Extracting an Empirical Intermetallic Hydride Design Principle from Limited Data via Interpretable Machine Learning

MH-PCTpro: A machine learning model for rapid prediction of pressure-composition-temperature (PCT) isotherms

We present a machine-learning powered Metal Hydride's Pressure-Composition-Temperature isotherm Predictor (MH-PCTpro) for metal compositions. To train the MH-PCTpro, an experimental database of PCT isotherms is built from published literature. The database comprises over 14,000 data points extracted from 237 PCT isotherms representing 138 distinct compositions. The dataset encompasses more than 25 elements and spans a broad spectrum of absorption temperatures (263-653 K) and hydrogen pressures (0.001-40 MPa). The model is validated on a wide range of alloy families and its predictions are consistent with experimental results. The model also captures temperature-dependent variations in plateau pressure, enabling determination of enthalpy and entropy of hydride formation through Van't Hoff plots. Hence, MH-PCTpro can be used as an ML tool for guiding PCT experiments, offering PCT isotherm predictions and valuable thermodynamic insights into materials suitable for solid-state hydrogen storage.

Tuning TiFe1-x Ni x Hydride Thermodynamics through Compositional Tailoring

In this study, we investigate how structural modifications induced by Fe substitution with Ni in the TiFe intermetallic alloy affect the thermodynamics of hydride formation and decomposition. The primary goal of substituting Fe with Ni was to reduce the plateau pressure of TiFe, a crucial parameter for reversible solid-state hydrogen storage applications under near-ambient conditions (below 150 °C and 50 bar). Alloy compositions TiFe1-x Ni x with x ≤ 0.30 were synthesized by arc melting. The structural and morphological properties were characterized using powder X-ray diffraction and scanning electron microscopy with energy-dispersive X-ray spectroscopy. The thermodynamic properties were investigated through volumetric measurements using a Sieverts' apparatus and calorimetric analysis with a high-pressure differential scanning calorimeter. We show that Ni incorporation effectively lowers the plateau pressure, stabilizing the hydride thermodynamics due to a more negative enthalpy of hydride formation. Moreover, the entropy of hydride formation increases with the Ni content, resulting in a linear correlation between the enthalpy and entropy values determined at different compositions. The enthalpy-entropy compensation effect was analyzed to determine whether it arises from statistical artifacts or is genuine to the system, as our findings suggest.

Machine Learning in Solid-State Hydrogen Storage Materials: Challenges and Perspectives

Machine learning (ML) has emerged as a pioneering tool in advancing the research application of high-performance solid-state hydrogen storage materials (HSMs). This review summarizes the state-of-the-art research of ML in resolving crucial issues such as low hydrogen storage capacity and unfavorable de-/hydrogenation cycling conditions. First, the datasets, feature descriptors, and prevalent ML models tailored for HSMs are described. Specific examples include the successful application of ML in titanium-based, rare-earth-based, solid solution, magnesium-based, and complex HSMs, showcasing its role in exploiting composition-structure-property relationships and designing novel HSMs for specific applications. One of the representative ML works is the single-phase Ti-based HSM with superior cost-effective and comprehensive properties, tailored to fuel cell hydrogen feeding system at ambient temperature and pressure through high-throughput composition-performance scanning. More importantly, this review also identifies and critically analyzes the key challenges faced by ML in this domain, including poor data quality and availability, and the balance between model interpretability and accuracy, together with feasible countermeasures suggested to ameliorate these problems. In summary, this work outlines a roadmap for enhancing ML's utilization in solid-state hydrogen storage research, promoting more efficient and sustainable energy storage solutions.

Assessment of fine-tuned large language models for real-world chemistry and material science applications

The current generation of large language models (LLMs) has limited chemical knowledge. Recently, it has been shown that these LLMs can learn and predict chemical properties through fine-tuning. Using natural language to train machine learning models opens doors to a wider chemical audience, as field-specific featurization techniques can be omitted. In this work, we explore the potential and limitations of this approach. We studied the performance of fine-tuning three open-source LLMs (GPT-J-6B, Llama-3.1-8B, and Mistral-7B) for a range of different chemical questions. We benchmark their performances against "traditional" machine learning models and find that, in most cases, the fine-tuning approach is superior for a simple classification problem. Depending on the size of the dataset and the type of questions, we also successfully address more sophisticated problems. The most important conclusions of this work are that, for all datasets considered, their conversion into an LLM fine-tuning training set is straightforward and that fine-tuning with even relatively small datasets leads to predictive models. These results suggest that the systematic use of LLMs to guide experiments and simulations will be a powerful technique in any research study, significantly reducing unnecessary experiments or computations.

