32^nd open access competition RESULTS

THE ALLOCATION COMMISSION DECIDED ON THE ALLOCATIONS WITHIN THE 32^nd OPEN ACCESS GRANT COMPETITION AS FOLLOWS:

Researcher: Miroslav Medveď            

OPEN-32-1       

Computational Modelling of Photo-Responsive Molecules    

Karolina CPU  Alloc=16900                    

A rapidly growing interest in the development of photo-responsive molecular systems stems from the human dream to precisely control properties and functions of materials at the atomistic level analogously to daily-life ON-OFF devices, which can be achieved by light. Compared to other external stimuli, the activation of photo-responsive materials by light is non-invasive and highly localized, which found countless applications in biomedicine and pharmacology (e.g., the photochemical control of drugs, drug delivery, bioimaging) as well as in nanotechnology and material engineering (e.g., healable and recyclable materials, reversible adhesives, data storage, energy storage, and light-guided self-assembly).  So far, the photo-responsive systems have mostly been optimized in terms of the position of absorption maxima rather than the absorption intensity. However, there are some applications (relying on low-intensity radiation sources) in which the effectiveness of absorption plays a crucial role.The scope of the project is to design of photo-responsive molecules optimized for efficient activation by low-intensity light in the visible spectral region using static quantum-chemical (QCh) approaches as well as non-adiabatic molecular dynamics (NAMD) simulations.

Researcher: Georgij Ponimatkin        

OPEN-32-10    

Learning Large Scale Object Manipulation Dynamics from the Video

LUMI-C  Alloc=500  LUMI-G  Alloc=8000          

The availability of large-scale video datasets featuring humans manipulating objects (such as Ego-Exo4D, Ego4D or EpicKitchen) opens the possibility of creating models that learn the dynamics of object interactions purely from the video. In this project, we propose to create a foundation model for predicting the dynamic object motion conditioned on the initial image of the scene and human description of the task. To achieve that, we will train the proposed model on multiple videos of human-object interactions with text captions. Such a model can then be used to guide robotic manipulation without large amounts of robotic manipulation data, since it will provide anticipated trajectories for the desired manipulation task given a textual caption describing the expected action. This could drive innovations in the field of industrial and home robotics, where robotic agents learn skills from the wealth of internet data and then more easily interact with the real world and complete complex multi-step tasks without any adaptation.

Researcher: Nikolaos Samaras         

OPEN-32-11    

Hydrodynamical simulations in MOND            

Karolina CPU  Alloc=25100                    

Although ΛCDM is widely accepted as the standard cosmological model, offering a plethora of explanations for the formation and evolution of galactic systems, it encompasses plenty of unresolved issues. The yet undetected Dark Matter and the Hubble tension are just a few of them. MOND or Milgromian Dynamics is an alternative theory able to show fundamental scaling relations of galaxies, arising naturally for gravitationally bound systems feeling low accelerations. We propose a detail hydrodynamical set of simulations which explore the emergence of the cosmic web and group of galaxies in a modified gravity scenario.

Researcher: Petr Kouba          

OPEN-32-12    

Machine Learning for De Novo Design of Multi-State Proteins              

LUMI-C  Alloc=1200  LUMI-G  Alloc=11500     

Despite the recent progress in AI-based protein design, producing functional proteins still remains a major challenge in biotechnology and biopharmaceutics due to the inability of state-of-the-art methods to account for protein dynamics. This project addresses this challenge by aiming to develop a diffusion generative model for de novo design of multi-state proteins. Designing a multi-state protein means creating a dynamic protein capable of exhibiting various structures as specified by the multiple states. Such capability of a single protein to dynamically modify its structure is often crucial for the function of the protein - especially in the case of enzymes, where the function is to perform a certain chemical reaction. Our method will therefore advance the state of the art for protein design by accounting for protein dynamics and will enable the development of next generation enzymes for plastic degradation and carbon capturing or biopharmaceuticals for the treatment of conditions such as stroke and Alzheimer's disease.

Researcher: Petr Miarka         

OPEN-32-13    

Meso-FEM approach to concrete fracture       

Barbora CPU  Alloc=4050        

The focus of recently developed construction materials is mainly considering the CO2 emissions reduction as the highest demand to be met, while keeping the similar mechanical properties. Unfortunately, the concrete’s highly heterogenous meso-structure poses a challenge for multiple fields of building industry. One of them is structural design, which is highly depended on certain mechanical properties and their correct estimation. Hence, the correct determination of stresses, which are responsible for damage and crack propagation in the materials is crucial.  The origin of the stress concentrations can have multiple causes from different loads at different age of the material. Focusing on the correct estimation of the stress at the aggregate-binder interface is crucial for improving structural safety, while boosting the application potential of these new construction materials.

Researcher: Anton Bushuiev

OPEN-32-14    

Accurate protein binder design via customized diffusion models       

Karolina CPU  Alloc=600  Karolina GPU  Alloc=5600  

Designing proteins that bind to specific target proteins has numerous pharmaceutical and biotechnological applications, including vaccine development and biosensor creation. Despite the virtually infinite theoretical space of possible proteins, only a limited number can be practically tested in a laboratory, often with low success rates for desired bioactivity. In this project, we will develop a machine learning approach to generate proteins that bind to target proteins with high affinity. Specifically, we will develop diffusion-based generative models that can be customized for proteins of interest, thereby enabling accurate prediction of binders. We will leverage a recent and highly successful paradigm in computer vision that guides diffusion models toward specific conditions, which, in our case, are the desired protein targets. Our developed approach will be applied to enhance protein binders for staphylokinase, a thrombolytic drug candidate, which will subsequently undergo wet lab testing.

Researcher: Oldřich Plchot   

OPEN-32-15    

Advancing Self-Supervised Speech Models by Integrating Target Speaker Clues        

LUMI-C  Alloc=1200  LUMI-G  Alloc=12000     

We will focus on advancing the capabilities of speech foundational models to provide target-speaker-related predictions. Inspired by human selective hearing, our primary goal is to improve the performance of various speech processing tasks when a speaker of interest is known beforehand, and they are present in challenging acoustic scenarios e.g., in a cocktail party problem. These tasks include applications such as automatic speech recognition (ASR), voice activity detection (VAD), speech enhancement (SE), etc..The current self-supervised learning (SSL) models are primarily optimized for single-source speech and tend to suppress non-dominant speakers as background noise, thus underperforming in mixture speech scenarios common in conversational contexts.  We intend to modify the pipeline of self-supervised speech representation learning by incorporating target speaker cues into existing SSL models, such as HUBERT, WavLM, wav2vec, etc.  We will start by attempting to leverage auxiliary target speaker clues when evaluating existing SSL models while keeping their parameters fixed. We will test the performance in target-speaker tasks such as target-speaker ASR, target speech extraction, and personalized VAD. Subsequently, we will shift to integrating target speaker information from diverse modalities into existing large-scale speech models in different ways: directly during training from scratch, further fine-tuning of existing pre-trained foundational models, including leveraging techniques like adapter-tuning to minimize training costs.

Researcher: Pavel Praks         

OPEN-32-16    

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

Barbora CPU  Alloc=1000  Barbora GPU  Alloc=500  Karolina CPU  Alloc=1600  Karolina FAT  Alloc=200 Karolina GPU  Alloc=1100        

Power systems are part of a critical infrastructure. Their operations are affected by various external sources which increase demands on the complexity of mathematical models used for simulating their behavior. Many such simulations need to be run by computer control systems to assess the consequences of control actions. To use the computational resources more effectively, advanced optimization 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, optimization 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 optimization problems. This project aims to develop and adapt specialized tools for high-performance computing infrastructure, providing recommendations for the configuration and control of analyzed energy systems.

