PRACE Preparatory Access – 37th cut-off evaluation in June 2019

Find below the results of the 37th cut-off evaluation of 3 June 2019 for the PRACE Preparatory Access.

Projects from the following research areas:

 

Effect of chemical skin penetration enhancers on the permeability of pharmaceutical ingredients

Project Name: Effect of chemical skin penetration enhancers on the permeability of pharmaceutical ingredients
Project leader: Dr Christian Wennberg
Research field: Biochemistry, Bioinformatics and Life sciences
Resource awarded: 100 000 core hours on Piz Daint
Description

The main barrier structure of the human skin resides in the intercellular lipid matrix of the skin’s outermost layer, the stratum corneum. Accurate skin permeation modeling requires a good understanding of the partitioning into the stratum corneum barrier structure by all the components present in topically applied formulations. An accurate model of the skin’s barrier structure, under influence of e.g. permeation enhancers, enables computational screening of possible formulation alternatives in order to enhance the delivery rate of existing drug formulations, and should be of vital importance during the development of new transdermal delivery methods for existing drugs.

top

Direct numerical simulation of the aeroacoustic sound produced by energy systems: scalability tests

Project Name: Direct numerical simulation of the aeroacoustic sound produced by energy systems: scalability tests
Project leader: Dr Valerio D’Alessandro
Research field: Engineering
Resource awarded: 100 000 core hours on MARCONI-KNL, 50 000 core hours on MareNostrum
Description

The long term goal of this project is to obtain a regular PRACE access in order to study the airfoil self-noise mitigation techniques for wind turbines applications. It is worth noting that the aeroacoustic noise radiated wind turbine blades, is a hot topic in the contemporary scientific literature. Indeed, the reduction of noise emission from most wind turbines mechanical and electrical components is a well-developed aspect, the reduction of the aerodynamic noise show several lacks of knowledge. The latter, which is predominantly generated by the rotating blades, is one of the main causes of annoyance for people living nearby wind farms, particularly in rural or suburban areas. The problem is also associated to small wind turbines, because the aerodynamic noise emission is related to a higher rotational speed. For this reason, within this wide research project, we have developed an open–source solver for compressible Navier–Stokes equations. Low-storage explicit Runge–Kutta (RK) schemes were adopted for time integration; on the other hand, in a finite-volume framewrok, Kurganov-Noelle-Petrova central-upwind scheme was used for convective terms and central rules for diffusive ones. Moreover a sponge-layer type non-reflective boundary treatment was adopted to avoid spurious numerical reflections at the far-field boundaries. Such techniques were selected and tested in order to allow the possibility of solving a broad range of physical phenomena with particular emphasis to aeroacoustic problems. The importance of having similar open-source solvers is to enable general public to perform similar computations. Indeed, numerical simulations of acoustic wave propagation phenomena are mostly limited to academic codes. The reliability, efficiency and robustness the of the solver was already proved by computing several benchmarks concerning far–field aerodynamic sound. We want to remark that, in some sense, the adoption of explicit time integration techniques can be considered a questionable choice but we have opted for this approach because it guarantees very good scalability performance. Moreover, after careful tests of several RK-approaches, we have found a good stability limit which corresponds to a maximum (in the whole domain) acoustic Courant number lesser than one. Preliminary tests on small Linux cluster have given very satisfactory results in terms of scalability. Hence with this preparatory PRACE project we want to evaluate our approach on cutting edge supercomputers in order to be ready to apply for the future PRACE regular calls in the next years.

top

Predicting Polymorphic Molecular Crystals with a Parallel Crystal Structure Search

Project Name: Predicting Polymorphic Molecular Crystals with a Parallel Crystal Structure Search
Project leader: Dr Adem Tekin
Research field: Chemical Sciences and Materials
Resource awarded: 50 000 core hours on MareNostrum, 50 000 core hours on Curie, 100 000 core hours on Hazel Hen, 20 000 core hours on JUWELS
Description

Predicting crystal structure of a material is one of the most important problems in computational material science. There are several computational methods (including simulated annealing, evolutionary algorithms, distributed multipole analysis, random sampling, basin hopping) that developed for solving crystal structure prediction (CSP) problems. Due to the huge number of local minima, employing global optimization tools for CSP is mandatory. In general, the objective function used in these global optimizations is the total energy of the crystal calculated by the density functional theory (DFT). Although this approach is very accurate, unfortunately it is not suitable for unitcells containing too many atoms. Because, in global optimizations, the objective function must be evaluated tremendously. Considering the fact that DFT calculations are very time consuming even if with a coarse setting, employing DFT energies as an objective function becomes impractical for the large crystalline systems. In order to circumvent this problem, we have developed a CSP approach called as CASPESA (CrystAl Structure Prediction via Simulated Annealing). CASPESA is suitable for fixed composition systems and it requires some structural information about the system such as the minimum distances between any species. These distances can be supplied from experiments, structure database or DFT calculations. The objective function used in these global optimizations is a distance of a specific atomic arrangement, which is expected to lower the energy of the crystal system. The aim of the global optimization is to maximize the number of these predefined structural arrangements. The selection of the objective function is a system specific parameter and needs to be carefully determined. The serial version of CASPESA was first used to find the crystal structures of metal borohydrides and metal amines. Recently, CASPESA was also successfully applied to determine the crystal structures of metal amine borohydrides and materials used in lithium batteries. Unfortunately, the serial version of CASPESA has a drawback when the search is conducting on all large search spaces. To improve this deficiency, we parallelized CASPESA in order to perform much deeper global searches. Our limited tests showed that the parallel CASPESA significantly outperforms the serial version of CASPESA. By using the preparatory access opportunity, we will able to observe the real performance of parallel CASPESA on Tier0 machines. Then, the scalability results will be used to prepare a proposal for project access call 19 or a later one.

top

Spin and Charge Excitations for Quantum Technologies

Project Name: Spin and Charge Excitations for Quantum Technologies
Project leader: Dr António Costa
Research field: Fundamental Physics
Resource awarded: 50 000 core hours on MareNostrum
Description

