TY - INPR A1 - Naik, Aakash A. A1 - Dhamrait, Nidal A1 - Ueltzen, Katharina A1 - Ertural, Christina A1 - Benner, Philipp A1 - Rignanese, Gian-Marco A1 - George, Janine T1 - A critical assessment of bonding descriptors for predicting materials properties N2 - Most machine learning models for materials science rely on descriptors based on materials compositions and structures, even though the chemical bond has been proven to be a valuable concept for predicting materials properties. Over the years, various theoretical frameworks have been developed to characterize bonding in solid-state materials. However, integrating bonding information from these frameworks into machine learning pipelines at scale has been limited by the lack of a systematically generated and validated database. Recent advances in high-throughput bonding analysis workflows have addressed this issue, and our previously computed Quantum-Chemical Bonding Database for Solid-State Materials was extended to include approximately 13,000 materials. This database is then used to derive a new set of quantum-chemical bonding descriptors. A systematic assessment is performed using statistical significance tests to evaluate how the inclusion of these descriptors influences the performance of machine-learning models that otherwise rely solely on structure- and composition-derived features. Models are built to predict elastic, vibrational, and thermodynamic properties typically associated with chemical bonding in materials. The results demonstrate that incorporating quantum-chemical bonding descriptors not only improves predictive performance but also helps identify intuitive expressions for properties such as the projected force constant and lattice thermal conductivity via symbolic regression. KW - Bonding Analysis KW - Machine Learning KW - Symbolic Regression KW - Chemical Understanding KW - Phonons KW - Thermal Conductivity PY - 2026 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-655150 DO - https://doi.org/10.48550/arXiv.2602.12109 SP - 1 EP - 28 PB - Cornell University CY - Ithaca, NY AN - OPUS4-65515 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - Robust Data Generation, Heuristics and Machine Learning for Materials Design N2 - My talk covered, among other things, robust data generation for machine learning. It showed how heuristics can be used within machine learning models and how they might also be extracted from machine learning models. Beyond this, I showed an automated pipeline for training machine learning potentials. T2 - Seminar Gruppe Stephan Roche CY - Barcelona, Spain DA - 22.01.2026 KW - Automation KW - Machine Learning KW - Materials Acceleration Platforms KW - Thermal Conductivity KW - Phonons KW - Bonding Analysis PY - 2026 AN - OPUS4-65428 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Ueltzen, Katharina T1 - Can simple exchange heuristics help us in predicting magnetic properties of solids? N2 - Environmental and scarcity issues of common functional magnetic materials for, e.g., permanent magnets have intensified the search for rare-earth-free alternatives. This challenge is increasingly met by machine learning of magnetic properties of transition-metal compounds. Surprisingly, bond-angle-derived features were not found to be relevant for magnetic structure prediction in previous studies using DFT-computed labels. This contrasts with the Kanamori-Goodenough-Anderson (KGA) rules of superexchange, present in every magnetism textbook. These semiempirical rules predict whether a nearest-neighbor magnetic interaction in insulators is FM or AFM based on the bond angle, orbital symmetry, and orbital occupancy. For some cases, the rules can be simplified further to only consider the bond angle of neighboring magnetic sites (KGA rules of thumb). We review magnetism—bond angle trends within the MAGNDATA database, the largest collection of experimentally determined magnetic structures. Observed trends follow the KGA rules of thumb, and exceptions can be rationalized. In contrast, bond angles in a popular theoretical DFT database show very different trends and do not depend on the magnetic ordering. Building on our analysis, we engineer heuristic-derived features for the machine learning of magnetic structures. We introduce a new, informative label for predicting magnetic structures that can be extended to magnetic sites and structures of arbitrary complexity. We show that features derived from the heuristic are of high importance for this machine learning task. Beyond this, our model enables the prediction of non-collinear magnetic structures. T2 - AI MSE 2025 CY - Bochum, Germany DA - 18.11.2025 KW - Magnetism KW - Machine Learning KW - Materials Design