Researcher: Paulo Miguel Guimarães da Silva         

OPEN-32-17    

Leveraging Large Language Models and Vector Databases for Enhanced Natural Language Querying      

Karolina CPU  Alloc=300  Karolina GPU  Alloc=700      Karolina CPU  Alloc=300  Karolina GPU    Alloc=700           

In this project, we aim to test and benchmark vector databases such as Milvus or pgvector, and experiment with multi-modal models and large language model queries. The specifics of this research will focus on leveraging BLIP [1] to generate vector representations of images and perform text-to-image queries. After creating the necessary vector embeddings and being able to query the vector database, this project will also benchmark the solution on larger vector and database sizes.

Researcher: Pavlo Yefanov    

OPEN-32-18    

Benchmarking Generalization of Imitation Learning Algorithms for Robotic Manipulation              

LUMI-G  Alloc=12000  

Imitation learning methods have shown great promise in teaching robots complex manipulation tasks by learning from expert demonstrations. These methods can provide a solution for agile robotics, enabling agents to adapt to new scenarios and remain robust in the face of environmental changes. For instance, a robot in a factory could adapt to a new facility layout or modified components without requiring reprogramming each time changes occur. This flexibility significantly enhances the efficiency and versatility of robotic systems in dynamic industrial and home settings. Recent works claim to improve data efficiency or the ability to generalize across different scenarios. However, a comprehensive open-source benchmark comparing these algorithms' generalization capabilities in simulation is lacking. Our project aims to fill this gap by conducting a thorough evaluation of recent state-of-the-art deep learning imitation learning algorithms. We anticipate that the results will provide valuable insights for achieving agile robotics in both industrial and home applications. We will compare algorithms such as Diffusion Policy, ACT, and RT-1 across a diverse set of manipulation tasks, including object grasping and fixed-base and mobile robot manipulation. We will use ManiSkill simulation environments with a diverse set of objects to manipulate.

Researcher: Dominik Legut  

OPEN-32-19    

Design of Magnetocaloric and Elastocaloric materials            

Barbora CPU  Alloc=50000  Barbora FAT  Alloc=250  Barbora GPU  Alloc=3000  DGX-2  Alloc=250  Karolina CPU  Alloc=32400  Karolina FAT  Alloc=100  Karolina GPU  Alloc=9700  LUMI-C  Alloc=8200                        

In this project we intend to study three types of effects: 1) Magnetocaloric (MCE), 2) Barocaloric (BCE), and 3) Pressure-tuned MCE: here the pressure will not be used to get BCE but just to tune the transition temperature to the desired value, which then will be followed by the MCE measurements. In this case, in contrast to systems with the first-order magnetic transition coupled to a structural transition, a structural transition is not necessary, since lattice entropy does not contribute significantly to the entropy change. Importantly, we want to study how magnetoelastic effects affect the value of MCE and BCE in a selected group of materials. Strong modification of the chemical composition will introduce a large disorder to the crystalline structure, which may result in interesting physical effects, but may make the interpretation of the results difficult. Magnetic and thermodynamic measurements as a function of temperature, external magnetic field, and pressure will allow us to determine values of magnetic entropy change, adiabatic temperature change, and refrigerant capacity.

Researcher: Carlos Manuel Pereira Bornes

OPEN-32-2       

Understanding Sn-zeolites under operando conditions using Neural Network Potentials

Karolina CPU  Alloc=34400  LUMI-G  Alloc=4400         

The transition from fossil to renewable feedstocks, such as biomass, for chemical and fuel production requires the development of new catalysts able to convert new feedstocks efficiently. Metasilicate zeolites, such as Sn-zeolites, have been suggested as active and selective catalysts in several liquid-phase reactions. Even though the current consensus suggests Sn is incorporated into the framework where it behaves as a Lewis acid site, the speciation, dynamical behaviour and reactivity of such sites are currently unknown. Conventional computational methods often rely on simplistic models where a single/few molecules interact with active sites modelled via static or very short dynamical calculations. We have recently developed a reactive neural network potential (NNP) that can readily model the behaviour of aluminosilicate zeolites over long equilibration runs while retaining the accuracy level of the training set, at (meta)GGA DFT level. This project aims to extend this methodology to train NNPs for Sn-zeolites able to perform large-scale simulations that can probe the stability, speciation of the active sites (open vs close sites), and dynamics taking place at temperatures, pressures, and surface coverages that accurately represent those used during a catalytic process, i.e. at operando conditions.

Researcher: Tomas Brzobohaty         

OPEN-32-20    

Deep learning for engineering shape optimization       

Karolina CPU  Alloc=110300  LUMI-G  Alloc=18100    

The rise of complex AI models such as LLMs has been made possible by the availability of sufficient data. On the other hand, the lack of data to create a good AI model for digital twin technology in engineering practice is undeniable. Exascale computing and AI are revolutionizing computational fluid dynamics (CFD) engineering by enabling unprecedented precision and efficiency. Exascale systems allow the simulation of complex fluid behaviors at thousands of shape modifications or operating states and deliver results in a reasonable time. AI algorithms, particularly machine learning models, can analyze CFD datasets generated by thousands of simulations, identifying patterns and, with a combination of evolutionary algorithms, optimizing processes that are too complex for traditional methods. AI-driven models can predict outcomes and suggest improvements, streamlining engineering projects' design and testing phases and allowing the creation of so-called digital twins. Combining exascale computing and AI in CFD engineering boosts computational speed and accuracy and fosters innovation, leading to more efficient and sustainable engineering solutions. This powerful combination is paving the way for advancements in aerospace, automotive, and environmental engineering, where precise fluid dynamics are critical. Our research aims to develop an AI model suitable for fast shape optimization and further enhances the capabilities of creating digital twin technologies in engineering practice. A large dataset for a specific task from external aerodynamics on a real-world scale will be created.

Researcher: Karel Tůma         

OPEN-32-21    

Effect of Diffusion on Formation of Omega Phase in Beta-Ti Alloys    

Karolina CPU  Alloc=1500  Karolina GPU  Alloc=200  

Modern aircraft are among the most sophisticated machines, with their efficiency and reliability stemming from decades of multidisciplinary research. Similarly, the surgical fixation of fractures and the replacement of hip or knee joints are some of the most significant advancements in medicine.Metastable beta-titanium alloys are known for their optimal combination of strength and ductility, which is obtained by complex thermomechanical treatment accompanied by several phase transformations. One of the most studied is the beta-to-omega transformation during which nano-particles of the omega phase are formed. This process is a result of two major mechanisms: elasticity and diffusion. However, it is rather impossible to design experiments that would separate these mechanisms and enable us to fully understand the formation of the omega phase.We propose to develop a phase-field model to describe the formation and evolution of the omega phase, considering both the athermal and the isothermal beta-to-omega transformation. The phase-field method is a powerful computational tool for the modelling of phase transformations and microstructure evolution problems.

