VÝSLEDKY 34. KOLA VEŘEJNÉ GRANTOVÉ SOUTĚŽE

Děkujeme všem žadatelům za podání projektů ve 34. kole Veřejné grantové soutěže IT4Innovations. V tomto kole jsme obdrželi 88 žádostí.

ALOKAČNÍ KOMISE VE 34. KOLE VEŘEJNÉ GRANTOVÉ SOUTĚŽE IT4INNOVATIONS ROZDĚLILA VÝPOČETNÍ ZDROJE TAKTO:

Řešitel: Anton Bushuiev  

OPEN-34-1     

All-Atom Any-Modality Molecular Diffusion Model for Biomolecule Design     

LUMI-G  Alloc=22300             

This project uses cutting-edge artificial intelligence to help design new medicines. We are building a computer model that can generate molecules—such as proteins or small molecules—that bind tightly and specifically to other molecules involved in disease, such as viral proteins or cancer-related signals. These binding molecules can become the basis for new therapies. Our model learns from detailed 3D structures of how biological molecules interact. Unlike earlier methods, our model works at atomic resolution and can handle a wide range of molecule types, including proteins with post-translational modifications, DNA, RNA, and drug-like compounds. The resulting machine learning model will be freely available for the scientific and biomedical communities, making it easier and faster to invent better treatments. We are also collaborating with experimental labs to test our model’s designs in the lab, potentially speeding up drug discovery and reducing costs.

Řešitel: Prashant Dwivedi             

OPEN-34-10  

Impact-Induced Damage Evolution in Alloyed Tungsten Surfaces for Fusion Reactors (IDEA)            

Barbora CPU  Alloc=15800;  Barbora NG  Alloc=4200;  Karolina CPU  Alloc=7100      

Tungsten (W) is a key material in fusion reactors because of its exceptional thermal stability and high melting point. However, it faces serious challenges in extreme plasma environments, where high-velocity dust impacts can cause erosion, cracking, and surface degradation. These effects are particularly critical for plasma-facing components (PFCs), which directly interact with the fusion plasma. To improve W’s resilience, researchers are exploring alloys with elements like chromium (Cr) and zirconium (Zr), which show promise in enhancing mechanical toughness and oxidation resistance.  This project uses cutting-edge molecular dynamics (MD) simulations to explore how W-based alloys respond to high-speed dust impacts, simulating realistic conditions inside fusion reactors. We investigate how variables like impact angle, temperature, and alloy composition influence crater formation, defect generation, and material damage. The study focuses on W-Cr and W-Zr alloys, aiming to identify combinations that best resist erosion and extend component lifespan.  The results will help design longer-lasting materials for fusion energy, reducing reactor downtime and boosting economic feasibility. Beyond fusion, the findings could also support applications in aerospace and high-performance manufacturing, where materials must endure extreme heat and stress.

Řešitel: Vít Svoboda          

OPEN-34-11  

Probing dynamically modified chirality during photodissociation          

Barbora NG  Alloc=9000;  Karolina CPU   Alloc=15100    

Chirality is a fascinating phenomenon that plays a crucial role in Earth’s biochemistry. Experimentally, chirality can only be probed using another chiral tool, such as circularly polarized light. Upon photoionization of a chiral molecule, scattered photoelectrons carry the chiral fingerprint of the molecule, which can be measured. Epichlorohydrin, a small chiral molecule of significant industrial interest, will serve as a prototypical molecule for investigating, both experimentally and theoretically, photo-induced dynamics leading to changes in chirality. In particular, we will explore chirality changes due to photodissociation, which leads to fragmentation of the molecule into methyloxirane residue, yet another chiral molecule. This will allow us to, for the first time, probe chirality change during ultrafast chemical transformation between two chiral molecules. Theoretically, we will model this whole photodissociation reaction using ab initio molecular dynamics and predict the resulting time-resolved chiral signals based on fully ab initio electron-molecule scattering calculations. These calculations will directly support our experimental results, which will be measured this June at the ELI Beamline Facility as part of the GAČR JUNIOR STAR project.

Řešitel: Sergiu Arapan     

OPEN-34-12  

Tunning the magnetic properties of transition metal dihalides with pressure.

Barbora CPU Alloc=3900;  Barbora GPU  Alloc=1000;  Barbora NG  Alloc=3200;  Karolina CPU Alloc=5400;  Karolina GPU  Alloc=1500;  LUMI-C  Alloc=1500   

Transition metal dihalides are a class of 2D van der Waals (vdW) materials presenting multiferroic order and non-collinear spin arrangements. This type of multiferroics with strongly coupled magnetoelectric effects have the potential to be integral components in functional vdW-based nanoscale devices. The non-colinear magnetism arises from frustration between the first nearest neighbor (NN) ferromagnetic and third-nearest neighbor antiferromagnetic (AFM) exchange within the triangular lattice planes. This non-collinear magnetic state may host a spin-induced ferroelectric polarization and that enhancing the interlayer interactions could help stabilize the multiferroic phase at higher temperatures. Given the high sensitivity of interlayer interactions to interlayer distance in vdW materials, the electronic and magnetic properties cab be tuned by hydrostatic pressure. Within this project we aim to study the effect of the hydrostatic pressure on the Heisenberg exchange interactions in nickel dihalides MX2 (M=Mn,Fe,Co,Ni; X = Cl,Br,I) by means of electronic structure calculations.

Řešitel: Jan Heyda              

OPEN-34-13  

Properties of liquids interacting with pressurized gases: from natural gas to hydrogen economy     

Barbora NG  Alloc=2600;  Karolina CPU  Alloc=4400;  Karolina GPU   Alloc=1600;  LUMI-C  Alloc=2700;  LUMI-G  Alloc=2000

As the world transitions to a low-carbon economy, hydrogen is expected to replace natural gas as a key energy carrier. However, challenges arise in its transportation and storage due to safety concerns. Liquid Organic Hydrogen Carriers (LOHCs), such as methanol, toluene, and dibenzyltoluene, offer a promising solution as they are compatible with existing fuel infrastructure. Additionally, natural gas contains methane and ethane, alongside BTEX compounds (Benzene, Toluene, Xylenes), which can crystallize and block pipelines, especially during Liquefied Natural Gas (LNG) production. Similarly, methanol and ethanol are used to prevent methane hydrates from clogging pipelines. Given their continued relevance in both natural gas and hydrogen-based industries, understanding the behavior of these compounds under industrial conditions is essential. Our research employs molecular simulations to investigate interactions of these transport liquids with pressurized methane, ethane, and hydrogen, providing insights critical for the future energetic stability of our society.

Řešitel: David Adamczyk

OPEN-34-14  

Knowledge Generalization in LLMs through Knowledge Injection            

Karolina GPU  Alloc=900;  LUMI-G Alloc=1000     

This research project investigates large language models' (LLMs) ability to generalize newly acquired knowledge. While current LLMs can work with explicitly provided facts, their ability to derive new knowledge from limited inputs remains insufficiently explored. We propose a \Knowledge Injection\ approach that overcomes traditional Retrieval-Augmented Generation (RAG) systems' limitations. Conventional RAG solutions merely retrieve and utilize existing facts. Our approach examines how to introduce knowledge into LLMs as general rules, enabling models to independently derive new facts. We employ structured knowledge graphs where, beyond individual facts, we define meta-rules describing relationships between relations.

Řešitel: Michal Kolar         

OPEN-34-15  

Simulations of a complete atomistic mitoribosome         

Barbora NG  Alloc=8800;  Karolina CPU Alloc=14800;  LUMI-C  Alloc=500;  LUMI-G  Alloc=500        

Ribosomes are large biomolecular complexes responsible for protein synthesis in all living cells. Simple organisms, such as bacteria, have smaller ribosomes than more complex organisms, such as humans. In eukaryotic cells, ribosomes are typically present in the cytosol in vast numbers, often reaching millions. Additionally, some organelles contain their own ribosomes, which differ from those in the cytosol. Over the past year, we have built an atomistic computational model of the mitochondrial ribosome – the first of its kind, to the best of our knowledge. In this project, we will perform molecular dynamics simulations in an explicit solvent and refine our model by validating it against available experimental data.

Řešitel: Frantisek Karlicky             

OPEN-34-16  

Many-body physics of van der Waals heterostructures from 2D materials II   

Barbora CPU  Alloc=36600;  Barbora FAT Alloc=340;  Barbora NG  Alloc=7500;  Karolina CPU Alloc=12600;  Karolina FAT Alloc=120;  Karolina GPU  Alloc=100;  LUMI-C  Alloc=3100          

Current needs of applied research on two-dimensional (2D) materials for flexible and ultrathin functional devices require computational predictions and computer-aided design. Stacking of two or more 2D materials leads to van der Waals (vdW) heterostructures, which are promising for engineering of its electronic and optical properties. Delicate physical effects in heterostructures will be studied by costly many-body methods because the usual density functional theory cannot describe the corresponding physics correctly. The project will contribute to the fundamental understanding of vdW heterostructures and boost their experimental research as well as technological applications.

Řešitel: Oldřich Plchot     

OPEN-34-17  

Cocktail SpeechLLM: A Speaker-aware Speech Language Model for Multi-talker Conversations with Instruction Tuning     

LUMI-G  Alloc=22300             

The recently emerged field of speech language models (SLMs) is a natural continuation on top of achievements with classical text-based large language models (LLMs), which have already penetrated into commercial applications offering solutions helping with text understanding, summarization, content creation, data analytics, etc. SLMs operate directly with speech modality and can perform tasks without explicit conversion to text, allowing them to leverage para-linguistic traits such as emotion, emphasis, sentiment, stress, etc. As a result, many natural language understanding tasks can be streamlined, and the model can generate answers (text or speech) with empathy, enabling more natural and versatile human-computer interactions. The ability to generate speech can also be leveraged and explored for tasks such as speech separation, enhancement, denoising, etc., which are typically handled by specialized models or standard signal processing techniques.    Within this research project, we focus on advancing the capabilities of SLMs by considering multi-talker scenarios. We will lay the groundwork that should lead to a model that is capable of better understanding natural human conversations as well as performing basic tasks related to these scenarios, such as multi-talker speech recognition, target speaker speech extraction, or speech mixture separation. 

Řešitel: Dušan Knop          

OPEN-34-18  

Algorithmic Theory Meets Practice: Experimental Studies in Graph Optimization and Computational Social Choice               

Barbora CPU  Alloc=1700;  Barbora NG  Alloc=900;  Karolina CPU  Alloc=1500            

Building on our recent theoretical work in algorithm design, we propose an experimental-computational project exploring several focused research directions. These include graph-theoretical problems, computational social choice, and potential connections to emerging AI-related algorithms. The project has three key objectives: (1) to advance theoretical and experimental understanding of these problems, (2) to engage students in hands-on research through computational exploration, and (3) to disseminate results via at least one peer-reviewed publication and arXiv preprints. By combining algorithmic theory with empirical validation, we aim to contribute new insights while fostering student research training.

Řešitelka: Elisabeth Christine Hehenberger    

OPEN-34-19  

Resolving the dynamics of endosymbiotic gene transfer in kleptoplastidic lineages 

Barbora NG  Alloc=7000;  Karolina CPU  Alloc=11700     

Plastid endosymbiosis has transformed our planet by introducing photosynthesis to eukaryotes and triggering the evolution of a massive diversity of plants and algae, yet we still do not know how this process works – mainly because most endosymbiotic events have happened such a long time ago. Kleptoplastidic lineages, lineages that steal and transiently retain plastids from their prey, are valuable models to study this process, as they potentially represent currently ongoing plastid endosymbioses. Different kleptoplastidic taxa retain plastids for different time spans and longer retention times may be correlated with increased genetic integration – in the form of gene transfers from the stolen plastid to the host nucleus. We propose here to determine the transcriptome-wide extent of genetic integration in kleptoplastidic dinoflagellates, a group of unicellular eukaryotes, with different plastid retention times. Correlating the number of detected gene transfers to the various retention times, will allow us to test the commonly accepted but untested assumption, that plastid endosymbiosis is coupled to a massive influx of genetic material and that the bulk of gene transfers takes place at the onset of endosymbiosis. In addition, the generation of a robust phylogenomic framework consisting of all available and newly generated dinoflagellate data will allow us to test various models of plastid evolution. Overall, our investigation will provide valuable new knowledge about the steps and their timing during this major evolutionary transition.

