Researcher: Radim Špetlík    


High-dimensional GCxGC ToF Mass Spectrograph Data Analysis 

Karolina GPU  Alloc=1800;  LUMI-G  Alloc=5000

Analyzing GCxGC-TOF-MS with modern machine-learning methods, especially deep learning-based approaches developed for computer vision, is crucial in analytical chemistry and beyond. GCxGC-TOF-MS provides comprehensive separation and identification of complex mixtures, but its vast and intricate data poses challenges for traditional analysis methods. Deep learning, including convolutional neural networks (CNNs), excels in handling such complexities, making it ideal for analyzing multidimensional chromatographic data.The benefits of using deep learning in this context are manifold. Firstly, it automates peak detection, deconvolution, and peak alignment in GCxGC-TOF-MS data analysis, reducing manual effort and human errors. Secondly, deep learning extracts features from raw chromatographic data, uncovering hidden patterns for improved compound identification and quantification. Additionally, deep learning models generalize well to new samples from large datasets, enabling robust applications in environmental monitoring, pharmaceutical analysis, and food safety assessment.Leveraging deep learning techniques from computer vision, such as transfer learning and data augmentation, optimizes limited GCxGC-TOF-MS data, potentially reducing expensive and time-consuming experimental runs. Integrating modern machine learning with GCxGC-TOF-MS enables advanced data analysis, facilitating exploration of complex chemical interactions and discovering novel compounds across scientific domains.


Researcher: David Zihala       


Extramedullary multiple myeloma: molecular pathogenesis and novel therapeutic targets           

Karolina CPU  Alloc=1000            

Multiple myeloma (MM) patients presenting with extramedullary disease (EMD) display significantly worse prognosis compared to those without EMD. This is, at least partially, due to the low efficacy of modern MM therapy, stemming from overall low understanding of EMD biology. MM is characterized by the clonal proliferation of plasma cells (PCs) inside the bone marrow (BM). In EMD, PCs become independent of the BM and colonize soft tissues distant from the bone. In comparison to MM, no detailed genomic or transcriptomic profiling in the EMD stage has been carried out and virtually no data describing EMD tumor microenvironment exist. In our study, we are combining DNA/RNA sequencing methods, including state-of-the-art single cell sequencing, and highly dimensional spectral flow cytometry to analyze the biggest cohort of EMD samples ever collected. In addition, we are further investigating our findings from real world patient’s samples using modern molecular and functional analyses using genetically modified cell lines.  By combining these advanced methodologies, we aim to shed light on the intricacies of EMD biology, paving the way for improved therapeutic strategies and better patient outcomes in this challenging condition.


Researcher: Daniel Krpelík   


Parameter optimisation of power system control software          

Karolina CPU  Alloc=4560            

Power systems are part of the critical infrastructure. Their operations are affected by various external sources which increase demands on complexity of mathematical models used for simulating their behaviour. Many such simulations need to be run by computer control systems in order to asses consequences of control actions. In order to use the computational resources more effectively, advanced optimisation algorithms use sophisticated techniques to iteratively search through action candidates, from which the best solution is selected afterwards. There are many such algorithms, each with a set of its own control parameters influencing how fast and whether can the algorithm find the optimal control action. Nevertheless, optimisation of power grid operations still poses a challenge for the algorithms because of the size and complexity of the set of possible configurations, which leads to the necessity of solving large mixed-integer topological optimisation problems. This computational experiment will compare several state-of-the-art algorithms on few selected real-world problems in order to provide recommendations for the configuration of grid control and decision support systems.


Researcher: Libor Šachl         


Towards the understanding of the subsurface planetary oceans using ocean-induced magnetic field

Karolina CPU  Alloc=1900            

The Earth is not the only body in the Solar System where water is present in its liquid state. Subsurface water oceans are also present on the Jovian satellites of Europa, Callisto, and Ganymede. The Earth’s oceans are home to abundant and diverse life. Similarly, the subsurface oceans of Jovian satellites are a promising candidate to host life (Blanc et al., 2020).The subsurface oceans of Jovian satellites were detected by interpreting the Galileo space probe’s electromagnetic (EM) measurements (Zimmer et al., 2000; Khurana et al., 2002; Kivelson et al., 2002). However, it is likely that the potential of EM methods has yet to be fully explored. Moreover, two space missions to the satellites of Jupiter (NASA’s Europa Clipper and ESA’s JUICE) are expected to bring new data in the early 2030s, providing unprecedentedly detailed information on the structure and temporal variations of the magnetic field in the vicinity of Europa, Ganymede, and Callisto.In this project, we propose to study the ocean-induced magnetic field (OIMF), which is generated by the circulation of conductive seawater in the presence of the ambient Jovian magnetic field via the process of EM induction. By inspecting various scenarios, we will estimate if the space probes can measure the OIMF and if the OIMF can constrain subsurface ocean dynamics.


Researcher: Petr Miarka       


Concrete Fracture: A meso-level approach          

Barbora CPU  Alloc=9216            

With the development of new, often more environmentally friendly, building materials, concrete fracture analysis attracts more research attention as it is crucial to understand the damage evolution in such materials prior to structural applications. This project aims to conduct a complex investigation of crack initiation in modern concrete types. For this, a meso-level numerical modelling approach is taken into account as it offers an opportunity to model the heterogenous inner structure of concrete, including aggregates, matrix and pores. This allows for a comprehensive understanding of crack formation and propagation under various loading scenarios. The obtained numerical results of the inner stress field can determine the precise location and external loading conditions of the beginning of the initiation of the eventually fatal damage evolution. This can be used in the future design of concrete mixtures that could use recycled materials with less cement content, which may lead to reducing CO2 emissions, while improving its durability and mechanical performance.


Researcher: Zdeněk Mašín   


Attosecond photoionization dynamics in atoms and molecular dimers   

Barbora CPU  Alloc=2400;  Barbora FAT  Alloc=592;  Karolina CPU  Alloc=2800    

Photoionization of molecules is an ultrafast process which takes place on attosecond timescales while the subsequent nuclear dynamics that it triggers unfolds on femtosecond timescales. Progress in development of laser systems capable of taking ``snapshots” of these processes allows us to obtain a detailed insight into the role of electron-electron interaction, coupled electron-nuclear motion and thus provides a pathway to their control. This project builds on and expands our previous work aiming to generate highly accurate ab initio results of attosecond time-delays and time-resolved photoelectron spectroscopy to support our collaborations with experimental groups. In this project we will follow the directions opened by our previous work and study larger systems (molecular dimers) and smaller systems (atoms) in new geometric and field configurations. In particular we will complete our studies of attosecond streaking in small molecules and atoms and follow hydrogen transfer on its natural time-scale using time-resolved photoelectron spectroscopy.


Researcher: Ivo Oprsal           


Near-surface geology and topography: Wave propagation, ambient vibrations and surface-to-reference amplification     

Barbora CPU  Alloc=3600;  Karolina CPU  Alloc=3000      

Earthquakes cause major social and economic impacts, especially in urban areas, with casualties and infrastructure damage worldwide. Strong ground motions originate from the earthquake source, travel through regional geology, and affect the local geological setting. The near-surface geology is critical in determining strong-ground motions and structural damage. Finite-difference (FD) modeling predicts 3D wave propagation effects in earthquake scenarios, enabling simulations of various ground-shaking scenarios and assessing their impact on structures in metropolitan areas. This aids in mitigating seismic hazards through anti-seismic structure designs, reinforcement of buildings, and sophisticated urban planning, yielding substantial societal and economic benefits during seismic events. A novel approach integrating geophysical methods, UAV photogrammetry, and geomechanical modeling studies rock monuments in Czechia. Teplice formation sandstones are prevalent in research sites within the Broumovsko Protected Landscape Area, including the Homole Cukru tower in the Adršpach Rock City and various locations in the Český Ráj (Bohemian Paradise) Protected Landscape, Figure 1. The study's methods are directly applied in anti-seismic designs, reinforcing structures, and urban planning, reducing damage and casualties during seismic events.


