Researcher: Jan Simkanin
Project: Thermochemically-driven convection and dynamos operating at low Ekman numbers
Allocation: 1 025 000 core hours
Abstract: The Earth's magnetic field is one of the most variable geophysical fields and provides us with effective protection against high-energy, electrically charged particles from the solar wind, solar flares and is a useful tool for navigation, not only for us, humans but also for animals. The Earth's magnetic field is generated by convective motions of an electrically conductive melt in the Earth's outer core and penetrates the surface of
the Earth. These generation processes are called Geodynamo for short. On the Earth’s surface, a large-scale, dipole-dominated field is observed. However, we have no direct information about the magnetic field in the Earth'score. For this reason, we model magnetohydrodynamic processes in the Earth's core numerically. Numerical modelling of Geodynamo has made enormous progress during last years thanks to the development
of computer technology. Numerical models produce magnetic fields that are close to the observed geomagnetic field. We are also able to reproduce its temporal changes, whether short-term or long-term ones (so-called secular variations). However, we cannot use parameters typical of the Earth's core in our models. The real parameter values cannot be used for computational reasons – so far there is no supercomputer in the world that could solve a Geodynamo model with such values. However, we are gradually approaching these values as the computers become more and more powerfull.

 

Researcher: Athanasios Koliogiorgos
Project: DFT study of electronic interaction and energy transfer between silicon and organic chromophores (ENTRANCE)
Allocation:  1 246 000 core hours
Abstract: Silicon constitutes the definitive material for solar cells and is also widely used in optoelectronic applications. Its often unsatisfactory electronic and optical properties can be enhanced with attachment of or proximity to organic molecules called chromophores, such as protoporphyrin IX (PPIX), which can improve its properties such as photoluminescence and efficiency. In this study, we will examine Si nanocrystals and bulk Si in interaction with PPIX molecules in different distances and with different orientations of PPIX in relation to Si. We will study their electronic structure with focus on the transition of the energy band gap of Si from indirect to direct caused by interaction with PPIX. We will also study their energy transfer phenomena under the Förster resonance energy transfer (FRET) model. Our study will yield conclusions as to which orientation and distance is ideal for enhancement of Si and will provide a computational paradigm for the ab-initio study of donor/acceptor systems where the acceptor is a semiconductor. 

 

Researcher: Jan Rezac
Project: Large-scale benchmarking of non-covalent interactions – extended coverage of London dispersion and other interactions
Allocation: 2 563 000 core hours
Abstract: To apply computational chemistry to real-world chemical problems, it is often necessary to work with large systems with thousands of atoms. This is especially true in the two currently most prominent research directions, in the applications of computational methods to biochemistry and to (nano)materials. This requires the use of approximate methods, often including empirical parameters. The development of such method then relies on accurate reference data that can be used for their parametrization and validation. Also, the emerging applications of machine learning to molecular systems require enormous amounts of data for their development. Here, we focus on non-covalent interactions, an effect of key importance in larger molecular systems. This proposal is a part of a larger project that aims to build a state-of-the-art database of accurate calculations that can serve this propose.  First three data sets covering hydrogen bonds had been already published [J. Řezáč, J. Chem. Theory Comput. 2020, 16(4); J.Řezáč, J. Chem. Theory Comput. 2020, 16(10)],  and made openly available at a dedicated website www.nciatlas.org. This proposal covers calculations needed to extend the coverage of other types of non-covalent interactions, namely extending the large data set of London-dispersion driven complexes (in preparation) to non-equilibrium geometries, benchmark calculations on halogen bonds and other sigma-hole interactions, and initial work on a data set of the interactions of biologically relevant cations.

 

Researcher: Diego Lopez
Project: Nanodiamonds for Photovoltaics (NaP)
Allocation: 2 150 000 core hours
Abstract: The transformation of sunlight into electricity represents one of the major tasks that faces the scientists and engineers of the 21st century.  However, the traditional silicon-based photovoltaic (PV) devices are not free of disadvantages, such as their expensive and polluting fabrication or their low performance in low-light conditions. In order to overcome these drawbacks, organic PV devices have been proposed, and many materials have been tested in order to achieve greater efficiencies. These devices are formed by an acceptor material placed close to a donor material, leading to the formation of hole/electron pairs. Nanodiamonds (ND), which can be described as fragments of diamonds, offer great promise withing that field. Compared with other materials, NDs provide great advantages for energy conversion, as they are stable, non-toxic and cheap. However, the PV behavior of those systems have been proven to be greatly affected by the size,11 and this might be of paramount importance on photovoltaics when matching the energy levels is required. This trend has been only described for NDs purely composed by carbon atoms with hydrogen passivation, and how doped NDs are affected is still unknown. Thus, the goal of NaP is to establish by means of density functional theory (DFT) the effect that the size of the ND and the dopants have on the energy of the molecular orbitals as well as in closely related features such as electron affinity and ionization potential.

 

Researcher: Michal Podhoranyi
Project: PaReTran6
Allocation: 237 000 core hours
Abstract: During the past decade, one of the main topics in reactive transport modelling has been the ongoing global search for strategies of safe nuclear waste disposal (Parhurst and Wissmeier 2015, Cochepin et al. 2008, Montarnal et al. 2007). Reactive transport modelling (Fig.1) can provide crucial information about the evolution of contaminant plumes over long time scales, and information about near-field processes, which are used to improve safety in the design of confining structures and containers (Parhurst and Wissmeier 2015). Globally, there are two commonly accepted disposal options – near-surface disposal and deep geological disposal (WNA 2017). Near-surface disposal at ground level, or in caverns below ground level (at depths of tens meters) is implemented in many countries including the Czech Republic.    
The aim of the project is to improve the possibilities of a potential risk analysis of environmental contamination due to the long-term radioactive species spread around a deep radioactive waste repository. This aim will be made by integrating HPC infrastructure into the process of reactive transport modelling.  The solution is complicated by the low accuracy of known rock medium parameters values in large depths and in the far future. The sensitivity analysis using numerical simulation computations is then very computationally consuming. Also, the individual reaction-transport simulations themselves are computationally consuming, and their parallelization and use of HPC (High-Performance Computing) allow the performing of the simulations in real size and available time.

 

Researcher: Jan Zemen
Project: Modeling of magnetic structure of perovskite/antiperovskite bilayers
Allocation: 1 549 000 core hours
Abstract: We propose to explore heterointerfaces of perovskite oxides and antiperovskite nitrides with non-collinear magnetic structure. We will focus on interface induced deviations from the triangular antiferromagnetic structure present in several bulk Mn-based antiperovskite nitrides that has been shown to host effects such as Anomalous Hall Effect (AHE), Magneto-optical Kerr Effect (MOKE), or Anomalous Nernst Effect (ANE) all of which originate in electronic structure with nonzero Berry curvature.  The main aim of the project is to find chemical and structural compositions of the heterointerface where the unique magnetic structure is preserved and can be manipulated via the ferroelectric polarization of the oxide layer or lattice strain. Using density functional theory (DFT) we will relax the geometry to find the equilibrium structure considering different atomic layer terminations at the interface. Subsequently, we will employ linear response theory to simulate magneto-optical (MO) spectra for the equilibrium canted magnetic structure determined in the previous step (assuming only the antiperovskite unit cell). We will build on the results and methodology of project OPEN-17-26 and OPEN-19-34 where we studied the sensitivity of magnetic ordering to chemical composition and lattice strain in bulk Mn-antiperovskite nitrides. Comparison to available MO spectra will allow us to identify samples with potential for applications in solid-state cooling, magnetic sensors and actuators.

 

Researcher: Jiri Brabec
Project: Development and application of machine learning models for strongly correlated systems
Allocation: 1 657 000 core hours
Abstract: We develop machine learning models for very fast estimation of correlation effects in the strongly correlated systems. The training set is formed from thousands of artificial systems composed from transition metals and various ligands, with different geometries and spin states. On this set we perform DMRG calculations, from which we extract energies as well as one- and two-orbital entropies. Additionally, we would like to test the performance and transferability on real challenging systems, including FeS clusters or polycyclic aromatic hydrocarbons.  

 

Researcher: Matus Kaintz
Project: STRUctural TUNing of band gap in doped-diamonds (STRUTUN)
Allocation: 1 593 000 core hours
Abstract: Diamond as a material of extreme properties has already proved to be a very promising candidate for a vast variety of electronic applications. As most of the research done so far was devoted to study just the influence of dopant on the electronic structure without any connection with the structural properties, this offers the way for further improvement of our knowledge. To this aim, the presence of dopant atoms in combination with vacancies will be considered in the diamond structure. The focus will be on understanding the coupled structural-electronic features governing the electronic properties of each system, and finally the bandgap. Density functional theory calculations will be used to identify such features, in order to obtain a detailed quantum mechanical description. Outcomes of the project will serve as a guide for the design of new electronic materials with targeted behaviour for both p- and n-type diamond semiconductors.