Researcher: Ivo Oprsal           

OPEN-32-22    

3D Modeling and Risk Assessment for Nuclear Facility Safety             

Barbora CPU  Alloc=1400  Karolina CPU  Alloc=1700  LUMI-C  Alloc=500                     

Earthquakes have major social and economic impacts, causing casualties and infrastructure damage worldwide. This includes urban areas and critical urban-service infrastructure, such as dams, power lines, pipelines, and, notably, nuclear waste disposal sites. This project is directly related to the proposed expansion of nuclear energy sites into the basin areas of the Czech Republic, highlighting the significance of local seismic amplification as a key factor influencing seismic hazard. The research aims to enhance planned approaches by focusing on numerical modeling of seismic responses across a range of selected sites with varying geological structures. The results will integrate with geological structure information and previous study outcomes into a regional-scale model, creating optimal methodologies for assessing amplification and screening all external seismotectonic hazards at the site. Advanced computational resources (IT4I) will be utilized for detailed seismic modeling, enabling more precise predictions and analyses of potential seismic events. Numerical modeling using computer resources will play a crucial role in simulating different seismic scenarios and understanding the impact of various geological conditions on seismic amplification.

Researcher: Jiří Jaroš

OPEN-32-23    

Closed-loop individualized image-guided transcranial ultrasonic stimulation III        

Barbora CPU  Alloc=5000  Barbora FAT  Alloc=50  Barbora GPU  Alloc=100  DGX-2  Alloc=50;  Karolina CPU  Alloc=600  Karolina FAT  Alloc=50  Karolina GPU  Alloc=700  LUMI-C  Alloc=200;  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: Marketa Paloncyova     

OPEN-32-24    

Graphene-based materials in nanomedicine  

Karolina CPU  Alloc=12500  Karolina GPU  Alloc=1100  Karolina VIZ  Alloc=50  LUMI-C  Alloc=1600  LUMI-G  Alloc=4800                        

Graphene derivatives are promising tools to nanomedicine, combining interesting electronic properties with biocompatibility. In this project we will focus on interactions of graphene derivatives with biomembrane models by means of multiscale molecular dynamics simulations. Reduced graphene oxide has been shown to be a functional electrode material and we will evaluate its biocompatibility and potential toxicity in contact with brain cell membrane models for its use as stimulating electrode for treatment of Parkinson’s disease. On the other hand, we will evaluate the interactions of graphene acid derivative with bacterial membranes, as these materials were shown to have antibacterial activity. In this project, we will develop protocols for combining all-atom and coarse-grained molecular dynamics simulations of nanomaterial-membrane interaction studies, that can be in the future extended to other biomembrane or nanomaterial models.

Researcher: Subhasmita Ray             

OPEN-32-25    

Magnetism in MXenes and its nanoscale control via phonon excitations        

Karolina CPU  Alloc=32700  LUMI-C  Alloc=8300         

MXenes, as two-dimensional materials composed of transition metal carbides and nitrides, unequivocally demonstrate exceptional electrical, thermal, and mechanical properties, positioning them as highly promising for future technology advancements. This study focuses on exploring MXenes' intrinsic magnetic properties with an innovative approach: controlling their magnetism at the nanoscale through phonon excitations, but exclusively through theoretical modeling and simulations. This research intends to delve into the interactions between phonons and magnetic moments within MXenes by leveraging theoretical models and simulations using VASP and Phonopy. Our goal is to precisely control magnetic phases and anisotropies by identifying and influencing specific phonon modes using simulations. This cutting-edge strategy could lead to a breakthrough in designing spintronic devices and magnetic storage systems, enabling the creation of high-performance, magnetically adjustable materials. Completing this project will not only deepen our understanding of phono-magnetic interactions in two-dimensional materials but also inspire new developments in nanoelectronics and materials science.

Researcher: Laëtitia Lebec   

OPEN-32-26    

Stability of a basal ocean in large icy moons of Jupiter and Saturn     

Karolina CPU  Alloc=8200        

Due to their subsurface water ocean and strong potential interactions between water and rocks, ocean worlds have become key objects for the planetary science community. These interactions are among the main criteria in exobiology for the appearance and sustainability of life. Among these objects, the most studied today are small ones, such as Enceladus, whose internal ocean is in direct contact with the core, offering a suitable environment for efficient exchanges of chemicals between both. Conversely, the habitability of large icy moons, such as Ganymede, has long been questioned due to the existence of a layer of high-pressure (HP) ice between the core and the ocean. This layer prevents direct water-rock interactions and could limit the chemical exchanges between both. However, several studies have shown that efficient heat and mass transfer can still occur by convection through this HP ice shell. Our last study [1] on this topic indicates that if the salts infiltrating the ice from the core densify the ice considerably, the salt/ice mixture will accumulate at the bottom of the HP ice layer. This could lead to the formation of a thin and highly salted basal ocean located between the core and the HP ice layer. The goal of our project is to model the evolution of this ocean to determine its stability over time and its possible effect on the overall dynamics and heat and mass transfer efficiency through the HP ice layer.

Researcher: Michal Vavrecka             

OPEN-32-27    

Procedural robotic environments with multi-policy algorithms           

Karolina CPU  Alloc=4000  Karolina GPU  Alloc=4700

This project aims to enhance robotic manipulation capabilities by adopting the myGym robotic simulator, integrating it with a procedural environment generator, and training custom multi-policy networks to solve long-horizon manipulation tasks. The primary objective is to develop and optimize these networks using advanced computational resources, including multi-CPU and multi-GPU nodes on a high-performance computing cluster.

Researcher: William Shakespeare Morton  

OPEN-32-28    

POL2PHASE     

Karolina CPU  Alloc=5800  Karolina GPU  Alloc=13100  LUMI-C  Alloc=9900               

Liquid-Liquid-Phase-Separation (LLPS) is key to cellular function, enabling the concentration and storage of proteins in specific regions. Proteins' intrinsically disordered regions largely drive this process, allowing cells to rapidly assemble or disassemble these clusters as needed. This process is similar to how we might store and retrieve winter clothes. Improper control of these condensates can result in neurodegenerative diseases or cancer. Recent experiments have manipulated protein condensate formation in vitro, a therapeutic potential for controlling abnormal condensates. While this is promising, the studies are laborious and limited by experimental techniques. Our study aims to replicate and expand upon these studies in silico using molecular dynamics (MD) across multiple scales. Our project will develop high-throughput coarse-grained methods to study how proteins colocalize during LLPS. We plan to use rational design and machine learning to create synthetic sequences or mutations that can either enhance or reduce colocalization among specific pairs of proteins. We will confirm our computational predictions by using experimental approaches like confocal and cryo-electron microscopy, small-angle X-ray scattering, and various other biochemical methods. The theories developed through this work will guide and set benchmarks for related in vitro research towards therapeutic applications.

Researcher: Raman Samusevich      

OPEN-32-29    

Prediction of enzymatic reaction mechanisms using machine learning          

LUMI-G  Alloc=6700  LUMI-G  Alloc=6700       

Terpene synthases (TPSs) generate the scaffolds of the largest class of natural products, including several first-line medicines. Yet terpenes and terpenoids are too complex to be efficiently synthesized industrially and are typically extracted from plants. Predicting reactions catalyzed by TPS enables screening for enzymes that produce compounds of interest. The amount of available protein sequences is increasing exponentially, and accurate computational prediction of their function remains an unsolved challenge.