Researcher: Kryštof Mráz     


Numerical Characterization of Polymeric Hollow Fibers Heat Exchangers

Karolina CPU  Alloc=4600            

Porous structures and products with a complex inner geometry are still considered as a great challenge for conventional CFD (computational fluid dynamics). Unlike classical CFD methods (e.g., the finite volume method), the Lattice Boltzmann has proven itself as a promising option for such complex computational domains. The aim of this project is to utilize the Lattice Boltzmann method for numerical simulation of flow through hollow fiber heat exchangers. These heat exchangers contain hundreds or thousands of hollow fibers with outer diameter approx. 1 mm. Such a complex geometry of a heat exchanger makes it impossible to simulate it by the conventional CFD. However, the comprehensive numeric simulations of many heat exchanger configuration is highly desirable, because it would fill the gap between rather simplifying analytical models and empirical experiments. The local and explicit nature of the Lattice Boltzmann method makes it more than suitable for a massive parallelization and high-performance computing. Lattice Boltzmann method results and effectiveness will be compared with a conventional CFD approach.


Researcher: Pavel Ondračka


Machine learning-enabled modelling of nanocomposite W-B-C 

Karolina CPU  Alloc=3500;  Karolina GPU  Alloc=300       

The bench-mark protective coating material utilized in the cutting and forming applications is TiAlN exhibiting high hardness as well as stiffness. Unfortunately, these favorable properties are associated with unwanted brittle deformation behavior resulting upon mechanical loading in the formation of cracks which limits the performance and lifetime of the coated tools. From a materials design point of view a rather unusual combination of properties - high hardness and stiffness together with moderate ductility - is therefore required for the next generation of protective coating materials. This project aims to design coating materials with high hardness and enhanced fracture toughness for cutting and forming applications with extended lifetime and performance by ab initio calculations on a novel class of nanocomposite tungsten-boron-carbide material.


Researcher: Martin Matys   


Gamma-ray flash generation using high-intensity laser  

Karolina CPU  Alloc=14004;  Karolina GPU  Alloc=40;  Karolina VIZ  Alloc=40         

The current development of multi-petawatt lasers opens the window to the quantum electrodynamics regime. Interaction of  these ultraintense lasers with matter results in gamma-photon emission mainly via the multiphoton Compton scattering process. The gamma-ray flashes are of a big interest for a wide portion of the scientific community, with theoretical predictions going along the same path as computer simulations. Generating intense gamma-ray flashes is of particular interest for astrophysical studies and many other fields like radiation chemistry, materials sciences, nuclear physics and applications such as gamma knife in medicine. In this work, we employ particle-in-cell (PIC) simulations to realize the effect of the laser incidence angle on the solid target regarding the yield and directionality of the resulting gamma-photon flash. By varying the target thickness in combination with the target electron density, the spectral and spatial properties of the  gamma-photon generation will be studied in advantageous schemes increasing the photon intensity. 


Researcher: Adam Pecina     


Semiempirical quantum-mechanical scoring function for reliable protein-ligand affinity predictions

Karolina CPU  Alloc=15000         

Accurate estimation of protein–ligand binding affinity is the cornerstone of computer-aided drug design. We have developed a universal physics-based scoring function addressing key terms of binding free energy using semiempirical quantum-mechanical computational methods. Here, we tend to evaluate our method against experimental “truth” in multiple diverse data sets and compare it to the state-of-the-art methods. Our project aims to answer several fundamental questions of computational drug design: on the existence of a universal yet computationally efficient physics-based scoring method for reliable affinity predictions, on the real performance of existing scoring methods when applied to known structures for which high-quality experimental data are available, on the performance when applied to real-world virtual screening scenario and on the necessity to invoke a higher level of theory and at what cost.


Researcher:  Jozef Hritz         


Fibrilization of Tau protein          

Barbora CPU  Alloc=20000;  Barbora FAT  Alloc=91;  Barbora GPU  Alloc=1000;  Karolina CPU  Alloc=3906;  Karolina FAT  Alloc=100;  Karolina GPU  Alloc=2800;  LUMI-C  Alloc=3600;  LUMI-G  Alloc=2000            

Although the molecular mechanism of Alzheimer’s disease is poorly understood, there is a strong agreement that pathologically altered tau protein plays an important role in the process. The main motivation of the proposed project is to study the fibrilization of Tau protein induced by truncation and phosphorylation.   The methodological challenge for the planned study is the large size of the simulation system and the fact that the tau protein belongs to the group of intrinsically disordered proteins (IDPs). The disordered character of the tau protein makes it difficult to study by current methods in structural biology and biophysics, which were developed and optimized for well-folded proteins.We will study tau protein variants by computational simulations based on molecular dynamics. Due to the need for an extensive conformational sampling of studied proteins, we will be also testing the performance of course-grain models and enhanced sampling methods. The obtained computational outcomes will be validated against experimental data obtained from nuclear magnetic resonance (NMR), small angular X-ray (SAXS), atomic force microscopy (AFM), and cryo electron microscopy (cryoEM). Most of these experimental data will be measured in our research group or in the labs of our collaborators, allowing for interactive approaches when needed. The direct interplay of computational and experimental techniques is the key point of this proposal.


Researcher: Dominika Mašlárová 


Optimization of laser-based electron acceleration            

Karolina CPU  Alloc=520;  Karolina GPU  Alloc=700          

Laser-based electron accelerators represent a promising concept of next-generation accelerators. Their main advantage is a remarkably short acceleration length, caused by a significant acceleration gradient (up to ~100 GV/m), about thousand-fold times higher than in the conventional radiofrequency accelerators. Such accelerators introduce a more compact and cheaper option, leading to better accessibility to electron accelerators in research, medical and industrial facilities. In one of the most popular methods, called laser wakefield acceleration (LWFA), electrons are injected into a plasma wave (wakefield), generated and dragged by a few-tens-of-fs, ultra-intense laser pulse in an optically transparent ionized medium (plasma). There are several crucial aspects that can determine the resulting quality of electron beams. The aim of this project is to investigate the influence of plasma density profile on final electron beam charge, divergence and energy by numerical kinetic simulations. The results of these simulations will be used to directly arrange the experiments for our collaborators in the Lawrence Berkeley National Laboratory (LBNL), Berkeley, California, USA.


Researcher: Daniel Bim         


Theoretical insight for designing new nickel photoredox calatalysts         

Karolina CPU  Alloc=17300         

Photoredox catalysis offers a potent method for organic synthesis through the conversion of photon energy into chemical potential, facilitating C–C/C–X couplings and C–H bond activations. This environmentally friendly approach remains underexplored due to the complexities of excited state dynamics and reactive intermediates. Our project aims to elucidate ground- and excited-state mechanisms in nickel-based photoredox catalysis using advanced theoretical calculations. By assessing reaction energetics, defining mechanisms, estimating redox activity of intermediates, and mapping excited-state potential energy surfaces, we seek to understand key electronic contributions to reactivity. Ultimately, this knowledge will aid in designing new ligand scaffolds for improved reactivity, validated through experimental testing.