 

Researcher: Ctirad Cervinka
Project: Glass transition temperatures of active pharmaceutical ingredients dispersed in low-molecular biocompatible materials
Allocation: 1493000 core hours
Abstract: Molecular modeling has emerged to the stage that new pharmaceuticals can be developed with computer-aided tailored-design of the active pharmaceutical ingredients (API) with respect to the desired pharmacological function of the API in the patients’ body at the atomic level. Real efficiency of numerous API is, however, also governed by their solubility in aqueous environment in organisms. Although being a simple thermodynamic property, the solubility can render many promising candidate molecules completely useless for pharmacological treatments. A majority of API is formulated as solids, mostly in the crystalline state. At low temperatures, a crystalline phase is always the thermodynamically stable and all other metastable phases (e. g. liquids, amorphous solids) tend to undergo a phase transition to the most stable state. Kinetic barriers depending on the material structure and outer conditions (temperature, pressure, purity, solvent, etc.) stabilize the metastable phases by impeding the stabilizing phase transitions. Thermodynamic rules imply that the metastable phases always are more volatile and better soluble than the most stable phase. As a consequence, metastable amorphous formulations of API seem to be a promising way how to increase efficiency of otherwise poorly soluble crystalline APIs. Molecular-dynamics (MD) investigation of beneficial roles of dispersing media (other compounds) is then crucial to optimize the solubility and stability of the amorphous dispersions of API.

 

Researcher: Robert Babjak
Project: Radiation generation with ultra-intense lasers
Allocation: 1 267 000 core hours
Abstract: For more than a century, particle accelerators have been revolutionizing science, medicine, and industry. One of the prominent directions towards societal applications is to use a relativistic electron beam coming from the plasma-based accelerators to generate radiation. Promising approach to generate electron beam with sufficient properties is to use the ultra-relativistic pulses interacting with plasma. In such interaction regime, the high-brilliance betatron X-rays are emitted by relativistic electrons that wiggle in the background field provided by the plasma wave. It is possible to generate even higher frequency radiation, all the way to the Gamma-ray range, through the interaction of ultra-short and ultra-intense lasers with relativistic charged particle bunches using the Compton scattering. The main scientific question to be addressed is thus on how to design a laser-plasma based source/target that can provide photon beams, with an emphasis on realistic target geometries, evaluating different target structures and resistivity components. The work will cover particle acceleration mechanisms, either in underdense or in overdense targets, also exploring novel physical mechanisms and the potential for their use as elements in compact radiation sources. These studies will be performed with the use of full-scale 3D particle-in-cell simulations.

 

Researcher: Jakub Sebesta
Project: The impact of final temperatures on the magnetism of transition metal alloys from ab-initio
Allocation: 1 493 000 core hours
Abstract: Common calculations employing the density functional theory deals with the ground state behavior only. However, physical properties strongly change with temperature. Regarding magnetic materials, temperature evolution modifies e.g. the magnetic ordering type or the strength and character of magnetic exchange interactions. Thus it is evident that the ground state calculations are insufficient for material description at real conditions e.g. at room temperature. However, the finite temperature description is challenging even for common metals, particularly due to the computational demands. In the proposed project, we would like to inspect the impact of final temperatures on the magnetic behavior of 3d magnetic elements and their alloys, employing solely first principle calculations. The obtained detailed insight into the magnetism of such materials at final temperatures can help one better understand how certain magnetic materials behave in real conditions, which is essential to design new outstanding materials with great application potential.

 

Researcher: Diego Lopez
Project: Extensive Pursuit of Singlet Fission Sensitizers (EPSFS)
Allocation: 6 885 000 core hours
Abstract: The transformation of sunlight into electricity represents one of the major tasks that scientists are facing nowadays. The traditional silicon-based photovoltaic (PV) devices are not free of disadvantages, such as their expensive and polluting fabrication or their low performance in low-light conditions. Improving the efficiency of PV devices will drastically reduce the cost of the electricity, and strategies such as the downconversion of short wavelength photons were proposed. One of the main phenomenons under investigation nowadays is called Singlet Fission (SF), a spin-allowed photophysical process in which one singlet excited state splits into two triplet states. This process might enable the generation of two low-energy excitons per absorbed photon. However, the development of this new technology is still impeded because the number of molecules in which SF was experimentally found is still small, and only a few new systems could be added the paradigmatic acene derivatives. Under those circumstances, it is captivating the approach recently published by Padula et. al. in which the Cambridge Structural Database (CSD) was computationally screened. At the end of this study, a final set of around 200 SF absorbers were highlighted among the initial list composed by 1 million candidates. In this project, 50 million molecules selected from the PUBCHEM database will be screened investigated by means of DFT in order to obtain the best candidates for SF.

 

Researcher: Christopher Heard
Project: Unravelling the deactivation pathways of encapsulated single atom catalysts
Allocation: 1 546 000 core hours
Abstract: Single atom catalysts (SAC) represent the maximum atom-efficiency of all heterogeneous catalysts, and have shown extremely high activity and selectivity in important industrial processes, from passive NOx adsorption in vehicle exhausts,[1] to selective catalytic reduction of ammonia.[2] Encapsulation into porous frameworks is a popular means to stabilise particles against deactivation via sintering, which is the major limitation that holds back utilisation of encapsulated SACs today.[3,4] However, little is currently understood about the processes of atom migration or particle growth in these systems, and therefore, strategies to limit catalyst deactivation remain unreliable and non-general. In this project we will investigate the binding, migration, sintering mechanisms and kinetics of catalyst deactivation of noble metal catalysts inside zeolite pores via two complementary methods: i) global structure optimisation (GO) at a quantum mechanical level and the elucidation of migration networks for Pt SAC in the industrially important zeolite CHA. ii) Development and application of neural network potentials to explore long-timescale diffusion and growth behaviour of small Pt clusters in a range of zeolite toplogies. We aim to determine the role of zeolite topology, channel dimensionality and confinement. This project will advance our understanding of the atomic-scale behaviour of SACs, and help predict deactivation routes for a class of industrially important catalytic materials.

 

Researcher: Martin Docekal
Project: Transfer Learning for Keyphrase Extraction
Allocation: 1 246 000 core hours
Abstract: Nowadays, we live in a world of vast amounts of text data. It becomes nearly unthinkable that a human would be able to go through a more significant number of documents and decide which are worthy of reading without using an automatic tool such as a search engine. However, even search engines can return a potentially large list of documents. In such cases, we can help the reader by providing keyphrases that are relevant to the given text.

 

Researcher: Tomas Karasek
Project: Aeronautics Large-scale Pilot of LEXIS project
Allocation: 1 317 000 core hours
Abstract: Avio Aero has launched a challenging research activity aimed at significantly improving the feasibility and exploitation of advanced numerical modeling capabilities for critical engine components. The synergy between new generation HPC platforms and Big Data management technologies will open new scenery for the design and optimization of aeroengines, enabling innovative investigation strategies and providing unprecedented levels of accuracy and detail.

 

Researcher: Michael Bakker
Project: Benchmarking MD/DFT Calculations of 31P Chemical Shifts in the C-Terminal Domain of RNA Polymerase II
Allocation: 888 000 core hours
Abstract: This project plans to compute 31P nuclear magnetic resonance (NMR) chemical shifts (CSs) for RNA Polymerase II (RNAPII CTD), an intrinsically disordered protein (IDP). IDPs are of particular interest due to their connection with neurogenetic diseases (e.g. Alzheimer´s or Parkinson´s). The C terminal of RNAPII is comprised of heptad repeats regulating transcription and cotranscriptional RNA processing, of great interest to oncology. Structural characterization of these proteins is prerequisite for understanding interactions amongst biomolecules. Unfortunately, flexibility inherent in IDPs is a barrier for investigation; traditional techniques (e.g. x-ray crystallography) are unfeasible and alternative techniques such as NMR spectroscopy become complicated. Computational methods, such as those proposed in this project, can be used to alleviate these problems. Snapshots taken from a molecular dynamics (MD) trajectory are employable to create sufficiently ergodic conformational ensembles better reflecting the flexibility in simulations. This project´s aim is to evaluate the implementation of MD and benchmark against lower-level QM simulations. Two particular heptad repeats were selected containing phosphorylated serine (pSer) and phosphorylated tyrosine (pTyr), both commonly phosphorylated aminoacids.

