9th Users' Conference of IT4Innovations

30 Oct 2025, 10:00 → 31 Oct 2025, 14:00 Europe/Prague

atrium (IT4Innovations)

Branislav Jansik (IT4Innovations), Karina Pešatová (IT4I), Tomáš Kozubek (IT4Innovations), Vít Vondrak (IT4Innovations, VSB-Technical University of Ostrava)

9^th Users' Conference of IT4Innovations will take place on 30 and 31 October 2025. All of our users, as well as research and project partners from various organisations, research institutions, and industry, are welcome to attend the Conference.

The submission deadline for contributions is September 14, 2025.

Registration is open until October 22, 2025.

Attendees will discover more about our future upgrade plans, listen to talks given by our prominent users, and can engage in discussions during the Users' Council meeting and a poster session.

Contribution types: Users' talks/Posters

Users' talks

Selected talks by our prominent users will be presented at scheduled times during the whole conference. Each talk is expected to take a maximum of 20 minutes (with discussion included):

• IT4I Project(s) (5 minutes): Dedicate approximately 5 minutes to provide insights into your computational approach and methods. Discuss software utilisation, job sizes, parallelisation techniques, and the extent of computational resources employed. Sharing your computational experience will provide valuable context for the audience.
• Research (10 minutes): Allocate around 10 minutes to delve into your research endeavours. Highlight key findings, address challenges encountered, and emphasise the significance of your work. This segment should offer a concise yet comprehensive overview of your research area.
• Discussion (5 minutes): Please reserve 5 minutes to engage with questions and discussions from the audience.

Posters

Please note that the required poster size is A1 portrait orientation. 

If you would like IT4Innovations to print and display your poster at the conference, please ensure you send your printing file to training@it4i.cz by October 17.

 

This conference is supported by the Ministry of Education, Youth and Sports of the Czech Republic through the e-INFRA CZ (ID:90254).

 

All presentations and educational materials of this course are provided under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license.

9_Users_Conference_IT4I_Programme.pdf

9_Users_Conf_list_posters.pdf

Presentations

• 00_branislav_jansik_user_conference_2025.pdf
• Access_to_the_EuroHPCJU_supercomputers_UserDay_IT4I_2025.pdf
• branislav_jansik_user_conference_2025_council.pdf
• Support from the High-Level Support Team_HLST.pdf

Presentations - Users' Talks

• 01_Valenta_Petr_Machine-learning optimization of laser-driven electron accelerators.pdf
• 02_Horny_Vojtech_Numerical Simulations at Karolina for advancing laser plasma research at echtreme light infrastructure.pdf
• 03_Svobodova_Petra_HPC-accelerated Deep Learning Pipeline for Microhologram dot pattern verification in secure ID document authentication.pdf
• 04_Vašina_Vojtěch_Electronic band structure calculations of crystals on noisy quantum simulators.pdf
• 05_Sklenak_Stepan_Supercomputers and synchrotrons_A new era of scientific Discovery.pdf
• 06_Sethia_Khyati_Microstructural Analysis of Aluminide Coatings using Deep Learning-Driven Image Segmentation and Synthetic.pdf
• 07_Cigarini_Luigi_Thermoelectric effects of structural defects in scandium nitride nanostructures.pdf
• 08_Cechova_Petra_Model complexity in all-atom plasma membrane simulations.pdf
• 09_Kolar_Michal_Translocation of short peptides trough carbon nanotubes.txt
• 10_Nedelnikova_Andrea_Atomic-Scale_Insights_into_Graphene-Based_Nanomaterials_Interacting_with_Brain_Cell_Membranes.pdf
• 11_Paloncyova_Marketa_Lipid nanoparticles in coarse-grained resolution_From internal organization to interaction with cells.pdf
• 12_Strakos_Petr_Bridging cosmological simulations and high-quality visualizations using HPC resources.pdf
• 13_Koliogiorgos_Athanasios_CrSBr-MoS2 heterostructures and the limits of VASP Parallelization.pdf
• 14_Nicolini_Paolo_Nanowear in Molybdenum Disulfide studied by molecural dynamics simulations.pdf
• 15_Legut_Dominik_Antiferromagnetic Exchange Coupling across Transition metal films with magnetic elements.pdf
• 16_Friak_Martin_Fine-tuning stability of La-Ni-Sn-H materials for hydrogen storage.pdf

Thursday, 30 October

10:00 → 11:00 Registration and networking

Convener: Branislav Jansik (IT4Innovations)

11:00 → 11:30 Welcome and IT4Innovations news

11:30 → 12:00 Keynote I

11:30 Machine-learning optimization of laser-driven electron accelerators

Laser-driven electron accelerators promise significant reductions in size and cost compared to their radiofrequency counterparts, unlocking opportunities for widespread use in hospitals, university labs, and beyond. Getting the most electron energy from such accelerators comes down to choosing just the right laser and plasma settings, which is a multi-parametric and highly nonlinear optimization problem. To address this challenge, we coupled computationally intensive particle-in-cell simulations with a Bayesian optimization algorithm. From the resulting data, we derived generalized scaling laws for electron energy, charge, and acceleration length as functions of laser energy. These scaling laws, together with the full set of input parameters, provide a practical framework for designing laser-driven electron acceleration experiments across a wide range of laser systems.

In this talk, we present the results obtained within the IT4Innovations Open Access Grant Competition projects OPEN-30-14 and OPEN-34-34, which have recently been published in [1, 2]. We discuss the use of the OPTIMAS library for Bayesian optimization (https://github.com/optimas-org/optimas) and the OSIRIS code for particle-in-cell simulations (https://github.com/osiris-code/osiris), highlighting our computational approaches, parallelization strategies, job scales, and the extent of resources employed at IT4Innovations.

This work was carried out in collaboration with Lawrence Livermore National Laboratory (USA), Princeton University (USA), University of Rochester (USA), and Kansai Institute for Photon Science (Japan). It was supported by the Defense Advanced Research Projects Agency (DARPA) under the Muons for Science and Security Program, by the project “e-INFRA CZ” (ID: 90254) from the Ministry of Education, Youth and Sports of the Czech Republic, as well as by the NSF–GACR collaborative grant (No. 2206059) and the Czech Science Foundation (Grant No. 22-42963L). A portion of the work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory (LLNL) under Contract No. DE-AC52-07NA27344, with additional support from the LLNL Institutional Computing Grand Challenge program.

[1] P. Valenta et al., Phys. Rev. Accel. Beams 28, 094601 (2025); https://doi.org/10.1103/knh7-hbr3.

[2] P. Valenta et al., Proc. SPIE 13534, 1353406 (2025); https://doi.org/10.1117/12.3058376.

Speaker: Dr Petr Valenta (Extreme Light Infrastructure ERIC / ELI Beamlines Facility)

12:00 → 13:00 Users' Talks I

12:00 Numerical Simulations at Karolina for Advancing Laser Plasma Research at Extreme Light Infrastructure - Nuclear Physics

The Extreme Light Infrastructure – Nuclear Physics (ELI NP) facility in Măgurele, Romania, operates the world’s most powerful laser system, capable of delivering two 10-PW pulses per minute. This facility is at the forefront of cutting-edge research in laser-driven particle acceleration and high-energy nuclear physics. To advance these fields, our group relies extensively on large-scale numerical simulations performed on the Karolina supercomputer at IT4Innovations.
We present an overview of the pivotal role of these simulations in supporting and guiding experimental research at ELI NP, with emphasis on three key areas:
1. Ion acceleration: from double-layer targets. Three-dimensional simulations reveal the sequence of physical effects, including pulse intensification in near-critical density plasmas via self-focusing, and the interplay of acceleration mechanisms such as hole-boring or light-sail radiation pressure acceleration, and target normal sheath acceleration. Together, these processes enable the acceleration of protons to energies exceeding 500 MeV, with energy spectra that fall off more slowly than exponentially.
2. Generation of monochromatic carbon ion bunches: We developed a laser-plasma source of narrow-band carbon ion bunches using the “peeler” scheme, in which an ultraintense laser pulse irradiates a thin foil from its edge. Such sources are of particular interest for applications in medical therapy.
3. Electron acceleration: We proposed a novel regime of laser wakefield acceleration in which nearly half of the laser pulse energy can be converted into broad-spectrum electron bunches. These beams can be employed directly or used to generate secondary bremsstrahlung radiation and tertiary particles such as neutrons.

The synergy between groundbreaking experiments at ELI NP and high-performance simulations at Karolina enables us to push the boundaries of knowledge in nuclear physics, laser-driven particle acceleration, and astrophysics. Numerical modeling not only validates experimental designs but also uncovers new opportunities and applications in fundamental research, industry, and medicine.

Speaker: Vojtěch Horný (Extreme Light Infrastructure - Nuclear Physics)

12:20 HPC-Accelerated Deep Learning Pipeline for Microhologram Dot Pattern Verification in Secure ID Document Authentication

Holographic elements are widely used in personal identification documents (e.g., ID cards, passports) and secure items (e.g., banknotes) to prevent counterfeiting. However, their reliable verification remains challenging due to lighting-dependent visual variability and the absence of scalable feature extraction methods.

To address this, we present a multi-stage deep learning pipeline for mobile, real-time verification of microhologram dot patterns—a physical security feature characterized by the random spatial distribution of embedded metallic dots. The proposed system first detects and rectifies the ID card, localizes the shield-shaped hologram containing microholograms, and segments individual microholograms using a U-Net architecture optimized for small-object detection. The resulting dot pattern is converted into a compact 64-bit perceptual hash code via difference hashing (dHash).

Similarity is evaluated using kernel density estimation (KDE) with Bayes-optimal decision boundaries for probabilistic classification. Multi-angle view fusion improves segmentation robustness against lighting variation, enabling accurate verification without specialized optical hardware.

Training and optimization of the pipeline required large-scale synthetic data generation, extensive hyperparameter tuning, and probabilistic model calibration, tasks accelerated by HPC resources. Experiments on both synthetic and real-world datasets achieve over 99.99% classification accuracy while maintaining low computational overhead at inference time.

By leveraging HPC resources for large-scale training and optimization, we make it possible to deploy state-of-the-art document verification techniques in real-world mobile applications. The resulting system offers a robust, scalable, and efficient approach to document verification based on microhologram patterns, providing both enhanced security and significant practical advantages for governments, financial institutions, and other stakeholders.

Speaker: Petra Svobodová

12:40 Electronic band structure calculations of crystals on noisy quantum simulators

Quantum computing is currently emerging as a possibly useful paradigm for solving highly complex computational problems. Current quantum computers are unfortunately too noisy to provide sufficient results, so quantum-classical hybrid algorithms emerged as a solution. Variational Quantum Deflation (VQD) has gained significant attention for addressing challenges in quantum chemistry, material science, etc. In our study we focused on the noise in current quantum computers and its impact on the calculations of the electronic band structure of silicon in the diamond lattice. The obtained results of our noisy-quantum computations have been compared with the classical diagonalization methods as well as results that were obtained from simulating the run of an ideal noiseless quantum computer. The quantum part of VQD ran on a classical simulator with imported noise models from real superconducting quantum computers from IBM.