Researcher: Bořek Patzák     



Karolina CPU  Alloc=19600         

MUSICODE (An experimentally-validated multi-scale materials, process and device modeling & design platform enabling non-expert access to open innovation in the organic and large area electronics industry) is a European collaborative research and innovation project led by a multi-disciplinary consortium. It aims to make high performance multiscale computational tools and modelling workflows readily accessible and customizable for the Organic and Large Area Electronics (OLAE), a rapidly emerging sector bringing disruptive technological revolutions  in Energy, Electronics, Transport, Photonics, Buildings, Lighting & Displays, Health, Wearables, IoT, Agriculture, etc.  MUSICODE will create an Open Innovation Platform for Materials Modelling (OIPMM) for the EU OLAE Academia and Industry, linked with state-of-the-art HPC infrastructures, with the goal to unleash the potential of OLAE.


Researcher: Timothée Emmanuel J. Rivel      


Antimicrobial Selective Peptides To Induce Cell rupture (ASEPTIC)           

LUMI-C  Alloc=19900    

Antibiotic-resistant bacterial strains pose a global health threat, demanding innovative solutions. Antimicrobial peptides (AMPs) offer a promising therapeutic avenue by disrupting bacterial membranes particularly through pore formations. In this project, we introduce a novel method to quantify the energetic cost of AMP-assisted pore formation, using molecular dynamics simulations. Our primary goal is to optimize peptide sequences by reducing the energy barrier of pore formation for bacterial membranes – thus increasing AMPs efficiency. Our secondary goal is to preserve a high barrier for mammalian membranes – thus reducing AMPs toxicity. We achieve this by tailoring peptides to exploit the distinctive lipid content in these both membranes and we validate our results by means of leakage assays on lipid vesicles with corresponding lipid contents.


Researcher: Azin Shahsavar 


Computational Studies of Molecular Nanoarchitectonics on Topological Insulator Surfaces

Barbora CPU  Alloc=11550         

Topological insulators have been attracting a great deal of attention thanks to their fascinating properties. As a consequence of the strong spin-orbit coupling, the bandgap in 3D topological insulators (TIs) gets inverted; this gives rise to a special type of surface states that have linear dispersion characteristic for Dirac fermions. In these states, the electron's spin is locked to its momentum, i.e., the Dirac surface states are topologically protected. This means that these states are robust against perturbations that maintain time-reversal symmetry., surface defects or chemical disorder. This property leads to a nearly dissipation-less current. Therefore, Tis possess an enormous potential for applications in spintronics and quantum computing. Apart from their topologically protected surface state, TIs also exhibit the topological magneto-electric effect, where an electric field generates a topological contribution to the magnetization in the same direction. This may bring novel means for electric-field control of magnetization. In this project, we aim to provide a theoretical understanding of the electrical tuning of magnetic coupling between magnetic centers within metal-organic networks synthesized on TI and the other substrates.


Researcher: Simona Chalupna     


The use of computationally expensive DFT functionals for salt-cocrystal systems investigation.  

Karolina CPU  Alloc=15000         

Nowadays, the portfolio of solid forms of pharmaceutical molecules is very wide. Pharmaceutical salts are one of the forms that can be used for formulation of active components.  About 50% of drugs used today are salts. Cocrystals are the other most dynamically developing pharmaceutical forms. The difference between salt and cocrystal is given only by the position of single hydrogen. Information about the exact position of this hydrogen is required in registration and patent documentation by regulation authorities. Since this difference between cocrystal and salt is so small, it is an essential task to develop new methods that could determine the exact position of responsible hydrogen. We are developing a computational method for such hydrogen position determination. The method does not require crystallographic data from high-quality monocrystal, and it works even with data from powder samples.


Researcher: Alexander Molodozhentsev       


Particle-In-Cell modeling of laser wakefield acceleration aiming high quality electron beam         

Karolina CPU  Alloc=25000         

A plasma-based acceleration scheme for particle acceleration by space charge wave was proposedby Y. Fainberg in 1956. This acceleration approach allows one to overcome one of the mostsignificant limitations in conventional accelerators - limited electric field gradient in radio frequencyaccelerating structures. Extreme laser-plasma accelerating gradients, demonstrated experimentallyby different teams, offer a path towards a compact laser-plasma accelerator (LPA). Such a compactaccelerator can be used as an electron beam driver needed in a broad variety of applications,including free electron lasers (FEL), Thomson sources and even electron-positron colliders with TeVenergy.Laser-plasma acceleration has been the subject of active research overdecades to generate highquality electron beams with up to GeV energies. A significant effort is being made to improve thequality, e.g., high charge, low energy spread, small beam emittance, and low divergence of theaccelerated electron beam. The electron beam quality in a laser-plasma accelerator (LPA) stronglydepends on the injection mechanism. In this study, we will investigate different injectionmechanisms and their influence on the accelerated electron beam quality. In particular, we willexplore the laser wakefield accelartion process using self-injection scheme, e.g., density downrampinjection, injection in a pre-formed plasma channel, and density tailoring in a capilary setup andionization induced injection mechanism e.g., injection from neutral gas in a single or multistageconfigurations. In this project, we will investigate all these possibilities using PIC simulation.


Researcher: Prashant Dwivedi            


Advancements in High-Velocity Impact Studies: Applications in Fusion Energy (AHVIST)

Karolina CPU  Alloc=17500         

In the field of nuclear fusion, the interaction of dust particles with plasma-facing materials (PFMs) poses a significant challenge [1-5]. This research focuses on the high-velocity impact of dust on polycrystalline tungsten (W) targets, a widely used PFM, with helium (He) bubble implantation. Utilizing molecular dynamics simulations, we will investigate the dust velocity impact on these targets, analyzing the resulting crater formation, volume displacement, and alterations in material properties. The study will also consider the effects of He bubble implantation, which can influence the material's response to dust impact [6-8]. The expected outcomes include a detailed understanding of the mechanisms underlying dust-induced damage, the role of He bubbles, and the influence of polycrystalline structure on the impact response. This research will contribute to the optimization of PFMs, enhancing their resilience and performance in nuclear fusion reactors, and providing insights into material behavior under extreme conditions.


Researcher: Igor Roncevic    


Electronic structure and magnetism in porphyrin nanoribbons  

Barbora CPU  Alloc=43890         

Atomically thick one-dimensional molecular nanoribbons based on fused porphyrins have attractive properties such as high electrical conductivity, efficient light-harvesting abilities, and ultrafast energy delocalisation. In most nanoribbons investigated so far, porphyrins host zinc ions, resulting in nonmagnetic systems. However, if zinc is exchanged with a magnetic metal (our blood is red because of iron porphyrins!), we can obtain porphyrin-based molecular ribbons which exhibit magnetism. This project deals with the computational investigation of such systems.In porphyrin nanoribbons, the metals are at least three times further apart than in a metallic solid, resulting in much weaker magnetic interactions. Such weak interactions are desirable for quantum computing, where each qubit (metal ion) needs to be individually addressable, which is not possible when interactions between qubits are very strong. In this project, we will quantify these weak interactions and identify the best candidates for multi-qubit systems.Introducing magnetism into porphyrin nanoribbons may also allow us to change their conductivity by applying a magnetic field, which is known as spintronics. Today’s hard drives operate on a similar principle, but on a much larger scale. We will investigate the possibility of using porphyrin ribbons as molecular spintronic devices and build an understanding of their operating principles.