Řešitel: Martin Šůstek      

OPEN-34-2    

Energy-based Models Training as an Alternative Method for Pre-Training Foundation Speech Models          

LUMI-C  Alloc=1600;  LUMI-G   Alloc=22300          

Self-supervised learning (SSL) has become the dominant approach for pre-training machine learning models. Before SSL, Restricted Boltzmann Machines (RBMs), a type of energy-based model (EBM), were commonly used for deep neural network pre-training. EBMs are generative models that learn data distributions without requiring labeled samples. Over time, research shifted from RBMs to more expressive EBMs capable of modeling highly complex distributions without constraints. However, these unconstrained EBMs often suffered from stability issues and high computational costs. Recent advances have addressed training instabilities and introduced ways to reduce computational costs, opening new possibilities for their application.   Our research explores whether EBM training can serve as an alternative pre-training method for foundation models in speech processing, replacing SSL. In addition to standard EBM training, we will investigate Joint Energy-Based Models (JEM), which combine EBM training with discriminative training. Popular speech SSL methods such as HuBERT or WavLM internally generate pseudo-labels, which are then used for discriminative training. We will leverage these pseudo-labels in the discriminative component of JEM, effectively introducing a hybrid EBM-SSL pre-training strategy. 

Řešitelka: Jana Pavlů       

OPEN-34-20 

Defect-tuning of TiO₂ properties      

Barbora CPU  Alloc=10700;  Barbora NG  Alloc=9400;  Karolina CPU  Alloc=15900;  Karolina GPU  Alloc=300       

Our society is confronting a critical challenge: addressing the degradation of microplastics, an omnipresent environmental threat. This degradation process can be photocatalyzed by colored mesocrystals, which absorb light across a wide range of wavelengths, from visible to infrared. Consequently, there is a growing demand for new materials with tunable light absorption. Among the promising candidates are TiO2 mesocrystals, which are superstructures composed of crystallographically aligned nanoparticles exhibiting collective and emergent properties. However, a key challenge remains: How to prepare the catalyst with the appropriate absorption wavelength? This project aims to unique insight into the anatase TiO₂ absorption-wavelength tuning, specifically examining the effects of structure defects (free surfaces, oxygen vacancies and manganese impurities). The proposed study will i) identify the most favorable free surfaces and ii) explore how free-surface type and concentration and distribution of oxygen vacancies and Mn impurities influence the structure and electronic properties of anatase TiO₂, including its band gap (related to the absorption-wavelength). This way, we will better understand the relationship between microstructure and electronic/optical properties. These experimentally unreachable relations will be studied using advanced quantum-mechanical approaches, leading to the development of novel photodegradation catalysts for combating microplastic pollution.

Řešitelka: Jana Precechtelova  

OPEN-34-21 

Analyzing the Dynamic Behaviour of Intrinsically Disordered Protein Binding through Dimensionality Reduction of the Binding Interactions and Transient Secondary Structures     

Barbora GPU  Alloc=3000;  Barbora NG  Alloc=600;  Karolina CPU  Alloc=900;  Karolina GPU Alloc=100;  LUMI-G  Alloc=16300             

Many proteins within the cell operate not as rigid machines, but as dynamic inter-converting molecules known as intrinsically disordered proteins (IDPs). These highly-flexible proteins encompass a large amount of known proteins, are implicated in a variety of disorders including cancer or neurodegenerative diseases such as Alzheimer’s or Parkinsons. Due to their diverse structures, they are capable of interacting with a wide variety of partners forming complexes and allowing them to operate. This investigation implements molecular dynamics to simulate these proteins and their interactions, and proposes new techniques to analyze their complex interactions to theorize new drug targeting strategies. The implementation of nonlinear dimensionality reduction on the inter-residue distance matrix allows future research into the field, and provides scientists with a more robust understanding of the nature of these proteins.

Řešitel: Lukáš Burget        

OPEN-34-22 

Leveraging Pre-trained Foundation Models for Speaker Diarization       

LUMI-C  Alloc=900;  LUMI-G  Alloc=10000

Speaker diarization is a fundamental component of conversational artificial intelligence to determine speaker turns in multi-speaker recordings. Despite recent advances, accurately identifying speaker segments in real-world scenarios with varying numbers of speakers and complex acoustic variations remains challenging, primarily due to the scarcity of annotated data.   We propose leveraging pre-trained foundation models, originally designed to generalize across a wide range of tasks and domains, to build a robust diarization system capable of operating in complex real-world environments. By adapting models such as Microsoft’s WavLM, we aim to overcome the limitations of current diarization approaches. Building on our prior works in this area, we will explore different end-to-end speaker diarization architectures built on top of different foundation models— Whisper, WavLM, NEST. We will also explore strategies for neural network pruning to develop an efficient and lightweight solution. The proposed approach is expected to significantly enhance and simplify speaker diarization systems in complex real-world applications.

Řešitel: Martin Nováček  

OPEN-34-23 

Extending PM6-ML with reaction data        

Karolina GPU  Alloc=1600   

We have developed a hybrid computational chemistry method based on semiempirical quantum mechanical (SQM) calculations and cutting-edge machine learning (ML). The current version of this method, PM6-ML, has already been published. PM6-ML outperforms both SQM methods and ML potentials in terms of accuracy, especially in large systems such as biomolecules.   We are interested in extending the applicability of the method to cover broader chemical space. One area previously missing from the training data is a description of chemical reactions and transition states. This proposal aims to remedy this by introducing additional data on 12k organic chemistry reactions.

Řešitel: Indranil Mal           

OPEN-34-24  Single-Molecule Optoelectronics on Ferroelectric Surfaces (SMOFS) 

Barbora CPU   Alloc=36600

As traditional silicon electronics are approaching their physical limits, scientists are exploring different approaches for future computing technologies, which can be faster, smaller, and more efficient. Our project investigates the powerful combination of two promising technologies: ferroelectric materials (which can store information through electrical polarization) and molecular electronics (where individual molecules act as tiny electronic components, like switches). Using advanced high-performance computing, our project will investigate how ferroelectric materials used as a substrate can enhance the performance and stability of molecular switches, making them reliable even at room temperature. These hybrid devices could revolutionize computing by enabling ultra-compact memory storage, quantum electronics, and powerful brain-inspired computing systems, while consuming far less energy than today's technology.

Řešitel: Mikulas Matousek            

OPEN-34-25 

DMRG for polaritonic chemistry     

Barbora CPU  Alloc=2000;  Barbora GPU  Alloc=500;  Barbora NG  Alloc=9800;  Karolina CPU  Alloc=16500;  Karolina GPU   Alloc=500;  LUMI-C  Alloc=500;  LUMI-G  Alloc=1400   

This project explores the new field of polaritonic chemistry, which lies on the intersection of  chemistry and quantum electrodynamics. We are aiming to uncover new insights into light-matter interactions at the quantum level. By applying advanced quantum chemical methods, we investigate how light can influence chemical reactions, materials properties, and catalysis in novel ways.  This new understanding will help develop more efficient, sustainable processes for chemical synthesis, paving the way for new materials with unique properties and accelerating reaction rates. This work could significantly impact industries like energy, materials science, and green chemistry, offering a path toward cost-effective and environmentally friendly solutions for modern chemical production.

Řešitel: Sumit Ghosh        

OPEN-34-26 

Ultrafast optical and thermal generation of non-trivial antiferromagnetic texture        

Barbora NG  Alloc=6900;  Karolina CPU  Alloc=11600     

Conventional magnetic memories are based on ferromagnetic domains, which can be switched with an electric current. Antiferromagnets provide an intriguing alternative to ferromagnetic memories due to their fast dynamics and efficient current-induced switching. Recently, a new memory scheme has emerged, which utilises non-trivial magnetic textures such as magnetic domain walls or skyrmions to store memory. Their low power consumption and fast operation have opened new possibilities for applications in conventional and neuromorphic computing. Antiferromagnets are particularly interesting in this respect, mainly for two reasons: (i) their ability to create atomically thin domain walls and (ii) their immunity to the detrimental skyrmion Hall effect which poses a severe hurdle for skyrmion racetrack memory. This project is focused on exploring the mechanism behind the ultrafast generation of such non-trivial magnetic configurations with external perturbations in antiferromagnets, a process crucial for antiferromagnetic texture-based memory. By thoroughly studying the optically and thermally induced magnetisation dynamics using the state-of-the-art hybrid quantum-classical time evolution scheme, we will explore the stability of different non-trivial textures far from equilibrium and identify emergent magnetic interactions that can result in new types of metastable textures, which cannot exist at equilibrium.

Řešitel: Martin Crhán        

OPEN-34-27 

Attosecond streaking in iodoalkane molecules   

Barbora NG  Alloc=12000;  Karolina CPU   Alloc=20100 

During the last two decades the measurement of a new observable within photoelectron spectroscopy has been unlocked. This is the  attosecond photoionization time-delay – the time it takes to ionize an atom or a molecule by absorption of photon. Several methods have been developed to measure this delay. The method of relevance for the present work is the so-called “Attosecond streaking” in which the oscillating electric field of a moderately strong IR field is used as a clock alongside a short (attosecond) photoionizing pulse. It has been suggested in the past that time-delays measured using this method could be used to investigate molecular structures with potentially high temporal resolution. However there are several experimental and theoretical hurdles present in the way of achieving this goal. From the experimental point of view an intriguing class of molecules are the iodoalkanes. The presence of the so-called “Giant Dipole Resonance” in the iodine atom allows experimentalists to overcome a number of experimental issues. However, from the theoretical point of view these molecules present a substantial challenge due to pronounced relativistic effects, strong electronic correlation and a large number of electrons. It is these challenges, among others, which we wish to investigate in this work.

Řešitel: Antonín Vobecký

OPEN-34-28 

Automotive Image Inpainting and Data Augmentation for EXA4MIND Project 

Karolina GPU    Alloc=5800 

This project, part of the EU Horizon 2020 EXA4MIND initiative, explores generative artificial intelligence for automotive system validation. Utilizing EuroHPC Karolina GPU nodes, it aims to augment real-world driving datasets by modifying weather conditions, lighting, and object appearances. The objective is to stress-test AI-based detection models, such as Yolo v11 and Yolo World, under rare or hazardous conditions that cannot be captured in real life. Key methodologies include Segment Anything Model (SAM) for object segmentation, Stable Diffusion Inpainting for object editing, and Instruct Pix2Pix for scene modifications. The project will generate extensive synthetic datasets to enhance Advanced Driving Assistance Systems (ADAS) and improve AI robustness. Findings will contribute to safer AI-driven automotive technologies while aiding the research community in understanding model performance across diverse datasets.

Řešitel: Vít Musil  

OPEN-34-29 

Graph-Based Foundation Model for Digital Pathology     

LUMI-G  Alloc=7000

Digital pathology (DP) involves digitizing glass slide-based histological samples and analyzing them using computer technology, primarily through image processing techniques. The application of AI in DP is rapidly expanding, with recent advances in large foundation models that aim to capture generic knowledge from whole-slide images (WSI) for more robust diagnostic solutions. However, existing foundation models are trained directly on WSIs, requiring vast datasets to account for staining variations and other differences. Our goal is to develop a foundation model trained on nuclei graphs—abstract representations extracted from WSIs by segmenting cell nuclei while preserving their shapes and spatial relationships. We hypothesize that these graphs retain significant diagnostic information, as many relevant tissue structures can be inferred from nuclei morphology and positioning. By training on nuclei graphs, we can reduce dataset size while maintaining diagnostic utility, though substantial computational resources are still required for self-supervised learning. Ultimately, our model aims to serve as a universally applicable feature extractor across diverse datasets, overcoming challenges posed by staining techniques and sample preparation variations.