Speaker: Vojtěch Vašina (Institute of Physics of Materials, Czech Academy of Sciences)

13:00 → 14:00 Lunch & Poster Session

14:00 → 15:00 Users' Talks II

14:00 Supercomputers and Synchrotrons: A New Era of Scientific Discovery

The integration of high–performance computing (HPC) with synchrotron light source facilities is reshaping the landscape of scientific discovery. Leveraging the capabilities of supercomputers for large–scale simulations and data–intensive analysis provides a powerful means to guide and complement experimental investigations. This presentation highlights the synergistic potential of combining synchrotron–based techniques with advanced computational resources to enable faster, more detailed, and accurate insights.
As a case study, we demonstrate the use of periodic density functional theory (DFT) calculations in conjunction with synchrotron diffraction experiments conducted on a beamline at the European Synchrotron Radiation Facility (ESRF). This integrated approach was employed to predict the structure and localization of bare Co(II) cationic sites within the ZSM–5 zeolite framework.[1]
By bridging experimental and computational methodologies, this workflow exemplifies a next–generation paradigm in materials characterization—accelerating discovery, enhancing beamline efficiency, and broadening access to complex, high–resolution analyses for the scientific community.

Reference
1. P. Rzepka, T. Huthwelker, J. Dedecek, E. Tabor, M. Bernauer, S. Sklenak, K. Mlekodaj and J. van Bokhoven, Science, 2025, 388, 423–428.

Speaker: Stepan Sklenak (J. Heyrovsky Institute of Physical Chemistry)

14:20 Microstructural Analysis of Aluminide Coatings using Deep Learning-Driven Image Segmentation and Synthetic Data Augmentation

Aluminide diffusion coatings are widely applied to steels in high-temperature environments such as steam turbines and solar systems, where they form corrosion-resistant layers that extend component lifetime. Accurate evaluation of these coatings is essential for quality control and accelerated materials development. Manual analysis of Scanning Electron Microscope (SEM) images to quantify features like layer thickness, pores, and chromium precipitates is challenging, time-consuming, and prone to human error due to noise, artifacts, and overlapping features. To address these limitations, this work introduces a deep learning-based approach for automated segmentation and metallographic evaluation of aluminide slurry coatings.

To overcome the limitation of a small manually annotated real SEM dataset, we generated a large corpus of high-fidelity synthetic SEM images together with their ground-truth masks using Blender’s procedural shading tools. Ground-truth labels for real SEM training data were created using the Trainable Weka Segmentation plugin in ImageJ, followed by manual refinement. A U-Net convolutional neural network was subsequently trained on a hybrid dataset combining real SEM micrographs with synthetic SEM images. Training employed high-performance computing (HPC) resources with distributed multi-GPU rendering for synthetic data generation and DistributedDataParallel training to accelerate model convergence. A combined Weighted Dice and Soft Cross-Entropy loss function was found to be most effective in handling class imbalance, particularly for pores and precipitates. Among the tested architectures (U-Net, Attention U-Net, DeepLabV3, and Swin UNETR), the baseline U-Net offered the best performance across most feature classes. The workflow achieved high segmentation accuracy across key microstructural features. On test data, the U-Net model delivered a Dice score of 98.7% ± 0.2 for the Fe2Al5 diffusion layer, 82.6% ± 8.1 for pores, and 81.5% ± 3.6 for chromium precipitates.

In addition to model development, synthetic SEM image generation was benchmarked on a high-performance computing cluster. Parallel rendering reduced synthetic dataset generation time compared to sequential rendering, while multi-GPU training further improved model scalability for larger datasets. Analysis of coatings produced with three slurry formulations revealed that samples prepared without a rheology modifier exhibited thicker Fe2Al5 layers due to dominant inward diffusion. In contrast, thinner coatings contained fewer pores and chromium-rich precipitates, independent of slurry composition.

This study streamlines coating characterization, improves quality control, and accelerates development by providing reproducible, high-accuracy segmentation. By bridging materials science and machine learning, this work demonstrates the potential of synthetic data augmentation and HPC-optimized deep learning workflows for advancing metallographic evaluation of protective coatings.

Speaker: Khyati Sethia (IT4Innovations National Supercomputer Center)

14:40 Thermoelectric effects of structural defects in scandium nitride nanostructures

The need for sustainable energy is nowadays well understood and multiple pathways are being explored, minimizing harmful by-products and energy waste in energy conversion processes. Thermoelectric effects have the potential to address these challenges, mainly by converting waste heat into reusable energy for industrial and domestic appliances [1].

Transition-metal nitrides (TMNs) are promising candidates for this purpose due to their favorable electronic and thermal properties. Among them, scandium nitride (ScN) is receiving considerable attention. The performance of bulk ScN is not sufficient for practical applications, but in recent years several techniques have been proposed to enhance its thermoelectric performance, primarily through the fabrication and use of nanostructured thin films [2,3].

The optimization of the thermoelectric performance of a material is a very complex task, as it requires acting separately on material properties that are strongly interconnected, such as electrical and thermal transport, maximizing the former while minimizing the latter. Achieving this goal requires a deep understanding of the mechanisms underlying these phenomena at the microscopic level.

Here we illustrate our work, in which we investigate the microscopic mechanisms underlying the variability observed in the thermoelectric response of ScN [4,5]. It is reasonable to hypothesize that this variability depends on the structural characteristics of the material, such as defects and the presence of heteroatoms in the crystal lattice. The Landauer approach allows us to relate the microscopic structures that compose the material to its macroscopic behavior, analyzing their contribution to the global properties [6].

Using this approach, we systematically analyze how different types of lattice imperfections—specifically oxygen impurities and nitrogen-site vacancies—affect electronic transport in ScN nanostructures. Defects are classified according to their symmetry (isolated, multiple contiguous, or associated with glide-type defects) and chemical nature. Our results identify two dominant defect classes with opposing effects on thermoelectric performance: (i) contiguous nitrogen vacancies, which enhance electrical conductivity but reduce the absolute value of the Seebeck coefficient, and (ii) oxygen substitutions coupled with nearby glide-type defects, which increase the absolute value of the Seebeck coefficient while hindering electrical conductivity [7].

[1]. J. He et al., Appl. Therm. Eng. 236, 121813 (2024).

[2]. P. Eklund et al., J. Mater. Chem. C 4, 3905 (2016).

[3]. B. Biswas et al., Phys. Rev. Mater. 3, 020301 (2019).

[4]. P.V. Burmistrova et al., J. Appl. Phys. 113, 153704 (2013).

[5]. J. More-Chevalier et al., Appl. Surf. Sci. Adv. 25, 100674 (2025).

[6]. S. Datta, Electronic transport in mesoscopic systems (Cambridge university press, 1997).

[7]. L. Cigarini et al., preprint (2025): https://tinyurl.com/2c4zw7rt

Speaker: Luigi Cigarini (IT4Innovations, VŠB – Technical University of Ostrava)

15:00 → 15:20 Session: Accessing EuroHPC Systems

Convener: Tomas Karasek (IT4Innovations, VSB-Technical University of Ostrava)

15:20 → 15:40 Support from the High-Level Support Team (HLST)

Convener: Joao Barbosa (IT4Innovations)

15:40 → 16:00 Coffee break

16:00 → 16:50 Users' Council

16:50 → 17:20 Keynote II

16:50 Model Complexity in All-Atom Plasme Membrane Simulations

Biological membranes play a crucial role in the ability of organisms to maintain cellular homeostasis in various environments, making use of the great chemical variability of lipids. However, this incredible diversity in membrane composition poses a great challenge for their use in molecular dynamics. To address this, we compared seven membrane models ranging in from a single-lipid POPC membrane commonly used in MD simulation, to a complex 18-lipid plasma membrane mimetic. We evaluated the impact of lipid composition on membrane properties and the conformational dynamics of an embedded TLR2 protein transmembrane fragment.

Our findings show that the main feature that governs membrane model properties is the presence or absence of cholesterol, and the subsequent membrane ordering. This, rather than the number or diversity of lipid species, had the major effect on membrane thickness, rigidity, protein conformational flexibility and its tilt in the membrane.

Speaker: Petra Čechová (Regional Centre of Advanced Technologies and Materials, Czech Advanced Technology and Research Institute (CATRIN), Palacký University Olomouc: Olomouc, CZ)

17:20 → 18:20 Users' Talks III

17:20 Translocation of short peptides through carbon nanotubes

We have been studying protein synthesis on ribosomes for many years. One of the open questions concerns the translocation of the nascent protein through the ribosome during synthesis (1). It is still unclear what forces drive the peptide from the catalytic center, buried deep within the ribosome, toward the ribosome surface.

To address this, we performed a series of non-equilibrium molecular dynamics simulations on a simplified system (2). Specifically, we examined how a peptide moves through a carbon nanotube under an applied mechanical force. We were particularly interested in the difference between pulling the peptide and pushing it in the same direction. Does it matter where the force probe is attached? And would that choice affect the conformational ensemble of the peptide?

Highlights
• It's fairly easy to move a peptide through a narrow channel when an external force is applied.
• Pulling the peptide makes its movement smoother than pushing, though in both cases the peptide's own flexibility adds complexity.
• Where the force is applied matters less than the peptide's sequence or how quickly the force is applied.

(1) Kolář et al.: Three stages of nascent protein translocation through the ribosome exit tunnel, WIREs RNA, 2024.
(2) Nepomuceno, Kolář: Sensitivity of peptide conformational dynamics in carbon nanotubes to directional mechanical forces, Phys. Chem. Chem. Phys., 2025.

Speaker: Michal H. Kolar (University of Chemistry and Technology)

17:40 Atomic-Scale Insights into Graphene-Based Nanomaterials Interacting with Brain Cell Membranes

Alzheimer’s and Parkinson’s diseases are progressive neurodegenerative disorders that currently have no cure. Deep brain stimulation (DBS), which delivers electrical signals to specific brain regions via implanted electrodes, has been used since the 1980s to alleviate symptoms. However, conventional metal electrodes often trigger inflammation, motivating the search for more biocompatible alternatives. Graphene-based materials have emerged as promising candidates due to their excellent electrical, mechanical, and potential biocompatibility properties.
Here, we present a comprehensive molecular dynamics study of the interactions between graphene derivatives and biologically relevant lipid membranes representing microglial cells, neurons, myelin sheaths, and commonly used in vitro cytotoxicity cell models. Our simulations reveal distinct interaction modes for GO and rGO, with surface oxidation playing a critical role in mediating membrane affinity, insertion depth, and disruption potential. Potential of mean force (PMF) calculations further quantify the energetics of membrane penetration, demonstrating that both the chemical nature of the graphene surface and the lipid composition of the target membrane significantly influence the interaction outcome.
These findings provide molecular-level insights into the design of graphene-based materials for neurointerfaces and establish key biophysical parameters for optimizing biocompatibility in future DBS electrode applications.