Researcher: Jakub Kopko      

OPEN-32-3       

Harnessing Synthetic Data for Machine Learning-Based Protein-Lipid Docking          

LUMI-C  Alloc=1000  LUMI-G  Alloc=13700     

Understanding protein-lipid interactions is vital for cell signaling, membrane dynamics, and drug discovery. Our project aims to enhance protein-lipid docking by examining how different data splitting methods influence the generalization capabilities of deep learning docking models. We will also create synthetic data that covers a wide range of interactions to boost prediction accuracy. This research will provide advanced tools for studying protein-lipid interactions, motivated by our goal to better understand Apolipoprotein-E (APOE). APOE plays an important role in the development of Alzheimer's disease, and the methods and knowledge gained from this project will be applied to analyze its properties more effectively.

Researcher: Roman Pleskot 

OPEN-32-30    

Molecular insight into calcium channels in plants      

Karolina CPU  Alloc=17500  Karolina GPU  Alloc=2100  LUMI-C  Alloc=9300  LUMI-G  Alloc=2900      

In plants, calcium is crucial for regulating various developmental processes, such as cellular signaling, cell wall growth, or maintenance of vacuolar pH. In all these processes, calcium channels play an indispensable role, contributing to calcium homeostasis in individual compartments of the plant cell. To this date, however, the molecular mechanism of calcium transport in plants is largely unknown. In this project, we aim to study the behavior of selected calcium channels in a model membrane environment, providing insight into channel-facilitated transport at the atomistic level of detail. The obtained results will allow us to understand better how plants regulate their growth, communicate throughout their body or control their response to pathogens and serve as the basis for developing next-generation crops.

Researcher: Denys Biriukov  

OPEN-32-31    

Design of Bacteria-Resistant Polymeric Implant Coatings      

LUMI-G  Alloc=19200  

This project will investigate molecular-level interactions between polymeric surfaces and bacterial saccharides to inform the novel design of implant coatings that minimize bacterial attachment. Functionalized polymers are commonly employed to coat metal implants, enhancing their characteristics. However, such functionalization can increase the unwanted bacteria affinity, leading to post-implantation infections. By utilizing molecular dynamics simulations, we will identify specific bacterial lipopolysaccharide moieties responsible for the binding of Gram-negative bacteria to functionalized polymeric materials. We will focus on systematically identifying the chemical groups that can effectively coat implant surfaces while exhibiting minimal interaction with bacterial polysaccharide sequences. The collected molecular insights will contribute to the development of knowledge-based functionalization strategies aimed at reducing bacterial attachment to implants.

Researcher: Petr Hyner          

OPEN-32-32    

Theoretical Insights Into Neural Networks       

DGX-2  Alloc=300  LUMI-G  Alloc=4580 

This project is focused on experimental evaluation of several theoretical ideas explaining behavior of neural networks. The first part of the project will deal with the reasoning abilities of neural networks, where by reasoning we mean the ability to compose facts which the network never processed in the same input during training. The second part of the project will deal with the analysis of the loss landscapes of neural networks and how they influence the training process. It is widely believed that the loss landscape of the overparametrized neural networks has a complicated structure with many singularities. Our aim here is to experimentally verify the structure of these landscapes together with the path traversed during training. The last part will deal with the sparsity of neurons activated inference. It is believed that when processing one given input, most of the neurons in large neural network are not active. We want to investigate this fact and design a neural network architecture that will activate only small subset of all available neurons during processing.

Researcher: Radim Špetlík   

OPEN-32-33    

Diffusion-based Neural Network Weights Generation for Style Transfer Problems     

LUMI-G  Alloc=6900    

This project explores an innovative approach to increasing learning speed of style transfer techniques of neural networks. Traditional style transfer methods based on neural networks often struggle with efficiency and adaptability, particularly when applying learned styles to new datasets. The research addresses this by leveraging a novel diffusion-based process to generate neural network weights. The result is a more versatile and powerful style transfer capability, opening new possibilities for artistic and practical applications in image and video processing.

Researcher: Jiri Klimes           

OPEN-32-34    

Accuracy and precision for extended systems XIII      

Barbora CPU  Alloc=36000  Karolina CPU  Alloc=15700  LUMI-C  Alloc=3700              

Computer simulations are nowadays widely used to understand different phenomena in our world. They are especially useful when experiments are difficult to perform or when we want to \play\" with the experimental systems, that is try what happens if we change in one way or another. One of such useful applications is prediction of crystal structures of a given molecular compound. In computer, we can construct a range of possible structures, obtain their energy, and compare which one is the lowest one thus the most stable. However, there is an issue, the method that we use to calculate the energy needs to be accurate so that the predictions are close to what happens in real world. This is difficult as we need to use quantum mechanics and the calculations tend to be computationally demanding. We thus need to use more approximate method and hope that the predictions will be close to the real world. Within our project, we want to go beyond this \"hope\" stage and understand which methods are reliable for predicting the energy ordering and which methods can give problematic results. We want to do this by taking clusters with two or three molecules and checking the predictions of the more approximate methods against reliable reference data that we will generate.

Researcher: Libor Veis            

OPEN-32-35    

Multi-scale modeling of 2D materials via QMC-quality Machine Learning Potentials and the Density Matrix Renormalization Group method    

Karolina CPU  Alloc=94100     

The research focuses on 2D materials. Using quantum Monte Carlo (QMC) methods, we have previously addressed atomic and electronic structure of perfect 2D crystals, mono- and bi-layers of phosphorene, transition metal dichalcogenides (TMDs), and h-BN and their property tuning via straintronics and layer engineering. Regarding a 2D monolayer as a basic “lego brick”, the properties can be perturbed by defects and perturbations introduced on purpose via “proximitizing” one 2D monolayer by another one, thus inducing desirable properties of one material into the other without destroying its own strong properties. An example being spin properties induced into graphene or phosphorene by proximitizing it by TMDs. New ways are explored to artificially stack van der Waals (vdW) 2D layers in the so-called 2.5D materials with unique physical properties. A new degree-of- freedom, the twist angle, characterising the mutual rotation of the monolayers may be explored, opening the field of twistronics. Here we plan to focus on defects in monolayers, their migration, and twistronics. Such studies are customarily conducted at the DFT or GW level. To significantly raise the quality bar, we propose use of the most accurate, albeit also the most computationally demanding many-body QMC methods. Due to the length-/time-scales involved, cost of brute force application of QMC methods is prohibitive. We propose a workaround, combining QMC with recently developed machine learning tools. Additionally, we aim to leverage the valuable high-quality QMC results and extend the applicability of the recently developed machine learning-enhanced Density Matrix Renormalization Group (DMRG) method to the aforementioned systems.