Řešitel: Michael Bakker   

OPEN-34-3    

De-Orphanizing Drug Targets: Molecular Dynamics and PROTAC Presents Novel Targets for Protein Elimination in Cells  

Barbora GPU   Alloc=1000;  Barbora NG    Alloc=1100;  Karolina CPU  Alloc=1800;  Karolina GPU   Alloc=1600;  LUMI-G  Alloc=22300

Many crucial proteins in our cells, called orphan nuclear receptors, are still mysterious, hindering the development of new medicines. Our research aims to shed light on these hidden targets, specifically NR2F6 (EAR-2), which has been linked to autoimmune disorders. We will use molecular dynamics simulations (MD) to understand how this protein, as well as a grand multitude of similar orphan nuclear receptors, works at a molecular level. Then we will explore a cutting edge drug discovery approach called PROTAC, which could selectively pave the way towards new and targeted drug therapies for aggressive cancer and other disorders. This research could unlock a new generation of treatments by targeting previously inaccessible proteins.

Řešitelka: Simona Kocour           

OPEN-34-30 

Privacy Options for 3D Maps             

Barbora NG   Alloc=100;  Karolina CPU   Alloc=200;  Karolina GPU  Alloc=2100            

3D mapping remains a challenge in Computer Vision, as it turns photos into digital models of the real world. These 3D maps are essential for Augmented and Virtual Reality, and autonomous navigation for cars and robots. The more detailed these maps are, the better they function, but the privacy concerns raise as the maps can reveal private information, especially when stored on cloud servers.  Traditionally, the focus has been on improving accuracy and detail, but as AI-driven mapping technologies evolve, privacy must become a priority. Right now, many apps (like Autodesk’s design tools, or Ikea’s room planner) require users to upload entire 3D scans, often without considering the risks of sharing personal details. Similar challenges exist in robotics, urban planning, and navigation, where sensitive data could be exposed if stored on cloud.  This project introduces novel privacy-preserving 3D mapping by giving users control over the level of detail stored in their maps.This novel 3D mapping framework enables users to choose what to keep and what to remove before storing or sharing their maps. By making 3D mapping more secure and customizable, this research paves the way for a future where AI-powered technologies respect user privacy without compromising functionality.

Řešitel: Pavel Krc 

OPEN-34-31 

MicroBUS simulations – initial phase          

Barbora NG  Alloc=16100;  Karolina CPU  Alloc=27100  

The PALM model 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. The goals of the TACR project MicroBUS (Microscale Based Urban Scenarios, TACR-SQ01010181) include development of new tools to assess the dispersion of traffic emissions in traffic-laden metropolitan areas at high temporal and spatial resolution. Moreover, additional model validation against measurements in wind tunnel is planned to test if important sizes of eddies, inside and above the urban area, are correctly captured. The results will be used by the Ministry of Environment and the municipal authorities to propose measures to achieve the air quality limits.  As the model is computationally demanding, all these simulations require a large amount of the parallel computation power which exceed the usual in-house resources. The supercomputer facilities in IT4I allow us to manage this challenge efficiently. During the proposed project we would like to test different configurations of the model and optimize our simulations which enable us to compute multiple runs of the final larger simulations. Results of this testing are necessary for the following multi-year IT4I call application.

Řešitel: Vladimir Petrik     

OPEN-34-32 

Generalization Studies of Robotic Manipulation Models

Barbora NG  Alloc=200;  Karolina CPU  Alloc=400;  Karolina GPU  Alloc=5800              

Vision-language-action models (VLAs) hold immense promise for achieving generalizable robotic manipulation, yet methods for rigorously evaluating and improving their generalization capabilities remain underdeveloped. This project addresses this critical gap by focusing on two key objectives: (a) developing a standardized methodology for assessing VLA generalization performance across diverse robotic manipulation tasks using a novel benchmark suite and defined metrics; and (b) enhancing VLA generalization through strategic data augmentation, leveraging the abundance of information available in uncurated sources like YouTube videos. By extracting and integrating relevant action sequences and visual information from these videos into existing training pipelines, we aim to improve VLA robustness and adaptability to novel environments, objects, and tasks. This project will contribute a framework for VLA evaluation, novel data augmentation techniques, and ultimately, advance the development of more generalizable robotic systems.

Řešitel: Alicia Moranchel Basurto           

OPEN-34-33 

3D-MHD models of magnetospheric accretion triggered by multipolar configurations: improve solution to spin-up stellar problem   

Barbora NG   Alloc=2800;  Karolina CPU  Alloc=4700;  Karolina GPU   Alloc=6900;  LUMI-C  Alloc=2900    

Dipolar magnetospheric accretion can provide a spin-down torque, which is exerted along the magnetospheric field lines connecting the star to the disk beyond the corotation radius. However, several observed astrophysical objects exhibit a constant period of rotation or spin-up/spin-down torque reversal. Therefore, an efficient mechanism of angular momentum removal is required.   Motivated by recent observational studies suggesting that magnetic fields at stellar surfaces may exhibit non-dipolar configurations. We will to study magnetospheric accretion triggered by multipolar magnetic configurations using three-dimensional (3D) non-ideal magnetohydrodynamical (MHD) simulations.  We will include most realistic boundary conditions on the star's surface called the perfect accretor, which will allow us to quantify the spin-up torque and its effect on the stellar rotation rate. Lastly, based on our numerical results, we will semi-analytically calculate an expression for the total torque on the star. Therefore, our results may provide a suitable solution to the stellar spin-up problem. 

Řešitel: Petr Valenta          

OPEN-34-34 

Bayesian Optimization of Laser-Driven Electron Accelerator

Barbora NG Alloc=13400;  Karolina CPU   Alloc=22600  

This project explores laser wakefield acceleration, a technique with the potential to shrink electron accelerators from kilometer-scale facilities to compact devices usable in hospitals, universities, and industry. We will employ Bayesian optimization combined with three-dimensional particle-in-cell simulations to identify the optimal operating regimes that, for a given laser pulse energy, maximize the cut-off energy of an electron beam accelerated via laser wakefield acceleration. Additionally, we will search for the potential scalability of the optimal parameters with respect to laser energy. Although the obtained results will be applicable to a wide range of laser parameters, the project will focus on optimizing electron beam characteristics for high-energy laser systems operating in less explored regimes. Particular attention will be given to systems such as the ELI L4 laser, which will soon deliver 1.5 kilojoules of energy in a single pulse. The findings will pave the way towards high-quality electron beams reaching 100 GeV and beyond for both fundamental science and practical applications.

Řešitel: Aleš Horák             

OPEN-34-35 

Slama - Slavonic Large Foundational Language Model for AI, high-quality training and fine-tuning

LUMI-G   Alloc=8500

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. This LUMI-G extension of the main Slama project on Karolina extends the training in several respects, a) supervised instruction fine-tuning, b) synthetic data generation for high-quality pretraining, and c) further continuous pretraining.

Řešitel: Jan Zemen             

OPEN-34-36 

Spin Dynamics of Non-collinear Antiferromagents and Ferrimagnets 

Barbora CPU   Alloc=32700;  Barbora GPU   Alloc=1200

Inter-site spin transfer torque in non-collinear antiferromagnetic and ferrimagnetic materials (ncSTT) was predicted in 2022. It has the potential to outperform spin orbit torque (SOT) widely used for ultrafast magnetization switching in research of non-volatile memory. In contrast to SOT, ncSTT does not rely on heavy elements (large spin-orbit coupling). The general aim of this project is to simulate ncSTT-driven magnetization switching in noncollinear Mn-based antiperovskite nitrides which were predicted to offer exceptionally large tunnel magnetoresistance (TMR) in a magnetic tunnel junction (MTJ) in 2023. Atomistic and continuum models of spin dynamics in the non-collinear magnetic material require input parameters such as exchange and anisotropy constants. The specific aim of the project is to perform Density Functional Theory (DFT) simulations and determine the required parameters as functions of chemical composition and lattice strain. The subsequent simulations of magnetization switching and domain wall motion driven by the novel spin transfer torque (by passing current) will be performed within the Czech Grant Agency project 25-18244S. Our results will contribute to the design of energy efficient memory and logic devices with read/write speed potentially surpassing those of current consumer random access memories.

Řešitel: Martin Dočekal   

OPEN-34-37 

Extraction of Software Mentions from Scientific Literature          

Karolina GPU  Alloc=2400;  LUMI-G  Alloc=4400 

With the increasing automatization of research, the importance of software and its proper citation in scientific papers grows. It is important not only to give credit to software authors but also to ensure the reproducibility of research results. However, the issue is that the software is often not properly cited.  For this reason, we want to automatically identify software mentions, using language models, in open-access scientific papers, validate identified software by authors, and register those software assets with persistent identifiers. We need reliable and cost-efficient systems to process tens of millions of scientific papers to achieve this goal.  The project is conducted in collaboration with partners (CORE, Software Heritage, HAL, Europe PMC) who operate existing open-access and software repositories. 

Řešitelka: Barbora Venosova    

OPEN-34-38 

Comparative study of the effect of mixed composition and vacancies on MXenes and MXene -based quantum dots              

Barbora CPU    Alloc=19200;  Barbora NG  Alloc=6700;  Karolina CPU  Alloc=11300;  Karolina FAT  Alloc=100;  Karolina GPU   Alloc=3500       

MXenes represent a wide and rapidly growing class of two-dimensional (2D) materials with a unique combination of properties such as high conductivity, chemical stability, magnetism, and versatile surface functionalization. These properties make MXenes very promising for a wide range of applications in nanoelectronics, spintronics, and photocatalysis. In recent years, their quantum dot derivatives (MXQDs), which combine the advantages of 2D materials with the additional effects of quantum confinement, have also received increasing attention. These nanostructures exhibit enhanced electronic, optical and magnetic properties that can be further tuned by targeted structural modifications.    Although interest in MXQDs has grown considerably in recent years, detailed understanding of their structural characteristics - such as surface functionalization and vacancy formation - remains limited. However, these structural features are expected to play a key role in defining the physicochemical properties of MXQDs. A deeper understanding and achieving precise control over the surface terminations and configurations of the vacancies could allow tuning their electronic, optical, and magnetic behavior, leading to the attachment of MXQDS with the desired properties.    In this context, systematic theoretical research will be carried out to model different types of MXQDs and to investigate how structural factors - such as vacancies and mixed surface functional groups - affect their electronic, optical and magnetic properties. The resulting findings will support the design of suitable types of MXQDs for potential applications in photoelectronics, spintronics and photocatalysis. 

Řešitel: Filip Jozefov          

OPEN-34-39 

DreaMS Plus: Unified mass spectra and molecular structures representation learning          

LUMI-C  Alloc=900;  LUMI-G  Alloc=16000

Metabolomics, the study of small molecules produced by cells, offers crucial insights into  living systems and reveals unique biological signatures. Although mass spectrometry is a  powerful tool for discovering these molecules, only about 10% of the generated spectra can  be confidently identified due to the inherent complexity of the data.  Recent advances in artificial intelligence provide new opportunities to overcome this  challenge. In this project, we will develop advanced deep learning techniques to extract  meaningful insights from complex mass spectrometry data. We plan to fine-tune our  DreaMS foundation model to enhance molecular structure recognition and develop a  unified model that aligns spectra with molecular structures in a shared latent space,  analogous to how CLIP connects images with text. Our approach leverages contrastive  learning to learn joint representations and integrates a latent diffusion component that “fills  in” missing spectral detail. This diffusion step refines the structure embeddings by  capturing subtle fragmentation patterns, effectively bridging the gap between coarse  molecular structure and the intricate details inherent in spectral data. Additionally, we will  create a diffusion-based de novo generation model that directly proposes molecular  structures from spectra. The expected outcomes are open-source tools and models that  accelerate molecular discovery in metabolomics, drug development, and environmental  analysis.