 

Researcher: Pavel Jungwirth
Project: Coalescence of solvated electrons in ammonia solutions: investigating nonmetal-metal transition in the context of percolation theory
Allocation: 789 000  core hours
Abstract: The present project focuses on the direct investigation of solvated electrons originating from dissolving increasing amounts of alkali metals in bulk liquid solvents through electronic structure calculations and ab initio molecular dynamics (AIMD). The goal is to provide insight into the transition from individual solvated electrons through dielectrons to the onset of more delocalized states. To begin with, we would like to investigate the possibility that the transition from blue to the bronze-coloured metallic solution and the associated increase of conductivity [1, 2] can be described in the framework of percolation theory. This amounts to checking whether the rise of delocalized states from distinct individual electrons and dielectrons can be modelled as a percolation transition. The power of percolation theory lies in the fact that it is capable of describing critical phase transitions of rich physical content, yet it may be formulated in terms of very simple geometrical concepts. It is a versatile model, with applications to such diverse problems as supercooled water, galactic structures, fragmentation, porous materials, earthquakes, and many more [3]. In percolation theory, a random system is described by a percolation threshold Pc representing the critical value of connectivity needed for the emergence of clusters spanning the whole system. Below the threshold, such connected components do not exist, while above it, a giant cluster of the order of system size can be observed [3]. In this project, we are proposing a hypothesis that the delocalized states in metallic ammonia solutions should differ in the percolation threshold value from the states of the distinct individual electrons and dielectrons. For this purpose, we plan to perform an extensive sampling of the electron density of alkali-metals ammonia solutions through ab initio molecular dynamics simulations on a hybrid DFT level utilizing advanced and heavily parallelizable simulation algorithms.

 

Researcher: Lukas Jeremias
Project: NMR and EPR Properties of Trimeric Copper(II) Complexes
Allocation: 879 000 core hours
Abstract: Nuclear magnetic resonance (NMR) spectroscopy is a widely employed method for characterizing the structures of new chemical compounds and biomolecular systems. However, this technique is generally less successful for open-shell molecules and metal complexes with unpaired electrons. Despite the difficulties (very unusual position of signals, signal broadening etc.), paramagnetic NMR (pNMR) spectroscopy is becoming very important in areas of research such as the structural characterization of paramagnetic metalloproteins, magnetic resonance imaging, or the development of molecular magnets. The interpretation of the experimentally obtained pNMR spectra rely mostly on DFT (Density-Functional Theory) calculations. For mononuclear complexes with one unpaired electron, the situation starts to be handled well especially for small molecules. The mononuclear complexes with more than one unpaired electron or complexes with more paramagnetic metal centers are still challenging for DFT calculations. The computational resources provided by IT4I infrastructure will be used for fully relativistic calculations of magnetic response properties of trinuclear paramagnetic complexes.

 

Researcher: Monika Stas
Project: Art of drug design – predicting binding affinity of ligands based on strain and solvation/desolvation energy
Allocation: 718 000 core hours
Abstract: In-depth understanding of protein-ligand interactions plays a key role in the drug design process. The strength of such interaction, denoted as binding affinity, depends on many factors and it is the main criterion for assessing small ligands as potential drug candidates. Although it has been extensively studied, the available approaches cannot predict in silico binding affinities correctly. Quite often, different experimental binding affinities are obtained from the similar or even the same interaction pattern of protein-binding ligands. It shows that there are also other factors affecting the ligand binding. One of these factors is the deformation/strain energy of the ligand. This can be envisaged as the energy penalty for the ligand to adapt its shape to fit the (semi)rigid protein active site when moving from the aqueous solvent to the protein. At the same time, the ligand undergoes a change in its solvation/desolvation energy. These two - strain energy and ligand solvation / desolvation were often overlooked in ligand optimization(s) - i.e. in the process to find a lead molecule for potential drug. The quantum chemical calculations for medicinally important systems: Cyclophilin A, Mitogen Activated Protein kinase 14 p38α, Factor A, HIV transcriptase, and HIV protease will allow us to approximate those energies. As a result, we will be able to estimate the contribution of those two energies in binding free energy based on the available experimental data. Overall, this could lead to further improvement and simplification in the computational approach to drug discovery. 

 

Researcher: Michail Kourniotis
Project: Resolving thermal instabilities and gaseous structures driven by massive star feedback in young clusters
Allocation: 704 000 core hours
Abstract: One of the intriguing open issues in observational astronomy concerns the evolution and composition of globular clusters, groups of metal-poor stars that are typically found in the surrounding of the Galaxy and other galaxies. Tightly bound by gravity, these massive (10^4 to 10^6 solar masses), spheroidal collections of stars are nowadays found to host multiple stellar generations showing significant deviations in age and chemical composition, particularly in light elements such as helium and nitrogen. One of the most promising theories of the globular clusters origin attributes such diversity to thermal instabilities occurring in winds of massive stars. This process can be efficiently studied by numerical methods. Throughout our recent run in IT4/Salomon (OPEN-18-41), we managed to successfully recover the gaseous seeds that can potentially induce the in-situ formation of new stars to be born within the existing stellar population. A side outcome of that study was the puzzling formation of disk structure in the central region of the cluster undermined though by limited resolution. Currently, we propose to undertake a detailed study in the central region of the young massive clusters in search of physical explanation of the disk structure against the possibility of a computational artifact and explore its dependence on the initial parameters and computational setup.

 

Researcher: Zdenek Masin
Project: Electron- and photon-induced processes in model molecules of the subunits of DNA and in small inorganic molecules
Allocation: 377 000 core hours
Abstract: Collisions of low energy (< 30 eV) electrons with molecules and molecular photoionization are processes that contribute to direct and indirect damage to DNA through, for example, dissociative electron attachment and dissociative photoionization. Studying these fundamental processes in constituent molecules of the DNA is therefore of practical interest. However, the constituent molecules are complex with dozens of electrons. Therefore it is desirable to gain insight into these processes as they occur in smaller molecules and their prototypes: in smaller systems the electronic dynamics can be studied more accurately. In this project we propose to study two topics. In the first topic we will perform calculations of electron collisions and single-photon photoionization of formic acid dimers that serve as a simple model for base-pairing in DNA. Concretely we will be studying the formation of temporary negative ion states and explore the effect they have on breakup reactions. Secondly we will study photoionization processes in smaller molecules induced by strong and ultrafast laser pulses in order to gain general insight into the role of multi-electron dynamics in photoionization. For this work we will apply state of art ab initio UKRmol+ and RMT suites of codes to study attosecond time-delays, generation of electron vortices and other multiphoton processes in smaller molecules such as CO, N2 and H2O and explore performing such high-quality calculations in larger molecules.

 

Researcher: Martin Kolisko
Project: Impact of lateral gene transfer on the evolution of diplomonads
Allocation: 357 000 core hours
Abstract: Diplomonads are a group of microbial eukaryotes that include medically and economically important parasites. Additionally, there are several putatively secondarily free-living species and host-associated commensalic species. Previous studies have suggested that lateral gene transfer (LGT) played important role in the evolution of diplomonads. However, the impact of LGT on microbial eukaryotic evolution remains unclear and controversial. Here we propose to identify LGT genes in 16 diplomonad isolates, 12 of which have been newly sequenced for our study. The results will address a fundamental question in evolutionary biology; does lateral gene transfer play important role in adaptation to different lifestyles – parasitic, commensalic or secondarily free-living. 

 

Researcher: Tomas Karasek
Project: Validation of exascale demonstrators
Allocation: 7 746 000 core hours
Abstract: This proposal complements the ExaQUte “exascale” project, which aims at constructing a framework to enable Uncertainty Quantification (UQ) and Optimization Under Uncertainties (OUU) in complex engineering problems using computational simulations on Exascale systems (http://exaqute.eu/ ). As it is well-known, UQ and OUU problems rely on the estimation of unknown quantities by performing many independent simulations on different scenarios. This poses an important challenge as many computational resources is needed to perform these analyses. Nonetheless, the independent sampling allows for high parallelization of the analyses in order to exploit HPC systems. The methods and tools developed in ExaQUte will be applicable to many fields of science and technology. The application chosen as a demonstrator focuses on wind engineering, a field of notable industrial interest for which currently no reliable solution exists. This will include the quantification of uncertainties in the response of civil engineering structures to the wind action, and the shape optimization taking into account uncertainties related to wind loading, structural shape and material behaviour. Wind also plays an important role in the assessment of citizens comfort, particularly when large constructions are to be made (e.g. re-modelling of city areas, construction of high-rise buildings), becoming this an application of public interest.

 

Researcher: Valeria Butera
Project: CO2 conversion into biofuels and fine chemicals: a DFT investigation
Allocation: 608 000 core hours
Abstract: CO2 capture and conversion into added-value products has two-fold benefits: first, it can contribute  to drastically reduce its content in the atmosphere and secondly, it can be used as carbon source for the production of biofuels as a green alternative to the limited fossil fuel resources. However, due to its high thermodynamic stability and low reactivity of CO2, catalysts are needed to reduce the activation energy of any kind of reaction in which it is implicated. This proposal addresses the development of two different types of efficient catalysts – homogeneous compounds and nanostructured GaN -  by means of DFT investigation for the conversion of CO2 into biofuels and other added-value products.