Speaker: Andrea Nedělníková

18:00 Lipid nanoparticles in coarse-grained resolution: From internal organization to interaction with cells.

Lipid nanoparticles (LNPs) have emerged as key vehicles for nucleic acid (NA) delivery, promising novel tool for vaccination, cancer and rare diseases therapy. LNPs are complex structures, composed of ionizable lipids (ILs), helper lipids, PEGylated lipids and cargos. ILs are responsible for efficient encapsulation of the NA cargo and for NA release during endosome maturation. The dynamic nature of LNP pH dependent behavior inside a target cell is a major challenge in experimental research of LNP behavior, therefore a full understanding of LNP structure and processes related to the cargo release are still missing.
Molecular dynamics simulations with multiscale resolution offer a powerful approach to explore LNP organization and function in various environments. All-atom simulations can describe the interactions of individual lipid functional groups with NAs and their mutual effect on their structure and stability, but are limited to either small models or a short simulation of a prebuilt LNP. On the other hand, coarse-grained (CG) simulations can be used to simulate the formation of a whole LNP in tens of nanometers and microseconds scale, predicting the internal LNP organization, distinguishing between lipid inverse hexagonal and lamellar phase. The lower computational costs allow CG simulations to study LNP in a systematic way, manipulating the composition and ratio of lipid species or in a desired bioenvironment in its path through the body, getting us closer to the description of the mechanism of the endosomal escape process.
The potentials and limitations of both the resolutions can be efficiently combined to a valuable workflow, advancing our understanding of LNP structure, stability and behavior. The provided insight can lead to a targeted in-silico design of next-generation delivery platforms, increasing the cargo delivery efficiency and decreasing the costs. Their integration into formulation workflows represents a promising direction for predictive, mechanism-informed design of therapeutic systems.

Speaker: Markéta Paloncýová (Regional Centre of Advanced Technologies and Materials, Czech Advanced Technology and Research Institute (CATRIN), Palacký University Olomouc)

18:20 → 21:00 Conference Dinner and Poster Session

18:20 Suitable boron-doped graphene substrate for glucose Raman signal enhancement

Surface-enhanced Raman spectroscopy (SERS) is a highly sensitive and selective technique. It greatly enhances the signal of an analyte compared to classical Raman spectroscopy, due to analyte-substrate interactions. A promising substrate for SERS is boron-doped graphene (B-graphene). At low boron concentrations of ∼1.39 at.%, it has been shown to enhance the Raman signal of simple organic molecules such as pyridine. The potential use of high-concentration B-graphene materials for SERS remains unexplored. Therefore, in our study, we investigate the influence of dopant concentration and relative adsorbate/substrate geometry on the effectiveness of B-graphene as a SERS substrate, with glucose as the analyte. We perform Density Functional Theory simulations using the PBE functional and the DFT-D2 van der Waals correction. By combining analysis of interatomic force constants and phonon eigenvector composition, we conclude that higher doping concentrations provide a larger enhancement to the Raman signal of glucose, while the molecule’s orientation relative to the surface plays a fundamental role in the Raman response. We suggest that 12.5 at.% B-graphene represents a potential substrate for SERS-based detection of glucose. Additionally, the phonon-based analysis can be promptly applied in the search for promising substrate materials for enhanced Raman response.

Speaker: Antonio Cammarata (Czech Technical University in Prague)

18:21 New potential for atomistic simulations in NiTi martensite

NiTi shape memory alloy has become the most widely used shape memory material in industrial, high-tech, and medical applications due to its unique thermal and mechanical properties, primarily represented by the shape memory effect and superelasticity. Due to these unique characteristics, this alloy has been used in numerous practical applications since its discovery in 1963. Despite having only a single active slip system, which theoretically limits the plastic deformability of individual grains in a polycrystalline material, NiTi martensite exhibits remarkable plasticity. The underlying mechanism behind this behavior remained unclear for a long time.
Recently, a novel deformation mechanism termed "kwinking" was identified [1]. The term originates from a combination of "kink" and "twin," reflecting the unique nature of this planar crystal defect. Kwinking exhibits characteristics of both twinning and kinking: it forms twin-related regions while simultaneously accommodating plastic deformation through geometrically necessary dislocation walls. It has been observed experimentally [2] that kwinking is the dominant mechanism of plastic deformation of the B19' martensitic phase and provides this phase with high ductility.
To understand the martensitic phase at the atomic scale, we will use the most modern methods of atomistic simulations that combine quantum-mechanical calculations using density functional theory (DFT) and molecular dynamics (MD) simulations. The quality of atomistic simulations in MD strongly depends on the accuracy of the available interatomic potentials. In the past decades, MD simulations have been employing semiempirical interatomic potentials but, recently, new techniques based on machine learning (ML) became available for potential generation. In our study, we will explore reliability of our new ML interatomic potentials by comparing their prediction with DFT data. In particular, we will calculate elastic constants of the martensite phase and the energy as a function of the monoclinic angle.

The ML potentials are constructed from the ab initio training set obtained either from the ab initio molecular dynamics simulations or from many ab initio static configurations by fitting free energies and forces acting on individual atoms. All data necessary for the construction of the training set were obtained using the ab initio software package VASP [3], [4]. For the fitting of the ML potential, we employ i) the VASP routine that collects the selected configurations occurring during ab initio MD simulations and includes them in the data set, ii) the neural network approach implemented in the code RuNNer [5] and iii) the atomic cluster expansion as employed in the code Pacemaker [6].

References
[1] H. Seiner, P. Sedlák, M. Frost and P. Šittner, Int. J. Plast. 168, 103697 (2023).
[2] O. Molnárová, M. Klinger, J. Duchoň, H. Seiner and P. Šittner, Acta Mater. 258, 119242 (2023).
[3] G. Kresse and J. Hafner, Phys. Rev. B 48, 13115 (1993).
[4] G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 (1996).
[5] J. Behler, J. Chem. Phys. 13, 170901 (2011).
[6] Y. Lysogorskiy, C. v.d. Oord, A. Bochkarev, S. Menon, M. Rinaldi, T. Hammerschmidt, M. Mrovec, A. Thompson, G. Csányi, C. Ortner and R. Drautz, npj Comput. Mater. 7, 1 (2021).

Speaker: Petr Řehák

18:22 Characterizing Nonlinear Dynamical System Behaviors Using Self-Organizing Maps with Clustering Techniques

This study proposes a novel framework for clustering dynamical systems to identify parameter combinations that elicit comparable responses of dependent variables under specified initial conditions. Using self-organizing map (SOM) clustering, trajectories are categorized into distinct behavioral regimes determined by parameter variations. A systematic sampling of parameter spaces produces well-defined clusters, revealing underlying patterns associated with excitability and oscillatory dynamics. The method enhances interpretability of complex datasets and supports the detection of critical system modifications. Analytical outputs, including parameter distribution maps, bar charts, and two-dimensional visualizations, are employed to characterize the clusters. As a case study, the FitzHugh–Nagumo model is analyzed, where clustering partitioned the dataset into twelve groups of parameter configurations exhibiting analogous dynamical responses, thereby uncovering fundamental patterns in system behavior.

Speakers: Ms Haiqa Ehsan (VSB - Technical University of Ostrava), Muhammad Zeerak Awan (VSB - Technical University of Ostrava)

18:23 On the role of point defects and metal clustering in the mechanical properties of titanium aluminium oxynitride coatings

Thermal stability and mechanical performance are crucial criteria in the design of next-generation protective coatings for cutting, drilling, and forming applications. Currently, TiAlN serves as the benchmark coating due to its excellent chemical, mechanical, and thermal stability. Titanium aluminium oxynitride (TiAlON), however, has emerged as a promising alternative, offering an increase in thermal stability of up to 300 °C [1].

This improvement in thermal stability is offset by a reduction in elastic modulus. We deposited TiAlON coatings by reactive high-power pulsed magnetron sputtering and observed a decrease in elastic modulus of ~20% at 15 at.% O as compared to TiAlN. In contrast, simple density-functional-theory-based models fail to capture this trend and instead predict a much larger reduction in the elastic constants, up to 50%. To address this discrepancy, we trained a machine-learning interatomic potential using the atomic cluster expansion formalism [2], enabling large-scale modelling of TiAlON. With this potential, we successfully reproduced the mechanical properties of TiAlON up to 25 at.% O in agreement with experiment. Moreover, we identified the initial stages of spinodal decomposition and the formation of vacancy–Frenkel pair complexes as the key structural features governing the elastic properties in the studied compositional range.

[1] D. M. Holzapfel et al., Acta Materialia 218, 117204 (2021).
[2] Y. Lysogorskiy et al., npj Computational Materials 7, 97 (2021).

Speaker: Pavel Ondračka (MUNI)

18:24 Membrane selectivity of Opi1 peptide derivatives

Membrane-binding peptides hold considerable promise for applications such as antimicrobials, biosensors, and extracellular vesicle purification. However, their successful implementation in biotechnology requires careful consideration of membrane selectivity. The Opi1 peptide has demonstrated a selective affinity for phosphatidic acid–containing membranes, as shown experimentally by Ernst et al. [1]. Luevano-Martínez et al. [2] further reported that the modulation of Opi1-related genes in yeast directly influences the cardiolipin content of membranes. This observation, together with the chemical similarities between phosphatidic acid and cardiolipin, suggests that Opi1 also exhibits notable affinity for cardiolipin, which is found in inner mitochondrial membranes as well as in some bacterial membranes. Nevertheless, only a limited range of bacteria contain substantial amounts of cardiolipin. Thus, peptides with selective affinity for other lipids, such as phosphatidylglycerol or phosphatidylethanolamine, are required to address the natural diversity of membrane compositions [3]. In this study, we use the Opi1 peptide as a reference framework to rationally design peptide derivatives with broader membrane selectivity. Our approach employs advanced computational methodologies for free energy calculations, including alchemical free energy transformations and umbrella sampling. Preliminary results indicate that Opi1 selectivity can be modulated through specific point mutations. The resulting mutants exhibit distinct patterns of selectivity toward human and bacterial membranes.