This is a multi-pronged theory project with the ultimate goal of unveiling the fundamental properties of the elementary excitations of strongly spin-orbit coupled nanostructures. I expect the results to pave the way to the development of quantum technologies in solid-state environments. Two main ideas motivate the focus on elementary excitations. The first is that their properties determine how nanostructures respond to external fields and how they relax towards equilibrium, thus controlling energy absorption and dissipation. The second is that elementary excitations themselves may be exploited as resources for the implementation of quantum technologies, playing the role of information carriers, quantum memories, quantum repeaters or information transducers. In this context, spin-orbit coupling has many roles. It controls the minimum energy for producing a spin excitation. It combines with certain lattice geometries to produce topologically protected states. Finally, it couples charge and spin in more than one way, giving rise to spin-Hall effects and hybrid excitations. Many of those aspects have been studied in isolation. This is only effective, however, if spin-orbit coupling is relatively weak. Moreover, by considering the interplay between all of them, the possibilities of new interesting and useful phenomena grow exponentially. I will investigate spin-orbit coupled heterostructures of nanoscopic dimensions by building upon a theoretical and computational toolbox I have been developing. It combines Density Functional Theory methods and realistic modeling of the electronic structure of the whole system, including interaction effects. By combining the predictive power of Density Functional Theory and the flexibility and clarity of atomistic modeling, my methodology provides results that can be related directly to experimental set ups and measurements, thus providing an unprecedented level of guidance for further experimental and technological developments.

top

Global dynamo simulations of massive stars

Project Name: Global dynamo simulations of massive stars
Project leader: Dr Fabio Del Sordo
Research field: Universe Sciences
Resource awarded: 100 000 core hours on MARCONI -KNL
Description

The convection zones in massive stars is the most probable context for the dynamo to occur. As this zone is confined to the inner part of the star it is still a matter of debate whether the magnetic field is of fossil origin or there is a mechanism able to transport it from the center to the top through a thick radiative zone (Cantiello & Braithwaite, A&A 534A 140, (2011); Bonanno, CoSka 48 154B, 2018). An essential step in this direction is to model the non-linear evolution of the instability of toroidal fields, known as Tayler instability, in spherical geometry. Previous investigations in this direction have come to conflicting conclusions about the role of a stable stratification. The key question is to understand if the instability of toroidal fields, known as Tayler instability (a necessary condition for the appearance of the -still unconfirmed- Tayler-Spruit dynamo) can indeed be stabilized if the Brunt-Vaisala frequency is greater than the Alfven frequency. In an ongoing, soon-to-be-published project we have been able to show, for the first time using direct numerical simulations, that a stable stratification can indeed stabilize the Tayler instability. The next logical step of this investigation is to discuss the role of rotation and differential rotation. Our setup is already able to investigate these issues which are clearly essential to understand the topology of magnetic field in the A and B stars and also in the radiative zone of the red giants. We will run a set of simulations that will investigate all the parameter space consisting in differential rotation, stratification, stellar mass and magnetic field intensity. This will allow to evaluate which are the thresholds for magnetic field to be transported to the surface of the star and to compare such values with the observed ones.

top

VARMER-ICE (VAriable Resolution Model to Evaluate the Regional Impacts of Climate changE)

Project Name: VARMER-ICE (VAriable Resolution Model to Evaluate the Regional Impacts of Climate changE)
Project leader: Dr Jean Iaquinta
Research field: Earth System Sciences
Resource awarded: 50 000 core hours on Curie
Description

This project aims at testing the feasibility of using of a next-generation numerical Earth System Model with a variable horizontal mesh refinement (typically representing a factor of 8 or better) over Europe and Scandinavia to evaluate the regional impact of events related to climate change. This model utilizes a variable grid refinement nested into a global 1° resolution grid, which allows for simulations at a fraction of the computational cost that a full high-resolution climate simulation would require, whilst maintaining consistency in the physics and dynamics between the targeted area and the larger global domain.

top

Explicit Modeling of Stellar Feedback with RAMSES: Scaling Tests

Project Name: Explicit Modeling of Stellar Feedback with RAMSES: Scaling Tests
Project leader: Prof Davide Martizzi
Research field: Universe Sciences
Resource awarded: 50 000 core hours on Curie
Description

In this project, the PI will perform scaling tests on his modified version of the open-source astrophysical simulation code RAMSES (Teyssier et al. 2002). This code combines N-body methods with adaptive mesh refinement methods to solve the dynamics of stars, gas and dark matter in cosmological volumes and in galaxies. In recent years, the PI has developed a series of methods to simulate the physics stellar feedback, i.e. the collection of high-energy events related to the presence of stars in galaxies, that strongly influence galaxy evolution. The PI has implemented these methods in the RAMSES code and is planning to generate a suite of extreme resolution simulations that will reveal unknown aspects of the physics of stellar feedback with unprecedented accuracy. The code has been tested to run lower resolution simulations on smaller computing clusters, but the time is ripe for the code to be tested on a bigger machine, which will allow the production runs to be performed in the future.

top

EMMI

Project Name: EMMI
Project leader: Dr Massimiliano Bonomi
Research field: Biochemistry, Bioinformatics and Life sciences
Resource awarded: 100 000 core hours on Piz Daint
Description

In this Preparatory Project I will obtain the preliminary data (scalability plots) in order to be able to apply to the PRACE 19th call for Project Access. In the main project, I will use metainference, an integrative computational and experimental approach that I have developed, to characterize the structure and dynamics of a system of extraordinary biological importance, the type 2 secretion system (T2SS) pseudopilus. Metainference is a Bayesian framework to model structural ensembles by integrating prior information on a system with noisy, ensemble-averaged experimental data. The metainference approach enables: i) automatically determining the accuracy of the input data; ii) optimally weighting multiple sources of information based on their relative accuracy; iii) ultimately modelling structural ensembles by improving the prior description of the system with experimental information. In this project, metainference will be used to integrate cryo-electron microscopy data with physico-chemical knowledge of the system under study.