 

Researcher: Georg Zitzlsberger
Project: Evaluation of Deep Neural Network based Urban Change Detection in Remote Sensing
Allocation: 670 000 core hours
Abstract: Since the dawn of satellite based remote sensing, land change detection significantly has evolved and is ubiquitous in many domains nowadays. To name a few examples, remote sensing based change detection is used to study the change of climate and vegetation, controlling the sprawl of urban settlements, or detecting and measuring human made ecological impact. In our work, we train a deep neural network to automatically detect urban changes over time using a novel approach to combine both Synthetic Aperture Radar (SAR) and multi-spectral observations to increase the accuracy, compared to multi-spectral only observations. We consider over two decades of satellite observations (missions ERS 1&2, and Landsat 5) of three different sites (Rotterdam, Liege, and Limassol) with different land cover characteristics. Our approach defines a framework with methods which provide an automatic process in order to apply our solution to other sites, satellite missions, such as Sentinel 1 & 2, and land cover types, with minimal manual intervention. Our work is based on the results of ESA’s Blockchain ENabled DEep Learning for Space Data (BLENDED) project, on which we build on. We further optimize and tune the suggested deep neural network to maximize the prediction accuracy. Our framework and results will be publicly made available for others to adopt.

 

Researcher: Petr Valenta
Project: Relativistic Mirrors in Laser-Plasma Interaction IV
Allocation: 610 000 core hours
Abstract: A relativistic mirror is an object that reflects incoming radiation while moving at velocity close to the speed of light. The theory of light reflection from such mirrors was formulated by Einstein in 1905. Within the scope of this project, we plan to study the physical realization of relativistic mirrors using strongly nonlinear Langmuir waves driven by an intense laser pulse in plasma. Second (counter-propagating) laser pulse undergoes a reflection from this mirror. The reflected radiation is compressed, amplified and its frequency is upshifted due to the double Doppler effect. This way, one may prepare a source of electromagnetic radiation with unique properties.
We will use large-scale computer simulations to numerically investigate the properties of laser-driven relativistic mirrors in plasma. We will be optimizing the reflection coefficient of the relativistic mirror as well as the factors of amplification and frequency upshift of the reflected electromagnetic wave. We will be working closely with the experimental team in order to design a setup for upcoming experiments using the state-of-the-art laser systems currently being built at the ELI Beamlines facility. The ultimate goal of our research is to develop compact and tunable source of coherent high-brightness radiation with wavelengths ranging from x-rays to gamma-rays.

 

Researcher: Ales Podolnik
Project: Flush-mounted probes for the COMPASS-Upgrade tokamak
Allocation: 421 000 core hours
Abstract: Operation of the future fusion devices is substantially affected by the physics of the outermost plasma layer, the edge plasma. To understand its behavior and to enable its control, proper plasma diagnostic capabilities are crucial. One of such diagnostic methods is the Langmuir probe and its various modification. The upcoming tokamak COMPASS-Upgrade will be equipped with a variety of the probe diagnostic and in its initial phase, the flush-mounted probes will be one of the chosen instruments to be operated in the parts of the device where the lower incident heat and particle fluxes will not be able to damage the probe material. Their construction is fairly simple and probe data are usually simple to be analyzed, however in the intended geometry, those probes will operate at the limits of the theoretical assumptions, under which the measurement results are inferred from the obtained data. To understand the operational conditions better, the particle-in-cell simulations are a suitable tool for such task. In this project, the simulation of current colection and mapping of the flux profile on the probe surface will be performed to aid the probe design and construction.

 

Researcher: Prashant Dwivedi
Project: HYPER-II (Hyper velocity impacts)
Allocation: 635 000 core hours
Abstract: Fusion power is probably the most sought-after technological goal in the pursuit of clean energy. Nuclear fusion, the nuclear reaction that powers the Sun and the stars, is a potential source of safe, non-carbon emitting and virtually limitless energy. Therefore, nuclear fusion would help curb greenhouse gas emissions and meet the increasing energy needs of future generations, and is part of the Horizon 2020 [1]. Replicating that process on earth at sufficient scale could unleash more energy than is likely to be needed by humanity, but the problem is creating the extreme conditions necessary for such reactions to occur, harnessing the resulting energy in a useful way, and controlling the reactions once they have been induced. Today fusion power is closer and closer. Harnessing fusion's power is the goal of ITER (35 countries involved), which has been designed as the key experimental step between today's fusion research and tomorrow's fusion power plants [2]. Still formidable many challenges must be overcame; Plasma-material interactions are a key issue in the realization of practical fusion power reactors, which is recognized since the beginning of magnetic fusion research. Controlling plasma-wall interactions is critical to achieving high performance in present day tokamaks, and this is likely to continue to be the case in the approach to practical fusion reactors. Tungsten (W) is the main candidate-plasma facing materials (PFM) for a fusion reactor and will be exclusively used in the ITER divertor [3]. Among other problems to be solved, the presence of high velocity impacts of "dust" (heavy atoms, clusters of them etc) against the PFM has been reported in several studies, with velocities being around 500 m/s to a few km/s. PFM´s is one of the research activities in the Road map to Fusion during Horizon 2020 through a Joint programme of the member of the EUROfusion consortium[4, 5, 6].

 

Researcher: Jaroslav Hron
Project: Patient specific geometry flow simulations IV
Allocation: 315 000 core hours
Abstract: The goal of carotid endarterectomy is prevention of ischemic stroke, one of the most common causes of morbidity or mortality in developed countries. The current indication criteria are primarily based on the grade of stenosis caused by the atherosclerotic plaque. The development and the character or the plaque are influenced by the hemodynamics in the carotid arteries. The goal of our project is to describe the relationship between the hemodynamic parameters and the character of the plaque. The main goal is to develop efficient numerical tools for reliable CFD analysis in patient specific cases. 

 

Researcher: Miroslav Voznak
Project: Advanced mobile networks data analysis and data volume handling
Allocation: 299 000 core hours
Abstract: Our goal is to deliver consistent data about population mobility retrieved from the mobile networks for future analysis by professional domains experts. This task requires ongoing and repetitive effort in search for the best calculation approach and robust, thus simple methods of big data volume processing. Therefore, we recalculate big volumes of data based on recent development to keep methods consistent. This would not be possible without HPC.

 

Researcher: Pavel Balaz
Project: Construction of phase diagram of a two dimensional skyrmion lattice
Allocation: 414 000 core hours
Abstract: Theory of topological phase transition, awarded by a Nobel Prize in 2016, predicts an unusual behavior of matter when the spatial dimension is reduced. It has been shown by Kosterlitz, Thouless, Halperin, Nelson, and Young that a novel phase, so called hexatic phase, can be observed in a system of two dimensional (2D) array of discs. Their theory, known as KTHNY, has attributed the observed second order phase transitions to the unbinding of topological defects, know as dislocations and disclinations. On the other hand, in more complex magnetic systems, featuring asymmetric exchange Dzyaloshinskii-Moriya interaction, formation of 2D topological vortex-like structures – so called skyrmions — has been predicted and observed experimentally. Under specific conditions (given by applied magnetic field and temperature) skyrmions can form a regular lattice resembling the one of solids but in two dimensions. Varying external conditions one can stabilize the skyrmion lattice or melt it into disordered liquid-like phase. In this project we shall study the intersection of the theory of 2D topological phase transitions and skyrmion systems. Since skyrmions in magnetic materials can be treated as 2D spherical particles, we shall apply the theory of topological phase transition in order to construct a phase diagram describing the phase transitions of the skyrmion lattice in a class of materials known as lacunar spinels. 

 

Researcher: Victor Camps
Project: Quantifying the effect of anisotropy on the energy cascade rate in compressible plasma turbulence.
Allocation: 404 000 core hours
Abstract: Turbulence is a non-linear phenomenon present in a great variety of physical contexts, from geophysical flows to the interstellar medium. It consists on the transmission of energy from large scale structures, such as whirls, to smaller scales, at which energy is dissipated. Thus, it provides a sources of heating for the different systems where it is present.
In hydrodynamics and plasma physics, the so called exact laws or Karman-Howarth-Monin-Politano-Pouquet (KHMPP) equations provide a quantitative description of the energy that flows from large to small scales due to turbulence, the energy flux between scales. These equations are directly derived from the basic equations describing the neutral or plasma fluids. However, it is still necessary to make assumptions on the anisotropy of the energy flux when analyzing the 1D data collected by most experiments.
The solar wind is the layer of the solar atmosphere that extends from 2000 solar radii over the Sun's surface up to the confines of the solar system. Turbulence in this plasma is affected by a large variety of parameters and the energy flux is considered to be highly anisotropized. By means of 3D Hall-MHD numerical simulations, we seek to study the energy flux derived from KHMPP equations in 3D space and evaluate how valid are the assumptions made about its anisotropy. We will mainly focus on large and “intermediate” plasma scales, at which the movement of protons is not entirely coupled with the mean magnetic field.