Řešitel: Martin Zelený       

OPEN-34-4    

First-Principles Study of Elasticity in Long-Period Modulated Ni₂MnGa Martensite    

Barbora CPU Alloc=8900;  Barbora NG   Alloc=25700;  Karolina CPU  Alloc=43300

Magnetic shape memory (MSM) alloys offer significant potential for applications in actuators, sensors, energy harvesters, and magnetic refrigeration systems due to the remarkable properties of their multiferroic martensite structure. Among these, the five-layered modulated martensite phase (10M) of the Ni-Mn-Ga alloy is the most extensively studied. The macroscopic deformation of these materials under an external magnetic field, known as Magnetic Field-Induced Strain (MFIS), results from the motion of highly mobile twin boundaries in magnetically ordered martensite, driven by a pronounced shear elastic instability in the lattice. Experimental studies suggest a complex interplay between lattice modulation, twin boundary mobility, and elastic properties. Recent neutron diffraction measurements reveal temperature-dependent evolution of the modulation function, leading to structural changes in the crystal lattice. However, ab initio calculations of elastic properties have not confirmed the observed shear instability in 10M martensite, likely due to their simplified lattice descriptions. This project proposes a novel ab initio model that accounts for the evolving modulation function and the high mobility of twin boundaries, providing a more accurate representation of the elastic properties of 10M martensite.

Řešitel: Athanasios Koliogiorgos             

OPEN-34-40 

Van der Waals Heterostructures for Electronic Applications (WHEA)   

Barbora NG   Alloc=28400;  Karolina CPU  Alloc=47700 

Van der Waals two-dimensional semiconducting heterostructures are under intense study due to their significant electronic properties that makes them suitable candidates for applications like spintronics, magnetic memories and field-effect transistors. One of the most promising vdW materials is CrSBr which is a magnetic semiconductor. The CrSBr/MoS₂ heterostructure is under study for its combination of the exciton-magnon coupling and magneto-optical properties of CrSBr with the tunable carrier polarity of MoS₂. The strong van der Waals interfacial coupling in this system provides a pathway for engineering p- and n-type 2D semiconductor channels, offering new possibilities for spintronic and optoelectronic applications. We will simulate different configurations of this heterostucture to elucidate the properties that emerge when combining these two materials. This computational project is directly connected to ongoing experiments.

Řešitel: Petr Hellinger       

OPEN-34-41 

Pressure-strain effect and non-gyrotropy in plasma turbulence               

Barbora NG   Alloc=12000;  Karolina CPU  Alloc=20100 

Turbulence is a ubiquitous phenomenon in space and laboratory plasmas. A nonlinear coupling between scales leads to a spread of energy over a wide range of scales, spectral energy densities exhibit a power-law behavior, and the energy flows/cascades from large to small scales. At small scales the energy is dissipated and plasma particles are heated. In usual fluids the heating is connected with the irreversible particle-particle collisions; however, in many space and astrophysical plasmas the collisions are too rare and the heating proceeds via reversible channels. The turbulent motion in weakly collisional plasma leads to a formation of complex (generally non-gyrotropic) particle distribution functions. To have a better insight to this complicated problem numerical simulation are necessary. We propose to study plasma turbulence in the context of one well-known example of weakly collisional turbulent system, in the solar wind (a flow of magnetized plasma from the Sun). We will use a three dimensional hybrid code (where electrons are assumed as a fluid but ions are treated as particles) to study properties of turbulent cascade towards ion scales, their heating/energization and its reversibility in connection with formation of non-gyrotropic distribution functions.

Řešitel: Adam Kurčina     

OPEN-34-42 

Towards chemically accurate neural network potential: effect of dataset on accuracy           

Barbora NG  Alloc=8700;  Karolina CPU    Alloc=14600;  Karolina GPU   Alloc=2000  

Computational chemistry plays a key role in elucidating and predicting complex chemical behavior, which is essential for the development of novel materials, bioactive compounds, and even in environmental research. The equivariant graph neural network (GNN) architectures (e.g. AIMNet2, MACE or NequIP) accurately model rotational/translational behavior of molecular properties. However, these GNNs are typically trained on limited datasets consisting of structures belonging to chemical space (commonly definited as stable/drug-like molecules). Although off-equilibrium structures are commonly included by employing molecular dynamics or normal mode sampling, these structures are not sufficient to cover chemical bond forming/breaking processes. We propose a paradigm shift in the construction of chemical datasets - we aim to cover various possible molecular arrangements. This comprehensive training dataset aims to explore the ability of GNNs to interpolate across various structural arrangements. We aim to enhance GNN accuracy even for unconventional chemical processes.

Řešitel: Jakub Hromádka

OPEN-34-43 

Modelling of the transport of the edge localized modes in the scrape-off layer of the COMPASS-U tokamak          

Barbora NG    Alloc=27900;  Karolina CPU  Alloc=47000

Magnetically Confined Fusion (MCF) plasma devices represent a promising way to a practically unlimited and environmentally friendly energy sources.  At present, number of MCF test devices, tokamaks, are under operation and some are under construction. One of such devices is the COMPASS-U, which will start operation in 2026 at the Institute of Plasma Physics of the Czech Academy of Sciences. COMPASS-U is a middle size, high magnetic field tokamak, where among others one of the main problems of MCF will be studied – plasma and power exhaust. The subject of the proposed project is predictive and fully kinetic modelling of the Edge Localized Modes (ELM) and their propagation in the scrape-off layer of the COMPASS-U tokamak, when highest power and plasma to the plasma-facing components are expected. Modelling of plasma transport in the COMPASS-U SOL will directly contribute to the machine design and optimization of the discharge scenarios and will provide a rare study of the ELM propagation in high density plasma. For the simulations we intend to use the 1D3V electrostatic particle-in-cell Monte Carlo code BIT1.

Řešitel: Martin Kolísko     

OPEN-34-44 

Evolution of parasitism in Parabasalids    

Barbora NG   Alloc=2700;  Karolina CPU   Alloc=4500     

Parabasalids are a diverse group of protists that contain not only the well-known pathogen Trichomonas vaginalis, but also commensals (i.e.: Pentatrichomonas) and secondarily free-living species like Lacusteria. Parabasalids studied on the genomic level are, with few exceptions, exclusively pathogenic. Parabasalids are an ideal group to study transitions between different lifestyles and host environments, as they possess evolutionarily intermingled parasites, commensals and secondarily free-living species. The goal of this project is to combine transcriptomic and genomic data to perform comparative analyses that will help us pinpoint specific genomic adaptations to different lifestyles and host environments.

Řešitel: Daniel Blasco Santana 

OPEN-34-45 

Relativistic Effects on Magnetic Response Properties     

Barbora NG  Alloc=1200;  Karolina CPU   Alloc=2000       

Nuclear magnetic resonance (NMR) is an extremely powerful spectroscopic technique for the structural elucidation of molecules based on the resonant absorption of radiofrequency waves by certain atomic nuclei when exposed to a strong external magnetic field. The presence of heavy atoms in the molecules, for example sixth-row elements, leads to remarkable modulations in the NMR parameters of other bound atoms due to a relativistic effect called spin-orbit coupling (SOC). The aim of this proposal is to unveil the role of SOC in the NMR parameters of transition metal (TM) hydride complexes. NMR shifts and magnetically induced current densities will be calculated at the fully relativistic level with the selective inclusion of SOC with the ReSpect program. The difference between the calculations including and excluding SOC is the contribution of the latter to the magnetic property. This project will have an impact on the development of NMR spectroscopy of heavy TM hydride complexes, which play central roles in relevant catalytic processes such as hydrogenation of alkenes in the petrochemical industry.

Řešitel: Marketa Paloncyova       

OPEN-34-46 

Phase preference of novel ionizable lipids              

Barbora NG  Alloc=13700;  Karolina CPU  Alloc=23000;  Karolina GPU   Alloc=2600;  LUMI-C   Alloc=2600;  LUMI-G  Alloc=22000

Nucleic acid medicine opened new possibilities in cancer and rare diseases treatment and vaccination strategies. The delivery of nucleic acid is limited by their fragility and they are therefore delivered encapsulated in lipid nanoparticles with pH responsive behavior. Ionizable lipids in lipid nanoparticles change their protonation state during endosome maturation and are therefore responsible for nucleic acid release. However, the design of lipid nanoparticle composition is still based on a trial-and-error approach, as their structure and mechanism of action are not fully resolved. In this project we will aim on developing a building and simulation protocol for lipid nanoparticles in order to investigate the face preference of lipid mixtures. The ability to routinely run MD simulations of lipid nanoparticles will open the field of targeted in-silico design of lipid composition tailored for specific use.

Řešitel: Pavel Jungwirth  

OPEN-34-47 

Interaction of fusogenic cell-penetrating peptides with curved lipid membranes       

LUMI-G    Alloc=22300           

Cell membranes are poorly permeable to most water-soluble molecules, including drugs. However, there are molecules that can facilitate transporting drugs into the cells. Cell-penetrating peptides (CPPs) are among the most promising of them. CPPs cross the membranes spontaneously while carrying the drug molecules as a cargo, although the exact mechanism of penetration remains unknown. They can also change the membrane shape and topology leading to formation of multilamellar structures and membrane pores. Although all these processes involve formation of the curved membranes, the role of curvature in them was never properly and systematically studied.     Our research aims to uncover the role of membrane curvature in the interactions of CPPs with membranes. Using molecular dynamics simulations and with advanced enhanced sampling we will elucidate the role of curvature in CPP permeation, CPP-induced membrane interaction and pore formation.     Computational findings will be correlated with the cryo-EM data to provide a deeper understanding of how CPPs interact with biological membranes. These insights could lead to the development of more effective drug delivery systems, improving treatments for various diseases. Furthermore, our study will contribute to fundamental biophysical knowledge, advancing the broader field of membrane research.

Řešitel: Zdeněk Futera     

OPEN-34-48 

Dynamics of Charge Transfer in (Bio)molecular Junctions           

Barbora CPU  Alloc=36600;  Barbora NG  Alloc=7500;  Karolina CPU   Alloc=12600  

The rapid development of biomolecular electronics in recent years, when single-molecular devices utilizing whole proteins have been manufactured and probed, raised fundamental questions related to charge transfer mechanisms in such components. The electronic-state misalignment on the biomolecule/metal contacts promotes the coherent tunneling mechanism, which is in drastic contrast to conventional electron hopping typical for native redox proteins. To elucidate the details of these processes, we plan to track the charge propagation using state-of-the-art quantum-dynamics computer simulation techniques. Performance of such challenging calculations, usually limited to small molecular systems only, can not only bring the desired insight into the charge transport processes but can also move forward limits of applied bio-simulation techniques.

Řešitel: Radim Špetlík      

OPEN-34-49 

Real-time Semantically Meaningful Video-to-Video Style Transfer        

LUMI-G   Alloc=2800

Recent progress in deep learning, especially in generative models and image-to-image translation, has led to video style transfer methods that can change a video’s look while preserving the video’s core content. This technology has applications in movies, virtual reality, gaming, video calls, and art, but it remains challenging due to high computational demands.    This project will develop a real-time video style transfer system using neural networks and multi-GPU training, ensuring that key elements like objects, people, and backgrounds are maintained while applying a requested style.