top

Cosmological hydrodynamical simulations in the ETHOS framework: the interplay between baryons and dark matter

Project Name: Cosmological hydrodynamical simulations in the ETHOS framework: the interplay between baryons and dark matter
Project leader: Prof Jesús Zavala Franco
Research field: Universe Sciences
Resource awarded: 50 000 core hours on MareNostrum
Description

The goal of the project is to perform scalability tests for cosmological hydrodynamical simulations within the ETHOS framework, which has been developed by the PI and collaborators. This framework generalises the theory of structure formation to contain a primordial cutoff in the power spectrum and strong dark matter self interactions. In this way ETHOS encompasses a broad range of dark matter models, including Cold Dark Matter (CDM), Warm Dark Matter (WDM), Self-Interacting Dark Matter (SIDM) and interacting Dark Matter (iDM). Our goal is to use these scalability tests as a part of a project access proposal with the scientific goal of performing full hydrodynamical simulations (using the AREPO code) to study the interplay between baryonic physics and dark matter physics in a variety of dark matter models. In particular, for the project proposal we are planning to perform a variety of simulations to explore the relevant parameter space in allowed dark matter/baryonic physics; with a flagship simulation of the Local Group with a resolution comparable to state-of-the-art simulations in CDM, and with the highest resolution available for alternative dark matter models. In this way, our team will be at the forefront of searches for clues about the dark matter nature in the galaxy population. Our team and collaborators have worked extensively in cosmological simulations. The project leader has ample experience in structure formation, particularly with different particle physics models, and has made significant contributions to bring the SIDM to a competitive level with CDM. Mark Lovell is an expert in WDM simulations, and has contributed significantly to put sterile neutrino dark matter (a viable WDM candidate) as one of the most competitive dark matter models. Mark Vogelsberger is a developer of the AREPO code and has made substantial contribution to the field of galaxy formation with hydrodynamical simulations, as well as being immersed in the resurgence of the SIDM model. The project leader and Mark Vogelsberger have been core members of the ETHOS framework from its inception. We are requesting computing resources to perform scalability test for simulations that include both baryonic physics and new dark matter physics. These tests will serve as an important component of our planned project access proposal. For the test we will use a similar setup to the Illustris project (~100 Mpc box/side box): three resolution levels to perform weak scaling, with ~40/300/2400 million elements (particles/cells). With the intermediate resolution ~300 million elements, we plan to perform a strong scaling test.

top

NANONP2AU –Nanoindentation of Nanopolycrystalline Nanoporous Au in Molecular Dynamics Simulations

Project Name: NANONP2AU –Nanoindentation of Nanopolycrystalline Nanoporous Au in Molecular Dynamics Simulations
Project leader: Prof Dan Mordehai
Research field: Chemical Sciences and Materials
Resource awarded: 100 000 core hours on MARCONI-KNL , 50 000 core hours on MareNostrum, 50 000 core hours on Curie, 20 000 core hours on JUWELS, 100 000 core hours on Piz Daint
Description

Nanoporous Au structures are considered as potential candidates in various emerging applications such as electrocatalyst, electrochemical actuators and sensors, owing to their large surface-to-volume ratio. As structural materials, one of the key issues is having light structures, but yet with desirable mechanical properties. Such, properties can be tailored, for instance, through geometry, owing to the fact that the ligaments are of a few nanometers and that strength is size-dependent, but also through the initial microstructure. Recently, a technique to manufacture ultra-fine grained nanoporous structures was proposed. Using nanoindentation, it was found that the hardness at room temperature is as high as that of ultrafine-grained bulk and it remains high even after heating by almost 300 degrees. The origin of these enhanced mechanical properties remains a riddle, which we aim at explaining in this project. We plan to perform multi-million atoms nanoindentation molecular dynamics simulations of nanopolycrystalline nanoporous Au structures, at various temperatures, to reveal the underlying dislocation mechanisms that govern mechanical properties of these structures. The large-scale simulations, which are planned based on this preparatory access, are intended to shed light on the effect of the initial microstructure and temperature on the deformation of nanoporous Au.

top

Lattice QCD on the Piz Daint system

Project Name: Lattice QCD on the Piz Daint system
Project leader: Dr Szabolcs Borsanyi
Research field: Fundamental Constituents of Matter
Resource awarded: 100 000 core hours on Piz Daint
Description

Quantum chromodynamics is the fundamental theory behind strong interactions. An ab initio non-perturbative treatment of the strong degrees of freedom, such as quarks and gluons is a prerequisite for precise predictions of the standard model. In this preparatory project we perform scaling studies on the Cray system using our MPI-CUDA implementation of the staggered hybrid Monte Carlo code. It will be used in quark gluon plasma research, the determination of the hadronic vacuum polarization as well as in searches for axion-like dark matter.

top

Role of Glycans in Epidermal Growth Factor Receptor Regulation

Project Name: Role of Glycans in Epidermal Growth Factor Receptor Regulation
Project leader: Mr Mykhailo Girych
Research field: Biochemistry, Bioinformatics and Life sciences
Resource awarded: 100 000 core hours on MARCONI , 50 000 core hours on Curie-KNL, 50 000 core hours on Curie-SKL, 100 000 core hours on Piz Daint
Description

Over the past decade, protein glycosylation attracts ever-growing attention in a variety of research areas, from biomedicine to biotechnology. Due to a strong causative link between protein glycosylation patterns and development of severe disorders, the study of protein glycans is of utmost importance. Epidermal growth factor receptor (EGFR) is a glycosylated transmembrane receptor that plays a pivotal role in a vast majority of cellular processes and involved in the development of several forms of cancer. Recently EGFR glycans have been sequenced, and we started to understand the involvement of protein glycans in EGFR functioning and cancer development. In particular, it was recently demonstrated that EGFR activity is mediated by association with membrane glycolipid GM3 through carbohydrate-to-carbohydrate interactions. However, the exact molecular mechanism of GM3 interaction with differentially glycosylated EGFR remains unexplored. Here we are going to elucidate the mechanisms behind the regulation of EGFR activity by GM3 glycolipid using an extensive all-atom Molecular Dynamics (MD) simulations with GROMACS package. Considering the size of fully glycosylation EGFR dimer which requires a large membrane for its accommodation, systems would need to go through proper scaling tests to maximize the efficiency of calculations. This project is aimed at benchmarking and fine-tuning the GROMACS performance for above described solvated EGFR systems.