 

Researcher: Petr Macha
Project: Simulation of tunnel probe
Allocation: 394 000 core hours
Abstract: Tokamak plasma diagnostics is essential for understanding the physics of the future tokamaks (ITER, DEMO). Especially, edge plasma physics has the key role in understanding of tokamak plasma equilibrium. One of the method how to measure edge plasma parameters is using electric probes. These probes are able to measure with high time and spatial resolution. Tunnel probe is an advanced type of Langmuir probe, which has been invented to measure parameters in plasma’s scrape-off-layer. The tunnel probe consists of two electrodes, tunnel, and back-plate, where the tunnel axis is parallel to magnetic field. Plasma flows into the open orifice and currents are distributed to the electrodes. By using 2D cylindrical particle-in-cell code PICCYL it is possible to model flow of charges inside the cavity and to calibrate the probe on the ion current ratio flowing between tunnel and back-plate. This ratio depends on electron temperature. Using this fact, it is possible to measure electron temperature with high time resolution by measuring ion current ratio after calibration is performed. Furthermore, because of its concave geometry, the electric field is contained inside the cavity and therefore there is no sheath expansion.

 

Researcher: Ondrej Chrenko
Project: Hydrodynamic interactions of planets with protoplanetary disks and the origin of close-in exoplanetary systems
Allocation: 1 521 000 core hours
Abstract: Recent astronomical observations have revealed that a substantial number of exoplanets exhibit relatively low masses (ranging between super-Earths and mini-Neptunes) and orbit close to their host stars. This subpopulation is usually referred to as close-in low-mass exoplanets. It has been hypothesized that close-in exoplanets formed early, while their natal protoplanetary disk was still around, and migrated towards the inner disk rim at about 0.1 au where they accumulated. If this is true, then close-in exoplanetary systems represent unique testbeds for our understanding of hydrodynamic processes of planet formation. Our aim is to perform hydrodynamic simulations of protoplanetary disks with embedded planets in order to identify processes that might have contributed to the formation of close-in exoplanets. First, we will study how low-mass planets migrate from moderate distances towards the inner disk rim. We will focus on previously unexplored phenomena such as the gravitational torques arising from a disk of pebbles. Second, we will study how multiple low-mass planets interact with the inner disk rim when their orbits become closely packed. Our results will contribute to the theory of planet formation, which will take us one step closer to understanding whether there can be life on other planets in our Galaxy. This project is directly related to the solution of a GAÄŒR Junior Star Grant (21-23067M).

 

Researcher: Martin Cadik
Project: Deep-Learning Approach to Topographical Image Analysis
Allocation: 394 000 core hours
Abstract: One of the fundamental tasks in computer vision and machine intelligence is information extraction and content analysis. When compared to superficial image analysis tasks, this requires the models to gain a deeper level of understanding of the underlying data. Our research focuses on the analysis and interpretation of images depicting natural environments. Specifically, this project deals with two main subtopics - visual geo-localization outdoors and perceptual analysis of natural models. Visual geo-localization is a research area focusing on the estimation of camera parameters (e.g., position and orientation) based on the content of visual media. The knowledge of such parameters enables numerous applications such as augmented reality, robot and unmanned vehicle navigation, or automatic organization of multimedia. Our research focuses on the visual geo-localization in natural environments, which is far less studied than camera pose estimation indoors or in urban areas. The perceptual analysis task aims at gaining deeper insight into procedurally generated natural objects, such as trees. The aim is to gather a wide range of tree species and ground-truth subjective perceptual quality scores. Consecutively, our goal is to computationally reproduce user-perceived quality (or "naturalness") of synthetic models of trees using recent machine learning methods.

 

Researcher: Olena Mokshyna
Project: Exploring the ligands binding to bystin target: continued study
Allocation: 375 000 core hours
Abstract: Bystin is a protein involved in the cell growth process and is overexpressed in human cancer cells. Despite bystin being a promising target for anticancer drugs and drugs against blackfan anemia, no site(s) of binding was previously identified. The only available crystal structures of bystin are those of pre-40S ribosomal subunit. No holo X-ray structures of bystin (i.e. bound to ligand) are available at the moment; and no ligands were previously described in the literature. This project continues the study of bystin and its interaction with small molecule ligands. Our previous findings have shown that bystin has two primary shallow binding sites. In the previous parts of this project, we explored ligand binding to the primary binding site using a range of state-of-art molecular simulation techniques (from replica exchange to metadynamics). We were able to establish stable binding poses for most of the ligands and distinguish two main groups of ligands with varying activity. The free energy calculations of ligand-protein systems were successful for the ligands synthesized in IMTM, and we were able to obtain the concise free energy profiles and observe ligand binding-unbinding path. The continuation of this project is going to be dedicated to several points of interest. The behavior of corticosteroids is one of those. Experimental findings show bystin to be one of their targets. We also intend to focus more on the group of ligands that bind to the secondary binding site. To get more robust computational confirmation we are going to use more complex collective variables, as well as simulation techniques such as funnel metadynamics.
Every part of this project will be conveyed in close collaboration with the experimental biologists.

 

Researcher: Michal Merta
Project: Development of BEM-based solvers IV
Allocation: 315 000 core hours
Abstract: One can choose from several numerical methods for modelling natural phenomena occurring in the real world, let us mention, e.g., the finite element method or the finite volume method. The main features of the boundary element method (BEM) make it well suited for problems stated on unbounded domains (such as sound or electromagnetic wave scattering) or shape optimization problems. Within the previous project, we have mainly focused on the development of the BEM-based solver for the time-dependent heat equation. The current project aims at optimization of its shared and distributed memory parallelization, development of new approaches to accelerate solution, and last but not least, porting part of the computation to GPUs. The global space-time approach leads to the possibility of parallelization both in space and time thus improves the scalability on current and future supercomputers. 

 

Researcher: Jiri Jaros
Project: Photoacoustic tomography of the breast V
Allocation: 315 000 core hours
Abstract: Photoacoustic tomography (PAT) is a biomedical imaging modality based on the photoacoustic effect. In PAT, non-ionizing laser pulses are delivered into biological tissues. Some of the delivered energy will be absorbed and converted into heat, leading to transient thermoelastic expansion and thus wideband ultrasonic emission in the low MHz range. The generated ultrasonic waves are detected by ultrasonic transducers and then analyzed to produce images. Since the optical absorption is closely associated with physiological properties, such as hemoglobin concentration and oxygen saturation, the PAT is used to visualize vasculature inside tumors with a very high resolution.
The purpose of this allocation is to reconstruct photoacoustic images of the breast of 20 patients scanned by the Pammoth imager.  The resolution, accuracy, sharpness, motion and noise artifact, and the depth of penetration will be investigated and optimized. This study is the final phase of the model validation and moves us towards the deployment in a real PAT system for breast mammography.

 

Researcher: Martin Mrovec
Project: Combination of metaheuristic and local optimization strategies on Grassmann Manifolds with applications in Electronic Structure Calculations
Allocation: 210 000 core hours
Abstract: A significant portion of supercomputing resources is spend by Quantum Physics and Chemistry calculations. There are several reasons for this. One is a persisting high computational complexity despite the theoretical progress that has been made in recent decades. Further, the results have several areas of applications such as pharmacy or the material engineering. There exist many software packages focused on such calculations. One of the usually solved tasks is a searching of minimal (Ground State) energy of a system of electrons and nuclei. The solution is searched by iterative methods where an initial guess is chosen and then a given procedure is repeatedly performed until the convergence is reached. Unfortunately, a common feature of many such methods is the lack of theoretical results proving their convergence. This is also reflected in practical calculations where users encounter situations, where the most commonly used methods fail to converge or they end up in a local minimum which does not represent the Ground State. Our research is focused on a development of new approaches and algorithms. On the one hand, we are focused on development of local optimization methods with a guaranteed convergence. On the other hand, we are interested in global exploration strategies that serve as tools for avoiding undesirable local minima. These strategies are based on the Swarm Intelligence and adapted to be suitable for solving Quantum Chemistry problems. In this project we plan to develop methods based on combining features of both approaches. 

