MPI parallelization when, ideally, each system is simulated on its own core. As a result, one obtains the values of measured thermodynamic parameters for all the temperatures and pressures included at once. In addition, thermodynamic properties depending on cluster configurations through the internal energy and volume (like the enthalpy or the heat capacity) will be calculated using multiple histogram methods, originally developed for the NVT ensemble and recently extended to the NPT ensemble in our group. Two dimensional, energy-volume histograms pre-calculated for each simulated system form an input to this algorithm and, as a result, classical two-dimensional density of states is obtained as a function of volume and energy. From the density of states, one can simply obtain the value of the thermodynamic quantity at an arbitrary temperature and pressure by a computationally cheap two-dimensional integration. If this integration is repeated for a sufficiently dense grid of temperatures and pressures, one obtains a smooth dependence of the calculated parameter on the temperature and pressure.


Project Title: Searching for new ceria oxidation catalysts: exploiting synergies between surface doping and surface modification
Project Leader: Dr. Michael Nolan, Tyndall National Institute, Cork, Ireland
Resource Awarded: 1 500 000 core hours on FZJ – JuRoPA

Dr Ganduglia-Pirovano Maria – Centro Superior de Investigaciones Científicas (CSIC), Spain
Cerium dioxide is a well studied and leading material in oxidation catalysis, e.g. for CO oxidation or oxidative dehydrogenation of alcohols. Key to this application of CeO2 is the relative ease with which it undergoes a reduction-oxidation cycle. Reduction can take place by removal of oxygen by a CO molecule to form CO2, with the missing lattice oxygen replenished by reaction with atmospheric O2. The reduction of Ce4+ to Ce3+ is relatively facile and there have been numerous studies examining the fundamentals of this process and how the reactivity of ceria can be tuned. This can be done in two ways (1) substituting another metal species at a Ce site, e.g. Ti, Zr (“doping”) and (2) modifying the surface with molecular sized metal oxide structures, e.g. VO2, where ceria acts as an active support. The applicant and collaborator have been active in studying these means of modifying the reactivity of ceria surfaces by doping and addition of VO2 to the (111) surface. In this project, we will study the synergy of these approaches to modify ceria to design new oxidation catalysts based on a rational design approach using large scale simulations. We will study VO2 at the ceria (111) and (110) surfaces, in which the surfaces have been modified with metal atom dopants that are known to improve the reactivity of the bare surfaces. This will allow us to develop new insights into the factors that drive the reactivity of modified ceria through a synergy between doping and surface modification. Simulations of oxidative dehydrogenation of methanol will be used to select new ceria materials that will be predicted to be improved oxidation catalysts for further investigation in experiments.


Project Title: The Influence of the interface between amorphous and crystalline silicon in core-shell nanowires on photovoltaic properties.
Project Leader: Dr. Michael Nolan, Tyndall National Institute, Cork, Ireland
Resource Awarded: 2 488 320 core hours on CINES – JADE-Harpertown

Dr. Fagas Giorgos – Tyndall National Institute, Cork, Ireland
The Si-Interfaces project will undertake first principles density functional theory (DFT) simulations of the technologically important interface between amorphous and crystalline silicon (aSi-cSi) in model core-shell nanowire structures. These structures are of great interest as leading candidates for third generation silicon photovoltaics (PV), whereby the light absorption and charge transport properties can be engineered by composition and structure to tune their response to incident light. The project will apply a heat and quench simulation approach, in which crystalline Si nanowires of 5 nm diameter are heated to above their melting temperature for different times and then quenched to 0 K to produce a region of amorphous silicon interfaced with crystalline Si, i.e. an aSi-shell-cSi-core. The structures resulting from this produce different thickness aSi, depending on the initial melting time and the crystal face and these will be relaxed with DFT. Addition of hydrogen passivates the aSi region. The electronic and optical properties of the aSi-cSi nanowires will be determined, in particular the valence and conduction band offsets and charge transport properties will be studied. The results of these studies will be used in collaboration with experiment to allow a deep understanding of how the structure of aSi-cSi core shell nanowires determines the optical properties, which will be crucial for developing high efficiency third generation PV devices.


Project Title: Simulation of Polymer Semi-Conductors
Project Leader: Prof. Vlasis Mavrantzas, University of Patras, Department of Chemical Engineering, Rio, Patras, Greece
Resource Awarded: 7 200 000 core hours on FZJ – JuRoPA

Dr. Orestis Alexiadis – University of Patras, Department of Chemical Engineering, Rio, Patras, Greece
Dr. Alexandros Anastasiou – University of Patras, Department of Chemical Engineering, Rio, Patras, Greece
Dimitris Tsalikis – University of Patras, Department of Chemical Engineering, Rio, Patras, Greece
Materials which exhibit ordered morphology at the nano-scale have drawn considerable attention in the last two decades due to their unique combination of optoelectronic properties, ease of preparation and low cost manufacturing. Systems like semiconducting polymers (e.g., semi-crystalline poly-thiophenes), micro-crystalline silicon and graphene have proved to be very promising candidates for a variety of applications like organic micro-electronics and photovoltaics.