Řešitel: Saltuk Mustafa Eyrilmez              

OPEN-34-5    

Quantum Mechanics-Enhanced Structural Data Generation: Building Foundations for Advanced ML Models in Drug Discovery    

Barbora NG   Alloc=5600;  Karolina CPU   Alloc=9400;  LUMI-G  Alloc=3600   

Modern drug discovery remains a lengthy and costly process despite computational advances. Machine learning (ML) models show promise but face serious limitations due to inconsistent and insufficient experimental data for broad generalization. Meanwhile, classical screening methods struggle with accuracy due to forcefield limitations and inadequate conformational sampling.  Quantum mechanical (QM) calculations provide incomparable accuracy in molecular interactions that classical approaches cannot achieve. In this project, we will generate a comprehensive dataset of QM-scored protein-ligand complex conformations for fragment-sized compounds (up to 10,000) against diverse protein binding sites using a hierarchical pipeline that includes classical docking, molecular mechanics (MM) geometry minimizations and rigorous QM scoring.  By leveraging IT4Innovations supercomputing, we will create a quantum-enriched database of molecular building blocks spanning multiple protein families. This computationally uniform dataset will enable the development of protein binding-site-specific ML models capable of faster processing multi-million compound libraries with QM-level accuracy orders of magnitude.  Unlike approaches relying on inconsistent experimental data, our computationally standardized dataset addresses the quality issues hampering current ML methods. This work establishes a robust bridge between quantum mechanics and machine learning, potentially transforming structure-based drug discovery with accurate predictions across therapeutic areas. 

Řešitel: Miroslav Kolos    

OPEN-34-50 

Exciton Dynamics in 2D Transition Metal Systems             

Barbora NG  Alloc=10300;  Karolina CPU    Alloc=17300;  Karolina GPU    Alloc=5000              

Excitons—electron-hole pairs bound by Coulomb attraction—play a central role in defining the optical characteristics of semiconducting materials. These quasiparticles are typically generated by photoexcitation and contribute to phenomena such as photoluminescence, with their binding energies directly influencing exciton lifetimes. Accurately capturing excitonic effects has long posed a theoretical challenge, but recent advances in many-body quantum methods and high-performance computing have enabled precise simulations using the Bethe–Salpeter equation (BSE). This formalism allows for the direct computation of frequency-resolved optical responses, providing insights into a material’s absorption features, photoluminescence spectra, and the nature of both bright and dark excitonic states. Such data are critical for the design of optoelectronic materials with tailored light-emission properties, enhanced frequency conversion, and efficient nonlinear optical behavior. Importantly, realistic modeling must also account for temperature-dependent effects, which arise from interactions between excitons (or charge carriers) and lattice vibrations. In the proposed work, we employ state-of-the-art computational approaches to explore these effects and deliver high-fidelity predictions of the optical responses in a new set of two-dimensional materials based on transition metals.

Řešitel: Jiri Klimes

OPEN-34-51 

Accuracy and precision for extended systems XIV             

Barbora CPU   Alloc=24200;  Barbora NG  Alloc=7700;  Karolina CPU    Alloc=13000;  LUMI-C  Alloc=2900

The development of new or improved pharmaceuticals makes a substantial use of methods of materials science. One of the examples is the formulation of the solid form of pharmaceuticals. A pharmaceutically active compound, typically a molecule, has often several functional groups that ensure that it ends up on the right place in body and does the right thing. This can make the molecules complicated and when they form crystals, these can have different structures. This is called polymorphism. Knowledge of these polymorphs is crucial for the pharmaceutical industry as some of them might have undesirable properties. Simple methods, often based on quantum mechanics, are used to identify the possible polymorphs so that the landscape of structures is known. However, our data show that the methods that are currently used rely on cancellations of errors. This means that the results can look fine but there can be situations in which unexpectedly large errors occur. We are trying to find out how large problem the error cancellation is by developing a large dataset of reliable energy differences between polymorphs. Moreover, we are getting so-called many-body energy contributions to the energies of the structures which allows one to understand the accuracy of the different theoretical methods in detail. This knowledge can be then exploited to develop methods with improved accuracy, reducing thus the uncertainty of theoretical predictions.

Řešitel: Uladzislau Yorsh

OPEN-34-52 

Virtualisation and Multimodal Exploration of Heritage on Nazi Persecution (MEMORISE)     

Karolina GPU   Alloc=200;  LUMI-G  Alloc=1800   

Heritage related to Nazi Persecution (HNP) appears in various forms. The memories of eyewitnesses are directly inherent in diaries, letters and testimonies (written or recorded), and registers (e.g., death records, deportation statistics) and historical photographs deliver significant contextual information. Vast amounts of such resources exist across Europe, but they are barely accessible to the general public. MEMORISE will conserve and present these memories by anchoring them in the places where the Nazi crimes happened.  In order to make these accounts more readily available for both the public and researchers, we in MEMORISE intend to adapt and train specialized models for segmentation, tagging and critical themes highlighting in these historical documents for easier reading and to guide users through transcripts of video evidences and a thesaurus. The intent of the work is to provide an accessible platform for anyone who wishes to explore and comprehend the experiences of Nazi Persecution witnesses.

Řešitel: Tadeáš Kalvoda  

OPEN-34-53 

Determination of pentapeptides with inherent secondary structure propensity by machine learning and density functional theory     

Barbora NG   Alloc=200;  Karolina CPU    Alloc=300;  LUMI-G  Alloc=6570        

At what level does the secondary structure of proteins start to appear? We plan to use machine learning and DFT calculation to identify possible pentapeptides with inherent secondary structure preference as isolated tripeptides in solution. Such pentapeptides, if they exist, may help to understand the protein folding process, by which proteins adopt their stable conformation, which is necessary for carrying out biochemical reactions. We will use machine learning tools to identify such pentapeptides, which then will be synthesized, and their secondary structure will be assessed experimentally using circular dichroism spectroscopy. Additionally, we will carry out a density functional theory calculation as a reference evaluation of peptide energies.

Řešitel: Maximilian Lamanec      

OPEN-34-54 

Systematic Computational Investigations of Small Molecule Adsorption on Hexagonal Molecular Surfaces          

Barbora CPU   Alloc=8500;  Barbora FAT    Alloc=340;  Barbora NG  Alloc=9300;  Karolina CPU   Alloc=15600;  Karolina FAT   Alloc=100;  Karolina GPU Alloc=6000;  LUMI-C   Alloc=2400;  LUMI-G   Alloc=4700    

Monitoring small-molecule gases is critical for environmental safety and public health. This project aims to develop new sensor technologies based on molecular adsorption on hexagonal surfaces such as graphene and hexagonal boron nitride (BN). We address two central questions. First, we assess the accuracy of dispersion-corrected DFT methods in computing adsorption energies by benchmarking against high-level PNO-LCCSD(T)/CBS interaction energies. We will investigate a set of small molecules—CH₄, CO₂, H₂O, HCN, and SO₂—adsorbed on polycyclic aromatic systems (coronene, circumcoronene) and their BN analogs. Second, we investigate how adsorption energy converges with increasing surface flake size. Preliminary studies on CO₂ and HCN indicate rapid convergence for CO₂ but unpredictable behavior for HCN, likely due to competing dispersion and electrostatic interactions. SAPT calculations will further decompose these interaction contributions, informing the minimum model size required for reliable predictions. This integrated computational approach will guide the design of effective gas sensors for hazardous molecules.

Řešitel: Raul Chametla   

OPEN-34-55 

Magnetorotational instability in 3D circumplanetary dusty-gaseous disks: unraveling angular momentum transport problem in moon formation            

Karolina CPU   Alloc=500;  Karolina GPU   Alloc=7100     

Circumplanetary disk (CPD) formation around Jupiter-like planets offers us a mini-laboratory for the analysis of satellite formation around a massive central object (i.e., a miniature solar system), as well as the possibility of understanding the mechanism of angular momentum transfer. The latter is important because governs the accretion rate of gas and dust in the final stage of massive planet formation, which in turn shapes the orbital dynamics of low-mass bodies forming into planetary systems. Several theoretical and numerical studies have addressed this problem. However, in most cases ah-doc description is introduced for the redistribution of angular momentum in the CPD, or local numerical models are performed using boundary conditions that constrain the gas flow through the CPD. Here, we will study by means of high-resolution global three-dimensional non-ideal magnetohydrodynamic (MHD) multifluid simulations, the redistribution of angular momentum self-consistently within the CPD. Since for the first time we will resolve the magnetorotational instability (MRI) within the CPD that is triggered by a magnetic field. This will also allow us to study the drift, fragmentation, and coagulation of dust in the CPD that leads to moon formation. Furthermore, the study the formation of CPD- and planet-driven outflows, which can be used to identify planets in observed protoplanetary disks. 

Řešitel: Ján Michael Kormaník   

OPEN-34-56 

Prediction of binding affinity of de novo designed metallopeptides       

Barbora NG   Alloc=4500;  Karolina CPU   Alloc=7500;  Karolina GPU   Alloc=600

Metal ions are necessary for proper function of many proteins. The metals hold the proteins’ shape and sometimes allow for catalysis of many chemical reactions, among which we can find difficult formation of bonds or redox reactions. This reactivity makes metal ions very interesting for applications in both chemical and biochemical industry, because they allow complicated reactions to run at room temperature and under standard pressure. Designing new short peptides, which would selectively bind metal ions and catalyze wanted chemical reactions, and then predicting their binding affinities and selectivity for certain metal ions could thus open up many possibilities for their future applications.

Řešitel: Michael Komm   

OPEN-34-57 

Comparative simulations of inverse sheath           

Barbora NG  Alloc=14900;  Karolina CPU   Alloc=25200 

Inverse sheath is a fundamentally novel type of plasma boundary, which has been recently proposed as a theoretical solution for strongly emissive plasma-facing components. However, It’s applicability to real plasma conditions e.g. in nuclear fusion devices such as tokamaks and consequences for operation of future thermonuclear reactors is a subject of intensive debate within the community. This project aims to use two particle-in-cell codes – BIT1 and SPICE2 to investigate the feasibility and stability of this regime.

Řešitelka: Dana Nachtigallova 

OPEN-34-58 

Modelling of photochemistry in UV-Vis and X-ray domains in complex molecular system     

Barbora CPU  Alloc=33300;  Barbora FAT   Alloc=340;  Barbora NG  Alloc=27100;  Karolina CPU  Alloc=45700;  Karolina FAT   Alloc=120;  Karolina GPU   Alloc=7100;  LUMI-C   Alloc=2800;  LUMI-G  Alloc=10000 

This project aims to develop an innovative and scalable theoretical framework for modeling X-ray spectroscopies and to improve interpretability and chemical insight obtained within the rapidly developing state-of-the-art X-ray experiments. The tools and framework will be utilized to address fundamental questions on the reaction mechanisms and photochemistry in complex molecular systems, such as photoswitches or catalytically active metal complexes, operating under realistic reaction conditions. We will emphasize the application potential of resonant Auger electron spectroscopy which is site and element-specific and provides unprecedented information density on structural and electronic structure properties of molecules.

Řešitel: Mario Vazdar        

OPEN-34-59 

Controlled Transport of Cell-penetrating Peptides across Biological Membranes      

Barbora NG  Alloc=8300;  Karolina CPU  Alloc=14000;  Karolina GPU   Alloc=2200;  LUMI-G   Alloc=2500

The controlled transport of cargo into cells is vital for life. Alongside the active ATP-driven process of endocytosis, which plays a pivotal role in transporting cargo across membranes, there is another promising method known as direct passive energy-independent translocation, particularly relevant for controlled drug delivery. Short, positively charged cell-penetrating peptides (CPPs) are commonly utilized vectors for this purpose. This proposal aims to investigate the mechanism of passive translocation of CPPs at the molecular level using advanced molecular simulations. Our focus will be on understanding crucial steps in this process, including adsorption, aggregation, and translocation across advanced model membrane systems. The insights gained from this investigation will lay the groundwork for designing innovative drug delivery systems by identifying crucial molecular interactions during the transport of matter across biological membranes.