 

Researcher: Jakub Vymola
Project: CPU and GPU scaling of DFT calculations, part II
Allocation: 158 000 core hours
Abstract: It is more and more evident, that GPU-accelerated nodes are a great tool to increase the performance of density functional theory (DFT) calculations. However, at IT4Innovations the GPU nodes are currently not heavily used.
In this study we promote scaling tests of Vienna Ab initio Simulation Package (VASP) for number of compounds, i.e. one containing f-electrons (U2C3), metallic one (PtN2 in fluorite structure) and an insulator (pyrite structure of PtN2) for both CPU and GPU nodes. As shown by some, the speedup achieved by using GPU enabled code for can be as high as 5-10x [1], depending on the number of GPUs per node. With this claimed gigantic performance increase, it is necessary to have solid data on the scaling behavior.
We intend to compare the speeds of calculations on different numbers of compute nodes. We aim to find and verify the fastest parallelization configuration of VASP for CPU nodes for each of the test cases. After that, we perform the very same tests on GPU nodes and compare the final results with respect to the total amount of resources consumed. We aim to provide solid insight into the use of GPU on IT4Innovations machines and to show how to utilize them efficiently. We are hoping to reduce time and resources usage and to lower the price for research that heavily relies on the use of VASP code. References: [1] https://www.nvidia.com/en-us/data-center/gpu-accelerated-applications/vasp/

 

Researcher: Martin Beseda
Project: Computations of [N2/He]+ potential energy surfaces 4
Allocation: 138 000 core hours
Abstract: This project is a direct continuation of OPEN-20-20, OPEN-20-34, DD-20-23 and DD-20-29. The computation time we ask for will be utilized solely to obtain several potential energy surfaces of [N2/He]+ collision complex together with its gradients and their further representation by artificial neural networks. As such, the reasoning of the former project will be re-used in this proposal.
Considering the former one , it was shown experimentally, that rare-gas plasmas are well-working in medical applications. To understand the healing properties of cold rare-gas plasmas, however, detailed knowledge of processes is of crucial importance. We are mainly focused on the interaction of He with N2, i.e. this part is a direct continuation of the OPEN-14-25 project. Considering the computational demands of [N2/He]+ complex, the project will contain a thorough convergence analysis, taking different active spaces, basis sets and point groups available for a  given geometry, into consideration.

 

Researcher: Ivo Oprsal
Project: Modeling of the seismic waves in 3D regional and local structures
Allocation: 237 000 core hours
Abstract: High impact of earthquakes in urban areas is a consequence of significant casualties and earthquake damage caused to infrastructure in a number of locations over The World, every year. An earthquake is manifested by strong ground motions, generated in earthquake source, travelling through regional geology, and finally reaching a local geological setting. The wave propagation can be modeled by Finite-differences (FD). Added to that, the FD can also simulate ambient vibrations as part of the local geology research. A large variety of numerically simulated ground-shaking scenarios represent all earthquakes potentially appearing in a respective region. The synthetic data is then used to estimate earthquake impact on structures in metropolitan area. One of our goals is to evaluate local site effects due to large regional and subduction earthquakes for the Osaka Bay region. The relevancy of the expected simulations lies in scientific output and in their direct potential to efficiently mitigate and fight seismic hazard via engineering anti-seismic structure design, re-enforcement of existing structures, sophisticated urban projecting, and disaster mitigation planning with important economical and societal impact.

 

Researcher: Luigi Cigarini
Project: First-principles investigation of the thermoelectric conversion efficiency of phosphorene and arsenene
Allocation: 158 000 core hours
Abstract: New pathways for energetic conversion could face the problem of greenhouse effects on climate change. Thermoelectricity converts energy from thermal sources to electricity (or the opposite). It could be exploited to improve new type of renewable generators, to recover heat dispersion or even for small power sources like those necessary for wrist watches from body temperature. But research of new materials is necessary and computers will help us in simulating their properties and suitability.

 

Researcher: Vladislav Pokorny
Project: Magnetic impurities on superconducting surfaces
Allocation: 159 000 core hours
Abstract: Superconducting nanoscopic devices like Josephson junctions already become a standard building blocks on various technologies including rapid single flux quantum (RSFQ) electronics and qubits for quantum computing. If a superconductor is coated with an insulating or semiconducting layer,  we obtain a hybrid system in which various quantum mechanical phenomena as superconductivity, electron correlations and quantum tunneling can be separately tuned to obtain the desired properties. Understanding the complex interplay of these phenomena is necessary step in developing a new generation of devices as quantum supercurrent transistors or monochromatic single-electron sources. Supercomputers are now a necessary tool that helps us to build our theoretical understanding and explain the available experimental results before these devices can become the tools to extend the abilities of the current silicon-based electronics.

 

Researcher: Dana Nachtigallova
Project: Photoinduced charge-separation of functionalized carbon allotropes — the root to development of materials for energy conversion and storage
Allocation: 4 140 000 core hours
Abstract: The carbon-based materials are strong candidates for energy conversion and storage, one of the most urgent tasks for future energy scenarios. The caged carbon allotropes, in particular, have been investigated over the last decade to explore their capabilities in this field [1,2]. Numerous studies have been reported on designing and functioning the so-called second-generation nanocarbons, i.e., functionalized carbon-based nanomaterials, in the photoinduced charge-transfer leading to charge-separated state. [3].  The course of the excited-state transfer depends on several factors, including the carbon allotrope's chemical content and structure, the light-absorbing functional groups' character, and their mutual interaction. Despite intense research on this process, a fundamental step in solar energy storage [1-4], reports on their computational studies are rare. The complex nature of the photoinduced processes, the size of the systems, and frequently their strongly electronically correlated character make the computational description very challenging. The proposed approach based on the use of the DMRG methods, capable of describing strongly correlated systems, combined with the DFT treatment, offers an efficient and accurate approach to provide deeper insight into the phenomena underlying this photoinduced reaction. With the experimental studies, which will be designed in line with the calculations, the proposed studies will help design new, more efficient materials.

 

Researcher: Ondrej Soucek
Project: Viscous relaxation of Enceladuss ice shell
Allocation: 186 000 core hours
Abstract: Saturn's moon Enceladus has become one of the primary targets of planetary research in the past few years thanks to the Cassini mission discovery of water plumes emanating from its south pole [1]. A series of follow-up measurements and studies confirmed the presence of a global subsurface ocean, whose long-term stability is only possible if a strong heating source is active inside the body [2]. Deformation by Saturn's tides due to Enceladus's eccentric orbit has been accepted as the primary energy source driving the internal processes within Enceladus. However, a debate goes on concerning the location and importance of various dissipative processes within the moon and their contribution to the overall heat budget of Enceladus. Among these, the dissipation in the porous core and dissipation in the ocean have been targeted as possible major heat sources ([3],[4]), while the study [5] showed that dissipation within the icy shell plays probably only a secondary role. Gravity field inversions of Enceladus's shape (e.g. [6]) suggest a significant thinning of the shell in the south-polar terrain, consistently with the observed plume activity. The non-hydrostatic shape variations are most likely Figure 1: Ice velocity at the ice-ocean interface. maintained by significant lateral heat flux variations in the ocean, with a probable source inside the porous rocky core ([3]). Cadek et al. have studied the associated viscous relaxation of such geometry and observed a surprising flow pattern [7]. The extreme viscosity variations with depth, especially in the south polar region, where the cold surface at a mean temperature of approximately 59K gives rise to a viscosity exceeding 10 20 Pas while melting conditions (273K) take place just a few km underneath, corresponding to viscosity of the order of 10 14 Pas. This seems to lead to development of a narrow layer flow structure, very specific for the icy-moon setting. This proposal is dedicated to a more detailed study of such flow patterns and their interpretation based on finite-element three-dimensional numerical modeling.

 

Researcher: Vojtech Cima
Project: Deep Learning for Diabetic Retinopathy Detection
Allocation: 99 000 core hours
Abstract: Eye retina provides a unique non-invasive direct insight into the human body's bloodstream. It is currently being used in medicine to diagnose a number of diseases. It has been proved that retina artifacts may help to discover the early stages of diseases such as diabetic retinopathy, age-related macular degeneration, glaucoma, and other common diseases. Nowadays, the diagnosis of these diseases is conducted mainly by highly-specialized medical professionals - doctors based on a thorough visual evaluation of various eye-segment screens. In recent years, deep learning-based approaches have surpassed human performance in many domains including image recognition, object detection, or image segmentation also applied in the context of medical data. With this domain progressing swiftly forward, we aim to research, design and develop a deep learning-based solution that makes the diagnosis more efficient, more accurate, faster than the current manual process.

 

Researcher: Martin Surkovsky
Project: Deterministic Road Traffic Simulator
Allocation: 79 000 core hours
Abstract: A usage of historical data observed from traffic combined with high performance computing resources for traffic simulation can optimize traffic flow in a macro scale level. A set of algorithms used for route planning has been developed in ADAS laboratory, including in-house traffic simulator and a scalable, distributed system for serving vehicle routing requests which is able to handle tens of thousands individual cars. The aim of this effort is to reduce number of non-deterministic events in the traffic simulator to increase the reproducibility of results. In this way we will be able to fine-tune the setting of routing algorithms without need of doing several experiments with a particular setting to get statistically significant results.