Researcher: Alvaro Patricio Prieto Perez         


Future air-quality changes over Europe driven by emissions        

Karolina CPU  Alloc=28000         

Air pollution is influenced by both emissions and also by meteorology and climate in general. This study aims to study future air-quality over Central Europe at a moderate resolution using the regional climate model Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) and different emissions scenarios based on the Representative Concentration Pathways (RCPs) while the climate is kept at present day conditions to isolate the effect of emission changes only. In our simulations planned, we consider both present-day and future, scenario based emissions according to RCP4.5 and RCP8.5. The domain of the simulations has 9km horizontal resolution and is centred over Prague, covering Central Europe for a ten year period in each simulation. This length of period ensures statistically robust results. We expect to asses the contribution of changes in emissions in future air-quality over Central Europe.


Researcher: Ondrej Chrenko


Exploring the influence of accretion luminosity and gas turbulence on planet migration

Barbora GPU  Alloc=1500;  Karolina CPU  Alloc=3000;  Karolina GPU  Alloc=21600            

Planets are born in protoplanetary disks and, early on, their orbits evolve under the gravitational influence of the surrounding gas disk material. The resulting drift of the planet towards or away from the central star is known as planet migration. Understanding the mechanism of disk-driven migration is essential to understand the origins of observed orbital configurations of (exo)planetary systems. Here, we propose to study migration of (i) planets that are luminous due to accretion of solids and thus they heat the surrounding gas; and (ii) planets that evolve in turbulent gas. Our aim is to describe how perturbations of the gas distribution in the outlined scenarios affect the orbital evolution of embedded planets. Our investigation will be based on 3D and 2D radiative hydrodynamic simulations launched on GPU clusters of IT4I. This project is directly related to the solution of the GAČR Junior Star Grant 21-23067M and it will also contribute to the international COST Action CA22133.


Researcher: Pavel Praks        


Stochastic and deterministic methods for optimisation of distribution networks in the energy sector V   

Barbora CPU  Alloc=10000;  Barbora FAT  Alloc=400;  Barbora GPU  Alloc=1000;  Karolina CPU  Alloc=4000           

Electrical power consumption is gradually increasing over the years. In combination with the ageing of distribution grids and the 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 society. However, the purchase price and the maintenance cost of the switching and monitoring devices are high, therefore discrete optimization must be employed to identify the 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 a parallel implementation. It is the third project for testing of HPC infrastructure of IT4innovations for the modelling and optimisation of Czech distribution networks. The novelty of the project includes testing large-scale distributed hyperparameter optimization tools for energy networks, for example, the Ray Tune. Moreover, symbolic regression will be used as a tool for finding fast explainable machine-learning models of selected thermochemical processes (for example, gasification).


Researcher: Michal Kolar      


Gating of the ribosome exit tunnel         

Karolina CPU  Alloc=10800;  LUMI-C  Alloc=14900           

Proteins are fundamental biomolecules that play pivotal roles in all biological processes. The synthesis of all proteins takes place on 25 nm large particles known as ribosomes. Deeply embedded within these particles lies the catalytic center of the ribosome, where protein synthesis occurs. As a result, the newly created protein emerges from the ribosome through a tunnel approximately 10 nm in length. Interestingly, within this tunnel, there exists a constriction – a narrow segment – whose purpose remains unclear. In this study, we propose to perform atomistic simulations of the entire bacterial ribosome. Our primary objective is to investigate the conditions under which the constriction can be naturally closed, uncovering its functional significance.


Researcher: Jiří Jaroš              


Closed-loop individualized image-guided transcranial ultrasonic stimulation II    

Barbora CPU  Alloc=5000;  Barbora FAT  Alloc=50;  Barbora GPU  Alloc=100;  DGX-2  Alloc=50;  Karolina CPU  Alloc=1000;  Karolina FAT  Alloc=50;  Karolina GPU  Alloc=1000;  LUMI-C  Alloc=500;  LUMI-G  Alloc=100

Disorders of the brain, including neurological and psychiatric diseases, affect one in four people. New treatment options are needed with enhanced efficacy and reduced side-effects, costs, and invasiveness. Neurostimulation techniques that modulate the electrical activity of the brain have evolved as an important class of second-line treatments for pharmacoresistant cases. What is needed is a non-invasive brain stimulation technique that can stimulate brain targets with high anatomical precision, unlimited penetration depth, full reversibility, and low risk-profile. This can be achieved using the newly emerging technique of low-intensity focused transcranial ultrasonic stimulation (TUS) for neuromodulation.This project focuses on the development of closed-loop individualized image-guided transcranial ultrasonic stimulation, under the Horizon Europe CITRUS project. The ultimate goal of the CITRUS project is to develop a fully functional prototype of a medical device that integrates an ultrasound transducer system possessing advanced 3D steering capabilities with a custom-built magnetic resonance receiver, enabling high-resolution transcranial neuromodulation with unprecedented flexibility and sensitivity. The computational resources will be used for preoperative MR-based brain imaging, personalized ultrasound treatment planning including temperature mapping, and validation of fast real time re-planning software based on advanced mathematical models and artificial neural networks.


Researcher: Christopher Heard          


Structure Elucidation of Cu2O Photocatalyst Surfaces via Machine Learning        

Karolina CPU  Alloc=34717         

Cuprous oxide (Cu2O) is a p-type semiconductor with a range of promising applications, including photo(electro)catalysis (e.g. CO2 reduction/HER) and solar energy production, owing to its convenient bandgap and high theoretical solar power conversion efficiency (~20%).[1] To understand and optimize the properties of Cu2O towards these applications, a precise atomistic picture of the surface must be established, including surface and subsurface defects, common reconstructions and the response to atmospheric conditions, in particular the presence of adsorbates (e.g. H2O). However, despite a sizeable literature based on single crystal measurements from both experiment and theory, recent LEED-IV experiments have shown that further research is needed to resolve the structures of the low energy planes. In this work, we will generate machine learning potentials at the metaGGA level, combined with an established hybrid-level delta learning scheme.[2,3] This will allow for dynamical simulations of realistic models over long time and length scales, at a degree of accuracy hitherto impossible to reach. Via these simulations, in cooperation with state-of-the-art experimental analyses, we will bring new insights into the atomistic structure of Cu2O low index surfaces under realistic conditions, including the effect of temperature.


Researcher: Ondrej Klimo     


Suppression of laser plasma instabilities in inertial fusion relevant experiments 

Karolina CPU  Alloc=35000         

The production of safe and clean energy is one of this century’s main challenges. Recent experiments at the National Ignition Facility in the US have demonstrated the plausibility of inertial confinement fusion (ICF) using lasers as a future energy resource. However, these experiments do not use a setup that is conducive to efficient reactor design and the energy gain requires increasing by more than an order of magnitude. Many challenges remain to turn these experiments into a viable solution for energy production. One of the key obstacles for the direct-drive inertial fusion scheme, an energy generating relevant ICF scheme, is the coupling of laser energy to the fuel capsule. When lasers ablate the surface of a fusion fuel capsule large amounts of hot material are blown off as plasma. The incoming laser energy can be redirected or poorly absorbed due to complex laser-plasma instabilities (LPI), which can prevent energy gain particularly in direct-drive ICF altogether. Two solutions to this problem will be explored in this project, the use of broad bandwidth lasers, and externally applied magnetic fields.There are several laser facilities around the world which are developing broadband configurations as a solution to LPI in ICF, and experiments using large magnetic fields for this purpose have already begun. Using the resources requested in this proposal we will perform large scale kinetic simulations of ICF relevant scenarios with the aim of planning experimental campaigns.


