Cyfronet supercomputers support scientific discoveries of international importance

Publication date: August 4, 2026

30 scientific articles and 9 conference presentations in seven months. They share the highest ranking on the ministerial list and the utilization of computational resources provided by ACC Cyfronet AGH within the PLGrid infrastructure.

Each year, Polish research teams publish results in journals and at conferences with high international prestige. The implementation of many of these projects requires access to advanced computational infrastructure, specialized software, and expert support.

This article presents an overview of publications and conference presentations that:

• obtained 200 points according to the ministerial list,
• were created using supercomputers or software provided by ACC Cyfronet AGH within PLGrid,
• were published between 1 January and 30 July, 2026.

From quantum physics to artificial intelligence

The analyzed achievements cover various research areas.

Molecular and Quantum Physics

Papers published in, among others, Physical Review Letters and Physical Review X concern ultracold molecules, ion-atom collisions, and phenomena beyond the Standard Model. The ab initio calculations and quantum dynamics simulations used in them require the capability for parallel execution of complex computations.

Chemistry and Functional Materials

Publications in Angewandte Chemie, Chemical Science, and Small describe, among others, covalent organic frameworks, luminescent ferroelectrics, MXene materials, and catalysts. DFT simulations play a crucial role in them, allowing the correlation of the electronic structure of materials with their properties.

Engineering and Industrial Processes

Research described in Tribology International and Chemical Engineering Journal combines numerical modeling with experiments. This approach allows explaining process mechanisms, evaluating various technological variants, and reducing the number of costly tests.

Life and Environmental Sciences

Works from Nucleic Acids Research, PLOS Biology, and Journal of Hazardous Materials concern, among others, tRNA structure, enzymatic CO₂ reduction, and the interaction of nanoplastics with pulmonary surfactant. Molecular dynamics enable observing the behavior of atoms and molecules over time and analyzing processes inaccessible to direct experimental observation.

Artificial Intelligence and Machine Learning

Publications from the ICLR, AAAI, and EACL conferences cover research on language models, text generation, and the explainability of chemical models. Machine learning is also used to accelerate classical computational methods.

Training and comparing models require access to GPU accelerators. They enable the evaluation of multiple configurations and a more precise study of the effectiveness and limitations of developed solutions.

Computation as part of the research method

The role of computational infrastructure is not limited to faster execution of the same operations. Simulations allow predicting material properties, explaining experimental results, identifying reaction mechanisms, tracking the behavior of atomic systems, and training machine learning models.

In the study of nanoplastics, molecular dynamics enabled the observation of changes occurring within the surfactant model. In works concerning light conversion and two-dimensional materials, calculations allowed correlating atomic structure with optical and spectroscopic properties. Modeling clay extrusion, on the other hand, helped explain the causes of tool wear and indicate possible design changes.

HPC infrastructure thus becomes part of the research method itself. It influences not only the pace of project implementation but also the range of questions scientists can ask.

Significance of the presented research

Nature Communications, Physical Review X, Angewandte Chemie International Edition, and Journal of the American Chemical Society belong to prestigious journals in the fields they represent. Recognized scientific conferences also play an important role in the field of artificial intelligence. The presence of Polish teams in these communication channels increases the visibility of their research in the international circuit.

Many works address issues of broad social, environmental, and economic significance. These include, for instance, CO₂ reduction, designing materials for energy conversion, the impact of nanoplastics on the respiratory system, or the development of more efficient industrial processes.

The discussed publications often arise in collaboration with foreign institutions. Access to supercomputing infrastructure enables Polish teams to contribute to such partnerships by providing the capability to perform demanding computations.

The role of ACC Cyfronet AGH infrastructure

ACC Cyfronet AGH, as a key operator of the national PLGrid infrastructure, provides resources necessary for conducting advanced research. The Ares, Athena, and Helios supercomputers support projects in the fields of physics, chemistry, engineering, life sciences, and artificial intelligence.

The significance of the infrastructure is not limited to computational power. Researchers also receive access to:

• data storage space,
• specialized scientific software,
• CPU processors and GPU accelerators,
• technical support and consultations,
• training to help efficiently utilize resources.

Making this infrastructure available to teams from various centers allows for more efficient use of resources that independent maintenance remains beyond the reach of many institutions. This is particularly significant when considering not only the high costs of purchasing and operating supercomputers but also the necessity of ensuring appropriate technical background (supporting infrastructure) and competency background (engineers, technicians, administrators, software specialists).