 

Researcher: Jan Premus
Project: Constraining co- and after-slip model of the 2014 South Napa earthquake
Allocation: 158 000 core hours
Abstract: Elastic energy on fault planes, accumulated due to the relative motion of tectonic plates, is released abruptly in the form of an earthquake. Part of the energy often releases after the event in the form of an aseismic slip (afterslip) observed over the longer time period in the areas surrounding the ruptured region. An example of earthquake with abundance of coseismic and afterslip records is the 2014 South Napa, California, earthquake. The earthquake produced a significant surface rupture over the length ~12 km, which occurred mainly aseismically within the first month after the main event. In this project, we employ our code FD3D_TSN that simulates the dynamic rupture propagation on a fault during an earthquake and the long-term development of afterslip using a joint framework of laboratory-derived rate-and-state friction. We will employ a Bayesian inversion to infer spatially varying controlling dynamic rupture parameters. As data we will use both seismograms capturing the co-seismic rupture and long-term geodetic observations of the aseismic slip. We note that such an inversion on the joint seismic and postseismic data in a unified dynamic source model has not yet been performed worldwide. The result will assess the validity of the employed friction law and provide a unique view of the rupture process in an unprecedented resolution.

 

Researcher: Jiri Malik
Project: Thermomechanical interaction between outer ice shells and deep oceans on icy moons of Jupiter and Saturn
Allocation: 158 000 core hours
Abstract: The main motivation for this project is the study of the long-time thermomechanical interaction of the outer icy shell and the sub-surface ocean on Saturn's icy moon Enceladus. Enceladus, as a part of the Saturn system, was studied during the highly successful Cassini-Huygens mission, which started in 1997 and lasted nearly two decades. Despite its size (500 km in diameter, approx. 15% of our Moon), Enceladus is in the spotlight of planetary science community. According to data acquired in the mission, there is a global water reservoir beneath the surface of many ice moons [1], which brings the question of existence of heat sources capable of supplying enough heat to the liquid ocean. Based on conclusions that the only mechanism bringing enough heat to the ocean is the dissipation caused by tidal forces [1]—[3], how can one explain a dramatic variance of the ice shell thickness, which is down to several km in the south pole region (with respect to its average value of approx. 20 km)? This may be caused by hot springs located on the surface of a non-consolidated stone core, whose presence was suggested in [4] and supported by a specimen of material containing mineral elements acquired by the Cassini-Huygens probe. This brings us to the main aim of this project: a qualitative study of the described thermodynamic interaction on various time scales. The interaction is governed by two opposing processes: crystallization/melting on the bottom of the shell due to (lateral) variations of heat flux from the moon's core, and the dynamic relaxation of a non-hydrostatic configuration created by the phase change.

 

Researcher: Ales Vitek
Project: Mercury clusters — thermodynamics II
Allocation: 158 000 core hours
Abstract: This project is the continuation of OPEN-19-48 project focused on detection of phase changes in small clusters of mercury atoms through photoabsorbtion spectra and on the computation of thermodynamic properties under wide interval of pressures and temperatures. In our previous computations we have focused our attention on photoabsorption spectra of very small mercury clusters [1] and on the medium size clusters [2]. Photoabsorption spectra of mercury clusters can be easily experimentally measured. In other paper we have computed thermodynamic properties of Hg8 cluster [3]. But all this clusters are small and have a restricted number of minima on their potential energy surface. In this project we will compute properties of larger mercury clusters of 9 — 15 mercury atoms and we try to find relations between changes of photoabsorption spektra and phase changes, both induced by non-zero temperature and pressure. The goal of our computations will be to find, how are the thermodynamic properties changes with the increasing of particles from small clusters towards to the bulk limit and how to detect phase changes of small sub-nanoparticles from experimentally easily measureable photoabsorbtion spectra.

 

Researcher: Vladimir Ulman
Project: Fiji Bioimage Informatics on HPC - Path to Exascale
Allocation: 75 000 core hours
Abstract: "Bioimage Informatics on HPC" allows IT4Innovations to be involved in research on a completely new topical area of big biological image data processing on HPC. This specific research is focused on parallelization of key steps in lightsheet microscopy data processing as well as analysis of big data generated from other microscopic modalities. Particularly, multi-dimensional microscopy acquisitions present one of the main primary data sources in modern biological sciences and deployment of HPC in these areas is vastly unexplored approach to obtain biologically meaningful conclusions in a reasonable time.
This project is a continuation of the previous OPEN-20-21 call, solving particular VP3 subtasks of the Path to Exascale project. In the previous period, SPIM Workflow Manager was developed and  published under Apache License. The project aims at further development and dissemination of HPC-aware plugins for the Fiji community. In this call we would like to specifically focus on utilization of a newly developed library for seamless parallel execution of SciJava plugins — SciJava Parallel and its connection to a new structured image data storage. As a new research topic we are working on a distributed version of a simulator of artificial time-lapse images from developmental biology — EmbryoGen.

 

Researcher: Libuse Horackova
Project: The solution of differential equations using artificial neural networks with stochastic optimization methods with the main focus on the Poisson equation
Allocation: 69 000 core hours
Abstract: The solution of differential equations is needed almost anywhere we look. Artificial neural networks can help us solve these equations effectively. The main goal of this research is the use of artificial neural networks, using stochastic algorithms for optimization, for solving differential equations. We want to focus especially on the Poisson equation. The Poisson equation is a partial differential equation of the elliptic type. We want to solve the equation first for special cases (spheres, etc.) and then for general shapes. Furthermore, we would like to apply the solution in plasma physics, or in simulations dealing with molecular dynamics.

 

Researcher: Ivan Kolos
Project: Numerical modeling of load of structures in quasi-static effect of wind
Allocation: 57 000 core hours
Abstract: The project is focused on numerical modeling of flow around objects in the atmospheric boundary layer. This issue is complicated mainly due to the atmospheric turbulence, which requires the use of advanced numerical models of the flow coupled with detailed computational mesh of the domain. This research will contribute to bigger efficiency in design of building structures.

 

Researcher: Martin Micica
Project: Supercomputing for spin-based terahertz photonic structures
Allocation: 39 000 core hours
Abstract: Terahertz (THz) photonics features the immense potential for medical, security and telecom applications. Therefore, it is desired to develop new sources of terahertz waves with fast response, high intensity, controlled polarization properties with easy and cheap implementation. This can be achieved by THz sources employing spintronic phenomena such as terahertz spintronic emitters based on spin-Hall effect and spin-laser based THz sources. Numerical simulations of ultrafast dynamics and spin-transport will allow the deep understanding of spin based generation processes. The goals of this project are to develop advanced theoretical tools for spatio-temporal modelling of 1D and 2D grating-based spin vertical-cavity surface-emitting laser (spin-VCSEL) and metallic multi-layer spintronic emitters for terahertz applications utilizing scattering matrix (S-matrix) based on HPC friendly model implemented in MATLAB. Further optimization of structures will be performed using the finite element methods (FEM) in CST Studio and COMSOL. Obtained theoretical results from multiphysics and multidisciplinary simulations, combining optics, spintronics, condensed matter physics and plasmonics will be confronted with experiments carried out in spectroscopic and ultrafast laboratories at IT4Innovations.

 

Researcher: Pablo Nieves
Project: Towards multiscale modeling of magnetoelastic phenomena
Allocation: 3 878 000 core hours
Abstract: Magnetoelastic interaction couples the motion of atoms in a magnetic material with atomic magnetic moments and allows to transfer mechanical and thermal energies between phonon and magnon subsystems. Magnetoelasticity is of great interest for applications, but also from a fundamental point of view. For instance, precise control of magnetization by mechanically exciting the motion of atoms in magnetic materials, and vice versa, has enabled the development of a wide range of technological applications such as sensors (torque sensors, motion and position sensors, force and stress sensors) and actuators (sonar transducer, linear motors, rotational motors, and hybrid magnetostrictive/piezoelectric devices). Similarly, the combination of magnetism and heat is exploited in many applications like heat-assisted magnetic recording (HAMR), thermally assisted magnetic random-access memory (MRAMs), ultrafast all-optically induced magnetization dynamics, magnetic refrigeration, and biomedical magnetic hyperthermia. However, many aspects of magnetoelastic phenomena are not fully understood yet due to the complexity of the materials at large scale. Advanced modeling techniques and associated numerical tools based on a bottom-up multiscale approach could help to get a better understanding of magnetoelastic phenomena in magnetic materials across length scales. In this sense novel atomistic simulations that couple spin and atom dynamics could play an important role linking the microscopic and macroscopic scales. In this proposal, aiming to explore this possibility, we will study the performance of these models at finite temperature, and under magnetic field and stress for a set of well-known magnetic materials.