Researcher: Miroslav Medveď         


Mechanistic Insights into Photophysical Processes in Carbon Dots via Quantum Chemistry Methods       

Karolina CPU  Alloc=35875         

Rapid development of precise and ultrafast spectroscopic techniques providing data with extremely high spatial and time resolution nowadays enable investigation of photophysical processes (e.g., internal conversion, intersystem crossing, fluorescence, phosphorescence) and photochemical transformations (e.g., photoswitching, photocatalytic processes) with unprecedently deep insights into the radiative and non-radiative transformations. Simultaneously, development and efficient implementation of new accurate theoretical approaches relying on the remarkable increase of the computational power offer crucial information about the molecular electronic structure and the nature of excited states. Combined theoretical and experimental effort enables not only consistent interpretation of the experimental data but also opens a path to rational design of new materials with attractive properties. In particular, the understanding of ultrafast photophysical processes in carbon dots can significantly help to rationally design new tailorable green photocatalysts facilitating artificial photosynthesis reactions.The main target of the project is to perform coherent theoretical analyses of various radiative and non-radiative processes occurring after the photoexcitation of carbon dots (CDs) using state-of-the-art quantum chemistry methods and programs including surface-hopping dynamics simulations.


Researcher: Lukas Neuman 


Inductive Bias of Deep Neural Networks for Computer Vision     

Karolina GPU  Alloc=1000           

Deep Neural Networks models have in the recent years dominated virtually all areas of Artificial Intelligence and Computer Vision. Their main advantage is that, given enough training samples, a training algorithm can automatically update network parameters to directly maximise given objective, such as image classification accuracy. Despite the recent success, the models are easily confused by trivial samples not present in the training set and even the largest models lack basic generalisation and reasoning abilities despite having hundreds of millions of parameters and despite being trained on millions of very diverse data samples -- suggesting that a fundamental piece of understanding is still missing.We propose that one of the missing pieces in current models compared to humans is an appropriate inductive bias -- the set of prior assumptions used to generalise and make a prediction based on a finite set of training samples. In this project, we want to exploit this observation and search for new inductive biases to incorporate them into modern Deep Neural Networks used in common Computer Vision tasks. This will result in Deep Neural Network models which require less parameters, which are more efficient, which are less confused by out-of-distribution data samples and which require less training data, as using an appropriate inductive bias is likely equivalent to even exponentially less training data.


Researcher: Marek Pecha     


ml4py – distributed machine learning tools (dev/stage) IV           

Barbora CPU  Alloc=2000;  Barbora FAT  Alloc=10;  Barbora GPU  Alloc=100;  DGX-2  Alloc=200;  Karolina CPU  Alloc=3000;  Karolina FAT  Alloc=10;  Karolina GPU Alloc=200;  LUMI-C  Alloc=2500;  LUMI-G  Alloc=400 

This project continues with the implementation and benchmarking of emerging tools for distributed machine learning called ml4py. Unlike standardly used machine learning kits, ml4py implements a mechanism for training predictors over cluster in its runtime using Message Passing Interface (MPI). Thus, no other frameworks such as Horovod or orchestration based on running multiple container instances of an application are required. Accelerators such as graphic cards are utilized using the OpenCL and CUDA technologies through the ml4py runtime.Current research related to machine learning commonly prefers deep learning techniques using (deep) neural networks. An associated training phase could be considered as a stochastic process that converges to an expected model. Our previous research on machine learning shows that determinism allows us to design training pipelines directly and explain the qualities of an attained learning model and numerical behaviour of the underlying solver more straightforwardly. However, these approaches are expensive in the sense of computational cost, and they can efficiently solve up to medium scale tasks by the state-of-the-art. Since ml4py is designed for massively parallel approaches, determinism can be incorporated into training pipelines for big data analysis. To speed the deterministic training phase up, stochastic optimization methods can be exploited. Basically, we find a good candidate for a solution using stochastic approaches expeditiously, and this candidate is used as a hint for deterministic and parallel predictor training subsequently. It helps to attain high-quality explainable models within a reasonable time.The ml4py framework is tested on datasets ranging from remote sensing, geoscience to computer vision.


Researcher: Ilia Ponomarev 


MoS2 Crystallization via Reactive Molecular Dynamics (CREAMD)            

Barbora CPU  Alloc=9000;  Karolina CPU  Alloc=38200   

Transition metal dichalcogenides (TMDs) are a family of layered materials with a wide range of applications, including tribological coatings, catalysis and materials for electronics. TMD thin films are commonly produced as an amorphous material. Tribological applications rely on the natural tendency of TMDs to crystallize and form superlubricious sliding surfaces in the course of exploitation. Catalysis applications benefit from generating specific defects. Electronics applications normally require very high crystallinity which sometimes needs to be achieved in relatively mild conditions to not destroy other elements of the system. Several computational studies of crystallization or melting of MoS2, a notable member of TMD materials family, have been published, utilizing either ab initio methods or reactive empirical potentials like REBO or ReaxFF. However, they were limited to small systems and/or single-layer setups due to the limitations of methods. In our study, we will search for the insights into MoS2 crystallization via reactive molecular dynamics, using recently developed force field. This parameterization matches ab initio description for a wide range of structures and allows to simulate crystallization of bulk MoS2. We will look into the mechanism of MoS2 crystallization and the effects of temperature and pressure as well as sliding and various impurities.


Researcher: Petr Šesták        


Machine learning potentials via PACEMAKER and VASP code for grain boundary segregation in nickel and iron    

Karolina CPU  Alloc=13000;  Karolina GPU  Alloc=1400;  LUMI-C  Alloc=4600       

This project combines ab initio simulations (VASP code) and atomic cluster expansion (ACE) to develop and test precise potentials for the study of grain boundary segregation in nickel and iron. Such approach, based on machine learning model, allows us to use the classic molecular dynamics with precision corresponding to ab initio methods, to employ large simulation cells with tens of thousands of atoms and extremely decrease computational resources. In general, this project helps to overcome some limitations of the present atomistic simulations and molecular dynamics.


Researcher: Dana Nachtigallova    


Advancing Hydrogen Production: Insights from Computational Study on Photocatalytic Water Splitting over TiO2 Surfaces              

Barbora CPU  Alloc=30000;  Barbora FAT  Alloc=931;  Karolina CPU  Alloc=90000;  Karolina FAT  Alloc=1000;  Karolina GPU  Alloc=28900;  LUMI-C  Alloc=9300

Hydrogen production via photocatalytic water splitting has garnered significant attention in pursuing sustainable energy solutions. Despite the immense efforts, most of the existing photocatalysts suffer from low activities, a narrow range of absorption, and low solar energy conversion efficiency. The main obstacle to obtaining an effective TiO¬2 surface variant is the lack of knowledge of the complex nature of different surface types, including their detailed geometry and electronic character. Our studies aim to fill this gap by collaborating closely with experimental partners using state-of-the-art techniques. This collaboration has the potential to contribute to the realization of the hydrogen economy with great applicability in various fields and recognize the challenges that must be dedicated the drive the area further.


Researcher: Pavel Krc             


Expanding climatological coverage of microscale urban air quality simulations   

Karolina CPU  Alloc=50000         

The ARAMIS project focuses on integrated assessments of air quality and its impacts on population as well as on the environment. One of its tasks aims to use the microscale atmospheric model PALM to create a climatologically representative set of realistic simulations that will allow to study and assess the urban air quality in metre-scale resolutions. Thus a computationally accessible set of fine-scale simulations will provide data on urban air quality in an unprecedented detail, allowing to asses development and mitigation strategies for the future.