top

Preparatory run for full scale wind turbine aeroacoustic simulations

Project Name: Preparatory run for full scale wind turbine aeroacoustic simulations
Project leader: Dr Guannan Wang
Research field: Engineering
Resource awarded: 20 000 core hours on JUWELS
Description

Investigating the mechanism aerodynamic noise emitted from a full scale wind turbine rotor and its mitigation method using a commercial Lattice-Boltzmann-Method based high fidelity CFD code. (PowerFLOW).

top

PULSAR-PIC

Project Name: PULSAR-PIC
Project leader: Dr Francois Courvoisier
Research field: Fundamental Physics
Resource awarded: 50 000 core hours on Curie
Description

This project is part of a larger project, ERC-funded project PULSAR. PULSAR aims at developing novel ultrafast laser-materials processing concepts and strategies. At present, even with high pulse energy, laser processing remains limited to high-speed scanning point by point removal of ultra-thin nanometric layers from the material surface. This is because the uncontrolled laser-generated free-electron plasma shields against light and prevents reaching extreme internal temperatures at very precise nanometric scale. PULSAR aims at breaking this barrier and developing a radically different concept of laser material modification regime based on free-electron plasma control. PULSAR ‘s unconventional concept is to control plasma generation, confinement, excitation and stability. An ambitious experimental and numerical research program will push the frontiers of laser processing to unprecedented precision, speed and predictability. PULSAR key concept is highly generic and the results will initiate new research across laser and plasma material processing, plasma physics and ultrafast optics. Our final aim with the present project is to perform numerical simulations using Particle In Cell (PIC) codes of the laser-plasma interaction where plasmas diameters are on the order of a few hundreds of nanometers and reach meanwhile overcritical densities. We have already experimentally identified regimes where the above-mentioned barrier is overcome using spatially-shaped ultrafast pulses. Preliminary results obtained in our group demonstrate we are understanding the basics of the phenomena but more analysis with higher resolution and number of particles is needed. Our objective is to simulate the plasma structuration under the input beams and mechanisms of the high absorption and energy transfer when shaped beams are used.

top

HiFiTurb_T161_Cenaero

Project Name: HiFiTurb_T161_Cenaero
Project leader: Dr Koen Hillewaert
Research field: Engineering
Resource awarded: 100 000 core hours on MARCONI -KNL
Description

The HiFiTurb project aims at providing high resolution DNS databases for the improvement and development of advanced turbulence models, in combination with machine learning techniques. The current preparatory project is submitted in support of future resource requests to perform the computations during the research.

top

FullSTuP

Project Name: FullSTuP
Project leader: Dr Kunal Puri
Research field: Engineering
Resource awarded: 100 000 core hours on MARCONI -KNL
Description

The project aims to perform high-fidelity LES/DNS computations of a low-pressure-turbine with end-walls, with the aim to understand the complex flow physics and interactions of the flow field. An additional goal is to use high-fidelity simulations as a means to generate accurate statistics that will aid in the development of low-fidelity models like RANS.

top

Proton powered motors: learning from biology with multi-scale modeling

Project Name: Proton powered motors: learning from biology with multi-scale modeling
Project leader: Prof Dr Gerrit Groenhof
Research field: Chemical Sciences and Materials
Resource awarded: 100 000 core hours on Piz Daint
Description

A molecular motor is a key component of a nanotechnology toolbox. Although various molecular motors have been developed that undergo unidirectional motion when fueled by light or chemicals, an efficient conversion of (photo-) chemical energy into mechanical work at the nano-scale remains challenging due to the random Brownian fluctuations of the environment. Rather than working against it, Nature have evolved ways to exploit Brownian fluctuations, culminating in the evolution of the ATPase rotary motor proteins that use a proton free energy gradient between two sides of a membrane (i.e. the proton-motive-force) to convert random thermal motion into torque with a very high efficiency. Therefore, this protein could be a source of inspiration for building artificial molecular motors. Unfortunately, however, the mechanism by which these proteins rectify the non-directional Brownian motion into unidirectional rotation is not yet understood in atomic detail. Instead, we only have phenomenological theories that are not easily inverted to design new systems. The purpose of this project is to use molecular dynamics simulations, in which the effect the proton-motive-force is included, to acquire the missing atomistic insights into the ratcheting of proton-powered ATPase motors. We will also use these insights to design new Brownian nano-motors that are based on the principles of ATPase.

top

PULSAR-PIC

Project Name: PULSAR-PIC
Project leader: Dr Francois Courvoisier
Research field: Fundamental Physics
Resource awarded: 100 000 core hours on MARCONI -KNL
Description

This project is part of a larger project, ERC-funded project PULSAR. PULSAR aims at developing novel ultrafast laser-materials processing concepts and strategies. At present, even with high pulse energy, laser processing remains limited to high-speed scanning point by point removal of ultra-thin nanometric layers from the material surface. This is because the uncontrolled laser-generated free-electron plasma shields against light and prevents reaching extreme internal temperatures at very precise nanometric scale. PULSAR aims at breaking this barrier and developing a radically different concept of laser material modification regime based on free-electron plasma control. PULSAR ‘s unconventional concept is to control plasma generation, confinement, excitation and stability. An ambitious experimental and numerical research program will push the frontiers of laser processing to unprecedented precision, speed and predictability. PULSAR key concept is highly generic and the results will initiate new research across laser and plasma material processing, plasma physics and ultrafast optics. Our final aim with the present project is to perform numerical simulations using Particle In Cell (PIC) codes of the laser-plasma interaction where plasmas diameters are on the order of a few hundreds of nanometers and reach meanwhile overcritical densities. We have already experimentally identified regimes where the above-mentioned barrier is overcome using spatially-shaped ultrafast pulses. Preliminary results obtained in our group demonstrate we are understanding the basics of the phenomena but more analysis with higher resolution and number of particles is needed. Our objective is to simulate the plasma structuration under the input beams and mechanisms of the high absorption and energy transfer when shaped beams are used.