 

Researcher: Ekaterina Grakova
Project: 
Allocation: 32 000 core hours
Abstract: One of the basic principles of operational research is the search for optimal solutions to the problem with the use of mathematical modeling. For example, the optimization problems are to be found in transportation, economy, business, machinery, and in the industry in general. Optimization problems are being solved by a large number of optimization methods and algorithms which are mathematically complex and represent NP-hard problem. There are heuristics, metaheuristics, exacts, which solve various variants of the Arc Routing Problem and their application to real cases for the current heterogeneous HPC infrastructure. The exact algorithms are typical mathematical methods researching whole state space in order to find general (global) optimum. The heuristic algorithms and metaheustistics algorithms as oppose to exact algorithms search only through small part of state space, overlook optimal solution but the sometimes it can offer suboptimal solution and much faster.  The quality of results from the algorithms depends on the adjustment of the configuration parameters of the algorithm. In this project, we will use the HyperLoom platform for optimal parameters setting for the optimization of the VRP algorithms. 

 

Researcher: Rajko Cosic
Project: Development of sampling tools for photo-electron circular dichroism calculations
Allocation: 35 000 core hours
Abstract: The project aims to develop and test the simulation routines for generating the initial conditions in the photo-electron circular dichroism (PECD) simulations of chiral molecules. These simulations, based on the non-adiabatic molecular dynamics, are very sensitive on the initial conditions and, hence, the reliable sampling method has to be used.

 

Researcher: Michaela Bailova
Project: Boundary value problems with p-Laplacian
Allocation: 39 000 core hours
Abstract: Many engineering problems lead to solving quasilinear boundary value problems with p-Laplacian operator. Our approach is based on the fact that, under certain assumptions, the weak solutions of such problems coincide with critical points of corresponding functionals. In this project, we focus on problems with a mountain pass and saddle point geometry using minimax and optimization algorithms.

 

Researcher: Stanislav Polzer
Project: Stratification of patients with abdominal aortic aneurysm in covid era
Allocation: 17 000 core hours
Abstract: Abdominal Aortic aneurysm (AAA) is a permanent dilatation of abdominal aorta. Its rupture is a life-threatening event. On the other hand the non-ruptured AAA does not bring any complication to patient in most cases. Therefore the effort is in operating only cases with AAA close to rupture especially in the era of covid-19 pandemic when all elective operations are postponed to free hospital capacities for covid patients. Therefore currently used criterion of the maximal AAA is not sufficient. In this project, clinical application of our previously developed probabilistic rupture risk criterion will be performed to perform less surgeries and yet not threating patient´s life. The supercomputer is necessary since our computations are time demanding while surgeons need to know how dangerous particular AAA is as soon as possible. The project will last until the covid pandemic is solved by vaccine.

 

Researcher: Pavel Praks
Project: Stochastic and deterministic methods for optimisation of distribution networks in the energy sector II
Allocation: 32 000 core hours
Abstract: The electrical power consumption is gradually increasing over the years. In combination with the ageing of distribution grids and required integration of new uncontrolled sources (wind and solar systems), a higher emphasis is placed on power flow control and monitoring elements to ensure continued supply and required quality of the provided electrical energy for the society. However, the purchase price and the maintenance cost of the switching and monitoring devices is high, therefore discrete optimization must be employed to identify optimal placement and operating mode of the control devices. The stochastic approach is very robust but extremely time-consuming. Fortunately, the performance of stochastic methods can be accelerated by parallel implementation. It is a second project for testing of HPC infrastructure of IT4innovations for modelling and optimisation of Czech distribution networks. The novelty of the project includes testing of the AtsPy package (automated time series models in Python), which requires NVIDIA GPU with CUDA.

 

Researcher: Radek Halfar
Project: Investigation of biological systems using chaos theory III
Allocation: 12 000 core hours
Abstract: This project is part of long-term research that deals with the application of dynamic systems methods for the analysis of biological systems. Special emphasis is placed on investigating the propagation of electrical signals in cardiac tissue. New methods for detecting physiological and pathological conditions of the heart are created using modern and classical methods of chaos theory, such as RQA and entropy calculations. Furthermore, the causes of heart rhythm abnormalities and the possibility of predicting the development of electrical signal propagation in the heart are investigated.

 

Researcher: Lukas Brodsky
Project: Land Cover Supercomputing (LCS)
Allocation: 8 000 core hours
Abstract: The EU Geo-harmonizer project No. 2018-EU_IA_0095 (granted by INEA — Innovation and Networks Agency) aims at reducing problems of national data with using seamless complex (geographical) data over the entire extent of the EU, and “opening data” through using Open Data licenses, enabling wider public access to the data by not only scientists and specialists, but also non-professionals, facilitating increased use of EC-funded data without imposing any expectations on users to possess specialised or costly infrastructure, working closely with the national authorities.
The idea is to develop an EU-wide automated mapping system for harmonization of Open Data based on FOSS4G (Free and Open Source Software for Geospatial) and Machine Learning. We use the Geo-harmonizer tools and data to generate decision-ready layers including the land cover. To calculate these value-added product, we use Earth observation (EO) images, topographic data (DEMs) and other covariates to map the full European continental land cover map.
This Land Cover Supercomputing project aims at evaluating the utility of IT4I supercomputers for processing large amount of satellite images. The ultimate goal is to map the land cover of the whole Europe ~6 MIO km2, which comprise over 10+ TB of input satellite images. Successful processing of the full European land cover would be provided to the whole community as a product with open license through the https://opendatascience.eu portal. 

 

Researcher: Florian Belviso
Project: TRibological Ab initio Study of Heterostructures (TRASH)
Allocation: 2 894 000 core hours
Abstract: Two-dimensional (2D) materials present a wide range of properties that can find applications in electronics, energy storage, sensors, thermoelectrics, and tribology1,2,14. Their thickness can be reduced down to a single atom, giving them the shape of thin sheets. The first synthesised example was graphene3, followed by its analogous, and Transition Metal Dichalcogenides (TMDs)4.
TMDs are a unique family of 2D materials of general stoichiometry MX2, one of the most extensively studied currently. They showed a wide  range of applicability in the last decade4, and are a promising candidate for developing photovoltaic devices, dry lubricants15,16 and nanoelectromechanical systems13. The design of TMD-based heterostructures  increases the possibilities even more5.
The fabrication of 2D TMD-based systems relies on obtaining mono or few-layered TMD films6,7. These films can be obtained by layer exfoliation of the bulk material. Designing heterostructures from this 2D films implies to find ways to manipulate the layers on top of each other. Improving this exfoliation and manipulation techniques calls for a better understanding of frictional behaviour at the nanoscale. Previous studies of the properties of bulk TMDs8-12 showed that modifying the electronic structure of TMDs by way of charge injection or electric field can impact their frictional behaviour. Here we outline guidelines for controlling the tribological behaviour of TMD-based heterostructures using an electrostatic field.

 

Researcher: Sergiu Arapan
Project: Toward multiscale modelling of intrinsic magnetic properties of materials at finite temperature: I. First-principle calculations of the Heisenberg magnetic exchange.
Allocation: 2 873 000 core hours
Abstract: Magnetic phenomena had fascinated humans since the ancient times. Magnetism proved to be also very useful to one of the most important human activity — traveling, with the discovery of the compass. Today, one can hardly imagine our life without various applications of magnetic phenomena: all our electronic devices exploit the magnetism in various forms and to certain degree. Yet, only relatively recently, with the development of quantum mechanics, we started to understand the magnetic nature of materials. In 1928 Heisenberg has shown that the spontaneous magnetization in materials arises due to quantum-mechanical exchange interaction between atomic spins. The knowledge of the magnetic exchange interactions allows to completely describe and predict the magnetic properties of a material. We have recently developed a new approach to calculate these exchange interactions and, in this project, we will apply it to accurately predict the magnetic behavior in various magnetic materials.  

 

Researcher: Stepan Sklenak
Project: Periodic DFT studies of zeolite-based catalysts
Allocation: 2 586 000 core hours
Abstract: Zeolite based catalysts are the most important industrial catalysts used mainly in petrochemistry (production of fuels, various hydrocarbons), and furthermore, for N2O decomposition and N2O selective catalytic reduction (SCR), for direct NO decomposition as well as for SCR of NOx using various reducing agents. In addition, zeolites are also employed as catalysts in synthesis of fine chemicals and as well as in the processing of biomass. Zeolites are also used as ion exchange agents (e.g. they replaced harmful phosphates in washing powder detergents). Zeolites are crystalline microporous aluminosilicates with a unique microporous nature, where the shape and size of a particular pore system exerts a steric influence on the reaction, controlling the access of reactants and products. Thus, zeolites are often said to act as shape-selective catalysts. Increasingly, attention has focused on fine-tuning the properties of zeolite catalysts in order to carry out very specific syntheses of high-value chemicals. Periodic DFT methods permit investigations of properties of zeolite-based catalysts which are needed for their fine-tuning. DFT calculations are complementary to experimental examinations and together they can provide more complex knowledge of the properties of the studied catalysts and the reactions they catalyze.