Researcher: Thibault Derrien             

OPEN-32-36    

First-principLe investigatiOn of sample oRiEntation in electron excitatioN of metal-oxide crysTals upon lINEar polarization (FLORENTINE)         

Karolina CPU  Alloc=75300  Karolina VIZ  Alloc=1000

As a member of FZU Institute of Physics, the department of Scientific Laser Applications (SLA) of HiLASE employs predictive quantum formalisms for enabling the development of novel applications based on usage of intense laser processing of solids and nanomaterials. In recent years, SLA established the time-dependent density functional theory (TDDFT) as an excellent prediction tool for supporting the design of future photonic & optoelectronic components along with their functionalization. Started in Jan. 2024, the national inter-institution MSMT-OPJAK project “Sensors for information society” SENDISO (2024-2028) aims at developing both fundamental and applied research for enabling production of prototypes of new kinds of sensors produced using intense laser light. In this context, the present HPC project focuses on describing the excitation of electrons in poly-atomic crystals that are relevant to the SENDISO project, when they are irradiated by intense linearly polarized light. 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: Róbert Babjak   

OPEN-32-37    

Direct laser acceleration as a high-yield gamma-ray source  

Karolina CPU  Alloc=31400     

Conventional electron accelerators can deliver electron beams with energies of a few tens of GeV after an acceleration distance of several kilometers. These tools are crucial for exploring particle physics and material science. Additionally, the accelerated electrons, wiggling in the accelerating field, serve as a radiation source for various imaging techniques and cancer treatments.A laser-based alternative to conventional accelerators has the advantage of being able to deliver up to a hundred times higher electron charge in only a few centimeters. This project aims to explore the potential of a laser-plasma-based acceleration mechanism called direct laser acceleration (DLA) as a radiation source. A unique property of this mechanism is its ability to shift photon energies to the gamma-ray part of the electromagnetic spectrum, specifically to hundreds of MeV.Massively parallel simulations of laser-plasma interactions in the regime of strong-field electrodynamics using the OSIRIS framework will support the analytical description of DLA electrons as a radiation source. Computational time will also be dedicated to supporting experimental campaigns investigating the DLA mechanism on lasers such as OMEGA-EP (USA) and PETAL (France). Furthermore, advanced schemes for enhancing the yield of radiation will be tested to explore the capabilities of multi-petawatt laser facilities that are either currently functioning or about to be commissioned in the near future.

Researcher: Jakub Gazdoš    

OPEN-32-38    

Particle-in-cell simulations of plasma in pulsar magnetospheres      

Karolina CPU  Alloc=2900        

Pulsating Radio Sources (Pulsars) are rapidly rotating neutron stars. They have an envelope of strong magnetic fields filled with hot plasma of charged particles, called the magnetosphere. At the magnetic poles, the relativistic and hot plasma emits beams of radiation that travel across the universe. Most of the time, the star rotational axis is not aligned with the magnetic field axis. Thus, the beam rotates with the star and, in the right circumstances, is detected as periodic radio pulses when the beam points toward and away from the Earth.The study of pulsars is one of the youngest fields of astrophysics, arising half a century ago. Many breakthrough discoveries have been made, even the Nobel Prize in physics was awarded. Still, numerous questions remain to be answered about the processes that generate and influence the radio emissions.In our work, we plan to study two physical phenomena of the radio beam generation. First, we aim to study the effects of the curvature of spacetime introduced by the massive neutron star on the particle kinematics, the electromagnetic (EM) fields as well as the interaction of the plasma with EM fields and the resulting radiation. Second, we are interested in how the matter released from the star surface can suppress the radio wave emission by studying the star surface temperature.

Researcher: Ievgeniia Korniienko     

OPEN-32-39    

Magnetic contribution to acoustic wave propagations in ferromagnetic crystals with strong magnetoelasic coupling in near ferromagnetic resonance region        

Barbora CPU  Alloc=20000  Barbora FAT  Alloc=300  Barbora GPU  Alloc=1000  DGX-2  Alloc=250  Karolina CPU  Alloc=25400  Karolina FAT  Alloc=100  Karolina GPU  Alloc=9100  LUMI-C  Alloc=6400          

In the last years the coupling between spin waves (magnons) and acoustic waves (phonons) has regained significant attention of scientists. In particular, spintronics on the basis of spinconversion (concept underlying spin mediated energy interconversion among electricity, light, sound, vibration, heat) is considered promising for nowadays challenges with requirement to reduce energy consumption. In ferromagnetic materials with strong magnetoelastic coupling, such energy conversion is possible between magnons and phonons. Thus, the possibility of exciting ferromagnetic resonance using surface acoustic waves was described in Ref. [1]. In our work, on the contrary, we will investigate the influence of the magnetic subsystem on the speed of acoustic waves in the frequency range close to the frequencies of ferromagnetic resonance.

Researcher: Sergiu Arapan    

OPEN-32-4       

A computational study of the effect of microstructure on the coercivity and Curie temperature in magnetic L10 structures          

Barbora CPU  Alloc=8700  Barbora FAT  Alloc=500  Karolina CPU  Alloc=5500  Karolina FAT  Alloc=100  Karolina GPU  Alloc=2200  LUMI-C  Alloc=1500          

Because of their large magnetocrystalline anisotropy energies, L10 magnetic alloys may play an important role in ultra-high density magnetic recording media and permanent magnet applications. Both, the magnetocrystalline anisotropy and the Curie temperature of the L10 alloys can be affected by the atomic ordering transformation. The details of the microstructure of these materials affect coercivity and exchange interactions in a complex way by the formations of various types of defects such as c-axes variants, grain boundaries, antiphase boundaries, and twins. The presence of such defects may either enhance or diminish the performance of these magnetic materials. In this study we will apply our recently developed supercell approach for calculating the Heisenberg exchange interactions for estimating the effect of the disorder on the Curie temperature, and the special quasi-random structure method for calculations the magnetocrystalline anisotropy energy.

Researcher: Srimanta Maity 

OPEN-32-40    

Particle-In-Cell Modeling of Multi-Stage Laser Wakefield Acceleration for Generating GeV-Class Electron Beams     

LUMI-G  Alloc=19200  

A plasma-based acceleration scheme for particle acceleration via space charge waves was firstproposed by Y. Fainberg in 1956. This innovative approach overcomes one of the major limitationsof conventional accelerators: the restricted electric field gradient in radio frequency acceleratingstructures. Extreme laser-plasma accelerating gradients, demonstrated experimentally by variousresearch teams, pave the way for the development of compact laser-plasma accelerators (LPAs).Such compact accelerators have the potential to serve as electron beam drivers for a wide range ofapplications, including free electron lasers (FELs), Thomson sources, and even electron-positroncolliders with TeV energy.Laser-plasma acceleration has been the focus of extensive research over the past few decades,aiming to generate high-quality electron beams with energies reaching up to GeV levels. The multi-stage method is crucial for achieving steady, high-energy, and high-quality electron beams in laser-plasma-based acceleration. In single-stage laser wakefield acceleration (LWFA), the maximumenergy an electron beam can gain is limited by (i) the pump depletion length and (ii) the dephasinglength. A multi-stage LWFA approach employs several independent laser drivers and shortertargets, thus overcoming the constraints imposed by depletion length and dephasing length.However, multi-stage LWFA depends on the efficient coupling of the injected electron beam intothe wakefield of subsequent acceleration stages, which is highly influenced by the parameters ofthe externally injected beam, target, and laser driver.In this study, we investigate the multi-stage acceleration process using Particle-In-Cell simulations.Specifically, we examine the charge coupling and acceleration of externally injected electron beamsin the booster stage. We aim to uncover the underlying physical mechanisms and identify criticalparameters that affect the trapping of the injected beam into the wakefield.