Řešitel: Luigi Cigarini        

OPEN-34-6    

Systematic approaches to chemical doping in scandium nitride for enhanced thermoelectric applications  

Barbora CPU   Alloc=7300;  Barbora FAT    Alloc=150;  Barbora NG Alloc=6000;  Karolina CPU  Alloc=10100;  Karolina FAT   Alloc=100      

Thermoelectric effects convert temperature differences into electric voltage and vice versa. Finding materials with excellent thermoelectric properties is important for technological applications, as it helps reduce costs and minimize negative environmental impacts of modern technologies. Optimizing the performance of thermoelectric conversion in a material is a difficult task, as it depends on conflicting properties. It is necessary to increase electrical conductivity while simultaneously reducing thermal conductivity. To achieve this goal, a deep understanding of the material structure and transport mechanisms is essential. This project aims to study the thermoelectric properties of scandium nitride, a promising compound for future thermoelectric applications, using computational models in the field of materials science.

Řešitel: Debashree Manna           

OPEN-34-60 

Computational Modeling of Solvent Effects on Non-Covalent Interactions     

Barbora CPU   Alloc=2700;  Barbora GPU   Alloc=100;  Barbora NG  Alloc=4400;  Karolina CPU   Alloc=7500;  Karolina GPU    Alloc=5800;  LUMI-C   Alloc=2300;  LUMI-G   Alloc=4500             

Non-covalent interactions play a crucial role in catalysis, biochemistry, and supramolecular chemistry. Their strong directional nature makes solvent effects particularly significant in non-covalent systems.1,2 However, accurately describing these systems theoretically is quite challenging. Conventional implicit solvation models often struggle to capture these directional effects, especially in hydrogen-bonding interactions. Therefore, explicit solvent models, particularly for the first solvation shell, are essential in these cases. While low-level DFT, QM-MM and semi-empirical approaches can be effective with a sufficiently large number of explicit solvent molecules, achieving high accuracy with more advanced ab initio methods is often difficult.3,4 To enhance accuracy in large systems, we propose utilizing recent advancements in Machine Learning Perturbation Theory. By integrating quantum chemistry and machine learning, this work aims to establish the foundation for next-generation computational solvation models for non-covalent systems. This approach is expected to deepen our understanding of non-covalent interaction dominated enzyme catalysis, molecular recognition, and functional materials.

Řešitel: Dominik Farka    

OPEN-34-61 

Effects of Functionalization and Adsorption onto Functional Surfaces of Conductive (Bio-Inspired) Polymers         

Barbora CPU    Alloc=1800  

Two parts comprise this project: a follow-up study on our recent joint experimental and theoretical work on polythiophene and polypyrazine based systems and secondary will also want to extend our investigation in terms of conductive polymers to the indole-ring system.  The first part, our follow-up study will allow for a more systematic, in-depth study of PEDOT (polyethylendioxythiophene), PEDTT (polyethylendithiathiophene) and PTP (polythioenopyrazine) polymers. We intend to use quantum chemical methods to investigate their frontier orbitals, conformational stability, and their charge-transport. We want to focus on the effects of different functionalizations and/or doping. Further, we want to elucidate the effects of interaction of these polymers with different surfaces, which we have shown experimentally to have tremendous impact. In our recent experimental work, we have investigated oxidative chemical vapour deposition (oCVD) of PTP and inquire to gain insight into the effects of various surfaces (see below). These are expected to act as deposition templates to our polymer.  The second part of the project will investigate the indole-system for its reactivity and potential as a conductive polymer. Substitution effects will be studied by applying the study to various serotonine-like molecules (serotonine, triptans, etc.). The same properties will be studied as in the abovementioned polymers. This is in part to support a submitted GAČR-project.  This project is meant to be a complement for/basis of the experimental research on these materials. 

Řešitel: Libor Dostál          

OPEN-34-62 

Pincer complexes with P-containing donor sites: DFT and ab initio computations     

Barbora NG   Alloc=19100;  Karolina CPU  Alloc=32200 

Modern chemistry necessitates the development of new transformations with excellent performance   enabling high yields under mild conditions. The catalytic reactions used currently require transition   metals, which are usually expensive, rare or toxic and thus their replacement by other elements with   benign properties would be beneficial. This task is, however, challenging, as the main group elements   usually lack various oxidation states that make the catalytic cycles possible and delicate interactions   between the metal center and the ligands are needed to overcome this problem. The goal of the   present project is to investigate the stability and reactivity of pincer complexes with phosphoruscontaining sites using state-of-the-art computational simulations. The newly synthetized complexes   that are the subject of this study exhibit a broad versatility regarding the type of central element   (several examples from the heavier main group elements), the possibility of functionalization at the P   center in the ligand and the other substituents. The conformational stability of these complexes will be   studied by DFT and ab initio calculations, and the interactions between the main group metals and the   pincer ligand will be characterized.

Řešitel: Vojtech Horny      

OPEN-34-63 

Particle beams acceleration and extreme plasma states with multi-PW laser pulses              

Barbora NG  Alloc=8500;  Karolina CPU   Alloc=14300;  Karolina GPU  Alloc=4400;  LUMI-G   Alloc=5000  

Ultra-intense laser pulses can generate beams of charged particles, including protons, electrons, and even exotic particles such as positrons and fast neutrons [2]. These high-energy beams have remarkable applications, from cancer treatment [1] to nuclear waste transmutation and the production of medical isotopes. The recent commissioning of multi-petawatt laser systems like ELI Beamlines (Czechia), ELI NP (Romania), and Apollon (France) is expected to push the boundaries of laser-driven particle acceleration, producing more energetic and numerous particles than ever before. This research aims to unlock the full potential of these new accelerators through advanced computer simulations. Using state-of-the-art high-performance computing, we model the complex physical interactions involved, ensuring high-precision predictions. These insights will not only advance fundamental science but also help optimize future laser-based technologies, making them more accessible for medicine, industry, and clean energy research.

Řešitel: Valeria Butera      

OPEN-34-64 

Designing salen-like TM complexes as efficient catalysts for CCU         

Barbora CPU   Alloc=11500;  Karolina GPU  Alloc=3100 

Due to the gromwing amount of CO₂ gas in the atmosphere that leads to profound global climate changes, growing attention has been addressed to CO₂ capture and utilization (CCU) processes. Two main approaches to convert CO₂ into fuels and fine chemicals have been considered: reductive conversions in which the carbon atom is reduced to its lower oxidation states, and non-reductive process, where the +4 oxidation state of carbon in CO₂ is maintained. However, due to the high stability of this inert molecule, both reductive and non-reductive conversions are only possible upon utilization of a catalyst.  Among homogeneous catalytic systems, salen-like transition metal complexes have gained growing attention due to their high stability and the ability to fine-tune their electronic properties by modifying ligands and coordinating sites to get enhanced catalytic performance. Within this project, we aim at performing calculations based on density functional theory (DFT) to investigate their catalytic activity towards two main CO₂ reaction conversion: 1) the reductive hydrogenation of CO₂ to methanol and 2) the non-reductive CO₂ insertion into epoxides to produce cyclic carbonate.

Řešitel: Jiří Pittner

OPEN-34-65 

Theoretical design and modeling of efficient BODIPY-based photosensitizers              

Barbora NG   Alloc=9000;  Karolina CPU  Alloc=15100    

Molecular dynamics in excited states including the non-adiabatic and spin-orbit effects is an important theoretical tool for the simulation of photochemical processes which play an important role in nature and technology, like e.g. photosynthesis, phototherapy, photovoltaics, etc. Its computational cost when ab-initio or DFT methods are employed is very limiting in both size of the molecules treated and the length and number of computed trajectories. Machine learning (ML) has recently become very popular thanks to its widespread applications in many areas of science, industry, and commerce. Recently the machine learning methods have been successfully employed to speed up the molecular dynamics in the ground state, and some progress was done also on the ML of excited states and non-adiabatic effects. The aim of this project is to apply MD  and ML methods to study halogenated derivatives of BODIPY , which are potentially applicable as photosensitizers in the photodynamic therapy. The application is related to solving the grant no. 23-06364S awarded by the Czech Science Foundation.

Řešitel: Myank Singhal    

OPEN-34-66  Exploring Chaos within Hierarchal four body set up          

Barbora NG  Alloc=3000;  Karolina CPU   Alloc=5100       

The heart of our galaxy, centred around the supermassive black hole Sagittarius A* (Sag A*), is a fascinating region teeming with dynamic activity, particularly the presence of two young star disks nearly mutually perpendicular to each other. These disks raise intriguing questions about star formation in the strong gravitational fields near Sag A*. Recent research highlights that stars within 0.03 parsecs of Sag A* display varied orbits, showcasing complex interactions that could reveal hidden mechanisms governing their behaviour. Our study builds upon previous work, exploring the dynamics of these star systems in non-secular regimes—situations where traditional models may fall short. Utilising advanced numerical simulations, we investigate how changing eccentricities and separations between orbits can unveil unexpected areas of stability. Our findings indicate that some regions exhibit greater stability than anticipated, which could reshape our understanding of how disk-like structures form in this dynamic environment. Through a thorough analysis of our simulations, we aim to provide insights that could explain the dual disk phenomenon observed in the Galactic Centre. This research not only enhances our comprehension of star dynamics near supermassive black holes but also opens new avenues for exploring the evolutionary history of our galaxy.

Řešitel: Jakub Podgorny  

OPEN-34-67 

Filling the gaps in our knowledge of black holes  

Barbora CPU  Alloc=27800;  Karolina CPU  Alloc=38200;  LUMI-C  Alloc=2200             

Black holes (BH) are undoubtedly part of contemporary physics, resulting e.g. in the Nobel Prizes of 2017 and 2020. An accretion disc of orbiting matter around BHs is often formed, which is illuminated by a hot X-ray-emitting corona. The location of the corona, the structure of the disc and its detailed high-energy micro-physics are for decades one of the main unsolved mysteries in astrophysics. With this project we aim to calculate the interaction of light and matter in the accretion disc's atmosphere, including known physical effects that are largely neglected in the solutions currently existing in literature. Observational X-ray polarimetry is causing a revolution in our BH understanding thanks to the first dedicated 2-8 keV polarimeter on board of the IXPE mission (NASA / ASI, observing since 2022). Using state-of-the-art Monte Carlo code STOKES, we aim to provide unique numerical estimates of polarization of X-rays reflected from and transmitted through the disc's stratified atmosphere, constructing a table model to be compared with the latest observations of BHs by IXPE.

Řešitel: Ondřej Kobza       

OPEN-34-68 

Advancing Secure and Reliable Coding LLM Solutions   

Barbora CPU  Alloc=300;  Barbora GPU  Alloc=1000;  Barbora NG   Alloc=100;  Karolina CPU  Alloc=200;  Karolina GPU   Alloc=1400;  LUMI-G  Alloc=500   

We introduce AlquistCoder, a cutting-edge architecture for trusted AI and safe code generation with dynamic adaptability. AlquistCoder features a core code-based large language model (LLM) supported by tools such as a Planning Tool for workflow management, an Intention Recognition Tool to detect unsafe code prompts, and code analysis tools to verify code safety and functionality.    Through the use of Retrieval-Augmented Generation (RAG), AlquistCoder provides real-time updates and comprehensive analysis. The LLM is fine-tuned to prevent unsafe code creation while enhancing coding capability, using prompt optimization to maximize performance.    Our LLM was trained on a combination of synthetic, scraped and publicly available data, making our model’s outstanding performance on standard coding benchmarks while being safe.    For evaluation, we propose developing a novel security dataset, conducting expert assessments, and creating an internal red-teaming tool. This comprehensive approach ensures that AlquistCoder remains secure, robust, and effective in producing safe code.

Řešitel: Martin Jirka            

OPEN-34-69 

Pair creation in multiple laser beam collision       

Barbora NG   Alloc=2900;  Karolina CPU   Alloc=4900     

With the advent of 10 PW-class laser facilities, a new regime of laser-matter interaction is opening since quantum electrodynamics effects, such as pair production and cascade development, start to be important and affect the interaction. To study this new regime of interaction, numerical simulations are required. With their help, it is possible to predict and study quantum effects which may occur in future experiments at modern laser facilities. The aim of this work is to study a particular interaction setup when multiple laser beams interact with a seeding charged particle. Due to the quantum nature of the interaction, new particles and antiparticles can be created during the interaction. The efficiency of conversion laser light into matter and antimatter depends on the interaction setup. Optimizing the spatio-temporal characteristics of colliding pulses can increase the number of created particles by orders of magnitude. 