top

Scalability testing for multinode neural machine translation

Project Name: Scalability testing for multinode neural machine translation
Project leader: Dr Kenneth Heafield
Research field: Mathematics and Computer Sciences
Resource awarded: 100 000 core hours on Piz Daint
Description

Machine translation quality has substantially improved in recent years and e.g. German to English translation is good enough that companies are showing output to their customers for localization. Translation models are based on neural networks. Training them is expensive enough that computation is often the limiting factor in improving quality, with GPU improvements indirectly feeding into more complicated models and higher quality. Competitive translation models can take 8 weeks of GPU time to train for one language pair (i.e. German to English) and this is only after searching for good parameters. Training costs have increased over time as distributed training becomes commonplace and models have improved. This project will test scalability of distributed training and inference on Piz Daint for neural machine translation systems, with the aim of both accelerating training and improving model quality.

top

Three-dimensional time-dependent joint dynamical and ionic evolution of the interstellar gas in a section of the Milky Way

Project Name: Three-dimensional time-dependent joint dynamical and ionic evolution of the interstellar gas in a section of the Milky Way
Project leader: Prof Miguel Avillez
Research field: Universe Sciences
Resource awarded: 50 000 core hours on MareNostrum, 50 000 core hours on Curie, 20 000 core hours on JUWELS
Description

The interstellar medium (ISM) is driven by supernovae (stellar explosions) and, to a small extent, by stellar winds, that generate shock waves which continuously compress, heat and stir the medium. Thus, the gas becomes turbulent and the injected energy cascades down to the smallest scales where it is dissipated. The heated gas expands and, rises into the Galactic halo where some of it cools, forming clouds that fall back onto the disk (intermediate and high velocity clouds; IVCs/HVCs) due to the gravitational pull. These clouds bear the same chemical composition and the ionic imprint of the original gas which was polluted by metals from the supernovae. Ions/atoms are a major component of the ISM whose charge state is determined by competing ionization and recombination processes. In general the plasma is not in ionization-recombination equilibrium (IRE) due to widely differing ionization and recombination timescales, except at temperatures in excess of a million degrees K. But even at these temperatures deviations from equilibrium have been observed in X-ray emission.More importantly, , the thermal and dynamical feedback due to expansion, compression, heating and cooling also contributes to deviations from equilibrium with drastic consequences in the plasma emission and history. We propose to carry out, for the first time to date, a complete detailed numerical study of the joint thermal and dynamical evolution of the interstellar gas in a section of the Milky Way (including the disk and halo) by using state-of-the-art MHD adaptive mesh refinement parallel code capable of tracking on the fly the formation and evolution of the ions/atoms (of the ten most abundant elements in nature), the populations of the ion/atom levels (we use a model of up to 70 levels), and line (due to transitions between levels) and continuum (free-bound, free-free and two-photon) emission. This work will have profound implications on the current understanding of the ISM by providing answers to many unresolved questions in spite of more than three decades of intense research in the field.

top

Functional lattice instabilities in naturally layered perovskites

Project Name: Functional lattice instabilities in naturally layered perovskites
Project leader: Prof João Pedro Esteves de Araujo
Research field: Chemical Sciences and Materials
Resource awarded: 20 000 core hours on JUWELS
Description

The here submitted computational project is grounded in the ongoing FLIP- Functional lattice instabilities in naturally layered perovskites project. FLIP is an international collaboration that aims to produce, model and probe Naturally Layered Perovskites, providing new efficient materials for sustainable energy conversion. It will explore the possibilities to adjust oxygen octahedra rotations to create acentricity in Ruddlesden-Popper phases and double perovskite oxides enabling room temperature magneto-electric coupling or negative thermal expansion. The project focus in the incorporation of multiple cations into Ca3(Mn,Ti)2O7 and AA’BMnO6. Using methods including non-equilibrium approaches such High-Pressure or Pulsed-Laser Deposition it engineers crystallographic phases with adjusted lattice, electric and magnetic interactions. The overall study is anchored the theoretical modelling that allows the design of new multiferroic systems and predict their properties. These are to be correlated with the materials properties accessed by conventional macroscopic techniques and by a combination of local probe ones. Significant impact on unveiling strategies for materials functionalization far beyond state-of-the-art is foreseen.

top

Simulating the evolution of human bipedalism using multi-body dynamics and machine learning optimisation approaches

Project Name: Simulating the evolution of human bipedalism using multi-body dynamics and machine learning optimisation approaches
Project leader: Dr Karl Bates
Research field: Biochemistry, Bioinformatics and Life sciences
Resource awarded: 50 000 core hours on MareNostrum
Description

How, when and why our ancestors began to walk upright on two legs are questions to which we still have no clear answers. Here we suggest that recently discovered functional variation in modern humans provides the key to reconstructing the evolution of human locomotion. Over the past two years we have collected an unprecedented anatomical and experimental data base on locomotor morphology and biomechanics from 50 human volunteers. In this application we plan to use this data set to build, drive and validate new computational models to analyse this morphological and functional variation and generate new founding principles to interpret fossil evidence. These simulations contain four novel aspects: an unprecedented level of subject-specific anatomical and motion information; and simulation of locomotion over a range of non-uniform substrates using different optimisation goals to examine how substrate impacts on neuromuscular gait control. Our models are driven by machine learning algorithms that autonomously generate optimal motions without biasing predicted gaits with subjective user input. This process is extremely computationally demanding but allows us to predict the optimal gaits of fossil animals and thus to ‘replay’ the key anatomical transformations seen during the evolution of human locomotion.

top

Isotope Mass Impact on Tokamak Performance Improvement (IMIToPI)