KW - Chemically Complex Materials PY - 2025 AN - OPUS4-64799 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Cakir, Cafer Tufan A1 - de Oliveira Guilherme Buzanich, Ana A1 - Radtke, Martin T1 - A Digital Twin Workflow for Optimizing X-ray Emission Spectroscopy (XES) Parameters in Material Science N2 - In this contribution, I present a digital twin-based workflow designed to optimize experimental parameters in X-ray emission spectroscopy (XES), with a focus on reproducibility, data integration, and alignment with FAIR (Findable, Accessible, Interoperable, Reusable) principles in materials science and engineering (MSE). The developed pipeline begins with automated retrieval of crystallographic information from the Materials Project database based on a given sample composition. This structural data is then used to simulate the corresponding XES spectra using FDMNES, allowing for accurate prediction of element-specific emission lines. The simulated emission lines are fed into an X-ray tracing (XRT) module, which builds a virtual replica of the experimental setup. This digital twin environment enables predictive modeling of spectrometer performance based on geometric configurations. The optimization focuses on two key parameters: the choice of the analyzing crystal and the distance between the sample and crystal (which also defines the crystal–detector distance due to Bragg condition constraints). An active learning algorithm is employed to iteratively adjust these parameters in order to achieve a desired energy-per-pixel (E/pixel) resolution with minimal intensity loss, enabling efficient, data-driven experimental planning. By integrating data-driven simulations with real-time optimization strategies, this workflow supports efficient experiment planning while minimizing resource consumption and human error. Furthermore, all stages of the process—from data collection and simulation to optimization and visualization—are structured to ensure traceability and interoperability, facilitating future reuse and collaborative research. I hope this contribution aligns well with the topic “Workflows for FAIR MSE Data” and offers a concrete example of how digital twins can be harnessed to improve the design, execution, and documentation of spectroscopy experiments in the MSE domain. T2 - Opportunities and Challenges of FAIR Data at Photon and Neutron Facilities CY - Bad Honnef, Germany DA - 12.10.2025 KW - Digital-Twin KW - XES KW - Optimization KW - Machine Learning PY - 2025 AN - OPUS4-64676 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - GEN A1 - Ueltzen, Katharina A1 - Naik, Aakash A1 - Ertural, Christina A1 - Benner, Philipp A1 - George, Janine T1 - Software and data repository: Can simple exchange heuristics guide us in predicting magnetic properties of solids? N2 - Software and data for the publication "Can simple exchange heuristics guide us in predicting magnetic properties of solids?" Release that corresponds to the first preprint version of the article. Full Changelog: https://github.com/DigiMatChem/paper-exchange-heuristics-in-magnetic-materials/commits/v1.0.0 KW - Magnetism KW - Machine Learning KW - Materials Design KW - Chemically Complex Materials KW - Sustainable Materials Design PY - 2025 DO - https://doi.org/10.5281/zenodo.16811104 PB - Zenodo CY - Geneva AN - OPUS4-64672 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - Robust data generation, heuristics and machine learning for designing sustainable materials N2 - Despite advances in computational materials design, simulating large systems—such as defects, interfaces, or amorphous states—with quantum-chemical accuracy remains a major challenge.[1] Machine learning (ML) methods are emerging as powerful tools to overcome these limitations, enabling scalable and accurate modeling beyond traditional quantum-chemical approaches.[2] They also open new avenues for discovering non-toxic, earth-abundant alternatives to existing materials and can be combined with self-driving labs. [3] There are nowadays robust data generation strategies that underpin the development and benchmarking of ML models. [4,5]atomate2 I will focus on such strategies for quantum-chemical bonding analysis and ML interatomic potentials in my talk. Quantum-chemical bonding descriptors can be effectively used in ML models to predict phononic properties. [6] ML interatomic potentials offer a powerful approach for predicting energies, forces, and stresses—but their performance hinges on high-quality training data. Our automated framework, autoplex, enables diverse and scalable training workflows, from random structure searches for general-purpose models to phonon-aware pipelines for high-accuracy predictions.