[1] Ernst, et al. J. Cell Biol. 2018, 217(9), 3109-3126. https://doi.org/10.1083/jcb.201802027
[2] Luévano-Martínez, et al. Fungal Genetics and Biology. 2013, 60, 150-158. https://doi.org/10.1016/j.fgb.2013.03.005
[3] Malanovic, et al. Biochimica et Biophysica Acta (BBA)-Biomembranes. 2020, 1862(8), 183275. https://doi.org/10.1016/j.bbamem.2020.183275

Speaker: Alejandro Hernández Tanguma (CEITEC-MUNI)

18:25 Atomistic Insights into the Structure and Dynamics of Lipid Nanoparticles

Lipid nanoparticles (LNPs) are widely used as drug delivery systems and have been successfully applied in vaccines such as Onpattro, Comirnaty, and Spikevax. Despite their global success, key questions remain regarding the mechanisms of endosomal drug release from LNPs. Numerous experimental and theoretical studies have addressed this topic. Most previous theoretical studies have been performed at a coarse-grained resolution. My work aims to develop an atomistic simulation protocol to gradually construct inverse hexagonal lipid assemblies containing RNA molecules at their core. The structural and dynamic properties of these systems would be analyzed and compared with experimental data. This inverse hexagonal organization is hypothesized to play a critical role in facilitating endosomal escape, and understanding its behavior may provide valuable insights into improving LNP-based drug delivery.

Speaker: Adéla Chadalíková (CATRIN, VŠB-TUO)

18:26 Running a mixture of ATLAS jobs with widely ranging resource requirements at IT4Innovation

For years, the ATLAS experiment is running selection of its workflows (mostly Monte Carlo simulations, which have small inputs and outputs and can run even on tens of cores for hours) at IT4Innovation. As the experiment and its computing evolves, so does constitution of its workflows. Recently, we are more often getting into a situation when there is insufficient number of these jobs while there are many other jobs from other workflows. The most common example would be event generation which has small or no input and small output but, unlike simulations, runs only on one core. Insufficient number of runnable jobs means leaving available resources idle. To address that, the submission system was updated to allow, for example, jobs using one CPU core running next to jobs requiring 30+ CPU cores. This allows running more workflows. With the help of HyperQueue, available ATLAS jobs can efficiently fill available resources.

Speaker: Michal Svatoš (FZU, AV CR)

ATL-SOFT-SLIDE-2025-551.pdf

18:27 Benchmarking GW Methods for Accurate Prediction of Defect States in Doped Diamond

Diamond is a wide-bandgap semiconductor with exceptional physical properties, making it highly attractive for applications in power electronics, photovoltaics, nanophotonics, and quantum technologies. Its functionality is often engineered through the introduction of defects such as dopants or vacancies; however, accurate description of defect states remains challenging. Density functional theory (DFT) provides a useful starting point but underestimates band gaps and misplaces defect levels. These errors are especially pronounced for deep-level states, such as those associated with color centers (e.g. the NV center), and they propagate into subsequent predictions, including optical absorption. Many-body perturbation theory, in particular the GW approximation, offers significant improvement over DFT, but its accuracy strongly depends on the method used to evaluate dynamical screening. In practice, plasmon-pole models are often sufficient for insulators and semiconductors, while full-frequency methods are required for metals. Doped structures constitute a “middle-ground” between insulators and metals that has received little attention in GW studies, and even when addressed, clear benchmarks for methodological choices are often lacking.

To address this, we performed GW calculations of diamond-based structures containing substitutional boron and phosphorus, as well as phosphorus-vacancy and boron-vacancy-boron colour centres. In this way, we were able to analyse positions of impurity states and width of the bulk band gap in a variety of scenarios ranging from degenerate semiconductors to semiconductors with multiple deep defect states. We compared the Godby-Needs plasmon-pole model with the contour deformation technique using a variety of frequency grids. Furthermore, we investigated the effect of different treatments of dynamical screening on charged defects.

Our work highlights the trade-offs between accuracy and computational cost of different GW methods for simulating doped wide-bandgap semiconductors. Beyond numerical comparisons, it provides practical guidelines for choosing suitable methodologies when modelling systems where the physics lies between that of simple insulators and metals.

This work was supported by the project “The Energy Conversion and Storage”, funded as project No. CZ.02.01.01/00/22_008/0004617 by Programme Johannes Amos Commenius, call Excellent Research. This work was supported by the Ministry of Education, Youth and Sports of the Czech Republic through the e-INFRA CZ (ID:90254). The access to the computational infrastructure of the OP VVV funded project CZ.02.1.01/0.0/0.0/16_019/0000765 “Research Center for Informatics“ is also gratefully acknowledged.

Speaker: Matúš Kaintz (Czech Technical University in Prague)

18:28 Analysis of Magnetic Spin Disorder in Ferrimagnetic Iron Oxides with DFT and Beyond-DFT Approaches

Ferrimagnetic iron oxides such as magnetite or maghemite find many applications in magnetic recording, spin-based electronics and biomedicine. However, when reduced to the nanoscale, their long-range spin ordering often collapses, giving rise to surface-driven magnetic disorder. This disruption originates from broken interactions, symmetry and reduced atomic coordination at facets and edges of the nanoparticles, leading to the stabilization of spin-flipped states. Configurations involving magnetic disorder at the surface reduce the net magnetic moment and may correspond to metastable, energetically accessible states. Interestingly, magnetic disorder has also been associated with enhanced T₁ relaxation times in MRI, pointing to a potential connection between local spin structure and biomedical functionality.
In this study, we analyse electronic structure of bulk and nanoscale models of iron oxides in both ferrimagnetically ordered and spin-disordered states using Density Functional Theory (DFT) and selective approaches beyond standard DFT.

Speaker: Valentína Berecová (Institute of Physics of Materials CAS, v. v. i.)

18:29 Benchmarks of composite quasi-harmonic models of polymorphism in benzophenone and sulfamerazine

Accurate prediction of physical properties is necessary when developing high-value materials such as pharmaceuticals. At the same time, methods needed to achieve required accuracy, such as periodic density functional theory (DFT), remain commonly used despite being computationally demanding. Given that interest in modelling even larger and more complex systems has only continued to grow, these factors have led to the development of cruder yet faster algorithms such as density functional tight binding (DFTB). For instance, its implementation in DFTB+ is reported to improve runtimes up to two orders of magnitude by exploiting parallelization and matrix algebra [1]. Still, approximations inherent in these semi-empirical techniques can cause properties to be predicted inaccurately.
This work explores the performance of DFTB-D4 for predicting thermodynamic properties of two common pharmaceutical precursors, benzophenone and sulfamerazine, within quasi-harmonic approximation (QHA). As DFTB-only QHA is burdened by unacceptable errors, we look at whether matching DFTB outputs with DFT signals improves DFTB-only predictions [2]. In this composite DFT/DFTB QHA framework, costly constant-volume DFT calculations are run only at the equilibrium volume, and the volume-dependence of the crystal’s static and dynamic degrees of freedom is obtained instead by running DFTB at various fixed volumes. Comparisons for α- and β-benzophenone are made to describe whether discrepancies observed between individual QHA levels relate to polymorphism. Finally, the utility of the hybrid approach is demonstrated through the stable Pna21 polymorph of sulfamerazine, where a volume-dependent description of DFT energies can be too costly to obtain otherwise.
[1] Hourahine, B., et al., (2020), J. Chem. Phys., 152, 124101.
[2] Ludík, J., et al., (2024), J. Chem. Theory. Comput., 20, 2858.

Financial support from Czech Science Foundation under the project No. 23-05476M and from COST action CA22107 (BEST-CSP) is acknowledged. Computational resources were supplied under the project e-INFRA CZ (ID:90254) provided by the Ministry of Education, Youth and Sports of the Czech Republic. R. G. acknowledges funding from the grant of specific university research No. A1_FCHI_2024_001.

Speaker: Reynaldo II Geronia (UCT Prague)

18:30 EasyDock 1.0: customizable and scalable docking tool

EasyDock 1.0 - an open-source and scalable Python-based tool for fully automated molecular docking. The current version supports popular docking programs, namely Autodock Vina, gnina, and smina. The tool automatically prepares ligands by removing salts, generating initial conformers and stereoisomers, using RDKit, and performing protonation with the open-source program MolGpKa. Ring sampling is implemented to improve docking of molecules containing saturated ring systems. All input data, settings, and results are stored in an SQLite database, enabling interrupted jobs to be resumed. EasyDock integrates Dask for distributed computation across multiple machines. A built-in model predicts docking times to optimize task scheduling and reduce total runtime. Special cases, such as boron-containing molecules, are handled by temporarily substituting boron with carbon during the docking process. The ProLIF package is integrated to calculate protein-ligand interactions. The current version is composed entirely of open-source modules.

Speaker: Ms Guzel Minibaeva (Institute of Molecular and Translational Medicine, Faculty of Medicine and Dentistry, Palacký University)

18:31 Optimisation of k-Wave Ultrasound Wave Propagation Simulations Through Fourier Transform Pruning

Wave propagation simulations are crucial for planning non-invasive medical treatments such as focused ultrasound therapy. However, these procedures often require multiple high-resolution simulations, leading to significant resource consumption and potential delays in critical treatments. This work presents a method to accelerate simulations performed using the k-Wave toolbox while maintaining an acceptable level of accuracy for pre-planning purposes. The k-Wave toolbox uses the k-space pseudo-spectral method with Fourier basis functions, where substantial computational time is spent on Fourier transforms. By substituting standard Fast Fourier Transform (FFT) computations with a pruned version of the algorithm, speedups of up to 1.7x for large simulation domains were achieved. The proposed approach, using the Acoustic Field Propagator with a bisection pruning algorithm, eliminates over 60% of spectral coefficients. Despite this reduction, focal point errors for a single transducer setup remain below 1%. Tests involving transcranial ultrasound and tumor ablation have shown consistent accuracy with only minor shifts in focus position. This improvement has the potential to significantly accelerate clinical workflows for treatment planning. While these principles primarily enhance the k-Wave toolbox, they could also benefit wave propagation simulations in other fields, particularly for weakly heterogeneous media.

Speaker: Ondřej Olšák

18:32 Deep Potential Molecular Dynamics Study of Au(111)/MX2/Si tip

Transition metal dichalcogenides (TMDs) on metallic substrates, such as Au, exhibit unique interfacial interactions that influence their structural and tribological properties. In this study, we developed a machine learning-based force field using Deep Potential Molecular Dynamics (DPMD) to model TMDs on an Au substrate and the Si tip. The generated force field was validated by computing the radial distribution function (RDF) scalar products of density functional theory (DFT) and DPMD results. Our analysis shows a strong agreement between the DPMD and DFT calculations, demonstrating the accuracy and reliability of the developed force field. The phonon calculations show that all the systems are dynamically stable. These findings contribute to the efficient modelling of TMD-metal interfaces, paving the way for advanced simulations in nanotribology and material design.