Project Name: Isotope Mass Impact on Tokamak Performance Improvement (IMIToPI)
Project leader: Dr Fabien Widmer
Research field: Fundamental Physics
Resource awarded: 50 000 core hours on MareNostrum
Description

International progress on magnetic fusion depends on our capability to understand and predict plasma confinement properties. Reducing or controlling the thermal loss of energy is necessary in order to optimize the thermal energy confinement time and provide CO2-free energy for civil usage at an industrial level. Actual fusion experiments run hydrogen or deuterium plasmas but the next generation reactors, such as ITER or DEMO, will use a mixture of deuterium and tritium. Therefore, knowledge of the mass scaling is needed to extrapolate experimental results from actual to future reactors. This is especially important since an improvement of magnetic confinement was experimentally observed moving from an hydrogen to a deuterium plasma. This confinement improvement is known as the isotope effect. It is not yet clearly understood and is one of the greatest challenge of turbulence transport theory. Assuming that the plasma is electrostatic, collisionless, without background flows, that the electrons are adiabatic and that transport processes are local results in a turbulent transport level proportional to the square root of the isotope mass, in contradiction with the experimental observations. Relaxing any of these assumptions can alter the mass scaling above, a key point to elucidate the isotope effect. Significant efforts in the past thirty years have been done to investigate how relaxing these assumptions modifies the isotope mass scaling of turbulent transport. Amongst other phenomena, it was very recently shown that at high plasma pressure, electromagnetic fluctuations can reverse the mass scaling of the Ion-Temperature-Gradient (ITG) instability, one of the more robust source of anomalous transport in fusion devices. Additionally, it is known that the ratio of the ion Larmor radius to the system size impact heat transport, the turbulence correlation length and avalanches, but no study of the implications on the isotope effect has been performed yet. Further investigations of these phenomena are needed before a consistent and comprehensive picture of the isotope scaling of turbulent transport could emerge. This is the goal of the present proposal. To scrutinize the isotope mass impact on both electromagnetic stabilisation and global electrostatic effects we combine two 5D state-of-the-art gyro-kinetic codes: GKW in the local gradient-driven framework, and GYSELA, utilizing a global flux-driven approach. GYSELA is one of the few codes able to model both the plasma core and edge in an experimentally relevant flux-driven framework with a kinetic electrons and ions description. Such capabilities allow to determine what is the impact of the isotope effect on the edge-core turbulence interplay and the global organisation and regulation of turbulence. Electrostatic local gradient-driven GKW simulations, based on GYSELA global simulations time averaged profiles, will be used to separate the local and global impact of the isotope effect on turbulent transport. Finally, GKW local gradient- driven electromagnetic simulations will provide a test-bed for advanced quasi-linear models attempting to capture electromagnetic stabilisation of turbulence. The rare combination of local gradient-driven and global flux-driven simulations is expected to provide valuable new insights to the puzzling isotope effect.

top

ElectroPulse – Electric field and Pulse propagation in Magnetised Plasma Turbulence

Project Name: ElectroPulse – Electric field and Pulse propagation in Magnetised Plasma Turbulence
Project leader: Dr Guilhem Dif-Pradalier
Research field: Fundamental Physics
Resource awarded: 50 000 core hours on MareNostrum
Description

With the development of large machines like ITER, controlled fusion makes a huge step forward towards mastering the energy of the stars for civil usage. Steady international progress regarding the achieved fusion performance relies on our ability to predict the confinement properties of the plasma. Turbulent transport is the key player in this matter, modelled through the 5-dimensional gyrokinetic description, as required by the low collisionality of hot and dilute plasmas. Aspects of magnetised plasma turbulence are multiscale yet despite orders of magnitude in spatial and temporal scales between injection/dissipation and transport, the confined plasmas feel their material boundaries. This is common knowledge experimentally yet a vastly unexplored territory computationally. Predicting the interplay between spatially distinct regions of the plasma (core and edge), on disparate timescales is one of the major current issues. This proposal wishes to tackle aspects of this important problem, and doing so address two nagging problems: (i) assess whereby the radial electric field builds up and endures and (ii) probe edge-core interplay, especially through the investigation of transient dynamics, i.e. cold and hot bursts. The PI has strong expertise in the field and will be fully committed to this activity.

top

Scalability analysis of Tsunami-HySEA and Landslide-HySEA models.

Project Name: Scalability analysis of Tsunami-HySEA and Landslide-HySEA models.
Project leader: Prof Manuel Jesus Castro Diaz
Research field: Mathematics and Computer Sciences
Resource awarded: 100 000 core hours on Piz Daint
Description

The main objective of this project is to check the scalability of the Tsunami-HySEA and Landslide-HySEA numerical models in Piz Daint cluster as requirement to apply for the next PRACE call. Tsunami and Landslide-HySEA numerical codes have been developed by EDANYA (UMA) and are designed to simulate earthquake and landslide-generated tsunamis, respectively. In both models, shallow-water assumptions are used to simulate the evolution of the tsunamis waves. In the first one, the shallow-water system is coupled with the Okada model to estimate the deformation generated by an earthquake in the ocean basin, and in the second one, a Savage-Hutter-type model is used to simulate the landslide evolution. Both models are discretized using explicit second order Finite Volume solvers. Domain decomposition is used and multi-GPU implementation is carried out to speed-up the computations. At this stage, both models have been audited in CHEESE EU project (Centre of Excellence for Exascale Solid Earth, project ID 823844, H2020-INFRAEDI-2018-1) and also have been exhaustively tested with laboratory and real data (Macías et al, 2015; Macías et al, 2017a, 2017b and https://edanya.uma.es/hysea/index.php/benchmarks). In particular, in this project we will check the scalability of both codes in some significant test cases over real topo-bathymetries in Piz Daint cluster. Our codes are involved in three Pilot Demonstrators to be developed in the framework of ChEESE project. PD2 related with Faster Than Real Time (FTRT) tsunami simulations, PD7 developing Probabilistic Tsunami Hazard Assessment techniques (PTHA) and PD8 dealing with Probabilistic Tsunami Forecast (PTF). The final goal in the framework of ChEESE project is to prepare our codes, develop these pilot demonstrators and enable the related final services to end-users for the upcoming pre-Exascale and Exascale supercomputers. References 1. Macías et al (2015). [doi: 10.13140/RG.2.2.22999.47527]. 2. Macías et al (2017a). PAGEOP, 1-37. [doi: 10.1007/s00024-017-1583-1]. 3. Macías et al (2017b). NTHMP report. [doi: 10.13140/RG.2.2.27081.60002]. A complete set of references for T-HySEA and L-HySEA codes available at https://edanya.uma.es/hysea/index.php/references