Results from the first seven months of 2026 show that HPC infrastructure is an important element of the competitiveness of modern science. The diversity of fields, presence in prestigious journals and at recognized conferences, and the international nature of projects confirm the active participation of Polish teams in the global scientific circuit.

Supercomputers are not merely technical research support. Increasingly, they become one of its fundamental tools, just like laboratories, measuring apparatus, and access to scientific literature. They allow testing hypotheses, analyzing complex systems, processing large datasets, and conducting computational experiments on a scale inaccessible to standard computers.

List of analyzed achievements

1. Aminal-linked Covalent Organic Frameworks for Light Energy Upconversion
2. Structural Insights Into CO2 Transport Pathways in a W-Formate Dehydrogenase: Structural Basis for CO2 Reduction
3. Green-Emissive Ferroelectric Optical Thermometer Based on Cyclometalated Dicyanidoplatinate(II) Ions
4. Laser-Induced-Structural Transformation in Ti3CNTx MXene Monitored by Raman Spectroscopy with DFT Insight
5. Harnessing inactive molybdenum species in ill-defined catalyst to work in propylene metathesis by propane pretreatment
6. Interactions of ultra-fine polystyrene nanoparticles with lung surfactant monolayers and bilayers: A combined molecular dynamics and experimental study
7. Analysis of selected aspects of improving the durability of ceramic roof tile forming tools with consideration of numerical modeling
8. Charge Exchange Dynamics in Cold Collisions of 40CaH+ and 39K
9. Adsorption Hysteresis Under Control: Tuning Host–Guest Interactions via a Genetic Algorithm
10. Charge Shift in Calcite before High-Pressure Phase Transition
11. A single viral enzyme drives tRNA-dependent hypermodification of DNA at adenine
12. A practical computational protocol for photocatalytic reactions beyond ground-state approximations
13. Taming boroloborinines: toward photostable polycyclic antiaromatic hydrocarbons
14. Gradients not needed: ML-driven propagation of nonadiabatic molecular dynamics without reference gradients
15. Merging platinahelicene and nanographene: a strategy for circularly polarized phosphorescence in the near-infrared (NIR)
16. Origin of class B J-domain proteins involved in amyloid transactions
17. Solid-state nanopore sensing reveals conformational changes induced by a mutation in a neuron-specific tRNAArg
18. Generalized Gross-Pitaevskii Equation for 2D Bosons with Attractive Interactions
19. Optical Excitation and Stabilization of Ultracold Field-Linked Tetratomic Molecules
20. Extremely High Excitonic g Factors in 2D Crystals by Alloy-Induced Admixing of Band States
21. Ultracold High-Spin ?-State Polar Molecules for New Physics Searches
22. Vehicle Classification Based on Multi-Frequency Impedance Magnetic Profiles in Distance Domain
23. Capsular specificity in temperate phages of Klebsiella pneumoniae is driven by diverse receptor-binding enzymes
24. Enhancing Chemical Explainability Through Counterfactual Masking
25. Teaching Small Language Models to Learn Logic through Meta-Learning
26. One-step Nonautoregressive Natural Language Generation with Shortcut Flow Matching Models
27. There and back again: On the Relation between Noise and Image Inversions in Diffusions Models
28. Universal Properties of Activation Sparsity in Modern Large Language Models
29. Attention, May I Have Your Decision? Localizing Generative Choices in Diffusion Models
30. Beyond Classification: Continual Learning for Multimodal Retrieval
31. Universality in Ionic Three-Body Systems Near an Ion-Atom Feshbach Resonance
32. Reducing a lift-off distance of a nitrogen-diluted hydrogen flame evolving in a dry and humidified ambient flow through suction-driven global instability: Insights from LES
33. hexABC Seeking the Physical Code of DNA
34. Lattice Reconstruction Strategy for Fast-Charging Plateau-Type Hard Carbon Anodes in Ultra-Long-Life Sodium-Ion Batteries
35. ZEUS: Zero-shot Embeddings for Unsupervised Separation of Tabular Data
36. Is a Document Educational or Just Wikipedia-Style? — Pitfalls of Classifier-Based Quality Filtering
37. Methane Capture from Mine Ventilation Diffusers for Improved Methane Utilization and Reduced Emissions
38. Efficient LLM Moderation with Multi-Layer Latent Prototypes
39. On the Role of Computation in Reinforcement Learning