Researcher: Suresh Ravisankar        

OPEN-32-41    

Nanofriction Control Study in Transition Metal Dichalcogenides using Ab Initio Molecular Dynamics and Machine learning force field (NaCSTAM)          

Barbora CPU  Alloc=27309  Karolina CPU  Alloc=29600  Barbora CPU  Alloc=27309  Karolina CPU  Alloc=29600     

This project aims to design tribological materials with tunable frictional properties controlled by external electric fields using advanced atomistic simulation techniques. We address the high computational costs of ab initio molecular dynamics (AIMD) by integrating machine learning Force Field (MLFF) to automate and accelerate force field creation, maintaining accuracy while improving efficiency. This approach is applied to study MoS2 monolayers interacting with metallic substrates (Au and Ag) and a silicon AFM tip. Comprehensive analysis including density of states (DOS), orbital polarization, charge density, phonon spectra, and their relation with the friction coefficient enables us to understand the relations between the coupled electronic/structural features and the friction response. This study is then expected to advance the knowledge of the processes governing the friction response at the nanoscale in van der Waals materials and to guide the design of new tribological materials with target response.The outcomes of this project will contribute to the advancement of the project “Robotics and advanced industrial production” (reg. no. CZ.02.01.01/00/22_008/0004590) co-funded by the European Union, and the project “Nanocontrol” funded by the Czech Science Foundation (project no. 24-12643L).

Researcher: Dana Nachtigallova

OPEN-32-42    

Computational Study of Sustainable H₂ Production by Photocatalytic Water Splitting over TiO₂ Surfaces        

Barbora CPU  Alloc=10000  Karolina CPU  Alloc=49600  Karolina FAT  Alloc=100  Karolina GPU  Alloc=11500;  LUMI-C  Alloc=8700  LUMI-G  Alloc=9100             

The production of hydrogen through photocatalytic water splitting has garnered considerable interest as a sustainable energy solution. However, despite extensive research, the majority of current photocatalysts demonstrate low activity, limited light absorption range, and poor solar energy conversion efficiency. A key obstacle in developing an effective TiO₂-based photocatalyst is the insufficient understanding of the complex nature of its various surface types, including their precise geometries and electronic properties. Our research aims to address this knowledge deficit by collaborating closely with experimental partners utilizing cutting-edge techniques. This interdisciplinary approach has the potential to significantly advance the hydrogen economy, offering broad applicability across multiple fields and addressing the critical challenges necessary to drive progress in this domain.

Researcher: Karel Sindelka  

OPEN-32-43    

Computer simulations of complex liquids       

Barbora CPU  Alloc=32000  Karolina CPU  Alloc=16900          

Aqueous solutions are omnipresent in nature, industrial processes, and daily life. Understanding their behaviour in inhomogeneous environments (self-assembled or confined systems) in equilibrium and non-equilibrium (e.g., shear flow) conditions is important in many applications from medicine to environmental protection. This project investigates interactions of structured fluids (micellar solutions, surfactant bilayers) with surfaces; these play important role in industrial or household products (e.g., fabric softeners, cosmetic products). We use mesoscopic simulations to provide molecular-level insights into chemical and physical behaviour of these systems in and out of equilibrium.

Researcher: Simona Sajbanova         

OPEN-32-44    

The use of computationally expensive DFT functionals for investigation of salt-cocrystal continuum area            

Karolina CPU  Alloc=9400        

Pharmaceutical solid forms such as salts and cocrystals play a crucial role in pharmaceutical applications. The difference between salt and cocrystal is given only by the position of single hydrogen (1), making it essential to develop precise techniques for identifying this position.  Differentiation between salt and cocrystal compounds holds significant importance within the pharmaceutical industry, both for regulatory purposes and overall quality control. The Food and Drug Administration has explicitly outlined in their 2018 guidelines the necessity for accurate identification of pharmaceutical phase (2). We are developing a computational method for such hydrogen position determination. The method does not require crystallographic data from high-quality monocrystal, and it work even with data from powder samples.

Researcher: Aleš Horák          

OPEN-32-45    

Slama - Slavonic Large Foundational Language Model for AI, extended training         

Karolina GPU  Alloc=3600  Karolina GPU  Alloc=3600

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. The current extension aims at further training with enlarged training data for more capable models.

Researcher: Jan Lehecka       

OPEN-32-46    

Multimodal Transformers for Low-Resource Languages          

LUMI-G  Alloc=5300    

The project's goal is to continue our computationally challenging research in the field of Artificial intelligence with a focus on speech technologies in low-resource languages. Our team has strong experience with training state-of-the-art (SOTA) models for many speech- and text-related tasks, such as Automatic Speech Recognition (ASR), Text-to-Speech (TTS), Document Classification, Dialogue systems, etc. We focus mainly on low-resource languages, such as Czech, Slovak, Vietnamese, or Ukrainian, as these languages are usually completely missing or dramatically under-represented in models released by large AI companies. Training such models requires collecting a large amount of data from speech and text modalities and a lot of computing resources. We plan to release all foundation models we will train during this project (as we did during the previous projects) publicly for the speech community.

Researcher: Martina Ćosićová           

OPEN-32-47    

Testing of diabatization using neural network approach on systems Rgn+ and (N₂He)⁺

Karolina CPU  Alloc=3100  Karolina CPU  Alloc=3100

In this project, I would like to focus on the use of neural networks for diabatization of electronic bases. This problem is part of my dissertation, which I am working on as part of a cotutelle agreement in collaboration with Université Toulouse III - Paul Sabatier, the project is supported by the international Barrande Fellowship program. The objective of this project is finding a numerically convenient (so-called diabatic) representation of the potential energy matrix that appears in the system of partial differential equations for the motion of atomic nuclei. This system is obtained by Born-Oppenheimer separation of the electron and nuclear degrees of freedom in time-dependent Schrodinger equation. Neural networks diabatization is a new and promising approach which could significantly reduce the computational demands of dynamical simulations of quantum systems.  In this project, I would like to test the developed method on systems Rgn+ and (N₂He)⁺.

Researcher: Abhiram Bindu Ramanathan    

OPEN-32-48

Effect of crystallinity on tribological and mechanical properties of MoS₂        

Barbora CPU  Alloc=28800  Karolina CPU  Alloc=20600  Barbora CPU  Alloc=28800

Molybdenum disulfide (MoS2) is one of the most extensively studied transition metal dichalcogenides (TMDs) due to its effectiveness as a solid lubricant. In addition to its low friction characteristics, MoS2 exhibits excellent mechanical properties such as high elastic modulus and fracture toughness. However, these properties are significantly influenced by the degree of crystallinity of MoS2. A comprehensive study on the effect of the degree of crystallinity on the coefficient of friction and mechanical properties has not been conducted before. To this end, we aim to explore the influence of the degree of crystallinity on the friction, elastic and fracture properties of MoS2 through molecular dynamics (MD) simulations. The study will provide insights into the mechanics and behavior of MoS2, which can be crucial for optimizing its synthesis and application in engineering and lubrication technologies.

Researcher: Ivan Kološ           

OPEN-32-49    

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

Karolina CPU  Alloc=700  Karolina CPU  Alloc=700     

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 Klajmon 

OPEN-32-5       

Toward the Computational Prediction of Acceptable Solvents for Sustainable Biopolymers

Karolina CPU  Alloc=29700  Karolina FAT  Alloc=200;  LUMI-C  Alloc=2500  LUMI-G  Alloc=2400      

The goal of this project is to develop a sophisticated computational protocol for identifying efficient and environmentally acceptable solvents for biopolymers such as lignin, cellulose, and chitin. The protocol will leverage state-of-the-art techniques, including molecular dynamics and Monte Carlo simulations, the quantum mechanics-aided COSMO-SAC thermodynamic model, and a synergistic combination of these methods. Special attention will be devoted to studying and understanding the hydration and solubilizing mechanisms. If successful, the protocol will determine appropriate green solvent media (potentially operating at lower temperatures), facilitating the further utilization of biopolymers from renewable sources or waste materials to create useful products (e.g., textiles, hydrophobic coatings, and non-petroleum-based platform chemicals). Additionally, the large datasets produced within this project could be valuable for training AI-based models in the near future. This study is related to our MŠMT project “AI-Supported Search for Environmentally Acceptable Solvents for the Dissolution and Stabilization of Biopolymers for Their Utilization as Sustainable Materials” (LUABA24070), in collaboration with the group of prof. Werner Kunz from Universität Regensburg.