top

Modelling protein dynamics by combining SAXS and molecular dynamics on GPU

Project Name: Modelling protein dynamics by combining SAXS and molecular dynamics on GPU
Project leader: Prof Carlo Camilloni
Research field: Biochemistry, Bioinformatics and Life sciences
Resource awarded: 100 000 core hours on Piz Daint
Description

The use of small-angle X-ray scattering (SAXS) in combination with molecular dynamics simulation is hampered by its heavy computational cost. The calculation of SAXS from atomic structures can be speeded up by using a coarse-grain representation of the structure. A hybrid multi-resolution strategy has also been implemented to perform SAXS restrained simulations at atomic resolution by calculating the virtual positions of coarse-grained beads on the fly and using them for the calculation of SAXS. The accuracy and efficiency of the method has been demonstrated by refining the structure of two protein–nucleic acid complexes. Here we want to test the scalability of our GPU implementation of this algorithm so to enable the use of SAXS restrained MD simulations for very large systems.

top

IMPACT – Retracing the formation of the Moon in the aftermath of the Giant Impact

Project Name: IMPACT – Retracing the formation of the Moon in the aftermath of the Giant Impact
Project leader: Dr Razvan Caracas
Research field: Earth System Sciences
Resource awarded: 50 000 core hours on MareNostrum
Description

We aim at characterizing the evolution of the protolunar disk generated in the aftermath of the Giant Impact from its formation until its condensation. This project was awarded an ERC Consolidator Grant, IMPACT, for the duration 2016-2021. Very little is understood of the physics governing the Giant Impact and the subsequent formation of the Moon. According to this model an impactor hit the proto-Earth; the resulting energy was enough to melt and partially vaporize the two bodies generating a large protolunar disk, from which the Earth-Moon couple formed. Hydrodynamic simulations of the impact and the disk are currently based on unconstrained models of equations of state and phase diagrams. Here we use large-scale ab initio molecular dynamics to determine vaporization curves, position the supercritical points, and characterize the sub-critical and supercritical regimes for the major rock-forming minerals. Then we obtain the thermal profile through the disk, the ratio between liquid and vapor, and the chemical speciation. Eventually we constrain the impactor, proto-Earth and plausible impact scenarios. In order to be able to answer these fundamental questions we are asking for resources not available at Tier-1 or Tier-2 centers: 32471040 hours on MareNostrum.

top

Scalability DL_MESO on multi-GPU cluster

Project Name: Scalability DL_MESO on multi-GPU cluster
Project leader: Dr Jony Castagna
Research field: Chemical Sciences and Materials
Resource awarded: 100 000 core hours on Piz Daint
Description

This project is a PRACE summer of HPC 2019. The intention of this project is to benchmark the current version of DL_MESO on a multi-GPU architecture. DL_MESO is a software package for mesoscale simulation based on Dissipative Particle Dynamics and Lattice Botlzmann solvers. The current version for multi-GPU is written in CUDA language and covers only the main options provided by the serial version. The main goal is to further extend the development and then to use the PRACE resources to benchmark it on large systems.

top

Type B: Code development and optimization by the applicant (without PRACE support) (6)

Benchmarking the Astrophysics Software Gadget PRACE-5IP T7.3.A UEABS

Project Name: Benchmarking the Astrophysics Software Gadget PRACE-5IP T7.3.A UEABS
Project leader: Prof Miguel Avillez
Research field: Universe Sciences
Resource awarded: 200 000 core hours on MARCONI , 100 000 core hours on MareNostrum, 200 000 core hours on Curie-KNL, 200 000 core hours on Curie-SKL, 50 000 core hours on JUWELS
Description

GADGET 3 is a cosmological, fully hybrid MPI + OpenMP parallelized Smoothed Particle Hydrodynamics code that is tailored to solve a wide range of astrophysical problems, e.g., structure formation, colliding and merging galaxies, studying the dynamics of the gaseous intergalactic medium, forming of the stars and its regulation. GADGET includes a tree-code module, a communication scheme for gravitational and SPH forces, a domain decomposition strategy, a novel smooth particle hydrodynamics (SPH) formulation based on entropy as an independent variable, and in addition of the “TreePM” functionality. The objective of this project is to benchmark GADGET 3, and measure the code performance overall, as well as, the internal performance metrics by timing certain parts of the code. This work is part of the PRACE-5IP WP7, Task 7.3.A.

top

Simulations of Wind Collision, Accretion, and Feedback in Massive Stellar Binaries

Project Name: Simulations of Wind Collision, Accretion, and Feedback in Massive Stellar Binaries
Project leader: Dr Amit Kashi
Research field: Universe Sciences
Resource awarded: 250 000 core hours on Hazel Hen
Description

We use high resolution 3D hydrodynamical simulations to quantify the amount of mass accreted onto the secondary star of the binary system Eta Carinae, exploring two sets of stellar masses which had been proposed for the system, the conventional mass model (M1=120M⊙ and M2=30M⊙) and the high mass model (M1=170M⊙ and M2=80M⊙). The system consists of two very massive stars in a highly eccentric orbit. Every cycle close to periastron passage the system experiences a spectroscopic event during which many lines change their appearance, accompanied by a decline in x-ray emission associated with the destruction wind collision structure and accretion of the primary wind onto the secondary. We take four different numerical approaches to simulate the response of the secondary wind to accretion, each affects the mass loss rate of the secondary differently, and in turn determines the amount of accreted mass. The high mass model gives for most approaches much more accreted gas and longer accretion phase. We find that the effective temperature of the secondary can be significantly reduced due to accretion. We conclude that the high mass model is better compatible with the amount of accreted mass, ≈3×10−6M⊙, required for explaining the reduction in secondary ionization photons during the spectroscopic event and compatible with its observed duration. Our simulations are of high resolution that can follow instabilities in the colliding winds and resolve small clumps and thin filaments. The accreting star is also resolved and the direction from where gas is accreted and angular momentum is gained can be identified. We expect to have strong ties between the amount of accreted mass and transported angular momentum and the mode of accretion and the formation of an accretion disk.