[7] While quantum chemistry excels in many domains, properties like magnetism and synthesizability remain elusive. Here, heuristics or leveraging experimental data for ML offer promising alternatives.[8,9] T2 - Advanced Materials Safety 2025 CY - Dresden, Germany DA - 04.11.2025 KW - Nano Particles KW - Machine Learning KW - Automation KW - Materials Design KW - Sustainability KW - Material Safety PY - 2025 AN - OPUS4-64599 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - High-throughput Computational Screening With Chemical Heuristics, Workflows, and Machine Learning N2 - Within my talk, I have introduced the audience to high-throughput materials discovery with the help of workflow tools, machine learning and chemical heuristics. I started with a general introduction to the Materials Project software infrastructure and databases such as the Materials Project and NOMAD. Then, I dived into machine learning tasks relying on chemical bonding information, machine-learned interatomic potentials, and experimental data. Target properties of the machine-learning tasks were thermal properties, phonons, magnetism and synthesizability. T2 - Seminar at the Inorganic Chemistry Laboratory in Oxford CY - Oxford, United Kingdom DA - 25.09.2025 KW - High-throughput KW - Sustainability KW - Machine Learning KW - Materials Design KW - Thermal Conductivity KW - Magnetism KW - Synthesizability KW - Materials Acceleration Platform PY - 2025 AN - OPUS4-64217 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - Introduction to computational methods with periodic boundaries N2 - I provided an introduction to density functional theory (DFT) with periodic boundary conditions. I repeated the derivation of DFT and the Bloch theorem. Then, I combined both concepts. Additionally, I delivered an outlook to automatic DFT for solids and machine learning in the field. T2 - Computational Methods in Crystallography CY - Lausanne, Switzerland DA - 09.09.2025 KW - DFT KW - Automation KW - Machine Learning KW - Bloch Theorem KW - Materials Design PY - 2025 AN - OPUS4-64044 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - INPR A1 - Liu, Yuanbin A1 - Zhou, Yuxing A1 - Ademuwagun, Richard A1 - Walterbos, Luc A1 - George, Janine A1 - Elliott, Stephen R. A1 - Deringer, Volker L. T1 - Medium-range structural order in amorphous arsenic N2 - Medium-range order (MRO) is a key structural feature of amorphous materials, but ist origin and nature remain elusive. Here, we reveal the MRO in amorphous arsenic (a-As) using advanced atomistic simulations, based on machine-learned potentials derived using automated workflows. Our simulations accurately reproduce the experimental structurefactor of a-As, especially the first sharp diffraction peak (FSDP), which is a signature of MRO. We compare and contrast the structure of a-As with that of its lighter homologue, red amorphous phosphorus (a-P), identifying the dihedral-angle distribution as a key factor differentiating the MRO in both. The pressure-dependent structural behaviors of a-As and a-P differ as well, which we link to the interplay of ring topology and structural entropy. We finally show that the origin of the FSDP is closely correlated with the size and spatial distribution of voids in the amorphous networks. Our work provides fundamental insights into MRO in an amorphous elemental system, and more widely it illustrates the usefulness of automation for machine-learning-driven atomistic simulations. KW - Materials Design KW - Amorphous KW - Machine Learning KW - Machine Learning Interatomic Potentials KW - Automation PY - 2025 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-640204 DO - https://doi.org/10.48550/arXiv.2509.02484 SP - 1 EP - 43 AN - OPUS4-64020 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Zimmermann, Yoel A1 - Bazgir, Adib A1 - Al-Feghali, Alexander A1 - Ansari, Mehrad A1 - Bocarsly, Joshua A1 - Brinson, L Catherine A1 - Chiang, Yuan A1 - Circi, Defne A1 - Chiu, Min-Hsueh A1 - Daelman, Nathan A1 - Evans, Matthew A1 - Gangan, Abhijeet S A1 - George, Janine A1 - Harb, Hassan A1 - Khalighinejad, Ghazal A1 - Takrim Khan, Sartaaj A1 - Klawohn, Sascha A1 - Lederbauer, Magdalena A1 - Mahjoubi, Soroush A1 - Mohr, Bernadette A1 - Mohamad Moosavi, Seyed A1 - Naik, Aakash Ashok A1 - Ozhan, Aleyna Beste A1 - Plessers, Dieter A1 - Roy, Aritra A1 - Schoeppach, Fabian A1 - Schwaller, Philippe A1 - Terboven, Carla A1 - Ueltzen, Katharina A1 - Wu, Yue A1 - Zhu, Shang A1 - Janssen, Jan A1 - Li, Calvin A1 - Foster, Ian A1 - Blaiszik, Ben T1 - 32 examples of LLM applications in materials science and chemistry: towards automation, assistants, agents, and accelerated scientific discovery N2 - Large Language Models (LLMs) are reshaping many aspects of materials science and chemistry research, enabling advances in molecular property prediction, materials design, scientific automation, knowledge extraction, and more. Recent developments demonstrate that the latest class of models are able to integrate structured and unstructured data, assist in hypothesis generation, and streamline research workflows. To explore the frontier of LLM capabilities across the research lifecycle, we review applications of LLMs through 34 total projects developed during the second annual Large Language Model Hackathon for Applications in Materials Science and Chemistry, a global hybrid event. These projects spanned seven key research areas: (1) molecular and material property prediction, (2) molecular and material design, (3) automation and novel interfaces, (4) scientific communication and education, (5) research data management and automation, (6) hypothesis generation and evaluation, and (7) knowledge extraction and reasoning from the scientific literature. Collectively, these applications illustrate how LLMs serve as versatile predictive models, platforms for rapid prototyping of domain-specific tools, and much more. In particular, improvements in both open source and proprietary LLM performance through the addition of reasoning, additional training data, and new techniques have expanded effectiveness, particularly in low-data environments and interdisciplinary research. As LLMs continue to improve, their integration into scientific workflows presents both new opportunities and new challenges, requiring ongoing exploration, continued refinement, and further research to address reliability, interpretability, and reproducibility. KW - Large Language Models KW - Machine Learning KW - Materials Design KW - Bonding Analysis KW - Phonons KW - Thermal properties PY - 2025 DO - https://doi.org/10.1088/2632-2153/ae011a SN - 2632-2153 VL - 6 IS - 3 SP - 1 EP - 34 PB - IOP Publishing AN - OPUS4-64019 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Grandel, Jonas T1 - Fine-tuning universal interatomic potentials for phonon properties without catastrophic forgetting N2 - Accurate phonon predictions are essential for evaluating material stability and thermal behavior. Traditional DFT-based methods are computationally intensive, driving interest in faster, machine-learning-based alternatives. In this work, we fine-tune the machine learning interatomic potential MACE-MP-0b3 (https://arxiv.org/abs/2401.00096) to improve the prediction of harmonic phonons and thermal properties. A key challenge is the need for highly accurate force calculations. While fine-tuning can enhance phonon accuracy, it may reduce generalizability to other properties. Despite this, our study shows that fine-tuned models can accurately predict phonon properties (including renormalization of imaginary phonons) at the same time as volume-dependent properties like the bulk modulus. We fine-tune the base model using small sets of rattled structures and introduce a novel fine-tuning method, benchmarking it against existing techniques. These results suggest that refined MACE-based universal potentials are a promising path for efficient and accurate phonon modeling. All implementations are included in the equitrain training code (https://github.com/BAMeScience/equitrain). T2 - Psi-k Konferenz CY - Lausanne, France DA - 25.08.2025 KW - Machine Learning KW - MACE KW - Phonon KW - Thermal properites PY - 2025 AN - OPUS4-64010 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Liu, Yuanbin A1 - Morrow, Joe D. A1 - Ertural, Christina A1 - Fragapane, Natascia L. A1 - Gardner, John L. A. A1 - Naik, Aakash A. A1 - Zhou, Yuxing A1 - George, Janine A1 - Deringer, Volker L. T1 - An automated framework for exploring and learning potential-energy surfaces N2 - Machine learning has become ubiquitous in materials modelling and now routinely enables large-scale atomistic simulations with quantum-mechanical accuracy. However, developing machine-learned interatomic potentials requires high-quality training data, and the manual generation and curation of such data can be a major bottleneck. Here, we introduce an automated framework for the exploration and fitting of potential-energy surfaces, implemented in an openly available software package that we call (‘automatic potential-landscape explorer’). We discuss design choices, particularly the interoperability with existing software architectures, and the ability for the end user to easily use the computational workflows provided. We show wide-ranging capability demonstrations: for the titanium–oxygen system, SiO2, crystalline and liquid water, as well as phase-change memory materials. More