Researcher: Petr Bardonek   

OPEN-32-50    

Automating Vertical Reuse of Portable Stimulus Models in Functional Verification with Static Analysis

Barbora CPU  Alloc=1500  Barbora CPU  Alloc=1500 

Ever-increasing demands on embedded and computer systems increase their design complexity, putting more pressure on their error-free creation. A typical way of ensuring this is through functional verification, which is becoming harder with the increasing complexity of designs. New approaches and methods have to be developed, such as the Portable Test and Stimulus Standard (PSS), which provides a higher level of abstraction and defines graph-based models of verification intent used to drive stimuli generation. It aims to support the use of modular and reusable concepts in simulation-based verification.In the realm of digital system design, achieving the desired functionality involves combining modular units. This project aims to follow this approach for functional verification by combining PSS models for the verification of design units to create a PSS model for verifying larger designs that include these units. The steps include: (1) developing and testing data flow analysis of RTL for design understanding, (2) developing and testing control flow analysis with SMT solvers for discovered conditions, (3) generating constraints for PSS model connections, and (4) connecting PSS models. Creating a publication covering the experiments is a key goal.

Researcher: Diana Csontosová          

OPEN-32-51    

Applications of multi-orbital Hubbard model - Dynamical mean-field study 

Karolina CPU  Alloc=12500     

Dynamical Mean-Field Theory (DMFT) is one of the most successful methods for modelling physical properties of materials with strongly correlated electrons. It has been used to study, e.g., metal-insulator phase transitions, magnetic ordering, or excitonic condensation. In this project, we intend to use DMFT to gain insight into materials with intriguing physical properties. We propose a novel approach to understanding the implications of valence changes in ACu₃Fe₄O₁₂ (A stands for lanthanides), materials that crystallise in quadruple perovskite structure and exhibit phase transitions sensitive to small changes in external parameters such as temperature, pressure and doping. We will also model CrSb, an altermagnetic material exhibiting spin polarisation. DMFT and our own post-processing methods will allow us to characterize these materials on several levels, understanding both single-particle behaviour and collective excitations.

Researcher: Rene Kalus         

OPEN-32-52    

Ternary recombination processes in cold rare-gas plasmas – phases III (heavy rare gases)

Karolina CPU  Alloc=4900  Karolina CPU  Alloc=4900

Following preceding computational projects (OPEN-28-24 and FTA-24-7), within which strategies for the numerical modeling of ternary recombination of atomic ions in cold rare-gas plasmas were developed, tuned, implemented in the form of optimized software packages, and tested in a specific case study of cold argon plasmas, we propose to use the developed tools to model similar processes for heavier rare gases like krypton and xenon. Primarily, the effect of the atomic mass, range of inter-atomic interactions, and the role of electronic relativistic effects will be investigated.

Researcher: Jan Kuneš            

OPEN-32-53    

X-ray magnetic circular dichroism in altermagnets     

Karolina CPU  Alloc=20300     

Antiferromagnetic (AFM) spintronics is a rapidly developing field of physics with a span from fundamentally new phenomena to technological applications. It is driven by discoveries of new materials and new physical phenomena. Recently, a new class of AFM materials – altermagnets – was identified. Unlike conventional AFMs, altermagnets host the anomalous Hall effect, linear magneto-optical effects or spin polarized bands, previously associated with ferromagnetism. Our recent joint theory + experiment study has demonstrated that x-ray magnetic circular dichroism (XMCD) provides powerful tool for investigation of altermagnets [3]. While the application to altermagnets is novel, the experimental technique itself is broadly used.We have developed a method for calculation of XMCD and other core-level spectra based on Anderson impurity model and dynamical mean-field theory. The method has been applied to a number of transition metal compounds and other materials. Its quantitative accuracy and predictive power have been demonstrated. A predictive theory of XMCD in altermagnets allows us to apply our method to a hot field in physics and to profit from our recently acquired experience.

Researcher: David Barina      

OPEN-32-54    

Verification of the Collatz problem      

LUMI-G  Alloc=22600                

One of the most famous problems in mathematics that remains unsolved is the Collatz conjecture, which asserts that, for arbitrary positive integer n, a sequence defined by repeatedly applying the function f(n) = 3n+1 if n is odd, or f(n) = n/2 if n is even will always converge to the cycle passing through the number 1. The terms of such sequence typically rise and fall repeatedly, oscillate wildly, and grow at a dizzying pace. The conjecture has never been proven. There are however experimental evidence and heuristic arguments that support it. As of 2024, the conjecture has been checked by computer for all starting values up to 1.5 × 2^70 [Barina2024]. Our project aims to extend the computational records to higher values, and possibly to find some interesting results (numbers with extraordinary expansion factors, climbing to extremely high values, a counterexample, etc.)

Researcher: Damien Lucien Michael Gagnier           

OPEN-32-55    

Common Envelope Evolution: Local 3D MHD Simulations of Intraorbital Dynamo    

Karolina CPU  Alloc=53900     

Common envelope evolution (CEE) is a critical phase in binary star systems, where two stars interact within a shared envelope, leading to various close binary systems such as cataclysmic variables, X-ray binaries, Type Ia supernova progenitors, and planetary nebula nuclei. It also precedes most stellar mergers detected by gravitational wave (GW) detectors. Despite significant efforts, hydrodynamic simulations have struggled to match the orbital properties of observed post-CEE binaries. These simulations model the cores as point masses with artificial gravitational softening, causing non-physical flow structures and resulting in final orbital separations too large for a GW merger within the age of the Universe. In the preliminary project OPEN-30-55, I examined the impact of gravitational softening and intraorbital grid resolution on the evolution of the orbital separation and intraorbital flow dynamics using hydrodynamical simulations. I found that inadequate resolution or excessive softening radii, which are almost always the case in current ab initio CEE simulations, lead to incorrect orbital contraction timescales and dramatically underestimate the complexity of gas dynamics in the intraorbital region. Building on this work, in my proposed project, I will conduct the first 3D magnetohydrodynamical (MHD) simulations with proper intraorbital resolution and adequate gravitational softening of the binary's potential. I expect significantly stronger magnetic field amplification than current MHD CEE simulations and the formation of bipolar jet-like outflows. This work will play a crucial role in reconciling observed and simulated final orbital separations.

Researcher: Alvaro Patricio Prieto Perez      

OPEN-32-56    

Studying the contribution of changes in meteorology and emissions in future air-quality over Central Europe     

Karolina CPU  Alloc=9500        

Emissions, along with meteorology and climate, influence air pollution. In this research, we aim to study future air-quality in Central Europe at a moderate resolution using the regional climate model Weather Research and Forecasting model coupled with Chemistry (WRF-Chem), along with different scenarios based on the Representative Concentration Pathways (RCPs). We consider both present-day and future scenario based meteorology and emissions according to RCP4.5 and RCP8.5 in the simulations planned. The domain of the simulations has a 9km resolution, is centred over Prague and, to ensure statistically robust results, each simulation period is ten years long. With this study, we expect to asses the contribution of changes in both emissions and meteorology in future air-quality over Central Europe.