Elucidating Primary Degradation Mechanisms in High-Cycling-Capacity, Compositionally Tunable High-Entropy Hydrides

The hydrogen sorption properties of single-phase bcc (TiVNb)100-xCrx alloys (x = 0-35) are reported. All alloys absorb hydrogen quickly at 25 °C, forming fcc hydrides with storage capacity depending on the Cr content. A thermodynamic destabilization of the fcc hydride is observed with increasing Cr concentration, which agrees well with previous compositional machine learning models for metal hydride thermodynamics. The steric effect or repulsive interactions between Cr-H might be responsible for this behavior. The cycling performances of the TiVNbCr alloy show an initial decrease in capacity, which cannot be explained by a structural change. Pair distribution function analysis of the total X-ray scattering on the first and last cycled hydrides demonstrated an average random fcc structure without lattice distortion at short-range order. If the as-cast alloy contains a very low density of defects, the first hydrogen absorption introduces dislocations and vacancies that cumulate into small vacancy clusters, as revealed by positron annihilation spectroscopy. Finally, the main reason for the capacity drop seems to be due to dislocations formed during cycling, while the presence of vacancy clusters might be related to the lattice relaxation. Having identified the major contribution to the capacity loss, compositional modifications to the TiVNbCr system can now be explored that minimize defect formation and maximize material cycling performance.

Machine Learning in the Development of Adsorbents for Clean Energy Application and Greenhouse Gas Capture

Addressing climate change challenges by reducing greenhouse gas levels requires innovative adsorbent materials for clean energy applications. Recent progress in machine learning has stimulated technological breakthroughs in the discovery, design, and deployment of materials with potential for high-performance and low-cost clean energy applications. This review summarizes basic machine learning methods-data collection, featurization, model generation, and model evaluation-and reviews their use in the development of robust adsorbent materials. Key case studies are provided where these methods are used to accelerate adsorbent materials design and discovery, optimize synthesis conditions, and understand complex feature-property relationships. The review provides a concise resource for researchers wishing to use machine learning methods to rapidly develop effective adsorbent materials with a positive impact on the environment.

Recent Development in Nanoconfined Hydrides for Energy Storage

Hydrogen is the ultimate vector for a carbon-free, sustainable green-energy. While being the most promising candidate to serve this purpose, hydrogen inherits a series of characteristics making it particularly difficult to handle, store, transport and use in a safe manner. The researchers' attention has thus shifted to storing hydrogen in its more manageable forms: the light metal hydrides and related derivatives (ammonia-borane, tetrahydridoborates/borohydrides, tetrahydridoaluminates/alanates or reactive hydride composites). Even then, the thermodynamic and kinetic behavior faces either too high energy barriers or sluggish kinetics (or both), and an efficient tool to overcome these issues is through nanoconfinement. Nanoconfined energy storage materials are the current state-of-the-art approach regarding hydrogen storage field, and the current review aims to summarize the most recent progress in this intriguing field. The latest reviews concerning H2 production and storage are discussed, and the shift from bulk to nanomaterials is described in the context of physical and chemical aspects of nanoconfinement effects in the obtained nanocomposites. The types of hosts used for hydrogen materials are divided in classes of substances, the mean of hydride inclusion in said hosts and the classes of hydrogen storage materials are presented with their most recent trends and future prospects.

Predicting hydrogen storage in MOFs via machine learning

The H2 capacities of a diverse set of 918,734 metal-organic frameworks (MOFs) sourced from 19 databases is predicted via machine learning (ML). Using only 7 structural features as input, ML identifies 8,282 MOFs with the potential to exceed the capacities of state-of-the-art materials. The identified MOFs are predominantly hypothetical compounds having low densities (<0.31 g cm-3) in combination with high surface areas (>5,300 m2 g-1), void fractions (∼0.90), and pore volumes (>3.3 cm3 g-1). The relative importance of the input features are characterized, and dependencies on the ML algorithm and training set size are quantified. The most important features for predicting H2 uptake are pore volume (for gravimetric capacity) and void fraction (for volumetric capacity). The ML models are available on the web, allowing for rapid and accurate predictions of the hydrogen capacities of MOFs from limited structural data; the simplest models require only a single crystallographic feature.

Artificial Intelligence Application in Solid State Mg-Based Hydrogen Energy Storage

The use of Mg-based compounds in solid-state hydrogen energy storage has a very high prospect due to its high potential, low-cost, and ease of availability. Today, solid-state hydrogen storage science is concerned with understanding the material behavior of different compositions and structure when interacting with hydrogen. Finding a suitable material has remained an elusive idea, and therefore, this review summarizes works by various groups, the milestones they have achieved, and the roadmap to be taken on the study of hydrogen storage using low-cost magnesium composites. Mg-based compounds are further examined from the perspective of artificial intelligence studies, which helps to improve prediction of their properties and hydrogen storage performance. There exist several techniques to improve the performance of Mg-based compounds: microstructure modification, use of catalytic additives, and composition regulation. Microstructure modification is usually achieved by employing different synthetic techniques like severe plastic deformation, high energy ball milling, and cold rolling, among others. These synthetic approaches are discussed herein. In this review, a discussion of key parameters and operating conditions are highlighted in a view to finding high storage capacity and faster kinetics. Furthermore, recent approaches like machine learning have found application in guiding the experimental design. Hence, this review paper also explores how machine learning techniques have been utilized to fasten the materials research. It is however noted that this study is not exhaustive in itself.