generally, our study illustrates how automation can speed up atomistic machine learning in computational materials science. KW - Automation KW - Machine Learning KW - Machine learning potentials KW - Amorphous materials KW - High-throughput KW - Ab initio KW - Materials property prediction PY - 2025 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-639882 DO - https://doi.org/10.1038/s41467-025-62510-6 SN - 2041-1723 VL - 16 IS - 1 SP - 1 EP - 12 PB - Springer Science and Business Media LLC AN - OPUS4-63988 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - INPR A1 - Ueltzen, Katharina A1 - Naik, Aakash A. A1 - Ertural, Christina A1 - Benner, Philipp A1 - George, Janine T1 - Can simple exchange heuristics guide us in predicting magnetic properties of solids? N2 - A popular heuristic derived from the Kanamori-Goodenough-Anderson rules of superexchange connects bond angles and magnetism in certain transition metal compounds. We evaluate the fulfillment of this heuristic on databases of magnetic structures. The heuristic is partly satisfied in magnetic structures from experiment, and exceptions can be rationalized. We further show that incorporating this heuristic into machine learning models for magnetism improves the prediction. KW - Magnetism KW - Transition Metal Compounds KW - Materials Design KW - Sustainable Materials Design KW - Machine Learning KW - Materials Understanding PY - 2025 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-639290 DO - https://doi.org/10.26434/chemrxiv-2025-xj84d VL - Version 1 SP - 1 EP - 43 AN - OPUS4-63929 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - VIDEO A1 - George, Janine T1 - Materials design using chemical heuristics, workflows, and machine learning N2 - Bonds and local atomic environments are key descriptors of material properties, used to establish design rules and heuristics, and serve as descriptors in machine-learned interatomic potentials and the general machine learning of material properties. Software implementations such as ChemEnv and LobsterEnv identify local atomic environments based on geometrical characteristics and quantum-chemical bonding analysis (here using Crystal Orbital Hamilton Populations as computed with LOBSTER). Fully automated workflows and analysis tools now enable large-scale quantum-chemical bonding analysis. The first part of the lecture will demonstrate how these tools help develop new machine-learning models and intuitive understandings of material properties. New universal machine-learned interatomic potentials, such as MACE-MP-0, have been developed. The second part of the lecture will showcase how these potentials, combined with DFT, can accelerate research. It will focus on the interplay between DFT and machine-learned interatomic potentials, presenting automated workflows for training, fine-tuning, and benchmarking these potentials, implemented in our software autoplex. Additionally, it will show how to train new interatomic potentials from scratch by exploring potential energy surfaces, with the potential to enhance current universal machine-learned potentials. The lecture will also discuss the trend toward automation in computational materials science and our recent contributions. T2 - FAIRmat Seminar CY - Berlin, Germany DA - 05.06.2025 KW - Materials Design KW - Machine Learning KW - Chemical Bonding KW - Batteries KW - Amorphous Materials KW - Workflows KW - Machine Learned Interatomic Potentials KW - Phonons KW - Magnetism KW - Synthesizability PY - 2025 UR - https://www.youtube.com/watch?v=Sfco48s1fpU PB - YouTube, LLC CY - San Bruno, CA, USA AN - OPUS4-63744 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - INPR A1 - Zimmermann, Yoel A1 - Bazgir, Adib A1 - Al-Feghali, Alexander A1 - Ansari, Mehrad A1 - Bocarsly, Joshua A1 - Brinson, L. Catherine A1 - Chiang, Yuan A1 - Circi, Defne A1 - Chiu, Min-Hsueh A1 - Daelman, Nathan A1 - Evans, Matthew L. A1 - Gangan, Abhijeet S. A1 - George, Janine A1 - Harb, Hassan A1 - Khalighinejad, Ghazal A1 - Khan, Sartaaj Takrim A1 - Klawohn, Sascha A1 - Lederbauer, Magdalena A1 - Mahjoubi, Soroush A1 - Mohr, Bernadette A1 - Moosavi, Seyed Mohamad A1 - Naik, Aakash A1 - Ozhan, Aleyna Beste A1 - Plessers, Dieter A1 - Roy, Aritra A1 - Schöppach, Fabian A1 - Schwaller, Philippe A1 - Terboven, Carla A1 - Ueltzen, Katharina A1 - Wu, Yue A1 - Zhu, Shang A1 - Janssen, Jan A1 - Li, Calvin A1 - Foster, Ian A1 - Blaiszik, Ben T1 - 34 Examples of LLM Applications in Materials Science and Chemistry: Towards Automation, Assistants, Agents, and Accelerated Scientific Discovery N2 - Large Language Models (LLMs) are reshaping many aspects of materials science and chemistry research, enabling advances in molecular