Researcher: Antonio Cammarata      


first-principles simulations for PHOTOTRIBOlogy (PHOTOTRIBO)

Karolina CPU  Alloc=59869         

In static or dynamic conditions, nanotribological effects take place in nanoscale device; as a consequence, proper design, control, and reliability of, for instance, nanoelectromechanical systems (NEMS) become challenging. For example, friction in wheels and belts in nanoconveyor belt systems may limit the nanoscale mechanical transmission. The ideal way to modify nanoscale friction is to do it in a reversible way, that is, by avoiding permanent changes in the atom topology. Recently, it has been proposed to use ultraviolet and visible light to control the frictional properties; moreover, we observed a friction reduction of 60% under UV illumination when sliding on TiO2-anatase surfaces. However, the microscopic mechanisms determining the frictional response in the presence of light are still unclear. To this end, the final goal of this project is to use quantum mechanical simulations to investigate on the combined electronic and dynamic features determining the friction response of tribological materials at the nanoscale. To this aim, we will select  transition metal dichalchogenides and TiO2 thin films as case studies, which have large applicability in many nanotribological devices.


Researcher: Martin Klajmon


Molecular Simulation Study of Vaporization Properties of Complex Organic Compounds: Benchmarking Various Force Fields and Simulation Methods

Barbora FAT  Alloc=500;  Karolina CPU  Alloc=35900       

Knowledge of the vaporization properties is one of the key factors for rational design of materials and technological processes. Instead of costly and labor-intensive experimental determination, prediction of the vaporization properties using computational thermodynamic approaches, such as molecular simulations, could significantly accelerate the tailoring of new materials with favorable properties for specific technological purposes and conditions. This project aims at a thorough evaluation of the predictive performance of both molecular dynamics (MD) and Monte Carlo simulation techniques for the vaporization enthalpy of more than 30 technologically important complex organic compounds such as polyaromatic hydrocarbons, (poly)heterocyclic compounds, or ionic liquids. These substances serve as tailored organic semiconductor platforms or solvents. The simulated data will be evaluated against reliable experiment-based values available in the literature. Within MD, different force fields (nonpolarizable/polarizable will also be tested, providing the community with a clear picture of the performance of the state-of-the-art simulation techniques for the vaporization enthalpy.


Researcher: Jan Geletič          


Tests of local-scale simulations of urban climate with turbulent resolving model

Karolina CPU  Alloc=62000         

PALM modelling system allows to perform detailed simulations of conditions in urban areas, mainly with respect to phenomena of urban heat island, thermal comfort and air quality. ICS team significantly contributed to the model development and validation, moreover we use PALM for testing the efficiency of urban climate adaptation measures. For a HORIZON project CARMINE (Climate-resilient development pathways in metropolitan regions of Europe) we prepared two large domains for complex simulations including thermal comfort and air quality. Both domains represent the whole metropolitan areas in tens of meters resolution and will simulate most of important sizes of eddies above the urban area. We would like to test different configurations of the model and optimize these large simulations to enable us to compute multiple runs of the final simulations. Results of this testing are necessary for the following multi-year IT4I call application.


Researcher: Vojtěch Patočka


Seismo-geodynamic modeling of the Hellenic subduction            

Karolina CPU  Alloc=12000;  Karolina FAT Alloc=26           

Modern physics of the Earth investigates long-term geodynamic and short-term earthquake processes mostly independently. In this project, the seismological and geodynamic approaches will be combined to elucidate the Hellenic subduction, accompanied by intermediate-depth earthquakes (depth 60-200 km). It has been debated globally that these types of events are caused by dehydration embrittlement, but the process has not yet been fully understood. To verify the feasibility of such a mechanism, we shall develop new geodynamic models, involving realistic rheology and mineralogy, particularly focused on dehydration in the subducting lithospheric plate. The models will clarify how the observed intermediate-depth earthquakes are related to temperature, stress, and water transport. The models will also address deformation processes in the overriding plate, where shallow earthquakes pose a major natural hazard. The project builds on our 20-year seismotectonic cooperative research in Greece, involving eleven joint seismic stations. Petrophysical aspects will be solved with experts in the Netherlands and the USA.


Researcher: Jan Pichl              


A Controllable Empathetic Conversational Intelligence Using Dialogue Management and Large Language Models              

Barbora GPU  Alloc=700;  Karolina GPU  Alloc=900          

The open-domain conversational systems aim to achieve social intelligence by conducting coherent and engaging conversations. With the recent advancements in generative AI, the models such as GPT-3 or GPT-4 have proven effective in handling diverse user utterances and generating meaningful responses. However, a notable limitation of generative models is their lack of explainability, which hinders understanding the reasoning behind their responses. In this research, we address this issue by focusing on fine-tuning the models for a particular task and using them in advanced dialogue management. The dialogue management component will leverage ML models trained using reinforcement learning to adaptively select a proper conversational model according to the various parameters stored in the user profile. Our goal is to enhance conversational capabilities and contribute to the advancement of explainable dialogue systems. The research is based on our previous successful conversational bot Alquist that competed several times in the Amazon Alexa Prize competition.


Researcher: Sergiu Arapan   


A DFT study of electronic structure and thermodynamic properties of LK-99       

Karolina CPU  Alloc=9800;  Karolina GPU  Alloc=1600;  LUMI-C  Alloc=16300       

Very recently an experimental team from South Korea claimed to have obtained a room temperature superconductor material LK-99 [1,2]. The formula of this material is Pb10-xCux(PO4)6O, and its structure is very similar to lead apatite. The authors believe that the substitution of lead with copper in Pb-apatite leads to extraordinary superconducting behavior. If their claim turns to be true, then basically everything that runs on electricity is going to be affected, which could lead to another transition point in human history. At the present little is known about the mechanism that could induce superconductivity in LK-99. In this project we aim to describe the electronic structure of Pb-apatite and LK-99 materials by using DFT calculations, as well as the thermodynamic properties at finite temperatures to understand if such or similar materials can exhibit superconductivity.


Researcher: Michael Matějka             


How changing meteorological conditions impact snow cover and glacier melt on James Ross Island, Antarctica?

Karolina CPU  Alloc=70000;  LUMI-C  Alloc=21900           

The project investigates the impact of changes in meteorological processes on snow accumulation and glacier melt in the vicinity of the Czech Antarctic Station on James Ross Island, Antarctica. We will use the CRYOWRF model, coupling a state-of-the-art Weather Research and Forecasting (WRF) atmospheric model combined with a detailed snow model SNOWPACK for simulations. We will simulate two years with contrasting weather to see how the snow and glaciers respond to variable meteorological conditions, e.g., intense or negligible summer snowfall events. The model will be run in very high horizontal resolution to capture the impact of local terrain and snow redistribution on the mass balance of glaciers.


Researcher: Aleš Horák         


Slama - Slavonic Large Foundational Language Model for AI        

Karolina GPU  Alloc=4400           

The proposed Slama project focuses on building a new foundational language model concentrated on main Slavonic languages (Czech, Slovak, Polish, ...). The project’s primary goal is to explore the performance differences between state-of-the-art pre-trained multilingual models (where English texts represent the majority of training data) and a model tailored specifically to the Slavonic language group. The research will focus on developing generative models whose training data are more balanced in favor of the Slavonic language group rather than English. Therefore it should provide better results when used in AI tools processing mainly Slavonic languages. The resulting foundational model can then be easily applied in a range of AI tasks.