Řešitel: Matěj Svidenský 

OPEN-34-7    

Modelling of losses of the neutral beam injection in tokamaks

Barbora CPU  Alloc=16300;  Barbora NG   Alloc=300;  Karolina CPU  Alloc=500           

The study of fast ions in plasma in fusion devices through simulations is of interest as they are the main source of heating for the plasma in tokamaks. In some cases, their losses can cause significant sputtering of the plasma facing material into the plasma and contaminate it. In COMPASS-U, they will originate mostly from neutral beam injection (NBI) heating. The study and modelling of NBI-born particle behaviour is of great relevance: it might influence future design of the system and its integration in the overall reactor design. This project will build on the existing simulation tools in order to achieve an effective integrated modelling workflow that can span the whole existing database of about 400 plasma scenarios of the future COMPASS Upgrade tokamak. We will evaluate the NBI losses in multiple plasma configurations and also scan the NBI geometry to achieve the highest possible effectiveness of the heating without damaging the walls of the tokamak causing plasma contamination.

Řešitel: Šimon Vrba           

OPEN-34-70 

Modelling of DD plasma ELM transport in the JET SOL    

Barbora NG   Alloc=30200;  Karolina CPU   Alloc=50900

One of the hot topics of the Magnetic Confinement Fusion research is the estimation and reduction of  power loads to the plasma facing components and the related erosion of these elements. Among the  most critical ones are power loads during the so-called Edge-Localized Modes in tokamaks, when a  significant amount of hot plasma is quickly released from the confined region into the scrape-off layer  where it directly propagates towards the divertor plates. The aim of this project is to perform a  modelling of the ELM transport in the JET tokamak and study the evolution of the ELM related divertor  power loads and the resulting W sputtering rates. Understanding the connection between simulations  and experiments allows us to develop more effective predictive models for the next generation of large  tokamaks.

Řešitel: Fatemeh Kayanikhoo     

OPEN-34-71 

Multi-dimensional simulations of super-Eddington accreting magnetized neutron stars as ultraluminous X-ray sources             

Barbora NG  Alloc=13400;  Karolina CPU Alloc=22600   

We aim to study ultraluminous X-ray sources (ULXs) powered by super-Eddington accreting neutron stars (NS). We will investigate the impact of the NS compactness on its apparent luminosity, outflow power, and spectrum. The project seeks to explore the mass gap between massive NS and small lack holes (BH) through the X-ray emission spectrum. Additionally, we will examine the equation of state (EOS) of the NS through the spectrum extracted from the simulations and compare it with the observed spectrum of ULXs. Furthermore, we aim to study NS-ULXs with tilted magnetic dipole through 3D simulations. These simulations are crucial tools for understanding the physics behind the pulsations and the extraordinary luminosity of NS-powered ULXs.

Řešitel: Martin Špetlík      

OPEN-34-72 

Generative models for bone dataset augmentation          

LUMI-G  Alloc=8000

Osteoporotic fractures are life-altering events that can potentially lead to death. Currently, one in three women over the age of 50 experience such fractures. Developing new treatments requires a thorough understanding of bone adaptation and dysfunction. Our goal is to elucidate how population variation in bone morphology, a poorly understood area, relates to bone remodeling, a crucial process for optimal bone function. We hypothesize that external mechanical forces influence bone morphology and mineral density variations, which can be detected through clinical CT scans. We aim to identify patients' physiological loading states from CT scans. This innovative approach could revolutionize osteoporosis treatment by providing crucial information for early and precise individualized treatment.  However, the insufficient number of CT scans limits the validation of the hypothesis and model. To address this, we plan to generate additional samples using generative machine learning models. Specifically, we aim to use a denoising diffusion model based on 3D convolutional neural networks and a vector-quantized generative adversarial network (VQGAN) for dimensionality reduction.

Řešitel: Jan Kuneš

OPEN-34-73 

X-ray magnetic circular dichroism in non-collinear magnets     

Barbora NG   Alloc=12300;  Karolina CPU Alloc=20800  

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, it was realized that some compensated magnetic materials can host sizeable the anomalous Hall effect, linear magneto-optical effects or spin polarized bands, previously associated with ferromagnetism. These include materials with magnetic moments orientated parallel to each other (altermagnets) and also some non-collinear magnets. Our work demonstrated that x-ray magnetic circular dichroism (XMCD) provides a powerful tool for investigation of such materials.  We have developed a method for calculation of XMCD and demonstrated its predictive power. A predictive theory of XMCD in altermagnets as well as non-collinear magnets allows us to address some key questions in one of the hot fields of current physics.

Řešitel: Debora LANČOVÁ            

OPEN-34-74 

Simulation of Accretion Discs in Black Hole X-ray Binaries         

Barbora NG  Alloc=9600;  Karolina CPU  Alloc=16100     

X-ray binaries are among the brightest and most energetic sources in the Universe, powered by the accretion of material on a central black hole. The matter from the companion standard star flows on the black hole and forms an accretion disc, which is heated to extreme temperatures, and up to 42 % of the rest-mass energy may be extracted in the form of electromagnetic radiation, making it the most efficient energy extraction process after annihilation. In the past 50 years, many of these objects have been discovered, and their behaviour and properties were modelled using analytical, semi-analytical and numerical approaches. However, even now, the accretion disc theory is incomplete, and many observed properties of X-ray binaries have not been explained. We aim to simulate an accretion disc during the peak of a black hole binary outburst. The gas forming the disc is then coupled with radiation, with a strong magnetic field stabilising the flow, and its distinguished vertical structure strongly alters the observable signal. Our previous results showed that in this regime, standard tools to extract information about the black hole from observed data may provide wrong values because the analytical accretion disc models strongly deviate from the simulation results. To accurately model the accretion disc, a method that solves the co-evolution of the matter, magnetic field and radiation in general relativity on a well-resolved grid and for an extended duration must be used.

Řešitel: Ota Bludsky          

OPEN-34-75 

Tailoring Mass Transport Properties of Porous Materials for Enhanced Adsorption and Catalysis    

Barbora NG Alloc=9000;  Karolina CPU Alloc=15100;  Karolina FAT  Alloc=120             

This project aims to enhance our comprehension of how the dynamic behaviors of host-guest systems in porous solids affect mass transport properties, crucial for adsorption and catalysis. It focuses on three key issues: (i) the role of composition and guest molecules in the mass transport properties of flexible materials like zeolites and ZIFs, (ii) gating effects in low-silica materials, and (iii) the influence of internal defects (silanols) on gas separation. Through the development of accurate models, the project intends to control these properties for various applications, tackling challenges in quantifying framework flexibility, the dynamics of extraframework cations, and the use of internal defects to enhance selectivity in gas separation. The proposed project incorporates both theoretical models and experimental validation, aiming to bridge the gap between complex molecular behaviors and practical applications in material science.

Řešitel: Miroslav Rubes   

OPEN-34-76 

Computational Modeling of Cu/ZnO/Al₂O₃ Catalysts: Unraveling Active Sites and Structure-Activity Relationships for Enhanced Hydrogenation and Dehydrogenation Performance

Barbora NG  Alloc=8600;  Karolina CPU  Alloc=14500;  Karolina FAT  Alloc=120           

This project focuses on the computational modeling of Cu/ZnO/Al₂O₃ catalysts for hydrogenation and dehydrogenation reactions, aiming to uncover the relationship between synthesis conditions, catalyst structure, and catalytic performance. The key challenge lies in understanding the embedding of active copper sites within the ZnO/Al₂O₃ matrix. Theoretical modeling will be used to simulate the atomic and electronic structure of the copper active sites and their ZnO/Al₂O₃ interface, providing insights into how variations in synthesis conditions influence the geometry, electronic environment, and catalytic behavior of these sites. Computational studies will focus on predicting how the local structure and bonding at the copper–support interface affect the activity, selectivity, and stability of the catalyst during hydrogenation of CO, CO₂, and furfural, as well as dehydrogenation of ethanol and cyclohexanol. The combination of in-situ experimental data with detailed theoretical modeling will enable a deeper understanding of the structure-activity relationship, guiding the design of more efficient Cu/ZnO/Al₂O₃ catalysts.

Řešitel: Martin Dračínský

OPEN-34-77 

NMR Chemical Shift Predictions for Solvated Biomolecules      

Barbora NG   Alloc=12000;  Karolina CPU   Alloc=20100

Accurate prediction of nuclear magnetic resonance (NMR) chemical shifts is crucial for understanding molecular structure and interactions in solution. However, traditional implicit solvent models fail to capture key solute-solvent interactions, such as hydrogen bonding, leading to significant inaccuracies in predicted NMR parameters. To address this limitation, this project proposes a novel computational framework that integrates molecular dynamics (MD) simulations with machine learning (ML)-based NMR predictions. By explicitly modeling solvation effects, this approach aims to bridge the gap between experimental observations and computational predictions. The objectives of this research are twofold: (1) to develop an accurate computational model for NMR predictions in diverse solvents, and (2) to extend the methodology to biomolecular systems. By enhancing the predictive power of NMR chemical shifts in solution, this study will provide critical insights into non-covalent interactions in complex molecular systems.

Řešitel: Jan Blahut              

OPEN-34-78 

Optimal-control NMR sequences for 14N nuclei using extended spin systems            

Barbora NG  Alloc=24800;  Karolina CPU   Alloc=41800 

Nuclear Magnetic Resonance (NMR) spectroscopy is a powerful tool for studying the structure and dynamics of molecules, but many of its applications focus on nuclei like 1H or 13C. In contrast, 14N—which has nearly 100% natural abundance—is rarely used in routine NMR experiments. The challenge lies in designing pulse sequences that can efficiently manipulate with its 14N nuclear spin which has strong quadrupolar character (i.e. not symmetric charge distribution in the nucleus), making sensitive 14N detection difficult.  Our project aims to overcome this barrier by applying optimal-control theory to develop entirely new NMR pulse sequences for quadrupolar nuclei in solid-state. Optimal-control methods allow us to explore a vast space of possible pulses and select those that maximize signal and minimize experimental imperfections. Early simulations indicate that our approach could yield roughly twice the sensitivity of current state-of-the-art techniques, paving the way for significantly more efficient 14N detection.  If successful, this breakthrough will have a broad impact on whole solid-state NMR of organic compounds remains underutilized. Enhanced sensitivity in 14N experiments could offer valuable insights into pharmaceutical formulations, shed light on the behaviour of light-sensitive materials, and even expand our understanding of insoluble proteins.

Řešitel: Antonio Cammarata      

OPEN-34-79 

SUbstrate-dependence of nanofriction of 2D BOROphene and hBN MONOlayers (SUBOROMONO)           

Barbora CPU   Alloc=22900;  Barbora NG  Alloc=7000;  Karolina CPU  Alloc=11800  

The tribological performance of 2D materials makes them good candidates towards reduction of friction at the macroscale. Superlubricity has already been observed for various 2D materials such as graphene, transition metal dichalcogenides or MXenes, but other materials are investigated as potential candidates for low-lubricity applications. Among the latter, χ6-borophene is predicted to have ultra-low friction while hexagonal boron nitride (hBN) is already used to reduce or tune friction properties of materials, which is confirmed by our experimental study performed in ultra-high vacuum at room temperature. With the present computational project, we plan to perform quantum mechanical calculations to study the frictional behaviour of our synthesised χ6-borophene and hBN monolayers deposited on different substrates, with the aim to elucidate the atomic origin of the different friction response. The results will suggest how to fine tune the tribological behaviour of the two materials at varying supporting substrates for nanomechanical application.