Speaker: Mr Suresh Ravisankar (Czech Technical University in Prague)

18:33 Quantum-mechanical analysis of H7LaNi5-xSnx

Investigating hydrogen sorption in hydride-forming materials is essential for advancing a wide range of industrial technologies. Among such materials, LaNi$_5$ is a common candidate due to its ability to accommodate multiple hydride phases. Upon full hydrogenation, LaNi$_5$ forms the stable hydride phase LaNi$_5$H$_7$. Its properties such as lowering absorption and desorption pressures and enhancing resistance to repeated cycling can be improved through partial substitution.
In this work, we focused on LaNi$_{5-x}$Sn$_x$ compounds. Quantum-mechanical approaches play a crucial role in investigating hydride formation, particularly given the difficulty of experimentally resolving the precise crystal structures of these compounds, which further complicates subsequent thermodynamic analyses. Experiments-complementing Density Functional Theory (DFT) calculations provide key insights into ground-state properties, including equilibrium lattice parameters, formation energies, electronic density of states, and charge density distributions.
DFT calculations revealed that the compositions LaNi$_{4.88}$Sn$_{0.13}$ and LaNi$_{4.38}$Sn$_{0.63}$ exhibit the highest thermodynamic stability among the LaNi$_{5-x}$Sn$_x$ series. Increased Sn content in the investigated structures further influences the magnetic properties of the phases. The phase with a lower Sn concentration retains its magnetic character, whereas the phase with a higher Sn concentration becomes non-magnetic. These optimized structures subsequently serve as the basis for modeling the formation of the type H$_7$LaNi$_{5-x}$Sn$_x$ hydrides. This provides insight into how partial substitution of Ni by Sn affects hydrogen uptake and the stability of the resulting hydride phases.
Additionally, computational aspects of a technical character, such as the computational time requirements were analyzed with respect to the number of computational nodes and the internal parameters of the applied code, offering a practical perspective on the efficiency and scalability of such simulations at the IT4I supercomputing infrastructure.

Computational resources were provided by the e-INFRA CZ project (ID:90254), supported by the Ministry of Education, Youth and Sports of the Czech Republic. These resources were utilized through IT4Innovations National Supercomputing Center, MetaCentrum as well as CERIT Scientific Cloud. This work was created as part of the project No. CZ.02.01.01/00/22_008/0004631 Materials and technologies for sustainable development within the Jan Amos Komensky Operational Program financed by the European Union and from the state budget of the Czech Republic.

Speaker: Kateřina Dočkalová (Institute of Physics of Materials, Czech Academy of Sciences, Brno, Czech Republic)

18:34 A Scalable Framework with Modified Loop-Based Multi-Initial Simulation and Numerical Algorithm for Classifying Brain-Inspired Nonlinear Dynamics with Stability Analysis

The main challenge in analyzing nonlinear dynamical systems lies in the repetitive and inefficient need to simulate each initial condition and parameter configuration separately. This approach not only increases computational cost but also limits scalability when exploring large parameter spaces. To address this issue, we developed a loop-based numerical methodology that automates the exploration of parameters and initial conditions within a unified framework. The method is applied to a brain-inspired nonlinear dynamical model with three parameters and multiple coupling strengths. This framework enables systematic categorization of system responses through statistical analysis and eigenvalue-based stability assessment, while accounting for multiple initial states. The results highlight clear distinctions between periodic, divergent, and non-divergent behaviors and demonstrate how variations in coupling strength, 𝑘𝑖𝑗, can induce transitions toward stable periodic dynamics across different conditions. By reducing redundancy and improving scalability, the proposed approach not only simplifies the analysis process but also provides a generalizable framework for studying broader classes of complex systems.

Speaker: Haseeba Sajjad

18:35 The magnetic exchange coupling in Co/Ru/Co trilayers: a study of the bilinear and biquadratic coupling parameters by electronic structure calculations.

Interlayer exchange coupling between two ferromagnetic layers across a spacer layer has been intensively investigated within the last few decades. It was discovered that the interlayer exchange coupling across most 3d, 4d, and 5d nonmagnetic metallic spacer layers oscillates between antiferromagnetic and ferromagnetic as a function of spacer layer thickness (1,2). This discovery enabled the control of antiferromagnetic coupling between two ferromagnetic films, which is now used in most spintronic devices (3).
Until recently, the greatest interest has been in the bilinear term of the coupling, where the energy per area is linear in the directions of magnetizations in both magnetic layers. With this form of the interaction, positive values of the bilinear coupling favor parallel alignment of the magnetizations and negative values favor antiparallel alignment. In our recent work (4), we demonstrated that electronic stricture calculation methods can accurately predict the values of the bilinear coupling in Co/Ru/Co systems. In this work we extend our study to the biquadratic term of the coupling, which favors perpendicular orientations of two magnetizations. Biquadratic coupling is believed to have a separate origin from the bilinear coupling, and produced by the presence of disorder (extrinsic coupling). We show, however, that the intrinsic biquadratic term of coupled magnetic layers with perpendicular magnetizations may be of similar magnitude as the extrinsic one, but have both positive and negative values of the coupling. This can lead to both enhancement or reduction of experimentally observed biquadratic coupling.

1. P. Grünberg, R. Schreiber, Y. Pang, M. B. Brodsky, H. Sowers, Layered magnetic structures: Evidence for antiferromagnetic coupling of Fe layers across Cr interlayers. Phys. Rev. Lett. 57, 2442–2445 (1986).
2. S. S. P. Parkin, Systematic variation of the strength and oscillation period of indirect magnetic exchange coupling through the 3d, 4d, and 5d transition metals. Phys. Rev. Lett. 67, 3598–3601 (1991).
3. R. A. Duine, K.-J. Lee, S. S. P. Parkin, M. D. Stiles, Synthetic antiferromagnetic spintronics. Nat. Phys. 14, 217–219 (2018).
4. K. Winther , Z. R. Nunn, J. Lisik , S. Arapan , D. Legut, F. Schulz, E. Goering , T. Mckinnon, S. Myrtle, and E. Girt, Antiferromagnetic coupling across nonmagnetic transition-metal films alloyed with ferromagnetic elements. Phys. Rev. Applied 22, 024058 (2024).

Speaker: Sergiu Arapan (IT4Innovations VŠB - TU Ostrava)

18:36 Developing Machine Learning Force Fields for Transition Metal Dichalcogenides with metallic substrate Ag and Si AFM tip

Since ab initio molecular dynamics (AIMD) simulations are computationally very expensive to study the properties of 2D layered materials, 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 four transition metal dicalgogenides (TMDs) monolayer (MoS2, MoSe2, WS2,and WSe2) interacting with a metallic substrate silver (Ag) and a silicon (Si) atomic force microscopy (AFM) tip. The initial structure of TMDs with the Ag substrate is optimized using density functional theory (DFT) calculations, and the optimized structure is then used as a training data to develop machine learning force fields (MLFFs) . Classical molecular dynamics (MD) simulations are performed using LAMMPS to further optimize the structure with MLFFs and compute the radial distribution function (RDF) to assess the accuracy of the developed force fields.

Keywords: Ab-initio, Transition Metal Dicalcogenides, Machine Learning Force Fields, DFT.

Speakers: Dr Antonio Cammarata (Department of Control Engineering - KN:G-204, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic), Ravikant Kumar (Department of Control Engineering - KN:G-204, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic)

18:37 Light Utilization for Matter Emergence

Very strong electromagnetic fields can induce electron-positron pair creation predicted by quantum electrodynamics. A possible way to demonstrate it is via the multiphoton Breit-Wheeler process in strong laser fields. The challenge in this setup is to hold seed particles in the hot-spot region of the reaction. This method using radially polarised laser pulses to trap seed electrons is studied in M. Jirka and S. Bulanov, PRL 133 125001 (2024). We simulate this process in the PIC code SMILEI (https://smileipic.github.io/Smilei/), which contains the respective models. The further advancement in this field is planned in Project OPEN-34-69. This project will consist of designs of more advanced optical traps. To reach this goal, we need implement some new features in the SMILEI code.

This poster details the simulations and methods, it also provides perspective for forthcoming studies.

Speaker: Jan Vábek (Czech Technical University in Prague, FNSPE)

18:38 Use of open data for training machine-learning interatomic potentials

With machine learning and its use in science rapidly growing in popularity, the need for high-quality training data is increasing. Most researchers however train their models either on their own data or on curated databases. With the growing emphasis on open science, a large amount of data from other researchers is now openly available, but such data often come without any guarantee of quality, thus its suitability for machine learning is uncertain.

In this work, we assess the quality of the data in NOMAD, the largest open materials simulation database, and its practical applicability for training machine-learning interatomic potentials for atomistic simulations. We present a workflow designed to tackle several challenges associated with the NOMAD data: automatically filtering out results with low numerical accuracy, deduplicating structures in the training data, and combining results coming from multiple DFT implementations with different total energy offsets.

With this workflow, we have successfully trained silicon-based potentials using only simulations from NOMAD as training data. The resulting potential predicts phase stability at a level comparable to state-of-the-art potentials, while also accurately describing large-scale atomic systems, even at high temperatures. This demonstrates that using open data can significantly reduce the time and costs required to generate suitable training datasets for machine-learning interatomic potentials.

Speaker: Šimon Kratochvíl

18:40 Response of twins in NiTi martensite to mechanical loading

Shape-memory alloys are unique materials capable of undergoing large reversible strains and exhibiting the shape-memory effect, which is driven by external changes of temperature. These remarkable properties are based on a martensitic transformation between austenite (high-temperature phase) and martensite (low-temperature phase). The NiTi shape memory alloy has become the most widely used shape memory material in industrial, high-tech, and medical applications due to its unique thermal and mechanical properties.
In this work we aim at fundamental understanding of the behavior of twins in the martensite structure during mechanical loading. In order to consider a sufficient number of internal degrees of freedom, we constructed relatively large supercells representing the perfect and twinned martensite and studied their responses to shear and tensile loading. Such calculations are computationally very demanding when using ab initio approaches. Therefore, we developed a novel machine-learnt (ML) interatomic potential, tailored specifically for the martensite structure. Computationally accessible predictions based on the ML potential were benchmarked from first principles.

Speaker: Miroslav Černý (CEITEC, Brno University of Technology)

18:41 Molecular Dynamics Refinement of Cytochrome P450 Docked Complexes

Docking and molecular dynamics (MD) are widely used to predict drug–protein interactions. While docking provides rapid starting structures, MD can equilibrate ligand binding poses at the cost of greater computational demand. Automated pipelines remain limited, particularly for systems with covalently bound cofactors or metals, or for flexible protein docking.
We present an automated workflow built on the GPU-accelerated OpenMM engine for large-scale refinement of docked complexes. The pipeline integrates seamlessly with both rigid and flexible docking methods and automates ligand parameterization for the majority of cases. In benchmark tests, automatic setup succeeded for ~90% of ligands with rigid docking engines, while RoseTTAFold All-Atom achieved ~30% due to ligand conformation artifacts.
This workflow enables high-throughput, GPU-powered refinement of cytochrome P450 complexes, making MD refinement feasible for large numbers of docked complexes.