Researcher: Jiri Novacek       


Development of the workflow for an on-the-fly data analysis and publication of the raw cryo-electron microscopy data 

Karolina GPU  Alloc=2000           

Electron cryo-microscopy (cryo-EM) has become an invaluable technique in Structural Biology allowing determination of near-atomic resolution structures of >100 kDa macromolecular complexes or membrane proteins. The structural data are collected by imaging tens of thousands of molecules that can be either purified for the structural investigation in vitro or imaged directly in the cellular context (in situ). Thus, cryo-EM has recently gained significant attention in both academic and pharmaceutical research. As an imaging technique, the raw cryo-EM data are large (typically 1-2TB per one dataset) which poses specific requirements on the data processing, preservation, and publication in compliance with FAIR principles. We are developing a pipeline where the raw data collected at the electron microscopy facility are directly transferred to the remote HPC center for on-the-fly data processing using widely used data analysis software packages. The data transfer to the HPC, data sharing, and data archival are managed by iRODS federated cloud solution, and EUDAT services are utilized for the data publication after the embargo period.


Researcher: Josef Jon             


Automatic adversarial attacks on NMT evaluation and quality estimation methods          

Barbora CPU  Alloc=5000;  Barbora GPU  Alloc=4000;  DGX-2  Alloc=300;  Karolina CPU  Alloc=10000;  Karolina GPU  Alloc=17800     

Improvements large language models, stemming from availability of large-scale accelerated grids, have dramatically changed Neural Machine Translation (NMT) and automated evaluation of its outcomes. Recently, we have proposed a novel approach [1] leveraging the power of genetic algorithms, optimizing a list of candidate translations towards arbitrary evaluation metric. This method not only improves the quality of translation, but also offers comprehensive insights into the weaknesses and biases of the chosen metric. Our approach is able to generate examples that score well in a selected evaluation metric, but are in fact not correct translations of the source sentence, without any prior assumptions on properties of such adversarial example. Given the increasing presence of machine translation and language processing systems in general in our lives, it is important to be aware of any weaknesses, blind spots and biases in the automatic evaluation of such systems.Both the processes of generating the preliminary candidate translations and their subsequent evaluation and modification based on the metric impose substantial computational demands. We aspire to improve MT quality, analyze weaknesses of the current evaluation and quality estimation metrics, as well as create a new metric, based on insights learned from the optimization process.


Researcher: Libor Veis            


DMRG-AC study of longer oligoacenes   

Karolina CPU  Alloc=45000;  Karolina GPU Alloc=3000   

The character of the electronic structure of acenes has been the subject of the long-standingdiscussion. The recent scanning probe microscopy experiment has revealed the inelastic signal for tridecacene (13-cene) at 126 meV. which has been attributed to the spin excitation from the singlet biradical ground state to the triplet excited state. In this project, we would like to perform a thorough computational study of n-acenes for n = 6 – 14 by means of the density matrix renormalization group method (DMRG), which will be further corrected for the dynamical electron correlation by the adiabatic connection (AC) technique. We expect that our calculations will help to explain the experimental observation mentioned above, i.e. prove the emergence of biradical character in acenes series starting for 13-cene.


Researcher: Amutha Subramani        


Theoretical design and optimization of molecular thin 2D Sn-based perovskite photo absorbers interfaced between low dimensional material flakes        

Barbora CPU  Alloc=25000;  Barbora GPU  Alloc=4000;  DGX-2  Alloc=1000;  Karolina CPU  Alloc=23700 

Despite a growing body of research on lead free perovskite photovoltaic technology is also one of the leading goals as listed by European union, thus there is an urgent need to expand the search for the idea of combining distinct functional 2D lead free perovskite crystal materials with low dimensional material flakes hybrid junction modulated with dimensional and compositional engineering were till date not much explored that provides unprecedented platforms for exploring new mechanism to advance the multifunctionalities in 2D perovskites. This research program aimed to fulfill one of the keys enabling technologies (KET) of European Union. In this Research program, we thus intend to pursue, low dimensional transport layer functionalized molecular thin 2D tin based perovskite, by combining advanced theoretical and experimental results. The rapid development of theoretical designing and modeling tools, the low dimensional material interfaced 2D lead free perovskite heterojunctions will be designed, optimized with further controllable, scalable, and programmed modeling followed by efficient synthesis techniques of high-quality molecular thin 2D lead free perovskite heterojunction, with extraordinary performance will be designed and fabricated, for the special requirement applications such as biocompatible photovoltaic devices, solar power desalination, agro-photovoltaic, building-integrated photovoltaics, portable electronics, and so on. Our results will have a dramatic impact on the field of photovoltaic technology and interdisciplinary science.


Researcher: Hermann Detz  


Graphene nucleation from an ethylene based CVD process in the presence of oxygen    

Barbora CPU  Alloc=5000;  Karolina CPU  Alloc=5000;  Karolina GPU  Alloc=4600

Graphene and other 2D materials are widely considered as the basis for future electronic and optoelectronic devices. Intense research during the recent decades revealed their exciting material properties and the potential to realize novel device schemes based on atomically-thin materials, resulting in proof-of-principle devices. The next logical step is their integration with established electronic platforms, e.g. by providing novel sensing capabilities to portable electronic devices. To allow an economic large-volume production of such hybrid chips, it is necessary to grow defect-free graphene on suitable catalytic substrates like platinum. This project studies the nucleation of graphene from ethylene from a theoretical perspective. The obtained results are immediately relevant as they provide feedback for experimental studies with in-situ microscopy of this growth process. In particular, the calculations will help to identify characteristic defects that arise from the unavoidable presence of impurities like O2 or H2O in the growth reactor. A deeper understanding of the growth reactions will lead to optimized fabrication of future hybrid silicon-graphene chips.


Researcher: Valeria Butera   


Single-Atom catalysts based on TM@MoS2 for the CO2RR          

Barbora CPU  Alloc=6000;  Karolina CPU  Alloc=6000;  Karolina GPU  Alloc=4500

Within this project we aim at investigating new potential materials for the CO2 capture and utilization (CCU) process. In this regard, our focus will address the use of Single atom catalysts (SACs) since they have showed promising catalytic activity and sustainability. In a previous work, we have demonstrated the high catalytic activity of Ru@MoS2 towards CO2 conversion reaction, in which one single Ru atom is anchored on the 2D MoS2 monolayer. Here, we aim to explore the potential utilization of other transition metals (TMs) anchored on MoS2, TM@MoS2, as promising SACs for the CO2 reduction reaction (CO2RR) by using first-principles simulations. Following the same approach used previously, we will first evaluate the stability of the so-formed SACs. Once that the most stable SACs will be identified, we will investigate the reaction paths leading to the production of CO, CH4 and CH3OH as main value-added products. Our results will contribute to shed light on the potential of TM@MoS2 as selective SACs for the CO2RR and the understanding of the complex reaction mechanisms, aiming at the designing of more efficient catalytic systems. 


Researcher: Marketa Paloncyova  


Biocompatibility of carbon nanomaterials           

Karolina CPU  Alloc=19200;  Karolina GPU  Alloc=5040;  Karolina VIZ  Alloc=40    

Carbon nanomaterials (CNMs) are promising tools for theranostics, uniting medical therapy and diagnostics. CNMs are largely biocompatible, with a low toxicity, however, the principle of their interactions with biomaterials is not yet fully resolved. In this proposal, we will use multiscale resolution in molecular dynamics (MD) simulations in order to describe the interactions of CNMs with biomaterials, such as lipid membranes, nucleic acids and proteins. Regarding CNMs, we will focus on carbon dots and graphene derivatives, altering their size, shape and composition. We will focus on the interactions of CNMs with the respective biomaterials, with a special attention on the atomistic resolution of the interactions, their thermodynamic nature and mutual effects of the nanomaterial and biomaterial on each other. This proposed project will require both coarse grained and atomistic resolution, with fine tuning of parameters and simulation protocols. The gained insight will help us to understand the behavior of CNMs in biological environment and assess the safety and biocompatibility of the newly designed nanomaterials.