Simulating these materials at the nanoscale is inefficient with Molecular Dynamic (MD) methods because of the problem of long relaxation times, especially when highly ordered structures form at low enough temperatures. In order to circumvent the MD drawbacks, it’s imperative to develop and implement efficient Monte Carlo (MC) techniques (Metropolis MC and Kinetic MC) to effectively simulate systems with large scale nanophase-separated structures and overcome obstacles related with large system sizes and sluggish dynamics. This presents a considerable problem when one wishes to simulate the system over a wide range of temperatures and especially at the lower ones where the crystalline nano-domains are formed, since for the simulation to be ergodic, the MC moves should be characterized by sufficiently high acceptance rates. To deal with such a problem, we resort to parallel versions of the MC technique.

We propose here to design two different editions of such a parallel MC algorithm and a novel kMC algorithm in order to simulate two very important families of systems today: polymer semiconductors based on alkylthiophenes such as poly(3- Hexylthiophene) (P3HT) and micro-crystalline silicon thin films. The interest in these materials stems from the demand nowadays for low-cost, large-area semiconducting devices for displays and photovoltaic applications which is also reflected in the value of the market for devices for use in microelectronics; according to the European Semiconductor Industry Association,, this exceeds 12 billion € today.


Plasma & Particle Physics (2)


Project Title: Landau gauge propagators and vertices in Lattice QCD
Project Leader: Dr. Orlando Oliveira, Universidade de Coimbra, Center for Computational Physics, Coimbra, Portugal
Resource Awarded: 5 265 000 core hours on PDC – Lindgren

Dr. Paulo Silva – Universidade de Coimbra, Center for Computational Physics, Coimbra, Portugal
This computational project is a step towards achieving a better understanding of the infrared sector of Quantum Chromodynamics (QCD). We will focus on the QCD fundamental propagators and vertices in Landau gauge via lattice QCD simulations. This requires the use of HPC facilities. One of the main goals is to be able to produce a few quenched and dynamical ensembles in sufficiently large lattices, with a physical volume above 10 fm per side, and to gauge fix the configurations to the Landau gauge. The use of large lattice volumes is crucial to control the finite volume effects. Moreover, the gauge fixing process is a very demanding numerical problem. After the generation of the ensembles, one can proceed and study various QCD Green’s functions, like the quark and gluon propagators, three and four gluon vertices, and the quark-gluon vertex. Hopefully, we will able to contribute for a better understanding of non-perturbative features of QCD, like confinement and dynamical chiral symmetry breaking.


Project Title: Non-perturbative Renormalisation for Hadron Physics NPR-LQCD
Project Leader: Dr. Mariane Brinet, LPSC, IN2P3, LPSC, Grenoble, France
Resource Awarded: 886 080 core hours on CINECA – PLX

Abstract of the project: If the project is successful this will be published on the PRACE website unless you mark it as confidential below. Please make this summary understandable to a general audience. (500 words) Quantum Chromodynamics is the theory of strong interaction, whose ambition is to explain nuclei cohesion as well as neutron and proton structure, i.e. most of the visible matter in the Universe. Its application domain is even wider, since QCD controls the structure and interactions of all hadrons: proton, neutron, hyperons, pions, kaons,…It is one of the most elegant theory of Science History (with General Relativity); it has only very few parameters and allows to give a physical interpretation to a very broad range of phenomena using a well defined and very compact formalism. The only systematic and rigorous method to solve this theory is Lattice QCD (LQCD) — whose principle is inspired by Statistical Physics. LQCD aims at providing solutions of this fundamental theory of matter, without uncontrolled hypothesis, and with accuracies which rival that of experimental data. LQCD procedure consists in discretizing space-time on a 4 dimensional grid with a typical length of 3- 5 fermi and lattice spacing of the order of 0.1 fermi. Gluons are represented by SU(3) matrices associated to each link of the lattice and quarks propagate from site to site. Calculations are based on stochastic methods and are very demanding in terms of computing time when they take into account virtual loops of light quarks. Those quantum fluctuations are crucial to respect the fundamental symmetries of the theory but the computing time needed to include them increases when quark masses decrease. And Nature contains two light quarks (u and d), which essentially constitute protons and neutrons. Discretizing QCD on a space-time lattice leaves a large freedom in the choice of discretized action, the only constraint being that its continuum limit must lead to QCD action. QCD simulations are being performed by several international collaborations, each using a different choice of discretization. Among those lattice collaborations, the European Twisted Mass Collaboration (ETMC) is particularly active. To date, this is the only collaboration having generated configurations with four dynamical quarks in the sea (u, d, s, c). Understanding hadron structure from first principles is a fundamental goal of LQCD. Light hadron dynamics has indeed always been crucial in understanding strong, as well as weak, interactions. Many important questions concerning Standard Model parameters, as well as special non perturbative QCD features, can find their answer in low energy hadron dynamics study. However, while LQCD has become a precision technique for many QCD observables, the calculation of nucleon matrix elements remains an open challenge. It involves in particular a careful estimate of the corresponding renormalization constants, essential to compare lattice results to values deduced from experiments. The possibility to perform a proper non-perturbative renormalization is an essential feature of lattice calculations. QCD discretization on a space-time lattice provides indeed an obvious regularization of the theory, by introducing the lattice spacing as a natural cut-off. However, any comparison with physical results requires a precise control of the continuum limit. Renormalisation allows, from bare quantities computed at finite lattice spacing, to obtain physical observables with the accuracy sought (typically of the percent level). Our project goal is to compute renormalization constants for non local fermionic bilinear operators with four dynamical quarks in the sea (Nf=4), involved in the computation of matrix elements used to extract nucleon structure functions, in which we are specially interested. A proper comparison of these matrix elements with experimental values represents both a challenge and an opportunity for lattice QCD.


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