Speaker: Przemyslaw Grenda (Department of Biophysics and Physical Chemistry, Faculty of Pharmacy, Charles University, Czech Republic)

18:42 Conformational preferences of ribosomal protein fragments

The ribosome is a macromolecular complex that catalyzes all protein synthesis. It is composed of ribosomal RNA and proteins, which contribute to both its shape and function. The catalytic center of the ribosome, the peptidyl transferase, is where peptide bond formation occurs, and ribosomal proteins were gradually incorporated into this core during evolution. Whether the proteins associated with the catalytic center needed to adapt, or naturally adopted specific conformations based on their native states regardless of ribosomal interactions, is still not fully understood. We used molecular dynamics simulations of short ribosomal-protein fragments close to the catalytic center in water. Dihedral angles were compared with experimental data. The dihedral angles of peptides in water closely resemble the ones experimentally obtained and are near the global energy minima found in the modern ribosome. This may hint that the conformational pre-arrangement could contribute to the selection pressure. As a negative control, we are conducting simulations with randomized amino acid sequences containing the same amino acid composition as the original fragments. The results may give some insight into how proteins and RNA came to work together in the ribosome.

Speaker: Jan Heblt (University of Chemistry and Technology)

18:43 Local vs. Global FFT Approaches for High-Performance Ultrasound Simulation on Multi-GPU Systems

Simulating wave propagation with the Fourier collocation method
is computationally intensive due to its reliance on discrete Fourier
transforms (DFTs). While DFTs enable near-minimal spatial dis-
cretization, they scale poorly on modern high-performance com-
puting systems. This work evaluates two multi-GPU strategies for
three-dimensional simulations: a Global FFT approach using dis-
tributed transforms and a Local FFT approach based on domain
decomposition with halo exchanges. Experiments were performed
on system with 8 NVIDIA A100 GPUs connected via NVSwitch.
Precision tests show that the Local FFT approach maintains errors
around 0.1% when the halo covers the local PML region. Perfor-
mance results demonstrate that the Local FFT approach achieves
lower runtimes and significantly reduced communication overhead
compared to the Global FFT approach, particularly for larger do-
mains. These findings indicate that Local FFT decomposition is a
promising strategy for scalable, large-scale multi-node ultrasound
simulations.

Speaker: Oliver Kuník

18:44 The role of non-canonical amino acids on the protoribosome

The protoribosome, carrying the peptidyl transferase center, represents an ancestral core of modern ribosomes. Fragments of several ribosomal proteins, so called rPeptides, extend toward the PTC and stabilize protoribosomal RNA (1).

In this work, we investigate the conformational stability of a protoribosome model from T. thermophilus upon replacement of lysine and arginine residues to their shorter prebiotic analogues, diaminobutyric acid (DAB) and diaminopropionic acid (DPR). We used all-atom molecular dynamics simulations with GROMACS. Our preliminary data show increased root mean squared deviation of two rPeptides, indicating looser binding, while other rPeptides and the RNA scaffold remained conformationally more stable. These results suggest that ancestral ribosomal proteins with electropositive residues with a shorter side chains may have reduced structural stabilization capacity. To deepen this understanding, future work will include additional prebiotically plausible amino acids, providing a more complete picture of protoribosome optimization and its role in the origin of life.

(1) Codispoti et al.: The interplay between peptides and RNA is critical for protoribosome compartmentalization and stability, Nucl. Acids Res. 2024.

Speaker: Michael Křivan (Vysoká škola chemicko-technologická)

18:45 Optimizing NiTi Interatomic Potentials Through Atomic Cluster Expansion

Atomistic simulations provide a way to observe atomic behavior at the nanoscale level. There are two main approaches: the first is based on quantum mechanics (ab initio simulations), and the second relies on Newton's mechanics (molecular dynamics). However, despite advancements in computer science, including quantum computing, both approaches remain limited. Ab initio simulations are constrained by the size of the simulation cell and the necessity to perform quasi-static simulations at T = 0 K due to their complexity and high computational demands, while molecular dynamics primarily suffer from the accuracy of interatomic potentials. Many of these limitations can be mitigated through machine learning, which enables the construction of interatomic potentials from training sets obtained in ab initio simulations.
In this work, we present an interatomic potential for NiTi based on the Atomic Cluster Expansion, developed using the Pacemaker software package. We validate this potential by comparing it against the results of simulations using other interatomic potentials, quantum-mechanical calculations, as well as our own experimental data. Our quantum-mechanical calculations utilize density functional theory (DFT) within the generalized gradient approximation (GGA) to determine the ground-state structural, electronic, thermodynamic, and elastic properties of NiTi in low-temperature (martensitic) phase. The target properties include elastic constants, phonon spectra calculations, and vacancy formation energy. Specifically, the stress-strain method was employed to compute the full tensor of the second- order elastic constants and assess the mechanical stability of the studied phases, ensuring that the results are consistent with those obtained using other established potential.

Speaker: Petr Sestak (Academy of Sciences of CZ, Institute of Physics of Materials)

18:46 Tuning the atomic scale friction of doped transition metal dichalcogenides heterostructures: First steps

Transition metal dichalcogenides (TMDs) are a class of layered materials in which weak interlayer van der Waals forces enable facile sliding, giving rise to an exceptionally low coefficient of friction which effectively vanishes (< 10$^{-3}$) when measured in vacuum. In humid environments, however, frictional properties degrade due to oxidisation at the sliding interface, which increases friction (~10$^{-1}$) and accelerates wear. Such degradation can be mitigated through doping with selected cations or anions, and/or by employing multilayer heterostructure design, both of which increase hardness and thus enhance the performance and lifetime of the low-friction state. Despite these advances, a fundamental understanding of the atomic-scale mechanisms that govern friction in TMDs remains limited, hindering the rational design of lubricants with tailored properties. To tackle this issue, our plan is to systematically investigate doped and heterostructured TMDs to understand how frictional energy dissipation arises from phonon-phonon scattering, and subsequently how it can be tuned via materials design. Here, we present the first steps along this path: we setup and benchmark simulations of phonon modes and phonon-phonon scattering in doped and heterostructured TMDs using a machine-learning approach. The accuracy of the method is demonstrated through convergence of lattice thermal conductivity calculations, providing a validated framework that will later be combined with analysis of phonon-phonon scattering to link atomic-scale phonon-based energy dissipation to macroscopic tribological performance.

This work is co-funded by the European Union under the project “Robotics and advanced industrial production” (reg. no. CZ.02.01.01/00/22_008/0004590), the Czech Science Foundation project “Superlubricity: Sliding of 2D Materials” No. 23-07785S, and by the Ministry of Education, Youth and Sports of the Czech Republic through e-INFRA CZ: (ID: 90254).

Speaker: Mr Elliot Perviz (Czech Technical University)

18:47 Preliminary ERO2.0 erosion and transport simulations of tungsten impurities for COMPASS Upgrade

COMPASS Upgrade will be a new tokamak-type device in Prague that is used to maintain plasma reaching extreme temperatures of several keV using magnetic fields of several teslas. Tokamaks (and other devices) aim to enable the possibility of clean, globally available and almost inexhaustible nuclear fusion energy. However, there are many obstacles in the way. This work focuses on erosion of the reactor's tungsten heat shield under the flux of energetic plasma particles and subsequent contamination of plasma by the shield material.

Simulations were performed with the 3D Monte Carlo erosion & particle tracking code ERO2.0 assuming preliminary plasma background data from the fluid code SOLPS-ITER. Although the results show tungsten concentration values ​​safely below 1e16 (the limit for fusion reactor operation), with a lack of reliable input data, the work further addresses the sensitivity to the input data and the possibilities for further tungsten concentration reduction. It is worth noting that even 1 mm of geometry adjustments can reduce the tungsten concentration by a factor of 2, which can be vital for a reactor.

Speaker: Samuel Lukeš (Institute of Plasma Physics of CAS, Prague, Czech Republic)

18:48 Spin-lattice model for computational study of elastic response to ultrafast demagnetization in fcc Ni

Advances in the development of ultrafast lasers have made picosecond ultrasonics a novel research field which alows to generate and detect acoustic waves with frequencies up to terahertz and wavelengths down to nanometers. Picosecond laser ultrasonics is successfully applied for experimentally study of nanostructures, adhesion of monolayers, and profile inhomogeneity [1, 2]. However, due to the complexity of studied phenomena, impressive experimental achievements [3, 2] face limited theoretical descriptions and modeling of the laser-induced elastic response. The situation becomes even more crucial in the case of ferromagnetic materials due to the simultaneous magnetic degree of freedom interaction with the laser radiation and coupling with the elastic degree of freedom. In other words, due to the laser-induced ultrafast demagnetization process, the magnetic subsystem can provide its own contribution to the elastic stress.

In our work we perform an approach to atom-resolved study on the basis of atomistic spin-lattice simulations for laser-induced elastic response in the feromagnetic fcc Ni, which alows us both to calculate the lattice elastic response including ultrafast thermal expansion and to characterize the magnetic contribution to stress in this material [4]. Among the advantages of the proposed 3D atomistic model is the ability to create close to realistic structures with atomic resolution, defects, interfaces, given crystal orientations in layers, and the possibility to obtain the full strain and stress tensor components, which makes the proposed theoretical approach useful for further interpretations of experiments in the picosecond ultrasonics, as well as for providing other required parameters (like ultrafast thermal expansion coefficient) in micromagnetic models of picosecond timescale.

Acknowledgement:
This work was supported by the projects e-INFRA CZ (ID No. 90254) and QM4ST (No. CZ.02.01.01/00/22_008/0004572) by the Ministry of Education, Youth and Sports of the Czech Republic, and also by Czech Science Foundation of the Czech Republic by Grant No. 22-35410K. P.N. acknowledges support by Grant No. MU-23-BG22/00168 funded by the Ministry of Universities of Spain. A.F. acknowledges funding from the Spanish Ministry of Science and Innovation (Grants No. PID2022-139230NB-I00 and No. TED2021-132074B-C32) the Diputación Foral de Gipuzkoa (Project No. 2023-CIEN-000077-01). Research was conducted in the scope of the Transnational Common Laboratory (LTC) Aquitaine-Euskadi Network in Green Concrete and Cement-based Materials. R.I. acknowledges financial support from the project MAGNES funded by the Principality of Asturias Government (Grant No. AYUD/2021/51822) and from the project RADIAFUS V, funded by the Spanish Ministry of Science and Innovation (Grant No. PID2019-105325RB-C32).

References:
[1] T. Saito, O. Matsuda, and O. B. Wright, Phys. Rev. B 67, 205421 (2003).
[2] M. Mattern, A. von Reppert, S. P. Zeuschner, et al.,  Photoacoustics 31, 100503 (2023).
[3] O. Matsuda, M. C. Larciprete, R. Li Voti, and O. B. Wright, Ultrasonics 56, 3 (2015).
[4] I. Korniienko, P. Nieves, A. Fraile, R. Iglesias, D. Legut, Phys. Rev. Research 6, 023311 (2024).