Researcher: Amina Gaffour             


Exploring the Intrinsically Disordered Domains in p53 Protein

Barbora CPU  Alloc=4000;  Barbora GPU  Alloc=4200;  Karolina GPU  Alloc=1000

The p53 tumor suppressor protein is a critical regulator of cell cycle arrest and apoptosis. Unfortunately, the molecular mechanisms underlying p53 function are not fully understood, particularly in intrinsically disordered regions (IDR) which have been observed to play an important role in the protein’s folding, stability, and most importantly, function. In this study, we will implement molecular dynamics to investigate the IDRs in p53, focusing on phosphorylation sites, ion concentration, and the tetramerization process on secondary structure and the function of the tumor suppressor proteins. Preliminary findings are promising, showing a well-established effect on small trajectories from the concentration of ions, as well as an alteration in how phosphorylated residues behave. Our findings will provide new insight into the inner workings of the protein which may provide key information for the development of new therapeutic strategies for treatment.


Researcher: Thibault Derrien              


FiRst-principle investigAtioN of crystalline materials exCitation upon linear and Elliptical polarization StatES (FRANCESES)    

Karolina CPU  Alloc=152000      

This proposal will support the EU project Horizon 2020 RISE “ATLANTIC” No. 823897 in its last term. This project aims at combining several theoretical formalisms together in view of improving the predictions capabilities for the development of applications based on intense laser processing of solids. In the 5-years EU project (2019-2024), the HiLASE Centre (FZU, Dolni Brezany) is driving the effort for describing the excitation of the electrons in various laser-irradiated materials, along with describing the transient change of optical properties of these materials. By transferring the insights gained from available first-principles microscopic descriptions [acquired from the time-dependent density functional theory (TDDFT)] to large-scale approaches, the project guides the invention of novel usages of intense laser light for modifying and functionalizing bulk and nano-materials. In this context, the present HPC project focuses on describing the excitation of electrons in crystalline materials by linear and elliptical states of light polarization. In the frame of ATLANTIC, this project will also support the training of young researchers in using advanced theoretical techniques adapted to the problems met in the engineering field of laser processing.


Researcher: Roman Bushuiev             


Generative transfer learning from DreaMS - a foundation model for mass spectrometry

Karolina GPU  Alloc=7400           

Advancing scientific knowledge in the life sciences and drug discovery heavily relies on discovering novel molecules. However, the current understanding encompasses only about ten percent of the chemicals present within the human body and the plant kingdom. Mass spectrometry, a fundamental technique for identifying new molecular structures, faces intricate challenges in interpreting complex data. Our hypothesis is that the main bottleneck currently is the scarcity of reference mass spectral data, which forms the basis for existing methods. In our previous project, enabled by Karolina high-performance computing, we introduced an innovative methodology that employs self-supervised deep learning to directly derive molecular representations from millions of raw unannotated mass spectra. We termed these representations DreaMS (Deep Representations Empowering the Annotation of Mass Spectra) and demonstrated their potential to revolutionize the understanding of mass spectrometry data. Building upon this foundation, our forthcoming project delves into an in-depth exploration of these outcomes through meticulous transfer learning. In particular, our ultimate goal is to fine-tune DreaMS for the accurate decoding of complete molecular structures from mass spectra. Successfully achieving this goal would imply an automated expansion of the chemical space discovered thus far.


Researcher: Oldřich Plchot   


Unlocking the Potential of Unstructured or Weakly Labeled Data              

Karolina CPU  Alloc=900;  LUMI-G  Alloc=9000  

In this proposal, we want to explore the concept of self-supervised and weakly supervised learning that will enable utilizing the data without any labels or leverage the coarse and often unstructured and incomplete labels that are typically associated with long media files. A typical example can be the source data for nowadays very popular dataset for speaker verification VoxCeleb – YouTube videos of various celebrities. The source video files can contain multiple celebrities (for whom weak labels are available) and even multiple other persons without any labels. Using the concepts of self-supervised or weakly supervised learning, one can build a powerful person embedding extractor or verification system by replacing the standard training paradigm (individual short segments have assigned single speaker labels) with one that uses all segments identified in the source material with multiple person labels belonging to such set of segments. The outcome of such training is not only a person embedding extractor but also a system that strongly models the persons re-occurring in the training data. Such a system can also have other practical uses for organizing old and new data, mining for higher-level information contained across the recordings, or more effective online processing.


Researcher: Jiří Chudoba      


Simulations for LHC experiment ATLAS  

Barbora CPU  Alloc=50000;  Karolina CPU  Alloc=150000;  LUMI-C  Alloc=8600   

Experiments measuring particle collisions in Large Hadron Collider at CERN require huge computing capacity for data analysis and Monte Carlo simulations. In cooperation with other projects, they developed and now use distributed environment called Worldwide LHC Computing Grid (WLCG). Basic needs for computational and storage resources for Run III which had started in July 2022 were covered by WLCG pledged resources. However significant fraction of resources is used from so called unpledged resources, which are provided by HPC centers, occasional cloud resources and resources from volunteers via BOINC. In 2021 and 2022, Vega supercomputer (part of PRACE network) almost doubled available computing resources for the ATLAS experiment. Karolina supercomputer was the second most significant HPC contributor for the ATLAS experiment in 2021.


Researcher: Varun Burde      


3D reconstruction for object manipulation          

Karolina CPU  Alloc=8000;  Karolina GPU  Alloc=7100     

3D reconstruction from images is a demanding and multidisciplinary domain that plays a critical role in various disciplines, including computer vision, computer graphics, virtual reality, and robotics. In the context of object pose estimation, which assumes that highly accurate 3D models of objects are available. In our previous work (currently under submission), we questioned whether current 3D reconstruction algorithms can be used to build such 3D models automatically. Our experiments showed that the current generation of algorithms needs to be significantly improved along multiple axes:1)Accuracy of the reconstruction/representation2)Reduction of runtime to make it feasible to run these approaches in a few seconds on constrained hardware, e.g., robots3)Robustness to using a minimal number of input images to reduce scanning times (and processing times)The research focuses on experimenting, prototyping, and developing a method leveraging volume rendering equations from NeRFs to speed up the process of object pose estimation with the low number of images, which may enable realtime deployment in robotics.


Researcher: Ivan Kološ          


Numerical modeling of load of structures in quasi-static effect of wind  

Karolina CPU  Alloc=1300            

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: Petra Sukova     


Revealing EMRIs with UFOs        

Karolina CPU Alloc=1300            

We will study how the repetitive transits of a star or black hole in the close neighbourhood of the supermassive black hole in the galactic nucleus can influence the process of accretion of plasma. We will use general-relativistic magneto-hydrodynamical code to follow the motion of the gas, while it is being pushed by the perturber along its trajectory. Some of the gas may be expelled from the accretion disc into the polar region, where a strong magnetic field ordered along the axis of rotation of the black hole accelerates the gas away from the center up to relativistic speed. On the other hand, the density waves caused by the motion of the star in the medium are spreading both downwards and outwards and hence influence the accretion rate on the black hole and the density distribution in the disc. To see the complicated structure of the outflow and inflow, robust 3D simulations are needed with resolution high enough to capture the details of the flow close to the black hole horizon while stretching far enough to see the outflow on larger scales. We will find the observational traces of the presence of the object close to the nucleus, which will help to find promising targets for the upcoming European space gravitational observatory LISA (Laser Interferometer Space Antenna).