property prediction, materials design, scientific automation, knowledge extraction, and more. Recent developments demonstrate that the latest class of models are able to integrate structured and unstructured data, assist in hypothesis generation, and streamline research workflows. To explore the frontier of LLM capabilities across the research lifecycle, we review applications of LLMs through 34 total projects developed during the second annual Large Language Model Hackathon for Applications in Materials Science and Chemistry, a global hybrid event. These projects spanned seven key research areas: (1) molecular and material property prediction, (2) molecular and material design, (3) automation and novel interfaces, (4) scientific communication and education, (5) research data management and automation, (6) hypothesis generation and evaluation, and (7) knowledge extraction and reasoning from the scientific literature. Collectively, these applications illustrate how LLMs serve as versatile predictive models, platforms for rapid prototyping of domain-specific tools, and much more. In particular, improvements in both open source and proprietary LLM performance through the addition of reasoning, additional training data, and new techniques have expanded effectiveness, particularly in low-data environments and interdisciplinary research. As LLMs continue to improve, their integration into scientific workflows presents both new opportunities and new challenges, requiring ongoing exploration, continued refinement, and further research to address reliability, interpretability, and reproducibility. KW - Automation KW - LLM KW - Machine Learning KW - Agent KW - Bonding Analysis KW - Materials Searches KW - Finetuning PY - 2025 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-631720 DO - https://doi.org/10.48550/arXiv.2505.03049 SP - 1 EP - 33 AN - OPUS4-63172 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - Data-Driven Materials Design N2 - Implementations and algorithms, such as ChemEnv and LobsterEnv, for identifying local atomic environments based on geometrical characteristics and quantum-chemical bonding analysis are now available. Fully automatic workflows and analysis tools have been developed to use quantumchemical bonding analysis on a large scale. The first part of the lecture will demonstrate how our tools, which assess local atomic environments and perform automatic bonding analysis, help develop new machine-learning models and a new intuitive understanding of material properties. Many new universal machine-learned interatomic potentials, such as MACE-MP-0, have been developed. The second part of the lecture will showcase how these potentials, in combination with DFT, could significantly accelerate our research. The focus will be on the interplay of DFT and machine-learned interatomic potentials, presenting new fully automated workflows for training, finetuning, and implemented benchmarking these potentials, in our software autoplex (https://github.com/autoatml/autoplex). Additionally, I will show how to train new interatomic potentials from scratch by exploring potential energy surfaces extensively, offering a method to enhance current universal machine-learned potentials. Beyond this, the general trend toward automation in computational materials science and some of our recent contributions will be discussed. T2 - AI4AM2025 CY - Donostia, Spain DA - 08.04.2025 KW - Automation KW - Machine Learning KW - Chemically Complex Materials KW - Thermal Conductivity KW - Bonding Analysis KW - Workflows KW - Machine Learned Interatomic Potentials PY - 2025 AN - OPUS4-62950 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - George, Janine T1 - Phonon workflow and applications N2 - This talk introduced the participants to phonons and how they are typically computed. Then, I introduced the participants to the harmonic phonon, the Grüneisen, and the quasi-harmonic workflow that allows the computation of properties related to phonons. I also had detailed examples of how these workflows can be used in practice. Beyond this, I showed how these workflows can be used to automatically benchmark interatomic potentials and use them for the development of such potentials. T2 - "Automated ab initio workflows with Jobflow and Atomate2" CECAM Flagship school CY - Lausanne, Switzerland DA - 17.03.2025 KW - Automation KW - Phonons KW - Thermal Conductivity KW - Machine Learning KW - Software Development KW - Machine Learned interatomic Potentials PY - 2025 AN - OPUS4-62949 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Benner, Philipp A1 - Grandel, Jonas T1 - A shortcut towards phonon predictions N2 - Phonon calculations with ab-initio methods