Speaker: Ievgeniia Korniienko (IT4Innovations)

18:49 Modular Multiscale Approach (MMA) to simulate high harmonic generation in gases and applications

We are developing an open-source numerical computational suite (https://github.com/HHG-modelling/MMA-HHG-pre-release-2) for modelling high-harmonic generation (HHG) in gaseous media. This process provides a tuneable source of highly coherent XUV radiation, forming attosecond pulses that allow the probing of atomic and molecular processes (such as chemical reactions) at their natural temporal scales. Our goal is to establish an open-source community codebase that will become standard in the field, unify various approaches, and provide interfaces and data standards. The initial release, which is now under finalisation, describes the multiscale problem of HHG in the gas phase in its full complexity: the macroscopic scale for field description and the microscopic scale for laser–matter interaction. The development is motivated and accompanied by laboratory applications.

The nature of the problem allows us to pipeline the subtasks: (a) the propagation of the driving laser (unidirectional pulse propagation), (b) the time-dependent Schrödinger equation (TDSE) for the microscopic response, and (c) the diffraction-integral approach to aggregate the macroscopic XUV field from the microscopic emitters. Because the numerical computations are expensive, all modules use MPI or multithreading parallelisation. The main computational routines are written in low-level C/Fortran, complemented by Pythonic high-level routines for pre- and post-processing. The computational interface is the HDF5 data structure, which reduces data handling for visualisation and post-processing. For example, the phase and intensity profiles of the incident laser might be masked by phenomenological semi-classical models approximating the most probable quantum trajectories, providing an estimate of the XUV signal. Complementary to the main pipeline, all the modules are available as stand-alone applications. Namely, the pulse propagation can be used to study, for example, laser filamentation; the TDSE is available as a dynamic library for Python; and the computation of the XUV diffraction integral comes with its own interface. The recent developments allow multiplatform deployability (Intel/GNU/AppleClang) on various supercomputers, now moving towards IT4I as one of the main computational resources for the developers’ team. Pythonic interfaces make the computational routines easy to use and present the data in an understandable form. The code is accompanied by Jupyter tutorials to make it more user-friendly.

The initial deployment of the code was tested during Project FTA-25-17. The official release, accompanied by a publication, is under preparation. Moreover, the model already provides fertile ground for various applications. For example, completed work includes optimisation of the HHG source in a pre-ionised gas (https://doi.org/10.1038/s41598-022-11313-6), and ongoing work concerns a monochromatising scheme investigated in Project OPEN-34-87 related to GAČR 25-17853S.

The contribution will provide a general overview of the code suite and then present various selected applications.

Speaker: Jan Vábek (ELI Beamlines Facility, The Extreme Light Infrastructure ERIC, Czech Republic; Czech Technical University in Prague, Faculty of Nuclear Sciences and Physical Engineering, Czech Republic)

18:50 Gating of the ribosomal exit tunnel

Ribosomes are large ribonucleoprotein complexes that drive protein synthesis, a fundamental process for all living systems. During translation, the nascent polypeptide traverses the ribosomal exit tunnel, a narrow passage shaped by both ribosomal RNA and proteins. The tunnel's architecture influences nascent chain folding, interactions, and even stalling (Kolář et al., 2024). The narrowest region of the tunnel, the constriction site (CS), is formed by the extended loops of proteins uL4 and uL22. We focus on the earliest stages of translation, when the nascent peptide has not yet passed the CS. The molecular details of how the peptide interacts with the CS and how the CS responds remain unclear. To address this, we performed all-atom molecular dynamics simulations of the entire ribosome with polyalanine and polyglycine peptides of varying lengths, as well as without a peptide. Our results show that the CS is highly flexible and can close to about 0.4 nm. It is narrowest in the absence of a peptide and, when a peptide is present, exhibits a preference for interaction with uL22. These findings provide new insight into the earliest events of protein biogenesis.

Kolář, M. H., McGrath, H., Nepomuceno, F. C., & Černeková, M. (2024). Three stages of nascent protein translocation through the ribosome exit tunnel. Wiley Interdisciplinary Reviews - RNA, 15(6). https://doi.org/10.1002/wrna.1873

Speaker: Hugo McGrath (UCT Prague)

18:51 Magnetoelastic Properties in MnPt magnetic systems

Magnetic materials represent a key component in the industry. One of the most common and used magnetic material parameters is magneto-crystalline anisotropy, which takes place in permanent magnets, storage devices, etc. In addition, magnetoelastic behavior is widely used in many applications, such as acoustic actuators, transducers, or sensors, providing desirable fast response and high efficiency. The study of this phenomenon is substantial and engaging from the point of view of material design.

Since magnetoelastic behavior of cubic transition metallic systems is weak, we are focusing on the Pt-based tetragonal system studying from first principles the impact of the magnetic ordered structure on the magnetoelastic coefficients. We show a substantial effect on the magnetoelastic properties including the volume and anisotropic magnetostriction depending on the assumed magnetic structure.
The obtained calculated results are used to explained experimentally measured magnetoelasic behavior.

Speaker: Dr Jakub Šebesta (IT4Innovations, VSB - Technical University of Ostrava, 70800 Ostrava-Poruba, Czech Republic)

18:52 Machine Learning Based Interatomic Potential for Martensitic NiTi

Machine learning based potentials represent a new powerful tool for accurate modelling of complex interatomic interactions in materials. In our work, we apply the high-dimensional neural network (HDNNP) methodology using the atom-centred symmetry function descriptors for the shape memory alloy NiTi in the martensitic phase.

NiTi is a system that poses significant challenges for both experimental and theoretical modelling, and an accurate atomistic scale model is required to fully understand all of it's mechanical properties. Recent experimental results from thermomechanical testing and microstructural analysis of polycrystalline NiTi shape memory alloys have provided a qualitatively new picture of the plastic deformation mechanisms of this most common shape memory material. In particular, they challenge the widely accepted paradigm that austenite is a phase more susceptible to dislocation slip and plastic deformation in NiTi than B19' martensite, despite the fact that only a single 100M dislocation slip system is observed in martensite. The simultaneous activation of 100M plastic slip and reorientation of martensite between two variants sharing the (010)M plane results in kink banding and twinning, forming the so-called "kwink bands.

We show that our newly developed interatomic is suitable for simulating the evolution of the martensitic microstructure, is close in accuracy to density functional theory, preserves the strong plastic slip anisotropy and has all the prerequisites to simulate the new kwink mechanism.

Speaker: Petr Jaroš (Ústav termomechaniky AVČR v.v.i.)

18:53 ElaStr: An Online Tool Linking Elastic Properties and Crystal Structures

Many existing tools analyze anisotropic elastic properties using the elastic stiffness tensor, but they often overlook the direct relationship between crystal structure and elastic behavior. Most operate in Cartesian coordinates, which work well for high-symmetry lattices but are less intuitive for lower-symmetry systems where crystallographic directions and planes—defined by Miller indices—are more suitable. For example, in tetragonal structures, the [101] direction is not perpendicular to the (101) plane, and the deviation depends on the c/a ratio.

ElaStr is a user-friendly online platform that bridges this gap by combining elastic stiffness tensor data with crystal structure information. It supports input from both computational and experimental sources. Users can provide an elastic tensor together with lattice vectors or a standard structure file, and ElaStr will compute elastic properties along chosen crystallographic directions or normal to selected planes. The tool evaluates properties such as Young’s modulus, linear compressibility, shear moduli, and Poisson’s ratios.

When a selected direction lies within a given plane, ElaStr also determines shear modulus and Poisson’s ratio for that specific combination. Beyond single-direction analysis, the platform can map these properties for all directions within a plane, providing intuitive polar plots as well as crystal-structure visualizations. For this purpose, ElaStr generates structure files that can be directly opened in the VESTA program, allowing users to explore elasticity in the context of the atomic arrangement.

ElaStr is freely available online at: https://elastr.fme.vutbr.cz, with all analyses performed through a simple web interface. It is designed to support both teaching and research by making complex elasticity–structure relationships more transparent and accessible.

Speaker: Martin Zelený (Faculty of Mechanical Engineering, Brno University of Technology)

18:54 Numerical study of oscillations in solar prominence threads excited by vortex shedding

We study transverse (kink mode) oscillations in threads of solar prominences, which are large dense, cool plasma structures suspended in the Sun’s corona by magnetic fields. These threads can show both collective and individual behaviour, when threads or groups of threads oscillate with their own periods independently from the rest of the prominence.

Several studies have suggested that these and some other oscillatory phenomena in the solar atmosphere may be explained by the phenomenon of vortex shedding, a process well known in classical fluid dynamics in which alternating vortices are shed from a body immersed in a flowing fluid. This phenomenon has been widely studied in hydrodynamics but has not yet been satisfactorily investigated in magnetohydrodynamic conditions, such as in the solar atmosphere.

We study prominence oscillations in association with vortex shedding numerically by the magnetohydrodynamic approach in three dimensions. We model the structures as groups of flexible cylindrical bodies. Specifically, we ran a simulation with a group of three vertically aligned cylindrical bodies, and three simulations with two horizontally aligned cylindrical bodies at varying distances to evaluate the influence of separation distance.

Speaker: Ms Sofya Belov (University of South Bohemia in České Budějovice, Faculty of Science, Department of Physics)

18:55 The role of emissions in future Central European air-quality scenarios

The aim of this study is to assess how changes in emissions can affect future air-quality in Central Europe, without consideration of changes in future meteorology conditions. In order to do so, two future scenarios were used (RCP4.5 and RCP8.5) for the periods 2026-2035 and 2046-2055. We simulated the current and future air pollution levels using the Weather Research and Forecasting model coupled with chemistry (WRF-Chem) version 4.4 and the Comprehensive Air Quality Model with Extensions (CAMx) version 7.20.
So far, we have performed five simulations: one with present-day conditions (2010-2019) and four with future emission scenarios. Due to the large computing time, these four future simulations were run in the Karolina CPU after obtaining computing time in the 29th Open Access Competition, with the proposal title “Future air-quality changes over Europe driven by emissions”.
The present-day conditions were validated by comparing the model output with observed data for key pollutants, such as oxides of nitrogen, ozone and particulate matter. This validations points us to limitations in the models -such as incorrect monthly and hourly emission profiles, among others- that helps in comparing present-day and future emissions simulations.
To analyse the contributions of emission changes in air-quality, we computed several pollution indicators and seasonal averages for both present-day and future emissions simulations, and we computed the relative differences of these indicators. The results of this analysis shows that, in general, both seasonal averages and other indicators decrease in both RCP scenarios. Although there are intrinsic uncertainties to the modelling process, this research can help us understand how reductions in emissions can aid in improving air-quality.