Řešitel: Carlos Manuel Pereira Bornes 

OPEN-34-8    

Understanding CO₂ adsorption mechanisms in amine-grafted sorbents via machine learning potentials 

LUMI-G   Alloc=2606

The concentration of greenhouse gases (GHG) in the atmosphere has reached unprecedented levels, with CO₂ being the most abundant. Carbon Capture and Storage (CCS) technologies offer promising solutions, with solid sorbents emerging as efficient alternatives to traditional liquid amine scrubbing. However, the design of effective CO₂ adsorbents requires meticulous attention to various factors, including cost, recyclability, kinetics, and selectivity. Porous silicas, particularly amine-grafted mesoporous silicas, are very promising due to their high selectivity, high surface area, chemical stability, and tunable pore sizes. These material qualities gathered the attention of the community, leading to intensive study on the adsorption mechanisms on silica surfaces.   Computational studies are vital for an atomistic understanding of the mechanisms that govern CO₂ adsorption at the amine-grafted silicas surface. However, current computational methodologies are either limited in accuracy to describe reactive events (classical MD) or by their high computational cost (ab initio MD). To address this, we propose the development of a reactive MLP with broad configurational and chemical space that can accelerate and explore reactive pathways and discover novel species and the reaction mechanisms. 

Řešitel: Semen Yesylevskyy         

OPEN-34-80 

Elucidating molecular mechanism of topological flipping of the LacY membrane protein    

LUMI-G   Alloc=22300            

It is widely accepted that membrane proteins always maintain a single orientation and topology in the cell membrane, but some unique proteins, such as bacterial sugar transporter protein LacY, reversibly change their orientation in response to changes in the lipid environment or charge of their domains. The molecular mechanism of such flipping process is poorly understood and the exact structure of the flipped protein is not known beyond the rough topological arrangement.   In this project we will study the process of topological flipping of LacY in the lipid membranes by enhanced sampling atomistic molecular dynamics simulations for the first time, which will complement simultaneous experimental studies by our partners. The influence of lipid composition and protein charge on the thermodynamics of the flipping process will be studied. The structure of the flipped protein will be elucidated.  Understating the flipping mechanism of membrane proteins may have broad practical applications in drug discovery, synthetic biology and applied molecular biotechnology. The detailed study of flipping LacY protein will open a perspective to these future directions of research and technological development.

Řešitel: Shubham Agarwal           

OPEN-34-81 

High-Throughput Discovery of Multicomponent Transition Metal Dichalcogenide Alloys for Advanced Friction Applications (HIT-TMDs)    

Barbora CPU  Alloc=36600;  Barbora NG  Alloc=7700;  Karolina CPU   Alloc=13000  

Transition Metal Dichalcogenides (TMDs) have emerged as promising materials for tribological applications due to their layered structures, exceptional lubricity, and tunable electronic and mechanical properties. High-entropy-stabilized TMDs present a novel strategy for designing friction- and wear-resistant materials with enhanced performance. However, their structural stability, electronic behavior, and tribological response remain largely unexplored. This project aims to investigate the structural stability, electronic properties, and tribological behavior of multicomponent TMD alloys composed of Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, and W in both 1T and 1H phases. Utilizing high-throughput density functional theory (DFT) calculations and machine learning (ML)-assisted ab initio molecular dynamics (AIMD) simulations, we will establish structure-property relationships that govern their stability and frictional response. These insights gained will advance the understanding of entropy-driven stabilization in TMDs and contribute to the rational design of next-generation solid lubricants and protective coatings. Access to supercomputing resources is crucial to enable the large-scale computations required for this study.

Řešitel: Ravikant Kumar  

OPEN-34-82 

Study of Nanofriction Control in Transition Metal Dichalcogenides using Density Functional Theory and Machine Learning Force Fields (NCTM-DMF)        

Barbora CPU   Alloc=33100;  Barbora NG  Alloc=8400;  Karolina CPU   Alloc=14100 

This project seeks to understand the effect of external electrostatic fields on the tribological properties of 2D layered materials of scientific and technological interest. Since ab initio molecular dynamics (AIMD) simulations are computationally very expensive, we utilize machine learning force fields (MLFF) to reduce these high costs. This approach improves the process of force field development without significantly diminishing the accuracy of the quantum-mechanical calculations. Our study focuses on six transition-metal dicalgogenides (TMD) monolayers (MoS₂, MoSe₂, MoTe₂, WS₂, WSe₂ and WTe₂) interacting with a metallic substrate silver (Ag) and a silicon (Si) atomic force microscopy (AFM) tip. With the help of the classical molecular dynamic simulations, we investigate the friction in the presence of electrostatic fields at the nanoscale. Our final objective is to understand the origin of friction in these TMD materials, thus supporting the next generation of materials whose tribological response may be appropriately tuned through the use of external fields.

Řešitel: Shivprasad Shivaram Shastri   

OPEN-34-83 

Effect of surface and interface vacancies on the photocatalytic activity of van der Waals heterostructures (VAC-PHCAT)       

Barbora CPU  Alloc=28400;  Barbora NG  Alloc=6500;  Karolina CPU  Alloc=11000

Photocatalytic solar-to-hydrogen conversion by water splitting is regarded as an important sustainable energy source. Two dimensional (2D) van der Waals heterostructures can be suitable candidates to this aim, thanks to favourable carrier migration paths and large surface area facilitating catalytic activity. However, vacancies are known to occur in the as prepared materials or due to aging which can act as electron or hole traps. Such defect states can enhance or degrade the efficiency.  In this project, we consider PtSSe and WXY (X,Y=S, Te, Se, X≠Y)  heterostructures and study the effect of anionic vacancy states on band alignment, charge transfer and optical absorption in order to assess the photocatalytic activity. The results will provide useful physical insights to suggest suitable materials for photocatalytic or other photovoltaic applications.

Řešitel: Tahir Wahab         

OPEN-34-84 

High Entropy Alloys Transition Metal Dichalcogenides for photocatalytic applications (HEA-TMDCs)

Barbora CPU Alloc=20800;  Barbora NG  Alloc=12400;  Karolina CPU  Alloc=20900 

High-entropy alloys (HEAs) combine multiple principal elements at a near-equal fraction to form vast compositional spaces to achieve outstanding functionalities absent in alloys with one or two principal elements [1]. Using first-principles calculations, we will design novel HEAs, systematically evaluate their structural, thermodynamic, and dynamic stability, as well as determine the optimal temperature range for their synthesis. Furthermore, we will analyze their electronic properties to assess their optoelectronic behavior. Finally, we will investigate their photocatalytic performance by examining band edge alignment and overall water-splitting efficiency. This study will provide fundamental insights into the potential of HEA-TMDCs as a next-generation photocatalyst for clean energy applications.

Řešitel: Masoud Shahrokhikhorneh       

OPEN-34-85 

Tailoring the Optoelectronic Characteristics of Transition Metal Oxides and Carbon Nitride Heterojunctions for Enhanced Photocatalytic Water Splitting   

Barbora CPU  Alloc=27500;  Barbora FAT   Alloc=340;  Barbora NG Alloc=9000;  Karolina CPU  Alloc=15100 

As the demand for clean energy increases worldwide, so does the interest in photocatalytic water splitting for the production of sustainable hydrogen. Transition metal oxides (TMOs) and carbon nitride (CN) based systems, among which graphitic carbon nitride (g-C₃N₄) holds a special place owing to its electronic and optical characteristics are among the most promising materials for this application. However, present efficiency is still poor owing to the high probability of electron-hole recombination and low light absorption efficiency. As part of this project, we aim to contribute to the understanding of how TMOs can enhance the performance of CN based heterojunctions for water splitting in photocatalysis. For this purpose, we intend to employ computational simulations to design and optimize TMOs/CN heterojunctions with enhanced charge separation and light absorption. The findings of this research are expected to facilitate the development of low cost and highly active photocatalysts for the production of hydrogen through sustainable processes.

Řešitel: Jan Boháček         

OPEN-34-86 

Machine learning-based surrogate modeling for conjugate heat transfer enhancement with complex surface geometry: CFD simulation

Barbora NG   Alloc=600;  Karolina CPU Alloc=1000          

The project focuses on utilizing high-performance computational resources for CFD simulations to optimize conjugate heat transfer in complex surface geometries. Advanced additive manufacturing enables the fabrication of intricate heat transfer structures with high surface-to-volume ratios, but their analysis using CFD is computationally demanding. To overcome this, extensive computational resources will be leveraged to generate high-fidelity simulation data, which will then be used to train machine learning-based surrogate models. These models aim to reduce computational costs while improving efficiency and accuracy in heat transfer optimization. However, research on surrogate modeling for such complex geometries remains limited, particularly regarding the availability of simulation and experimental datasets. This project seeks to bridge that gap by combining large-scale CFD simulations with machine learning, advancing data-driven approaches for heat transfer enhancement.

Řešitel: Jan Vábek

OPEN-34-87 

Coherent EUV light source for compact lens-less microscopy and precision metrology        

Barbora CPU   Alloc=8400;  Barbora NG  Alloc=7500;  Karolina CPU Alloc=12600      

The goal of our project is to develop a novel technique for compact extreme ultraviolet (EUV) radiation sources [1] based on high-harmonic generation, which occurs during the interaction of an intense driving laser pulse with gas. Our approach uses a precisely tailored driving pulse together with a semi-periodic gas profile. As a result, it allows us to optimize the resulting wavelength of the source.    During the project, we plan to shorten the wavelength to achieve 20-nm resolution in a lensless microscopy setup by using a 13.5 nm source. Such a resolution is crucial for EUV lithography machines in the semiconductor industry. As another application, we plan to reach the “water window,” a spectral range that provides excellent contrast for studying living tissues in their natural environment.    The main outcome of the project shall be direct experimental implementation supported by our recent pioneering study [2]. Numerical simulations are indispensable for designing the experimental setup because they allow us to find optimal generation conditions and to buy the necessary laboratory equipment. Moreover, the range of conditions accessible in simulations exceeds experimental capacities, potentially leading to spin-off theoretical studies. 

Řešitel: Miroslav Cerny    

OPEN-34-88 

Exploring the mechanisms of plasticity in nitinol

Barbora NG  Alloc=6600;  Karolina CPU Alloc=11100;  Karolina FAT  Alloc=120;  Karolina GPU  Alloc=700;  LUMI-C   Alloc=3100    

Shape memory alloys are unique materials capable of undergoing large reversible strains and exhibiting the shape memory effect, which is driven by external temperature changes. These remarkable properties are based on a martensitic transformation between austenite (high-temperature phase) and martensite (low-temperature phase). Due to these unique characteristics, these alloys have been used in numerous practical applications since their discovery in 1959. The aim of this project is to gain a fundamental understanding of the advanced deformation processes in the martensitic structure of NiTi at the atomistic level—using molecular dynamics with new, advanced potentials based on machine learning models derived from quantum mechanical calculations. In particular, as there is only one possible shear system in NiTi martensite, it raises the question of why martensite can undergo plastic deformation so easily. Therefore, within this project, we will enhance our current machine-learned interatomic potential to develop advanced and highly precise atomistic models of martensite.

Řešitel: Pavel Ondračka 

OPEN-34-9    

Machine learning interatomic potential for TiAlON            

Barbora NG  Alloc=1200;  Karolina CPU  Alloc=2000;  Karolina GPU  Alloc=100           

The thermal stability of protective coatings is a key factor in determining their performance in cutting, drilling, and forming applications. TiAlN has become the benchmark coating for these applications due to its exceptional hardness and wear resistance. However, studies have shown that incorporating oxygen enhances the oxidation resistance of TiAlN thin films, further improving their thermal stability. This project aims to develop a machine learning interatomic potential to simulate the high-temperature elastic properties and diffusion processes in cubic titanium aluminum oxynitride (TiAlON). By doing so, we seek to understand the role of oxygen and point defects in shaping the high-temperature functionality and preformance of this material.