top

Improving scalability of ABCD Solver with replicated partitioning, new load balancing, and communication minimization techniques

Project Name: Improving scalability of ABCD Solver with replicated partitioning, new load balancing, and communication minimization techniques
Project leader: Mr Philippe Leleux
Research field: Mathematics and Computer Sciences
Resource awarded: 100 000 core hours on MareNostrum, 250 000 core hours on Hazel Hen, 50 000 core hours on JUWELS
Description

The hybrid block row-projection method ABCD solver (http://abcd.enseeiht.fr) is a sparse linear solver designed to solve large sparse unsymmetric systems of equations on distributed memory parallel computers. The method implements a block Cimmino iterative scheme, accelerated with a stabilized block Conjugate Gradient [1]. An augmented pseudo-direct variant is also proposed to overcome convergence issues [2]. Both methods are included in the solver with a hybrid parallelization scheme. The parallel performance of the ABCD solver will be subject to improvements in the first non-beta release, version 1.0, which we finalize in this project. Novel algorithms for the distribution of partitions to processes are introduced with minimization of the communications as well as balancing of the workload in mind. Furthermore, the Master-Slave approach on each sub-system, inherent to the hybrid scheme of the solver, is also subject to improvements in order to achieve higher scalability through run-time placement of processes. Finally, the bottleneck of the iterative block Cimmino scheme, its slow convergence for some problems, is tackled with the use of row replication strategies. We call a partitioning as replicated when the partitions are not necessarily pairwise disjoint. Intelligent replication, using a graph-based smart technique, can accelerate the convergence of the block Cimmino iterative method. [1] Ruiz, D. (1992). Solution of large sparse unsymmetric linear systems with a block iterative method in a multiprocessor environment. CERFACS TH/PA/9, 6. [2] Duff, I. S., Guivarch, R., Ruiz, D., & Zenadi, M. (2015). The augmented block Cimmino distributed method. SIAM Journal on Scientific Computing, 37(3), A1248-A1269.

top

HPC optimisation of SDLPS distributed simulator

Project Name: HPC optimisation of SDLPS distributed simulator
Project leader: Dr Pau Fonseca i Casas
Research field: Engineering
Resource awarded: 100 000 core hours on MareNostrum
Description

The project was awarded in the 7th call of the PRACE-SHAPE programme. The current application aims to have access to the compute resources required for finishing the SHAPE project.

top

I/O limit testing and usage of different I/O strategies

Project Name: I/O limit testing and usage of different I/O strategies
Project leader: Dr Boris Gudiksen
Research field: Universe Sciences
Resource awarded: 200 000 core hours on Curie
Description

During this project, we will try to further develop the I/O from the already parallelised stellar atmosphere code Bifrost. Earlier awarded PRACE projects has revealed that the I/O from the massively parallel jobs on Marconi-KNL intermittently produced I/O errors. The source of these errors and ways to circumvent them will be explored.

top

Global Acoustic 2D/3D

Project Name: Global Acoustic 2D/3D
Project leader: Dr Noriyuki Kushida
Research field: Earth System Sciences
Resource awarded: 200 000 core hours on Curie, 100 000 core hours on Piz Daint
Description

This project aims at modelling acoustic wave propagation on a global scale, mainly in the atmosphere and the ocean. The International Monitoring System of the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) has been monitoring hydro-acoustic and infrasound waves over the globe. Because of the complex natures of the oceans and the atmosphere, computer simulation can play an important role in understanding the observed signals. In this regard, the performance of computer simulations to depict observed signals, particularly one which is derived from a principal equation, needs to be investigated. The applicant has been developing a solver of the time-dependent scalar wave equation on a background flow. This solver employs the two-dimensional (2D) finite difference method (FDM) on the Yin-Yang grid so that a long-term time integration can be achieved and the singularity problem, namely the derivative of the governing equation cannot be determined and an approximation is usually introduced, is avoided [1]. So far, by comparing to a real-life classical hydroacoustic experiment performed in 1960 [2], the solver correctly models the kinetics of diffractions, which is not modelled in a simplified method such as ray-tracing. Diffractions are naturally taken into account in solving the wave equation, and therefore, a good agreement on the travel time of the diffracted waves is observed. Frequencies up 0.01Hz were modelled. However, these points are raised by domain expertss for further realistic modelling:; (1) Higher frequency waves should be modelled to enable the direct comparison of observed waves and modelling. (2) Three-dimensional (3D) effects, including sound speed profile and material properties, should be taken into consideration. (3) Attenuation and dispersion effects should be taken into consideration in order to discuss the source energy and observed signals. (4) Boundary conditions which can model energy loss should be implemented. All the above points lead to a high demand for computer resources, and therefore, a supercomputing facility in PRACE is necessary to achieve realistic modelling of acoustic wave propagation on a global scale. In this project, (1) and (2) are particularly targeted. [1] N. Kushida, and R. Le Bras: Acoustic wave simulation using an overset grid for the global monitoring system, Abstract [S44A-04] presented at 2017 Fall Meeting, AGU, New Orleans, LA, 11-15 Dec, 2017. [2] J. Northrop, and C. Hartdegen: UNDERWATER SOUND PROPAGATION PATHS BETWEEN PERTH, AUSTRALIA, AND BERMUDA: THEORY AND EXPERIMENT, Technical Report 585 of NAVAL OCEAN SYSTEMS CENTER SAN DIEGO, CALIFORNIA, 1980.