Using alchemical compression and machine learning to describe high-entropy alloys
A new method developed by the NCCR MARVEL laboratory of Michele Ceriotti at EPFL allows modelling alloys containing up to 25 transition metals, matching a remarkable accuracy with a manageable requirement of data and computational resources. The group validated the model and used it for computational experiments on three representative alloy compositions. In the future the model will be expanded to predict the catalytic behaviour of the surfaces of high-entropy alloys.
Crucial role of inter-site Hubbard interactions for the correct energetics of spinel Li-ion cathode materials
NCCR MARVEL researchers have applied DFT with extended Hubbard functionals (DFT+U+V) to the study of two candidate cathode materials belonging to the class of lithium-manganese-oxide spinels. They used a new approach to determine the Hubbard U and V parameters entirely from first principles and found that the method accurately predicts the geometrical properties, oxidation states, band gaps, and voltages of the two materials. The article was included in the "2023 PCCP HOT Articles" collection.
Temperature dependence of energy band gap suggests ZrTe5 is a weak topological insulator
Researchers including NCCR MARVEL’s Professor Ana Akrap at the Department of Physics at the University of Fribourg used Landau-level spectroscopy to determine the temperature dependence of the energy band gap in zirconium pentatelluride (ZrTe5). They found that the band gap is non-zero at low temperatures and increases monotonically when the temperature is raised, adding to evidence that ZrTe5 is a weak topological insulator.
Graph neural network parametrized potentials describe intermolecular interactions
An ETHZ team led by Prof. Sereina Riniker, Associate Professor of Computational Chemistry at the Department of Chemistry and Applied Biosciences, has developed an alternative strategy for parametrizing intermolecular interactions. Described in the paper “Regularized by Physics: Graph Neural Network Parametrized Potentials for the Description of Intermolecular Interactions,” recently published in the Journal of Chemical Theory and Computation, the approach accelerates and simplifies the parametrization process of classical force fields and can take advantage of large data sets. Used in combination with machine learning-based techniques, the model allows researchers to take advantage of the best of both, and access a universal optimization toolkit combined with robust and physically constrained models.
Solids that are also liquids: elastic tensors of superionic materials
NCCR MARVEL researchers have applied for the first time first-principles molecular dynamics with the Parrinello-Rahman method to characterize the elastic properties of superionic conductors. Their study highlights the importance of the quasi-liquid dynamics of the Li ions in the elastic response, and showcases a significant softening of the predicted moduli compared to previous static approaches. The approach, complemented with the computation of the statistical errors, provides accurate information and reference results for the elastic constants and moduli of three benchmark oxide and sulﬁde solid-state electrolytes, and paves the way for a better understanding of the mechanical properties of the materials that may be used in the next generation of solid-state-battery technologies.
Theory-guided design identifies single-phase high entropy alloys with appealing properties
Prof. William Curtin, leader of NCCR MARVEL’s Pillar 1, Design and Discovery of Novel Materials and member of the executive committee and colleagues including Dr. Christian Leinenbach, Head of the Advanced Processing & Additive Manufacturing of Metals at Empa have used theories and expanded thermodynamic tools to sift through a huge materials space to identify high entropy alloys that satisfy a range of necessary performance properties. They validated the approach against data from recent literature, identified a promising new material, fabricated it at Empa, and then confirmed several predicted properties through experiment. The method is general and can be adapted to to other performance requirements, making it a reliable and efficient means of discovery for the next generation of high-temperature alloys.
Ceriotti’s perspective on integrated ML models for materials published in MRS Bulletin
NCCR MARVEL’s Michele Ceriotti, EPFL professor and head of the School of Engineering’s Laboratory of Computational Science and Modelling, has published a perspective, “Beyond potentials: Integrated machine learning models for materials,” in the October edition of the MSR Bulletin.
DFT+U+V for accurate electronic properties of olivine-type Li-ion cathode materials
Understanding at the atomistic level how the properties of transition-metal elements lead to efficient electrochemical processes is critically important in the development of new cathode materials for lithium-ion batteries. While density-functional theory (DFT) calculations with local and semilocal exchange-correlation functionals can play an important role in producing first-principles predictions for these materials, they can also produce unsatisfactory results because of self-interaction errors. In the paper “Accurate Electronic Properties and Intercalation Voltages of Olivine-Type Li-Ion Cathode Materials from Extended Hubbard Functionals,” recently published in Physical Review X Energy, NCCR MARVEL researchers and colleagues carried out a comparative study of four electronic-structure methods for selected olivine-type cathode materials. They found that the DFT+U+V method clearly outperforms the others and is able to describe the interactions accurately, opening the door for the study of more complex cathode materials as well as for a reliable exploration of the chemical space of compounds for Li-ion batteries.
New organocatalysts database to drive reaction optimization methods in organic synthesis
Simone Gallarati and a team led by Professor Clemence Corminboeuf, head of the Laboratory for Computational Molecular Design at EPFL, have created a dataset containing thousands of organic molecules that have been mined from the literature and from the Cambridge Crystallographic Database, and enriched with species generated in a combinatorial fashion by recombining molecular building blocks. The result is a map that could help researchers navigate organocatalyst space and enable informed catalyst design. It’s also the starting point for a multitude of fragment-based reaction optimization methods.
Data-driven “sorting hat” ranks synthesizability of hypothetical zeolites, suggests likely chemical compositions
Zeolites are nanoporous frameworks that can be applied in a number of industrial processes, particularly in separation and catalysis. While much effort has been put into identifying and synthesizing new zeolite structures, success has been largely theoretical. While massive databases of hypothetical zeolites have been generated, containing millions of new framework structures, none has been made in the lab. In the paper “Ranking the synthesizability of hypothetical zeolites with the sorting hat,” NCCR MARVEL’s Michele Ceriotti, EPFL professor and head of the School of Engineering’s Laboratory of Computational Science and Modelling, and colleagues present a data-driven procedure for distinguishing known structures from hypothetical ones and categorizing them into compositional classes. The approach helps identify promising synthetic candidates and suggests likely chemical compositions, providing a sort of recipe for materials chemists.
OSSCAR supports teaching, fosters computational thinking with interactive approach
A paper describing the Open Software Services for Classrooms and Research (OSSCAR) platform has recently been published in the journal Computer Physics Communications. Developed through a collaboration between NCCR MARVEL and the Centre Européen de Calcul Atomique et Moléculaire (CECAM), OSSCAR provides an open collaborative environment for developing and accessing educational resources through web applications. Associated tools are easy to use, and create a uniform and open environment that can be used by a large academic community and are meant to facilitate learning and help people avoid duplicating efforts in the creation of teaching material. Contributions to expand the educational content of the OSSCAR project are welcome.
Polarons free from many-body self-interaction in density functional theory
Polarons can affect numerous phenomena in a material but have been difficult to model correctly. In the paper “Many-Body Self-Interaction and Polarons,” recently published jointly in Physical Review Letters and in Physical Review B as an Editor’s Suggestion, researchers Stefano Falletta and Professor Alfredo Pasquarello of the Chair of Atomic Scale Simulation at EPFL advance the conceptual understanding of the self-interaction problem in density functional theory, paving the way to efficient calculations of polarons in large systems, in systematic studies involving large sets of materials, and in molecular dynamics evolving over long time periods.