Speaker: Alvaro Patricio Prieto Perez (Charles University)

Friday, 31 October

09:00 → 09:30 Keynote III

09:00 Bridging Cosmological Simulations and High-Quality Visualizations using HPC resources

Modern cosmological simulations produce large and complex data sets, often containing tens of billions of particles spread across terabytes of files. Although these simulations are essential for deepening our understanding of galaxy formation, large-scale structures, and cosmic evolution, their scientific impact is limited without effective methods of visual analysis and communication. Bridging the gap between raw numerical outputs and high-quality film visualizations remains a significant challenge, especially when working in high-performance computing (HPC) environments.

In this talk, we will present an integrated workflow and toolkit designed to make cosmological data more accessible, interactive, and visually appealing. Our workflow focuses on three new components: BSpace, a Blender add-on that provides a user-friendly interface for interacting with astrophysical datasets; CyclesPhi, a modified version of the Blender Cycles renderer optimized for HPC clusters; and SpaceConverter, a standalone software application for converting particle-based outputs from simulation codes such as HACC, ChaNGa, and OpenGADGET into volume formats suitable for high-quality rendering.

The BSpace add-on allows users to connect Blender directly to local or remote HPC resources, select regions of interest, configure parameters, and import volumetric data files into Blender scenes for inspection and animation. Renderer CyclesPhi extends the Cycles engine, enabling distributed path tracing across multiple GPUs, and integrates seamlessly with BSpace for interactive remote rendering. The SpaceConverter application forms the backbone of the workflow, offering efficient parallel voxelization and conversion of SPH and N-body simulation data into OpenVDB or NanoVDB volumes. Its design supports both offline command-line processing and interactive integration with Blender, allowing researchers to balance automation, flexibility, and performance.

We demonstrate the ability to generate visually striking renderings of gas flows, dark matter structures, star distributions, and galactic magnetic fields, emphasizing both scientific detail and cinematic clarity. By combining robust HPC data processing with artist-friendly visualization tools, this framework lowers the barriers for astrophysicists to create publication-quality images and animations, while opening avenues for broader public engagement in cosmological research.

Speaker: Petr Strakos (IT4Innovations)

09:30 → 10:10 Users' Talks IV

09:30 CrSBr-MoS$_2$ Heterostructures and the Limits of VASP Parallelization

CrSBr-MoS$_2$ heterostructures combine the properties of the antiferromagnetic semiconductor CrSBr and the TMD semiconductor MoS$_2$, yielding a promising platform for spintronic and optoelectronic applications. In connection with experimental work, we present a comprehensive computational investigation of this specific heterostructure. Using density functional theory (DFT) with both PBE and hybrid HSE06 functionals, we explore the structural, electronic, magnetic and optical properties of CrSBr-MoS2 systems, where CrSBr is 3-4 layers thick and MoS$_2$ a monolayer. We observe induced magnetization in the MoS2 layer via proximity effect, strongly dependent on the interlayer distance. The density of states and band structure analysis indicates that the heterostructure has the properties of a small bandgap semiconductor in the ground state, but the density of states analysis indicates a weakly coupled system, retaining the characteristics of the separate parts. The calculation of the absorption coefficient via the independent particle approximation (IPA) reveals distinct peaks corresponding to the CrSBr and MoS$_2$ layers, in qualitative agreement with the experimental photoluminescence spectrum. Our calculations exploit the full potential of the parallelization capabilities of VASP installed in the Karolina HPC cluster (up to ~2500 CPU cores) and show the computational limits of VASP for large systems involving demanding jobs like band structure, hybrid functionals and optical properties.

Speaker: Athanasios Koliogiorgos (Charles University)

09:50 Nanowear in molybdenum disulfide studied by molecular dynamics simulations

Molybdenum disulfide (MoS$_2$) is a layered material that has been used as a solid lubricant for decades[1]. The experimental verification of superlubricity in the early nineties for crystalline MoS$_2$ coatings in ultrahigh vacuum conditions[2] sparked again the interest for such a material. Since then, a lot of effort has been devoted in order to understand (and possibly control) the complex phenomena taking place during sliding in such a system, e.g. the formation of crystalline layers from molecular precursors[3] or from amorphous material[4]. On the contrary, the understanding of abrasive wear processes on such a material is still in an early stage.
In this study[5], we performed molecular dynamics (MD) simulations in order to shed light on the elementary steps in the wear process of MoS$_2$ with atomistic detail. A rigid diamond tip has been used to indent a six-layer thick system, and then the tip has been dragged in order to scratch the material. The position and the force acting on the tip has been followed for more than 20 ns, together with the damage done on the substrate. Different MoS$_2$ orientations and normal loads have been considered. The trajectory analysis revealed that the tip moved in a stick-slip fashion, somehow recalling the dynamics that takes place during sliding, but with the slip events occurring on a much longer length scale. Remarkably, it turned out that about only one fifth of the input energy is irreversibly spent in surface damage. The system has been studied also with atomic force microscopy (AFM) experiments, where two different regimes were identified in terms of distributions of the force drop during the slip phase: a generalized extreme value distribution (at high loads, also observed in MD simulations), and a Gaussian distribution (at low loads). These findings provide new insight into the fundamental mechanisms of friction and wear in layered materials and establish a framework for precision nanomachining of van der Waals solids, relevant for next-generation devices at sub-micrometer scales.

References
[1] T. Spalvins, J. S. Przybyszewski, NASA Technical Note D-4269 (1967).
[2] J. M. Martin et al., Phys. Rev. B, 48, 10583 (1993).
[3] S. Peeters et al., Appl. Surf. Sci., 606, 154880 (2022).
[4] P. Nicolini et al., ACS Appl. Mater. Interfaces, 10, 8937 (2018).
[5] P. Koczanowski et al., currently under review in Small (2025).

Speaker: Dr Paolo Nicolini (FZU - Institute of Physics of the Czech Academy of Sciences)

10:10 → 10:35 Coffee break

10:35 → 11:15 User's Talks V

10:35 Antiferromagnetic exchange coupling across transition metal films with magnetic elements

Magnetic materials represent a key component in the industry. One of the most common and used magnetic material parameters is magneto-crystalline anisotropy, which takes place in permanent magnets, storage devices, etc. In addition, magnetoelastic behavior is widely used in many applications, such as acoustic actuators, transducers, or sensors, providing desirable fast response and high efficiency. The study of magnetic exchange interactions is substantial and engaging from the point of view of material design. Interlayer exchange coupling has been intensively investigated for over thirty years and is incorporated in almost all spintronic devices, e.g. into the hard data disk storages. However in past, research was focused on the coupling between magnetic layers across a nonmagnetic spacer layer. Here we find that antiferromagnetic interlayer exchange coupling can be achieved across spacer layers containing over sixty atomic percent of magnetic materials. X-ray magnetic circular dichroism measurements reveal that the magnetic atoms in the spacer layer have a large magnetic moment. Magnetic atoms in the spacer layer can double the coupling strength, leading to the largest antiferromagnetic bilinear coupling observed in magnetic multilayers deposited with the industrial technique of choice, magnetron sputtering. Electronic structure calculations predict a dependence of interlayer exchange coupling on both the magnetic material concentration and the spacer layer thickness. The DFT calculations also offer insight into the role of magnetic atom interactions within the spacer layer in influencing interlayer exchange coupling, the role of the interface as well as the role of various magnetic transitional metals in spacer[1-2].

Acknowledgement:
The authors acknowledge financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC), the Czech Science Foundation, Project No. 22-35410K and the projects e-INFRA CZ (ID:90254)
and QM4ST No. CZ.02.01.01/00/22_008/0004572 by the Ministry of Education, Youth and Sports of the Czech Republic.

References:
[1] K. Winther, Z. R. Nunn, J. Lisik, S. Arapan, D. Legut, F. Schulz, E. Goering, T. Mckinnon, S. Myrtle, and E. Girt, Physical Review Applied 22, 024058 (2024).
[2] S. Arapan, J. Priessnitz, D. Legut, A. Kovacs, H. Oezelt, D. Boehm, M. Gusenbauer, and T. Schrefl, "Effect of interface on magnetic exchange coupling in Co/Ru/Co trilayer: from ab-initio simulations to micromagnetics", https://doi.org/10.48550/arXiv.2501.08724
[3] Z. R. Nunn, J. Lisik, P. Omelchenko, S. Koraltan, C. Abert, D. Suess, and E. Girt, J. of Applied Physics 133, 123901 (2023).

Speaker: Dominik Legut (IT4I)

10:55 Fine-tuning stability of La-Ni-Sn-H materials for hydrogen storage

The La-Ni-Sn compounds are considered promising candidates for future hydrogen-storage applications. Some phases from the La-Ni-Sn system have been insufficiently studied so far, and some critically important data are missing. We have, therefore, employed quantum-mechanical calculations combined with experiments to assess the stability of several La-Ni-Sn-H materials. Our study covers (i) binary La2Ni7 as well as La5Ni19 in the hexagonal and trigonal structure, (ii) ternary LaNi5-based ternaries La(Ni,Sn)5 with several Sn concentrations and a partial sublattice disorder, as well as (iii) hydrogen-containing quaternaries HLaNiSn and H2LaNiSn. We found La2Ni7 stable thermodynamically and mechanically but not dynamically due to the presence of imaginary-phonon modes. The two La5Ni19 phases are also thermodynamically and mechanically stable, and their structural, electronic, thermodynamic, magnetic, and elastic characteristics are very similar. Regarding the LaNi5-based ternaries La(Ni,Sn)5, we identified a miscibility gap with respect to the tin content, which was confirmed experimentally. Lastly, the two hydrogen-containing quaternaries HLaNiSn and H2LaNiSn were computed thermodynamically, mechanically and dynamically stable featuring specific high-energy hydrogen-related phonon states.

Acknowledgments
This study is a part of the project No. CZ.02.01.01/00/22_008/0004631 entitled “Materials and technologies for sustainable development” within the Jan Amos Komensky Operational Program financed by the European Union and from the state budget of the Czech Republic. Computational resources were made available by the Ministry of Education, Youth and Sports of the Czech Republic under the Project e-INFRA CZ (ID:90254) at the IT4Innovations National Supercomputing Center, the MetaCentrum and CERIT-SC. Access to the CERIT-SC computing and storage facilities provided by the CERIT-SC Center, under the programme "Projects of Large Research, Development, and Innovations Infrastructures" (CERIT Scientific Cloud LM2015085) and access to CESNET storage facilities provided by the project “e-INFRA CZ” under the programme “Projects of Large Research, Development, and Innovations Infrastructures” LM2023054) is acknowledged.

Speaker: Martin Friák (Institute of Physics of Materials, Czech Academy of Sciences, Brno, Czech Republic)

11:15 → 12:00 From Supercomputing to Quantum Computing: Introduction of VLQ

Convener: Branislav Jansik (IT4Innovations)

12:00 → 13:00 Lunch