Researcher: Sofya Belov        

OPEN-32-57    

Dynamics of the Solar Corona in the Era of Data Intensive Observations        

Barbora CPU  Alloc=26500  Karolina CPU  Alloc=43900          

Plasma physics examines fundamental astrophysical processes using the solar corona as a model. Creating and confining plasma on Earth is costly and challenging, increasing interest in natural plasma systems like the Sun’s atmosphere. Solar plasma studies offer insights into various plasma conditions. The solar corona, a high-temperature, fully ionized part of the atmosphere, is key for understanding fusion reactors and solar-terrestrial relations.In the proposed research we address outstanding questions of modern solar physics connected with dynamic phenomena in the solar atmosphere summarised below. The key common theme linking the research are magnetohydrodynamic (MHD) waves which are a ubiquitous feature of solar atmospheric dynamics.Solar flares, caused by rapid magnetic energy conversion, result in sudden brightness increases. Despite extensive data, flare energy releases remain unclear. Quasi-periodic pulsations are common but underexplored features of flares. Another key question is about the hot temperature of the corona. From basic physics, one would expect a decreasing temperature as one moves away from the energy source at the core of the Sun. However, the solar atmosphere of the Sun shows temperatures well above 1 million Kelvin.Coronal mass ejections (CMEs) are powerful and crucial for space weather forecasting and understanding exoplanet habitability. Despite advances in CME studies, questions remain about their morphology, kinematics, and aerodynamic drag force. Upcoming missions like ESA’s Proba-3 promise significant advancements in CME research.

Researcher: Taoufik Sakhraoui          

OPEN-32-58    

Magnetism in MXene and graphitic carbon nitride 2D materials          

Barbora CPU  Alloc=20000  Karolina CPU  Alloc=9400             

MXenes, as members of 2D materials family, are composed of transition metal atoms (M = Ti, V, Cr, Mo, W, Sc, Hf, Zr, Ta, Y, La or Nb), X (carbon or nitrogen), and surface terminal groups (T = -O, -OH, -H, -F, -Cl, -Br, -S, -I , -Te, -NH, -BrI, and -ClBrI) and have the general formula Mn+1XnTx, n=1, 2, 3, 4. In some cases, the transition metals may be well combined and further extending application potential of these promising materials. Moreover, graphitic carbon nitride (CxNy) 2D material consists of C, N, and some impurity H, connected via triazine-based patterns. It possesses electron-rich properties, basic surface functionalities and H-bonding motifs due to the presence of N and H atoms. It is thus regarded as a potential candidate to complement carbon in material applications. Due to their high stabilities, MXenes and CxNy have attracted significant attention over the past decade.On the other hand, accurate theoretical description of 2D materials remains challenging. Due to the presence of multiple configurations and magnetic alignment for each of them, and the need for accurate predictions of their properties, supercells with hundreds of atoms are required.

Researcher: Ales Vitek           

OPEN-32-6       

Two-dimensional density of states as a complete information of thermodynamic properties of atomic and molecular systems 

Karolina CPU  Alloc=1100        

Microscopic systems consist of dozens of particles, atom or molecules, exhibit different properties than we have experience with bulk limit of matter. If we fixed e. g. system volume and temperature (NVT), the pressure could fluctuate, and system may behave slightly differently in comparison if we fixed temperature and pressure (NPT, and volume can fluctuate now). Our goal is performed simulation in constant -pressure and temperature parallelly at different temperatures and pressures, collect energy-volume histograms and through the two-dimensional multiple histogram method compute two-dimensional density of states. From this we would like to compute results in NVT, NPT and NVE ensembles and compare results directly with the simulations obtained from NVT and NVE simulations.

Researcher: Luigi Cigarini     

OPEN-32-7       

Strategies of chemical doping for high-efficiency thermoelectricity in transition metal nitrides

Barbora CPU  Alloc=20000  Barbora FAT  Alloc=150  Barbora GPU  Alloc=700  DGX-2  Alloc=100  Karolina CPU  Alloc=19600 Karolina FAT  Alloc=100  Karolina GPU  Alloc=4700  LUMI-C  Alloc=4100  LUMI-G  Alloc=16100     

Thermoelectricity means converting temperature differences into electric voltage and vice versa. Finding materials with excellent thermoelectric properties is important for technological applications, such as enhancing energy efficiency in industrial plants, automobiles, and household devices. This project aims to study the thermoelectric properties of scandium nitride and chromium nitride, two promising compounds for future thermoelectric applications, using computational models of material science. We seek to develop effective strategies to improve the performance of the future cost-effective energy conversion devices, which could significantly impact technological advancements in energy efficiency.

Researcher: Alicia Moranchel Basurto          

OPEN-32-8       

Magneto-hydrodinamyc interactions of circumestellar disks and intermediate- mass mergers stars  

Karolina CPU  Alloc=6900  LUMI-C  Alloc=5800           

The magnetic field of a rotating star can have a strong influence on the matter in an accretiondisk leading a complex types of accretion. Additionally, the topology of the stellar magneticfield plays an important role in disk accretion and in the photometric and spectralappearance of the stars. Being able to model numerically the accretion process in differentscenarios gives us the possibility to better understand the dynamics of accretion disks arounddifferent objects. However, this is not an easy task due to the large number of freeparameters. Therefore, the use of observational parameters reported on different stellarobjects is essential.Here, the goal of this project is to study the effect of the strong magnetic field on the gasaccretion rate and the inner disk structure for a particular type of objects called FS CMa stars,which observationally exhibit an intense magnetic field that can reach thousands of Gauss(for instance, for star with a mass of solar masses). Currently, this type of star isthought to arise from a post-merger event, implying that there may be a very complex non-stationary magnetic field configuration acting on the accretion disk. Therefore, we will use inour numerical models multipole configurations of the magnetic field, which has never beenstudied for stars with such intense magnetic fields.

Researcher: David Zihala      

OPEN-32-9       

Advancing Risk Stratification and Treatment in Multiple Myeloma Through Genomic Analysis of Circulating Tumor Cells     

Karolina CPU  Alloc=2100        

Multiple myeloma (MM), the second most common hematologic malignancy, is marked by the uncontrolled proliferation of abnormal plasma cells (PCs) primarily within the bone marrow (BM). However, some of these cells escape the BM and enter the bloodstream. The presence of circulating tumor cells (CTCs) is emerging as a critical prognostic indicator in MM, with higher levels correlating with a poor prognosis. Currently, only a few studies have explored the biological mechanisms underlying the presence of CTCs in MM patients. Notably, previous genomic research has focused exclusively on patients with high CTC levels. In this project, we will analyze genomic and transcriptomic data from patients with varying CTC levels, including those with no detectable CTCs. This approach will enable us to understand the progressive nature of this significant risk factor. Additionally, we aim to demonstrate the feasibility of using whole genome sequencing (WGS) of CTCs from peripheral blood (PB) to achieve accurate genetic risk stratification. Such minimally invasive diagnostics from PB could provide a more convenient, practical, and less harmful approach to managing MM patients in the future.