are computationally expensive. The use of universal machine learning models reduces the cost, but raises concerns about prediction quality. Fine-tuning with only a few structures, improves predictions of phonons, thermal properties and especially diffusive thermal conductivity, while reducing computational cost by a factor of 10 in average compared to DFT methods. T2 - DPG Frühjahrstagung CY - Regensburg, Germany DA - 16.03.2025 KW - Machine Learning KW - MACE KW - Phonons PY - 2025 AN - OPUS4-62770 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - INPR A1 - George, Janine A1 - Naik, Aakash A1 - Ueltzen, Katharina T1 - Reflections from the 2024 Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry N2 - Here, we present the outcomes from the second Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry, which engaged participants across global hybrid locations, resulting in 34 team submissions. The submissions spanned seven key application areas and demonstrated the diverse utility of LLMs for applications in (1) molecular and material property prediction; (2) molecular and material design; (3) automation and novel interfaces; (4) scientific communication and education; (5) research data management and automation; (6) hypothesis generation and evaluation; and (7) knowledge extraction and reasoning from scientific literature. Each team submission is presented in a summary table with links to the code and as brief papers in the appendix. Beyond team results, we discuss the hackathon event and its hybrid format, which included physical hubs in Toronto, Montreal, San Francisco, Berlin, Lausanne, and Tokyo, alongside a global online hub to enable local and virtual collaboration. Overall, the event highlighted significant improvements in LLM capabilities since the previous year's hackathon, suggesting continued expansion of LLMs for applications in materials science and chemistry research. These outcomes demonstrate the dual utility of LLMs as both multipurpose models for diverse machine learning tasks and platforms for rapid prototyping custom applications in scientific research. KW - Large Language Models KW - Materials Design KW - Materials Properties KW - Phonons KW - Bonding Analysis KW - Machine Learning PY - 2024 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-622113 DO - https://doi.org/10.48550/arXiv.2411.15221 SP - 1 EP - 98 AN - OPUS4-62211 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - INPR A1 - Amariamir, Sasan A1 - George, Janine A1 - Benner, Philipp T1 - SynCoTrain: A Dual Classifier PU-learning Framework for Synthesizability Prediction N2 - Material discovery is a cornerstone of modern science, driving advancements in diverse disciplines from biomedical technology to climate solutions. Predicting synthesizability, a critical factor in realizing novel materials, remains a complex challenge due to the limitations of traditional heuristics and thermodynamic proxies. While stability metrics such as formation energy offer partial insights, they fail to account for kinetic factors and technological constraints that influence synthesis outcomes. These challenges are further compounded by the scarcity of negative data, as failed synthesis attempts are often unpublished or context-specific. We present SynCoTrain, a semi-supervised machine learning model designed to predict the synthesizability of materials. SynCoTrain employs a co-training framework leveraging two complementary graph convolutional neural networks: SchNet and ALIGNN. By iteratively exchanging predictions between classifiers, SynCoTrain mitigates model bias and enhances generalizability. Our approach uses Positive and Unlabeled (PU) Learning to address the absence of explicit negative data, iteratively refining predictions through collaborative learning. The model demonstrates robust performance, achieving high recall on internal and leave-out test sets. By focusing on oxide crystals, a well-characterized material family with extensive experimental data, we establish SynCoTrain as a reliable tool for predicting synthesizability while balancing dataset variability and computational efficiency. This work highlights the potential of co-training to advance high-throughput materials discovery and generative research, offering a scalable solution to the challenge of synthesizability prediction. KW - Materials Design KW - Materials Discovery KW - Synthesizability KW - Machine Learning KW - Cotraining KW - New Materials KW - Materials Acceleration Platforms PY - 2024 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-622104 DO - https://doi.org/10.48550/arXiv.2411.12011 SP - 1 EP - 39 AN - OPUS4-62210 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER -